Integration of Renewable and Distributed Energy Resources in Power Systems
Integration of Renewable and Distributed Energy Resources in Power Systems
Renewable and
Distributed
Energy Resources
in Power Systems
Edited by
www.mdpi.com/journal/energies
Integration of Renewable and
Distributed Energy Resources
in Power Systems
Integration of Renewable and
Distributed Energy Resources
in Power Systems
Editors
Tomás Gómez San Román
José Pablo Chaves-Ávila
MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin
Editors
Tomás Gómez San Román José Pablo Chaves-Ávila
Comillas Pontifical University Comillas Pontifical University
Spain Spain
Editorial Office
MDPI
St. Alban-Anlage 66
4052 Basel, Switzerland
This is a reprint of articles from the Special Issue published online in the open access journal
Energies (ISSN 1996-1073) (available at: https://www.mdpi.com/journal/energies/special issues/
integration energy resources).
For citation purposes, cite each article independently as indicated on the article page online and as
indicated below:
LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. Journal Name Year, Article Number,
Page Range.
c 2020 by the authors. Articles in this book are Open Access and distributed under the Creative
Commons Attribution (CC BY) license, which allows users to download, copy and build upon
published articles, as long as the author and publisher are properly credited, which ensures maximum
dissemination and a wider impact of our publications.
The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons
license CC BY-NC-ND.
Contents
Gregorio Fernández, Noemi Galan, Daniel Marquina, Diego Martı́nez, Alberto Sanchez,
Pablo López, Hans Bludszuweit and Jorge Rueda
Photovoltaic Generation Impact Analysis in Low Voltage Distribution Grids
Reprinted from: Energies 2020, 13, 4347, doi:10.3390/en13174347 . . . . . . . . . . . . . . . . . . . 1
Kostas Hatalis, Chengbo Zhao, Parv Venkitasubramaniam, Larry Snyder, Shalinee Kishore,
Rick S. Blum
Modeling and Detection of Future Cyber-Enabled DSM Data Attacks
Reprinted from: Energies 2020, 13, 4331, doi:10.3390/en13174331 . . . . . . . . . . . . . . . . . . . 29
Ingo Liere-Netheler, Frank Schuldt, Karsten von Maydell and Carsten Agert
Simulation of Incidental Distributed Generation Curtailment to Maximize the Integration of
Renewable Energy Generation in Power Systems
Reprinted from: Energies 2020, 13, 4173, doi:10.3390/en13164173 . . . . . . . . . . . . . . . . . . . 57
Fernando Postigo Marcos, Carlos Mateo Domingo, Tomás Gómez San Román and
Rafael Cossent Arı́n
Location and Sizing of Micro-Grids to Improve Continuity of Supply in Radial
Distribution Networks
Reprinted from: Energies 2020, 13, 3495, doi:10.3390/en13133495 . . . . . . . . . . . . . . . . . . . 97
Kaan Ozgun
Towards a Sustainability Assessment Model for Urban Public Space Renewable
Energy Infrastructure
Reprinted from: Energies 2020, 13, 3428, doi:10.3390/en13133428 . . . . . . . . . . . . . . . . . . . 119
Morsy Nour, José Pablo Chaves-Ávila, Gaber Magdy and Álvaro Sánchez-Miralles
Review of Positive and Negative Impacts of Electric Vehicles Charging on Electric
Power Systems
Reprinted from: Energies 2020, 13, 4675, doi:10.3390/en13184675 . . . . . . . . . . . . . . . . . . . 183
v
About the Editors
Tomás Gómez San Román is Professor of Electrical Engineering at the Engineering School
of Universidad Pontificia Comillas in Madrid, Spain. Currently, he is the Director of Instituto de
Investigación Tecnológica (IIT) at Comillas. He obtained the Degree of Doctor Ingeniero Industrial
from Universidad Politécnica, Madrid, in 1989, and the Degree of Ingeniero Industrial in Electrical
Engineering from Comillas in 1982. He joined IIT in 1984. From 1994 to 2000, he was also
Director of IIT, and from 2000 to 2002, the Vice-Rector of Research, Development, and Innovation
of Comillas. Prof. Gómez has a wealth of experience in industry joint research projects in the field
of Electric Energy Systems in collaboration with Spanish, Latin American, and European institutions.
He has served as project manager and/or principal investigator in more than 80 research projects.
His areas of interest are in operation and planning of transmission and distribution systems, power
quality assessment and regulation, and economic and regulatory issues in the electrical power
sector. He has published more than 100 articles in different specialized journals, such as IEEE
PES Transactions, in addition to conference proceedings and has co-authored the book “Electricity
Economics: Regulation and Deregulation” in Wiley-IEEE Press. He is a Senior Member of IEEE. He is
President of the Council of the Power System Computation Conference and serves or has served
on the technical committees of other conferences such as Probabilistic Methods Applied to Power
Systems and IEEE Power Tech. He has been Visiting Researcher at the Energy Analysis Department of
the Lawrence Berkeley National Laboratory in California. From May 2011 to October 2013, he served
as commissioner at the Spanish Energy Commission (CNE).
José Pablo Chaves-Ávila is Research Professor at the Institute for Research in Technology (IIT)
at the Engineering School (ICAI) of the Comillas Pontifical University. Since October 2016, he has
been leading and operating agent of the Academy of the International Smart Grids Action Network.
From August 2020, he was appointed as member of the Expert Advisory List of the Comisión Regional
de Interconexión Eléctrica (CRIE). In 2008, he obtained his bachelor’s in Economics from University
of Costa Rica (Costa Rica). As part of the Erasmus Mundus Master in Economics and Management
of Network Industries, he holds a master’s in the Electric Power Industry from ICAI and master’s
in Digital Economics and Network Industries from Paris Sud-11 University (Paris, France). In 2014,
he obtained the Erasmus Mundus Joint Doctorate on Sustainable Energy Technologies and Strategies
Degree at Delft University of Technology (The Netherlands), a joint program with Comillas Pontifical
University and the Royal Institute of Technology (KTH), Sweden. His areas of interests are energy
economics, integration of renewable resources and distributed energy resources in the electricity
sector, smart grids, and regulation of the electricity and gas sectors. He has more than 30 national
and international publications in journals and conference proceedings on these topics. José Pablo has
been Visiting Scholar at the European University Institute (Italy), the Lawrence Berkeley National
Laboratory (USA), and the Massachusetts Institute of Technology (MIT), USA. During his postdoc,
he participated in the Utility of the Future project, a joint project between IIT and MIT.
vii
Preface to ”Integration of Renewable and Distributed
Energy Resources in Power Systems ”
Decarbonization, digitalization, and decentralization are transforming our power systems
towards a more sustainable energy system. The integration of large amounts of renewable
generation is required to achieve an almost 100% decarbonized power system. Many of the new
renewable generators, wind and solar, are connected to distribution grids. Digitalization allows
changing the traditional paradigm of passive grid operation and planning to smart grids with
higher levels of automation and self-control, maintaining the security of supply and economic
efficiency. Communications and cybersecurity are required to interconnect millions of sensors and
equipment connected not only on the network but also on the customer side. In this Special Issue,
we have the opportunity to delve deeply into some of the advances that are required to achieve
this transformation. Microgrids are a new concept to operate, in a clever manner, thousands of new
control equipment located on customer premises, such as flexible demand, renewable generation
sources, and storage installations. In addition, microgrids is an alternative to create more resilient
systems against extreme weather conditions or massive cyberattacks. New planning techniques
for urban districts and grid development tackling uncertainty regarding load growth and flexibility
are also needed. Finally, new operational and planning criteria are explored to integrate a massive
penetration of distributed generation, electric vehicles, and controllable loads.
ix
energies
Article
Photovoltaic Generation Impact Analysis in Low
Voltage Distribution Grids
Gregorio Fernández 1, *, Noemi Galan 1, *, Daniel Marquina 1, *, Diego Martínez 1, *,
Alberto Sanchez 2 , Pablo López 2, *, Hans Bludszuweit 1, * and Jorge Rueda 2, *
1 Fundacion CIRCE, Parque Empresarial Dinamiza, Avenida Ranillas 3-D, 1st Floor, 50018 Zaragoza, Spain
2 Grupo Cuerva, C/Santa Lucia, 1 K. Churriana de la Vega, 18194 Granada, Spain; [email protected]
* Correspondence: [email protected] (G.F.); [email protected] (N.G.); [email protected] (D.M.);
[email protected] (D.M.); [email protected] (P.L.); [email protected] (H.B.);
[email protected] (J.R.)
Abstract: Due to a greater social and environmental awareness of citizens, advantageous regulations
and a favourable economic return on investment, the presence of photovoltaic (PV) installations
in distribution grids is increasing. In the future, not only a significant increase in photovoltaic
generation is expected, but also in other of the so-called distributed energy resources (DER), such as
wind generation, storage, electric vehicle charging points or manageable demands. Despite the
benefits posed by these technologies, an uncontrolled spread could create important challenges for
the power system, such as increase of energy losses or voltages out-of-limits along the grid, for
example. These issues are expected to be more pronounced in low voltage (LV) distribution networks.
This article has two main objectives: proposing a method to calculate the LV distributed photovoltaic
generation hosting capacity (HC) that minimizes system losses and evaluating different management
techniques for solar PV inverters and their effect on the hosting capacity. The HC calculation is based
on a mixture of deterministic methods using time series data and statistical ones: using real smart
meters data from customers and generating different combinations of solar PV facilities placements
and power to evaluate its effect on the grid operation.
Keywords: photovoltaics; distributed energy resources (DERs); grid impact; power quality;
low-voltage distribution network; inverter regulation
1. Introduction
The interest in renewable energies has grown considerably all over the world in recent time.
Among renewable energies, solar photovoltaics is by far the most widespread technology adopted
for distributed power generation. The global photovoltaic energy generation capacity is increasing
rapidly due to efficiency improvements and cost reductions in recent years. It is expected that by 2022,
solar PV will be the largest renewable energy source worldwide [1].
Not too many years ago, photovoltaic systems were considered suitable only for isolated systems
in remote locations, because solar PV was costly and there were serious concerns regarding negative
impacts on the grid [2]. Large cost reductions, together with improved inverter technology and new
standards and regulations have made solar PV also attractive for grid connected applications, covering
a wide range from utility-scale facilities to small, distributed installations.
PV systems can be connected to the grid at any voltage level and can be classified into three main
categories according to their size (power generation) [1]:
• Large-scale (or utility-scale) systems have an installed power higher than 100 MWp and are
typically connected to high-voltage transport grids.
• Medium power systems, from 1 MWp to 100 MWp, are connected to medium voltage
distribution grids.
• Small systems, with a generation power below from 1 MWp, are distributed along low-voltage
grids and very small systems (below 10 kWp) are typically connected to single-phase residential
systems. The number of small systems in self-consumption applications has been increasing rapidly
in recent years as a way to reduce electric energy bills for residential and commercial customers.
In fact, a new term has emerged as a consequence of this phenomenon of self-consumption:
“prosumers” [3], which indicated the important evolution of the end user from a passive consumer
to a more active producer and consumer. This paper is focused on small and very small PV systems
power level, connected always in prosumers’ facilities, and the related impacts on LV grids.
The electric power system (transmission and distribution) was designed for unidirectional power
flows, from generation to consumption points. This classical conception of the electric system limits
system flexibility and adaptability to new technologies, as distributed generation or storage in LV grids.
Consequently, new requirements and technologies become necessary to deploy in a safe manner a
massive integration of renewable energy sources in distribution grids, both medium and low voltage [4].
These new requirements and technologies are incorporated in the concept of the “smart grid”.
As explained in the previous paragraphs, PV generation integrated in AC LV grids is a popular
and accessible source of renewable energy for many potential consumers. Solar resource availability in
most areas of the world makes this option for electric generation widely adoptable. Furthermore, it can
provide benefits to the grid in terms of reduced losses or overload compensation, as well as improve
voltage control and management.
Despite the potential benefits of distributed PV, solar radiation variability is an important issue
which may lead to instability at the common connection point, with respect to voltage and frequency
variations [5]. A high penetration of distributed generators (prosumers) could create different failure
possibilities, not only in the LV grid, but also in the medium voltage (MV) grid. Intermittent renewable
energy integration presents significant technical and economic challenges. So that, an optimal strategy
to operate and control the distribution grid in presence of high photovoltaic generation levels is
essential. Main issues of solar PV distributed generation are well known and are listed here below [6,7]:
• Voltage deviations. Energy injection of distributed resources along LV lines can change voltage
profiles, generating overvoltage and undervoltage whose value can change along the day.
• Phase imbalances. In most scenarios, LV photovoltaic facilities are single-phase, generating
imbalances respect to the other phases, this phenomenon causes an increase in neutral current
and, therefore, losses increase.
• System losses. Total line losses decrease as photovoltaic penetration increases and energy generated
is consumed locally. However, if the active power injected by the PV systems to the connection
point exceeds the load demand, losses increase as not all the energy can be consumed and it is
exported to the grid.
• Reverse power flow. Electric grids have been designed to withstand unidirectional power flows,
so a power flow in the opposite direction could generate problems, affecting grid components as
transformers, especially if protection devices are not designed to deal with reverse power flows.
• Harmonic injection. Harmonic currents generate voltage harmonics in LV grids that distort the
waveform affecting connected loads. Some PV inverters can introduce harmonics into the grid,
which means losses increase in transformers that cause heating in its windings and also heating in
protection systems, leading them to malfunctions.
• Short-circuit currents. High photovoltaic penetration levels cause greater short-circuit currents,
which could lead to greater damage to the grid equipment.
To evaluate the effect of renewable generation systems, the penetration level and its limit in the
shape of hosting capacity are used. For the first term, the adopted definition in this article is the relation
between the total installed peak power and the total power contracted by consumers, kWp/kW in a
2
Energies 2020, 13, 4347
given LV grid [8]. For the second one, hosting capacity, there are many definitions as shown in [9]
or [10], but the approach used in this paper is the maximum penetration level of distributed solar PV
generation that could be installed without exceeding grid safe operating limits (thermal and voltages)
and reducing energy losses.
As it can be seen, the calculation of the hosting capacity of a grid is not trivial and is influenced
by many parameters as the grid components and its characteristics, the consumption profiles of the
customers connected, the generation resources and their location in the electrical system [9,10]. In [8]
a deep analysis of hosting capacity tools and methods is made and the next main HC assessment
techniques classification is proposed:
• Deterministic methods
Constant generation methods, that assume that the output of DERs is fixed during the
evaluated period. These techniques require small amounts of input data, are simple and
fast but the results are approximated and the scenarios tested are reduced.
Time series methods, in these techniques the constant value of DERs, the main drawback
of previous methods, is avoided using generation profiles. This approach is accurate but
need big data sets.
• Stochastic methods introduce the variability or ignorance of many factors, as DERs power or
location [11], in the calculation process. Stochastic techniques do not need large amounts of data
and can provide accurate results but can be slow and complex
• Optimization based methods. DERs are placed and sized as a result of an optimization process.
These methods are exact for the optimal case, not for others.
• Streamlined methods. In these methods the scenarios evaluated are reduced to an amount that
represent the possible grid states. These techniques have approximated results.
There are many works calculating hosting capacities for an electric grid and others that propose
and evaluate different techniques to increase this penetration level without jeopardizing the system or
increasing losses. For example, Reference [12] lists and describes current and new solutions to enable a
large-scale penetration of solar PV systems in distribution grids. The authors classify the solutions
according to the provider: DSO, prosumer and interactive solutions. Proposed solutions are: network
reinforcement, STATCOM use, prosumer storage, increase of self-consumption using tariff incentives,
curtailment of power feed-in to the grid, active power control of PV inverter depending on the grid
voltage (P(U)), reactive power control of PV inverter depending on the grid voltage or generated active
power (Q(U) and Q(P)) or demand response managed by local price signals are some examples.
In References [13–15], as in other many sources, the use of distributed storage is proposed to
increase the penetration of solar PV generation reducing its impact on the grid. To reach a widespread
adoption of this solution, storage costs need to be further reduced. In several countries, government
support is given to accelerate mass adoption with the objective of the desired cost reduction.
On the other hand, Reference [16] proposes and analyses the use of Set Voltage Regulators (SVR) to
reduce voltage fluctuations in distribution systems with high solar PV penetration levels. This solution,
although interesting, involves the renovation of expensive assets.
In Reference [17] a deep analysis of the negative effects in electric distribution grids with high
solar PV penetrations is carried out. To mitigate these negative effects, the connection of PV generators
as three-phase systems is proposed. This solution is proved to reduce losses and voltage issues, but it
is difficult to be applied in real grids as many consumers have single-phase grid connections and a
mixture of single phase and three phase solar PV generation systems is more than expected.
Reference [8] proposes the installation of single-phase solar PV generators in specific phases
to reduce grid issues and increase PV penetration. To choose the proper phase, a deep analysis of
load along time has to be carried out. This solution is also promising, but the installation of solar PV
3
Energies 2020, 13, 4347
facilities depends mainly on the consumers and some country-specific normative make the limitation
of the installation of PV generators in public grids difficult.
The main objective of [18] is to evaluate how prosumers can support the distribution system
operator in the management of the grid and focuses on the changes and demand management, that the
customer can offer to benefit both, prosumer and DSO. It also shows the importance of photovoltaic
installations near consumption points in order to obtain a better balance between generation and
consumption. A big effort is being made through different projects, as in [19] or in [20], and platforms
to design, develop, test and foster these local flexibility markets.
Reference [21] shows a different approach as it includes economic and regulatory aspects in the
analysis of the maximum PV penetration. The effect of self-consumption policies (economics) and
regulations in PV facilities size is analysed to evaluate the PV penetration level in the electric grid
and its effect on grid operation. This paper proposes proper regulation, technical and economical, as
an alternative to grid reinforcement in high PV penetration grids. Reference [22] continues with the
economical point of view to propose solutions to improve solar PV penetration without hampering the
electric system in Latvia.
In Reference [23], a reactive power control is implemented in PV inverters managing reactive power
according to voltage in the grid connection point. It also reduces the active power generation if the
reactive power consumption does not produce the desired decrease in terminal voltage. Reference [24]
proposes three reactive power controls in order to reduce overvoltage, two of them based on reactive
power management depending on active power generated, Q(P) and the other to grid voltage, Q(U).
Following these ideas, the objective of this article is twofold: on the one hand, to propose a
method to calculate the photovoltaic hosting capacity that minimizes system losses, and on the other,
to evaluate different management techniques for solar PV inverters and their effect on the hosting
capacity aforementioned. As it is shown in the next section, the proposed HC calculation method is a
mixture of the deterministic methods using time series data and the statistical ones. This union aims to
obtain a simple and fast-running method that provides accurate results and covers multiple possible
scenarios for the deployment of distributed generation. Other works as [11] or [25] propose similar
methods, but in the research shown in this document real data from users is used, solar PV facilities are
installed always in consumers facilities (single or three-phase), to evaluate the effect of self-consumption,
and its power is variable and related to the contracted power of the possible prosumer.
This paper is organised as follows: Section 2 describes the grid used to test different PV inverter
control and the methodology used to calculate the optimal hosting capacity and the effect on it
of the solar PV inverters management. Section 3 shows the main results of the tests carried out.
All calculations shown in this document have been carried out on a real low-voltage distribution
grid in Granada, southern Spain. Finally, Section 4 discusses the main conclusions derived from the
simulations and proposes new working lines.
4
Energies 2020, 13, 4347
Next, the methodology to obtain the optimal distributed photovoltaic penetration is detailed.
This analysis consists in increasing the photovoltaic penetration with steps of 10 kW, to reach the
maximum level of PV penetration that could be installed by the clients randomly. In the Figure 1,
it is shown a flow chart of the methodology to apply. The first step of the flowchart is to define
different variables:
(a) maximum number of scenarios for each level of penetration (Max_num_scenarios). For this study
30 scenarios have been chosen.
(b) maximum penetration level (Max_Penetration). For this study 400 kW,
(c) the minimum level of penetration (Install_level). For this study it starts with 10 kW.
5
Energies 2020, 13, 4347
After defining the model parameters, the scenario counter is initialized (Scenario = 0). This variable
is a counter increase (scenario = scenario + 1) until reaching 30 scenarios at each level of penetration.
These 30 scenarios locate PV facilities uniform distribution in the network at existing customer
connections, applying a Monte Carlo approach. The total power installed with PVs is set as the level of
penetration. The size of each PV system is assumed to be equal to the customer’s contracted power and
the phase to which it is connected. The total PV power installed is defined by the level of penetration.
After completing all scenarios, penetration level is increased by 10 kW. This process is repeated until
maximum penetration level is reached.
Networks should not have congestion problems or voltage deviations. Limits are set by each
facility. In this case it is Grupo Cuerva that stablishes that voltage limits are 1.07 p.u. and 0.93 p.u.,
while the congestion limit of lines and transformer is 100%. In order to verify if there are any problems
of congestion or voltage deviations, a quasi-dynamic study is executed each scenario, which consists in
a power flow for each hour of a day.
If no problems are detected, total losses are calculated and stored for each of the 24 h. Once there
are 30 valid scenarios, the penetration of PVs is increased (Penetration = Penetration + 10 kW) until the
maximum penetration is reached (Max_Penetration). After that, the curve “PV penetration-losses”
is obtained. Finally, the optimal PV penetration level is derived from the minimum of the PV
penetration-loss curve. According to the model definition, network losses decrease with increasing PV
penetration until a certain level, where losses start to increase again.
After obtaining the optimal penetration value that minimizes systems losses, ten random uses
cases are created with the optimal renewable penetration (240 kW), this choice is based on the study
made to obtain the minimum losses in the system. The ten use cases are created with different PVs
location (more information about the network topology in Appendix B). These ten use cases include
congestion or voltage deviations. The control strategies are then evaluated in terms of ability to solve
voltage problems caused by distributed PV generation. The controls tested are:
• Constant power reactive control (Q = constant): PVs consume or generate reactive power regardless
of the active power installed. An amount of reactive power to be consumed or generated must be
set. In order to evaluate efficiently this control strategy, six different values for Q are studied:
PF = 0.95 inductive
PF = 0.95 capacitive
PF = 0.9 inductive
PF = 0.9 capacitive
PF = 0.85 inductive
PF = 0.85 capacitive
• Power factor control depending on the active power generated (PF = f(P)) [26]: PVs consume reactive
power as a function of the active power generated at each moment. In this case the power
factor depends on the active power generation. For this control strategy, two parameters are
defined: (1) the minimum active power, that indicates which is the value where reactive power
6
Energies 2020, 13, 4347
consumption starts; (2) the minimum value of the power factor. Figure 2 shows the PF variation
curve used to set control strategies such as:
PF = 0.95
PF = 0.9
PF = 0.85
In order to test this methodology, a network model has been developed in PowerFactory
DIgSILENT. A real model provided by Grupo Cuerva has been chosen, this real model is a district of a
rural town. This model contains 266 consumers, from which 45 have a three-phase connection and
221 single-phase. The total installed power of all customer is 1, 320 MW approximately. This network
is divided in 11 feeders (Figure A1 in Appendix A, each of the feeders is indicated in different colours)
which have different lengths and different numbers of consumers. Figure 3 shows the LV network
topology (More information about the network topology in Appendix A).
7
Energies 2020, 13, 4347
3. Results
The first part of this section provides the results of the study of optimal PV penetration and in the
second part results from the different use cases are presented for each control strategy.
8
Energies 2020, 13, 4347
Figure 4. Total grid losses from all scenarios and PV penetration levels.
Figure 5 shows, in orange, the scenarios with problems in the network (congestion and/or voltage
deviations) that have been obtained by Monte Carlo method, while trying to reach the 30 scenarios
(30 scenarios are defined as Max_num_scenarios in Figure 1, i.e., the 30 scenarios for each level of
penetration in which losses are calculated), without problems (blue).
As it can be seen in Figure 5, as distributed PV generation increases, more scenarios were found
with problems of voltage deviations, which complicated the finding of the 30 scenarios without
voltage deviations. For example, in the scenario with a penetration of 240 kW of distributed PV,
12 scenarios with grid problems have been obtained, this means that to find 30 scenarios without
problems, 42 scenarios had to be analysed.
9
Energies 2020, 13, 4347
3.2. PV Control
In the following paragraphs, the control strategies explained in Section 2 are compared identifying
the advantages and disadvantages of each control. To highlight the advantages and disadvantages of
each option, three indicators have been used:
• Losses
• Voltage deviations (undervoltage and overvoltage)
• Congestions
To test each of the controls, ten use cases are created with different PV locations (see Appendix A
for the locations of the PVs in each use case) but all of them with the same penetration 240 kW (total
installed power for the PVs). In each of the use cases, a quasi-dynamic study (hourly load flow) will be
performed over the period of one day in which the three indicators mentioned above will be calculated.
Figure 6. Constant power reactive control: Losses (More information see Table A3 in Appendix C).
The second key indicator in this analysis is voltage deviation. In this case, no undervoltage event
happened but overvoltage events occur. These overvoltage events are visualized in Figure 7.
When generating more reactive power (green and blue lines in Figure 7) there are more overvoltage
events than when reactive power is consumed (grey line in Figure 7). For example, in use case 4, when
reactive power is generated at 30% of active power, there are 25 overvoltage events in the network.
However, if the reactive power is consumed at 30% of the active power, only one overvoltage event has
been observed. Finally, the key congestion indicator has been analysed and the shows no congestion in
any use case.
10
Energies 2020, 13, 4347
Figure 7. Constant reactive power control: Voltage deviations. (More information see Table A4 in
Appendix C).
Figure 8. Reactive power control with constant power factor: Losses (More information see Table A5 in
Appendix C).
As can be seen in Figure 9, results regarding overvoltages are also very similar to the first control
strategy with fixed reactive power control. In comparison, a slightly higher number of overvoltage
events are observed for fixed PF control. This effect can be explained due to the fact that reactive power
is highest when active power peaks, which are the hours where reversed power flows can be observed.
11
Energies 2020, 13, 4347
Figure 9. Reactive power control with constant power factor: Voltage deviations (More information see
Table A6 in Appendix C).
Finally, the result of this congestion indicator shows zero as there is no congestion in any of the
use cases.
The results obtained evaluating the losses for this control strategy can be seen in Figure 10. As it
can be seen in Figure 10, the loss values obtained applying the different control parameters are very
close to each other. Thus, the effect on network losses is very limited.
Figure 10. Reactive power control with constant power factor: Losses (More information see Table A7
in Appendix C).
12
Energies 2020, 13, 4347
Overvoltage events are shown in Figure 11. Similar to the other two control strategies,
no under-voltage is observed. The number of overvoltage events in this case is considerably lower,
which is mainly due to the fact that no capacitive reactive power has been considered.
Figure 11. Power factor control depending on the active power generated: Voltage deviations
(More information see Table A8 in Appendix C).
In Figure 11, it can be seen that the control strategy of reactive power generation with 0.5 p.u.
of active power and PF = 0.9 p.u., is the one where less overvoltage events are obtained. Therefore,
if only the overvoltage events are considered, this would be the optimal strategy.
PF = 0.95 p.u.
PF = 0.9 p.u.
PF = 0.85 p.u.
The first key indicator to analyse is losses. Figure 12 shows network losses for this control strategy
of PF = 0.95 p.u. (grey dots), PF = 0.9 p.u. (yellow dots) and PF = 0.85 p.u. (green dots).
Figure 12 shows that the lowest losses are obtained when the power factor is 0.95, i.e., when less
reactive power is delivered per voltage level.
13
Energies 2020, 13, 4347
Figure 12. Reactive power control depending on terminal voltage: Losses (More information see
Table A9 in Appendix C).
The number of overvoltage events is shown in Figure 13, where is becomes clear that with
the lowest PF (0.85) limit, overvoltages can be reduced most. This is not surprising, as the more
reactive power is allowed to be consumed, the higher is the effect on terminal voltages. Nevertheless,
it should be highlighted here, that this reduction of overvoltages comes with the downside of increased
network losses.
Figure 13. Reactive power control depending on terminal voltage: Voltage deviations (More information
see Table A10 in Appendix C).
As happened in the previous control, in the use case there are no congestions and therefore the
last key indicator, congestions, is zero.
14
Energies 2020, 13, 4347
In the Figure 14, identifying the optimal control is difficult because there is no clear tendency in
the controls. For example, in the constant value control of Q, a very small value of losses is observed
if the reactive power is low and a very high value of losses if the reactive power consumed is high.
The same situation occurs when the constant PF control is applied. However, the three dots (green dots
in Figure 14) that define the PF = f (P) are close and the losses are small enough in order to select this
control as optimal control.
Figure 15 shows the number of overvoltage events in each use case for each control strategy.
In Figure 15 it can be seen the effectiveness of each control. It is represented by the number of terminals
in overvoltage. The results show that Q = f (U) (yellow line Figure 15) is the most effective control as in
all use cases has the lowest number of overvoltage events.
For the selection of the optimal control, it is necessary to take into account both key indicators
(losses, voltage deviations) and to reach an agreement on which level of losses and overvoltage events
are considered as valid. As a conclusion, the optimal controls would be Q = f (u) and PF = f (p) since
they are the controls that cause the least overvoltage events and the loss values are sufficiently small.
15
Energies 2020, 13, 4347
4. Conclusions
In this paper, two main topics have been analysed: on the one hand, to propose a method to calculate
the photovoltaic hosting capacity that minimizes system losses, and on the other, to evaluate different
management techniques for solar PV inverters and their effect on the hosting capacity aforementioned
In a first step, optimal PV penetration levels have been derived using a Monte Carlos approach,
considering optimal PV penetration when network losses are minimised, that is to say, the HC has
been obtained in each node where self-consumption can be introduced, which minimizes the losses
of the system.. This approach is very promising for planning purposes, if a detailed network model
is available. Similar methods are proposed in the works of [11] or [25], however in this article the
scope has been extended using real data from the smart meters, the installed PVs can be installed in
single phase or three-phase (depending on the installation of the consumer), the size of the installed
PV will depend on the contracted power of the consumer, with this it is possible to evaluate the effect
of PV self-consumption in a distribution network. Some improvements can be proposed for future
studies: (i) it has been used a standard day and it is proposed to use a whole year of smart meter data
to evaluate the seasonal effect (autumn, spring, winter and summer), (ii) it has been assumed that PV
facilities power is equal to the contracted power and it could be interesting modelling the variability
of the installed PV facilities and for the last (iii) it will be evaluated other methods to expand the
penetration of renewables such as distributes storage technologies and / or demand response policies,
either through direct signals or through a local market flexibility.
The second topic was dedicated to the evaluation of 4 different control strategies of solar PV
inverters, considering the possibility of reactive power control to mitigate excessive voltage deviations.
In addition, the effect on total network losses have been documented. Congestion problems were also
expected, but were not observed in this case study, given the restrictive condition which limits installed
solar PV power at consumer premises to their respective contracted power. The effectiveness of this
limit could be demonstrated hereby.
In the tested grid, 240 kW of installed PV power has been found as the optimal penetration level
in terms of technical losses. The 4 control strategies have been tested for this penetration level, and ten
use cases were generated in which overvoltage occurred. Control strategies with constant reactive
power or PF setpoints were outperformed by more dynamic approaches considering generated power
to modify PF and terminal voltages to modify reactive power.
16
Energies 2020, 13, 4347
Author Contributions: Conceptualization, N.G., G.F. and A.S.; methodology, N.G. and G.F.; software, D.M.
(Daniel Marquina) and D.M. (Diego Martinez); validation, P.L.; formal analysis, G.F. and N.G.; investigation,
D.M. (Diego Martinez) and P.L.; resources, D.M. (Daniel Marquina) and G.F.; writing—original draft preparation,
G.F. and N.G.; writing—review and editing, D.M. (Daniel Marquina), D.M. (Diego Martinez), A.S., P.L., J.R. and
H.B.; visualization, G.F., N.G., D.M. (Daniel Marquina) and D.M. (Diego Martinez); supervision, N.G.; project
administration, G.F. All authors have read and agreed to the published version of the manuscript.
Funding: This research has received funding from the European Union’s Horizon 2020 Framework Programme
for Research and Innovation under grant agreement no. 864319, project PARITY.
Acknowledgments: We thank CUERVA Group for supporting the research carried out and to provide information
about their distribution grids.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A
In this section the main characteristics of the network model used for the developments are
detailed. Figure A1 shows a single-line diagram of the network, in which each of the feeders that
conform the network can be seen in different colours.
The cables used to model the network are listed in Table A1, this table shows the characteristics of
the type of cable which are: voltage of the conductor (Voltage), maximum allowed current (Current),
the cable is type, resistance (R), reactance (X) and inductance (L).
17
Energies 2020, 13, 4347
To characterize the network, real measurements of the clients obtained from the smart meters
have been used, in Figure A2 an example of the consumption of two customers profiles is shown.
The voltages of the secondary substation are shown in Figure A3, In Figure A3 the voltage of
phase 1 (phase R) can be seen in blue, the voltage of phase 2 (phase S) can be seen in orange and the
voltage of phase 3 (phase T) can be seen in grey.
The characteristics of the secondary substation are described in the Table A2.
18
Energies 2020, 13, 4347
Appendix B
Use case 1:
Figure A4 shows the location of the photovoltaic installations in case of use 1, the locations of the
photovoltaic installations are shaded in blue.
Use case 2:
Figure A5 shows the location of the photovoltaic installations in case of use 2, the locations of the
photovoltaic installations are shaded in blue.
19
Energies 2020, 13, 4347
Use case 3:
Figure A6 shows the location of the photovoltaic installations in case of use 3, the locations of the
photovoltaic installations are shaded in blue.
Use case 4:
Figure A7 shows the location of the photovoltaic installations in case of use 4, the locations of the
photovoltaic installations are shaded in blue.
20
Energies 2020, 13, 4347
Use case 5:
Figure A8 shows the location of the photovoltaic installations in case of use 5, the locations of the
photovoltaic installations are shaded in blue.
Use case 6:
Figure A9 shows the location of the photovoltaic installations in case of use 6, the locations of the
photovoltaic installations are shaded in blue.
21
Energies 2020, 13, 4347
Use case 7:
Figure A10 shows the location of the photovoltaic installations in case of use 7, the locations of the
photovoltaic installations are shaded in blue.
Use case 8:
Figure A11 shows the location of the photovoltaic installations in case of use 8, the locations of the
photovoltaic installations are shaded in blue.
22
Energies 2020, 13, 4347
Use case 9:
Figure A12 shows the location of the photovoltaic installations in case of use 9, the locations of the
photovoltaic installations are shaded in blue.
23
Energies 2020, 13, 4347
Appendix C
This section details the results of each use case for each of the control strategies. The tables show
the results of voltage deviations and losses.
Constant power reactive control
Table A5. Reactive power control with constant power factor: Losses.
24
Energies 2020, 13, 4347
Table A6. Reactive power control with constant power factor: Voltage deviations.
Table A7. Power factor control depending on the active power generated: Losses.
Power Factor Control Depending on the Active Power Generated (PF = f (P)):
Use Cases
P = 0.5 & PF = 0.9 P = 0.5 & PF = 0.95 P = 0.65 & PF = 0.9 P = 0.65 & PF = 0.95
Use Case 1 0.0164 0.0155 0.0150 0.0146
Use Case 2 0.0180 0.0171 0.0166 0.0163
Use Case 3 0.0218 0.0211 0.0207 0.0204
Use Case 4 0.0210 0.0203 0.0201 0.0198
Use Case 5 0.0174 0.0165 0.0161 0.0158
Use Case 6 0.0201 0.0192 0.0189 0.0185
Use Case 7 0.0189 0.0179 0.0174 0.0170
Use Case 8 0.0189 0.0181 0.0177 0.0174
Use Case 9 0.0177 0.0169 0.0165 0.0162
Use Case 10 0.0190 0.0181 0.0177 0.0173
Table A8. Power factor control depending on the active power generated: Voltage deviations.
Power Factor Control Depending on the Active Power Generated (PF = f (P)):
Use Cases
P = 0.5 & PF = 0.9 P = 0.5 & PF = 0.95 P = 0.65 & PF = 0.9 P = 0.65 & PF = 0.95
Use Case 1 0 0 0 0
Use Case 2 2 2 2 2
Use Case 3 9 12 12 12
Use Case 4 1 2 2 3
Use Case 5 5 5 5 5
Use Case 6 0 0 0 0
Use Case 7 8 10 10 11
Use Case 8 5 5 5 5
Use Case 9 0 0 0 1
Use Case 10 3 3 3 3
25
Energies 2020, 13, 4347
Table A10. Reactive power control depending on terminal voltage: Voltage deviations.
References
1. Mansouri, N.; Lashab, A.; Sera, D.; Guerrero, J.M.; Cherif, A. Large photovoltaic power plants integration:
A review of challenges and solutions. Energies 2019, 12, 3798. [CrossRef]
2. Munkhchuluun, E.; Meegahapola, L. Impact of the Solar Photovoltaic (PV) Generation on Long-Term Voltage
Stability of a Power Network. In Proceedings of the 2017 IEEE Innovative Smart Grid Technologies—Asia
(ISGT–Asia), Auckland, 2017; pp. 1–6. Available online: https://ieeexplore.ieee.org/abstract/document/
8378456/authors#authors (accessed on 20 April 2020).
3. Eguía, P.; Torres, E.; Etxegarai, A. Impact of Photovoltaic Self-consumption on Power Losses and Voltage Levels
of MV Distribution Networks. 2017. Available online: https://ieeexplore.ieee.org/abstract/document/8260161
(accessed on 20 April 2020).
4. Tavakkoli, M.; Pouresmaeil, E.; Godina, R.; Vechiu, I.; Catalão, J.P. Optimal management of an energy storage
unit in a PV-based microgrid integrating uncertainty and risk. Appl. Sci. 2019, 9, 169. [CrossRef]
5. Aziz, T.; Ketjoy, N. PV Penetration Limits in Low Voltage Networks and Voltage Variations. IEEE Access
2017, 5, 16784–16792. [CrossRef]
6. Patil, A.; Girgaonkar, R.; Musunuri, S.K. Impacts of Increasing Photovoltaic Penetration on Distribution
Grid—Voltage Rise Case Study. In Proceedings of the 2014 International Conference on Advances in
Green Energy, (ICAGE), Thiruvananthapuram, India, 17–18 December 2014; pp. 100–105. Available online:
https://ieeexplore.ieee.org/abstract/document/7050150 (accessed on 20 April 2020).
7. Gandhi, O.; Kumar, D.S.; Rodriguez-Gallegos, C.; Srinivasan, D. Review of power system impacts at high
PV penetration Part I: Factors limiting PV penetration. Solar Energy 2020, 1–21, In Press, Corrected Proof.
Available online: https://www.sciencedirect.com/science/article/abs/pii/S0038092X20307118 (accessed on
20 April 2020).
26
Energies 2020, 13, 4347
8. Wajahat, M.; Khalid, H.A.; Bhutto, G.M.; Leth Bak, C. A Comparative Study into Enhancing the PV Penetration
Limit of a LV CIGRE Residential Network with Distributed Grid-Tied Single-Phase PV Systems. Energies
2019, 12, 2964. [CrossRef]
9. Ul Abideen, M.Z.; Ellabban, O.; Al-Fagih, L. A review of the tools and methods for distribution networks’
hosting capacity calculation. Energies 2020, 13, 2758. [CrossRef]
10. Bollen, M.H.; Rönnberg, S.K. Hosting capacity of the power grid for renewable electricity production and
new large consumption equipment. Energies 2017, 10, 1325. [CrossRef]
11. Klonari, V.; Toubeau, J.; Lobry, J.; Vallée, F. Photovoltaic integration in smart city power distribution:
A probabilistic photovoltaic hosting capacity assessment based on smart metering data. In Proceedings of
the 2016 5th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS), Rome,
Italy, 23–25 April 2016; pp. 1–13.
12. Mateo, C.; Frías, P.; Cossent, R.; Sonvilla, P.; Barth, B. Overcoming the barriers that hamper a large-scale
integration of solar photovoltaic power generation in European distribution grids. Sol. Energy 2017, 153,
574–583. [CrossRef]
13. Yuan, J.; Weng, Y.; Tan, C.W. Quantifying Hosting Capacity for Rooftop PV System in LV Distribution Grids.
2019, pp. 1–10. Available online: https://arxiv.org/abs/1909.00864 (accessed on 20 April 2020).
14. Hong, B.; Li, Q.; Miao, W.; Yan, H.; Liu, J. Energy Storage Application in Improving Distribution Network’s
Solar Photovoltaic (PV) Adoption Capability. Energy Procedia 2018, 145, 472–477. [CrossRef]
15. Denholm, P.; Margolis, R.M. Evaluating the limits of solar photovoltaics (PV) in electric power systems
utilizing energy storage and other enabling technologies. Energy Policy 2007, 35, 4424–4433. [CrossRef]
16. Jeong, Y.C.; Lee, E.B.; Alleman, D. Reducing voltage volatility with step voltage regulators: A life-cycle cost
analysis of Korean solar photovoltaic distributed generation. Energies 2019, 12, 652. [CrossRef]
17. Torres, I.C.; Negreiros, G.F.; Tiba, C. Theoretical and experimental study to determine voltage violation,
reverse electric current and losses in prosumers connected to low-voltage power grid. Energies 2019, 12, 4568.
[CrossRef]
18. Bremdal, B.A.; Mathisen, G.; Degefa, M.Z. Flexibility offered to the distribution grid from households
with a photovoltaic panel on their roof. In Proceedings of the 2018 IEEE International Energy Conference
(ENERGYCON), Limassol, Cyprus, 3–7 June 2018; pp. 1–6. Available online: https://ieeexplore.ieee.org/
abstract/document/8398848 (accessed on 20 April 2020).
19. PARITY Project. Available online: https://parity-h2020.eu/ (accessed on 20 April 2020).
20. FLEXCoop Project. Available online: http://www.flexcoop.eu/ (accessed on 20 April 2020).
21. Mateo, C.; Cossent, R.; Gómez, T.; Prettico, G.; Frías, P.; Fulli, G.; Meletiou, A.; Postigo, F. Impact of solar PV
self-consumption policies on distribution networks and regulatory implications. Sol. Energy 2018, 176, 62–72.
[CrossRef]
22. Sauhats, A.; Zemite, L.; Petrichenko, L.; Moshkin, I.; Jasevics, A. Estimating the economic impacts of net
metering schemes for residential PV systems with profiling of power demand, generation, and market prices.
Energies 2018, 11, 3222. [CrossRef]
23. Ren, B.; Zhang, M.; An, S.; Sun, X. Local Control Strategy of PV Inverters for Overvoltage Prevention in Low
Voltage Feeder. In Proceedings of the 2016 IEEE 11th Conference on Industrial Electronics and Applications
(ICIEA), Hefei, China, 5–7 June 2016; pp. 2071–2075.
24. Zhang, Y.; Wang, X. PV reactive voltage regional autonomy control strategy for Q (U) improvement in
distribution network. In Proceedings of the 2019 4th International Conference on Intelligent Green Building
and Smart Grid (IGBSG), Yi-chang, China, 6–9 September 2019; pp. 497–501.
25. Duwadi, K.; Ingalalli, A.; Hansen, T.M. Monte Carlo Analysis of High Penetration Residential Solar Voltage
Impacts using High Performance Computing. In Proceedings of the 2019 IEEE International Conference on
Electro Information Technology (EIT), Brookings, SD, USA, 20–22 May 2019; pp. 1–6.
26. German Standard VDE-AR-N 4105. Available online: https://www.vde.com/en/fnn/topics/technical-
connection-rules/power-generating-plants (accessed on 20 April 2020).
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
27
energies
Article
Modeling and Detection of Future Cyber-Enabled
DSM Data Attacks
Kostas Hatalis 1,∗ , Chengbo Zhao 1 , Parv Venkitasubramaniam 2 , Larry Snyder 3 ,
Shalinee Kishore 2 and Rick S. Blum 2
1 Department of Electrical and Computer Engineering, Lehigh University, Bethlehem, PA 18015, USA;
[email protected]
2 Faculty of Electrical and Computer Engineering, Lehigh University, Bethlehem, PA 18015, USA;
[email protected] (P.V.); [email protected] (S.K.); [email protected] (R.S.B.)
3 Faculty of Industrial and Systems Engineering, Lehigh University, Bethlehem, PA 18015, USA;
[email protected]
* Correspondence: [email protected]
Abstract: Demand-Side Management (DSM) is an essential tool to ensure power system reliability
and stability. In future smart grids, certain portions of a customer’s load usage could be under the
automatic control of a cyber-enabled DSM program, which selectively schedules loads as a function
of electricity prices to improve power balance and grid stability. In this scenario, the security of DSM
cyberinfrastructure will be critical as advanced metering infrastructure and communication systems
are susceptible to cyber-attacks. Such attacks, in the form of false data injections, can manipulate
customer load profiles and cause metering chaos and energy losses in the grid. The feedback
mechanism between load management on the consumer side and dynamic price schemes employed
by independent system operators can further exacerbate attacks. To study how this feedback
mechanism may worsen attacks in future cyber-enabled DSM programs, we propose a novel
mathematical framework for (i) modeling the nonlinear relationship between load management and
real-time pricing, (ii) simulating residential load data and prices, (iii) creating cyber-attacks, and (iv)
detecting said attacks. In this framework, we first develop time-series forecasts to model load demand
and use them as inputs to an elasticity model for the price-demand relationship in the DSM loop.
This work then investigates the behavior of such a feedback loop under intentional cyber-attacks.
We simulate and examine load-price data under different DSM-participation levels with three types
of random additive attacks: ramp, sudden, and point attacks. We conduct two investigations for the
detection of DSM attacks. The first studies a supervised learning approach, with various classification
models, and the second studies the performance of parametric and nonparametric change point
detectors. Results conclude that higher amounts of DSM participation can exacerbate ramp and
sudden attacks leading to better detection of such attacks, especially with supervised learning
classifiers. We also find that nonparametric detection outperforms parametric for smaller user pools,
and random point attacks are the hardest to detect with any method.
Keywords: demand side management; demand response; cyber-physical systems; dynamic pricing;
load forecasting; attack detection
1. Introduction
Demand-Side Management (DSM) is an essential component in smart grids for planning,
monitoring, and modifying consumer load levels. Furthermore, cyber-enabled DSM will allow smart
grids even higher levels of automated decision-making capabilities to selectively schedule loads on
local grids to improve power balance and grid stability. Such a cyber approach relies heavily on
real-time, two-way communication capabilities between a central controller and various adaptable
loads. Research into the reliability of the cyberinfrastructure that enables DSM is therefore vital.
The main concerns in ensuring DSM security lay in the feedback mechanism of real-time electricity
pricing and distributed DSM controllable loads. For instance, in residential grids, a single home
consumes only a small amount of electric power, and if it becomes compromised, then it might not
have a noticeable impact on the power grid. However, a carefully planned or even random cyber
attack might impact other loads not under attack by taking advantage of the feedback mechanism of
load management.
Two-way communication capabilities of Advanced Metering Infrastructure (AMI) enable a utility
or Independent System Operator (ISO), in the retail power markets, to collect high-resolution energy
usage from consumers and enable dynamic pricing to adapt to consumer demand [1]. AMIs provide
an efficient way for ISOs to schedule prices and then communicate them to consumers for automatic
DSM control of certain portions of a consumer’s load. AMIs can also provide practical ways for ISOs
to set DSM goals, such as reducing peak or decreasing aggregate load levels through price influences.
However, there are several vulnerabilities in AMIs that present noteworthy security issues since they
are directly accessible by users [2]. Due to the large scale deployments of AMIs, ISOs encourage the
utilization of marginally cheaper hardware, resulting in constrained computational resources to allow
for robust security capacities, such as intrusion monitoring.
DSM programs also utilize demand response, a specific tariff or program to motivate customers
to respond to changes in price or electricity availability over time by altering their regular electricity
use habits. We take this a step further and look at future cyber-enabled DSM programs [3] that
will autonomously control household loads such as water heaters and HVAC units based on Real
Time Pricing (RTP). As part of the reliable implementation of this future cyber-DSM, it is crucial to
understand the dependency between dynamic pricing and automatic demand response as well as the
risks. We hypothesize that cyber-DSM programs can be particularly vulnerable to cyber attacks such
as false pricing information or direct load manipulation, especially when the participation rate in DSM
is high.
Our work is thus motivated to study these vulnerabilities in DSM. We provide a mathematical
formulation of the feedback between utilities and DSM systems, and then simulate, analyze, and test
different detection methods for attacks on such feedback. This relationship between load and price
is shown visually in Figure 1. As prices go up, demand naturally responds by decreasing. However,
if AMIs are hijacked, and false lower prices are reported to DSM systems, there will be an inappropriate
increase in demand. A similar effect happens if an attacker directly controls user loads. Then, a higher
load usage by the attacker may inadvertently lead to higher prices for the rest of the grid. We present a
standard of how attackers can exploit such a dependency.
We propose a mathematical framework of the feedback between price setting and DSM systems to
study how attacks can be structured and how to detect them. Similar frameworks have been considered
in previous work. In [4], the author applied a profitable attacking strategy to a price-load dependency
model. Another control-theoretic approach to derive the fundamental conditions of Real Rime Pricing
(RTP) stability under integrity attacks is described in [5]. The authors in that work use an elasticity
model between price and load and apply an autoregressive model to obtain the forecast load and price.
The main contributions of our approach can be summarized as follows:
1. We formalize the demand–price relationship using a nonlinear elastic demand model to achieve
DSM goals. We simulate load and pricing data in a DSM system, in particular under a strategic
conservation scheme.
2. We propose two modes of attacks on DSM systems: false pricing data injection and direct
load manipulation. We prove their equivalence and highlight three types of attacks that could
be undertaken by each mode. We then empirically show how an increased use of DSM can
exacerbate attacks.
30
Energies 2020, 13, 4331
3. We simulate these attacks and compare sequential change-point and machine learning methods,
as well as a deeper dive in parametric vs. nonparametric methods, for detecting DSM attacks.
In our main results, we demonstrate the impact of DSM on detection performance and identify
what kind of detectors are effective for different attacks.
In Section 2, we provide a literature review on DSM and important DSM strategies on real-time
pricing and load forecasting. In Section 3, we present a block bootstrap technique for simulating the
non-DSM load distribution of a micro-grid of N homes from template residential load time series.
We then propose dependency models for the feedback nature of load and prices in Section 4, where we
also showcase simulations of residential load and electricity prices when an automatic DSM program
controls certain portions of consumer demand as a function of price. In Section 5, we present two
modes of cyber attacks, direct load manipulation attacks, and price data injection attacks that can have
a significant influence on the feedback of load and price. We prove these two attacks are equivalent.
In Section 6, we review several sequential detection and supervised learning methods. In Section 7,
we present our experimental results that compare performance under different DSM participation rates
when they are under three types of attacks: ramp, sudden, and point attacks. We find that point attacks
are hardest to detect, and supervised learning methods can provide better results than sequential
detection methods. To provide a deeper analysis, we also conduct a comparison of parametric and
nonparametric detectors. We conclude in Section 8 with possible directions of future work.
3ULFH +LMDFNLQJ
3ULFH
7LPH
3ULFH $OWHUV
'HPDQG
'HPDQG
1 2 3 7LPH
Figure 1. Feedback effect between price and DSM demand. As prices go up, demand decreases.
However, if prices are hijacked and false prices are fed to DSM systems, then a false low price increases
demand, and a false high price can decrease demand. The same is true if demand was altered by an
attack. If load usage is increased by an attacker, then prices would increase and vice versa.
31
Energies 2020, 13, 4331
2. Background
Time
Strategic Conservation
Load Shifting
Demand
Demand
Time
Time
Demand Side
Valley Filling Management Flexible Load Shape
Demand
Demand
Time
Our study focuses on modeling and simulating the DSM goal of strategic conservation due to
its simplicity and essential use in the smart grid. This goal also makes it easier to study attacks
on DSM by modeling strategic conservation as a general reduction in load. Attacks then could
stand out more visibly than under other goal schemes like flexible load shaping. More specifically,
this DSM goal aims at reducing aggregate load demand through directed reduction of electricity
consumption. The successful implementation of strategic conservation programs usually requires some
combination of financial incentives to customers, energy-efficient building standards, and appliance
efficiency improvement. We envision that AMI and smart appliances in residential DSM programs will
automatically control specific portions of consumer load as a function of real-time electricity prices to
achieve the goal of strategic conservation.
32
Energies 2020, 13, 4331
T
min ∑ ( Lt − Lt )2 ,
Pt t =1
where Pt is the RTP at time t, Lt is actual load, and Lt is the target load level the ISO is interested in
achieving via DSM. The aim is to choose a price Pt for each time step such that the actual load would
reach as close as possible to the target level. In [7], a DSM strategy is proposed based on a heuristic
optimization strategy to shape the load curve close to the desired shape. A heuristic-based evolutionary
algorithm was used to solve the above minimization problem. A multi-agent game-theoretical DSM
approach is proposed in [8]. The authors use game theory and formulate an energy consumption
scheduling game, where the players are the users, and their strategies are the daily schedules of their
household appliances and loads. In [9], the minimization problem is solved by utilizing a feedforward
neural network to map the nonlinear relationship between price and load. Recently, the DSM problem
is addressed in [10] as a multi-objective optimization problem that also seeks to balance other merit
functions such as energy production cost, customer preferences, and other constraints.
33
Energies 2020, 13, 4331
(a requirement for most time series methods), excels at modeling nonlinear dependencies, and can fit
large data sets. Disadvantages for most machine learning models are that most hyperparameters are
continuous (difficult to tune), they require extensive feature engineering, and may get stuck in local
minimums. Models for LF include support vector machines [15], feedforward neural networks [16],
recurrent neural networks [17], random forests [18], and ensemble learning [19].
3. Microgrid Simulation
Before modeling the load-price dependency of a DSM system, we first need to obtain some ground
truth data of what load data from a residential micro-grid look like without the presence of DSM,
where we assume the elasticity of demand to price is very low. To do this, we use the power time series
from several homes as templates for our grid and generate artificial N household datasets. There are
several ways to generate artificially residential load data, such as using power grid simulators like
MATPOWER [22] or GridLAB-D [23]. Our object is to create a model-free real-data driven alternative
to grid simulators. We can generate unlimited but plausible univariate load data to serve as the base
demand for sample households before applying a DSM system.
1 60 i
60 i∑
E(kWh) = t(hr) × P(kW ) (1)
=1
34
Energies 2020, 13, 4331
Figure 3. Process of block bootstrap simulation of a new home power usage (bottom) from a
template home (top). Example simulation samples are taken from four days from the template series
with replacement.
4. Dependency Model
dL P
d = · . (2)
dP L
35
Energies 2020, 13, 4331
PED shows the responsiveness of the load L demanded of electricity to a change in Price (P).
Recent work [9] in demand response, particularly in the context of cybersecurity, has utilized this
approach. Our procedure differs in that we use real demand data to determine the “unmodulated”
pattern of load data. An absolute value of PED = 1 shows unitary elasticity. For instance, when d = −1,
then a 1% change in the price will have a 1% change in the load demanded. As prices increase, the load
will decrease. When absolute PED falls between 0 and 1, this signifies that the load demand is inelastic,
while a value greater than 1 says that the demand is elastic. When |d | = 0, the demand is perfectly
inelastic. A change in price does not affect the load, while d = ∞ represents perfect elasticity. If d is
constant, the whole demand curve is then
L = a · P d (3)
where a is a scaling constant. An example demand curve estimated from Equation (3) can be seen in
Figure 4. The figure also showcases the nonlinear relationship between load and price where as the
price, the independent variable, increases, the load demanded, the dependent variable, decreases.
Figure 4. Load as a function of price with arbitrary price range $1–$100, α = 10, 000, and d =
−0.1, −0.2, −0.4, −0.6, −1, −2.
where Pt is the RTP for time t, Pc is constant of the retailer’s market costs which does not vary with RTP,
θt,i ≥ 0 is a scaling factor representing the stochastic process that determines the user load, and t,i
d is
36
Energies 2020, 13, 4331
the elasticity coefficient for the individual customers sensitivity to price changes. It can vary over time
but, without DSM incentives, most users have a fairly inelastic PED.
For experimental purposes, in modeling individual user loads, we set φt,i equal to simulated
bootstrapped user load profiles defined in Section 3.2. We assume that prices, user preferences θt,i ,
and t,i
d have been absorbed in the calculation of the simulated load series. Thus, we use φ as a
t,i
reference point to show how much electricity a user wants to consume without the influence of a DSM
program. We model the task of the demand response administrator as modifying φt,i to some desired
load levels.
Realistically, only a certain portion κt,i ∈ [0, 1) of customer i’s power usage will be under control
of a cyber DSM program. There will always be some stochastic and, more importantly, an inelastic
component of power usages, such as using a microwave oven or electric hairdryer. We model the DSM
modulated load (including the inelastic portion) as follows:
dsm
lt,i = (κt,i φt,i ) Pt t,i + (1 − κt,i )φt,i , (4)
dsm
where the first part (κt,i φt,i ) Pt t,i is the load level customer i that allows the DSM program to determine
as a function of price Pt and the DSM’s elasticity to price t,i dsm . Price elasticity of DSM dsm may vary
t,i
over time and customer and affects how much power usage should be affected by price. The second
portion (1 − κt,i )φt,i of a customers load is the stochastic component. The DSM portion parameter
κt,i is defined as a function of time where users can add or remove house loads under DSM control
over time. The term κt,i can be modeled as a random variable (e.g., Uniform or Gaussian) or as a fixed
constant for users. Once each customer’s load is determined, total load (modified by a cyber-DSM
program) for time t for N customers is calculated by
N
Lt = ∑ lt,i . (5)
i =1
N
Φt = ∑ φt,i , (6)
i =1
which represents the total demand had there not been a DSM program for a time period t (IE κt,i = 0, ∀i ).
where inputs to the prediction model are past base load Φ and price P values from time t − k to
t − 1, and other predictor variables X such as time-of-day and weather information. Various types of
37
Energies 2020, 13, 4331
prediction models can be used for f pred such as neural networks as reviewed in Section 2.2. We now
define RTP based on the formulation in Equation (3) as such
1/ˆ dsm
L∗
Pt =
Φ̂t
. (8)
L∗ = a · Pt ˆ dsm
Goal 1: L∗ = Lt
. (9)
Goal 2: L∗ = Lt + ( Lt−1 − Lt−1 )
The component L∗ adjusts the RTP based on two goals the ISO may have. The first goal is to adjust
the price to push power usage directly to the target level Lt with the assumption that there is near
100% DSM participation by all customers. The expectation is that, if demand for time t is Φ̂t , then a
price point is set to push load usage to Lt . Of course, if participation is less than 100%, which is more
likely, the target level Lt will not be reached by Pt . To push the aggregate load from all users, as close
as possible to the target load level with an unknown amount of participation, a penalty would need to
be added to Pt . We model this as Goal 2, where the idea is to affect the power usage of those under
cyber-DSM control even more than goal 1 to compensate users who are not participating in DSM.
Some users will not be participating, or only have a small portion of their power usage under
control by the cyber-DSM program. We model their remaining power usage as inelastic to RTP. Thus,
to push aggregate load to a target level, taking into account that some load usage is inelastic, we need
to push RTP much higher or lower to have a bigger effect in pushing DSM controlled load closer to the
target load. This is what Goal 2 attempts to do, with the component L∗ = Lt + ( Lt−1 − Lt−1 ) taking the
target load level for time t and adding the difference from the previous target load Lt−1 and realized
load Lt−1 as a penalty to adjust RTP to compensate for the difference. If L∗ < 0, then we set L∗ = 10
or to some arbitrary small target value. By subtracting the difference between the previous load and
target level, we make up for users not participating in DSM by forcing DSM users to have a higher
price to push their load even lower.
The term ˆ dsm in Equation (8) is an estimate of the price elasticity of DSM of the whole grid;
if individual user coefficients t,idsm are unknown, then ˆ dsm can be estimated from observing past values
of price and load under different levels of DSM control. Alternatively, the ISO can define dsm for
all household cyber DSM programs. The formulation in Equation (8) sets prices by comparing the
adjusted target load for time t to the forecasted base demand Φ̂t for the same time. This demand Φ̂t
would be the level if no load were under DSM influence, thus to influence and alter it, Φt needs to be
estimated as accurately as possible.
In our approach, if the aggregate load is above the target load, RTP is set higher to decrease
demand. If the aggregate load is lower than the target load, the price decreases to increase demand.
Thus, as also can be observed in Equation (8), there is direct feedback between price and load
Equation (4). The block diagram in Figure 5 also outlines this feedback that showcases the relationship
between the utility and the grid. Generation sets the target load based on the price and supply of
power, and the controller sets the price signal and the elasticity of demand coefficient for DSM systems.
The price is then communicated into the grid into DSM systems, which adjust load usage appropriately.
The bold red lines in Figure 5 highlight the feedback relationship between price and demand. The scope
of work is in the mathematical modeling of the controller and DSM system relationship. With such a
model, we present in the next section how attackers can exploit it.
38
Energies 2020, 13, 4331
External
Factors
X X
ܮ௧
ܮ௧ ߳ௗ௦ ߳ௗ௦
߮௧
߮௧ ܲ௧ ܲ௧
ܮᇱ௧ DSM System
Controller
ܮᇱ௧ ܲ௧
Generation
Figure 5. Block diagram highlighting the feedback between the utility and grid.
Simulation Parameters
N 200 homes
dsm −1
Ltarget 200 kWh
39
Energies 2020, 13, 4331
shown in Figure 6a, we see that prices range from $0.01 to $0.03 per kWh. As expected with no DSM
program to influence demand, the observed load is equal to the base load. The same is observed
in Figure 7a when the simulation is run with the Goal 2 RTP model. The only difference under this
scenario is the price range which is exceptionally high ranging from $0.01 to $0.6 per kWh. This is
expected in this scenario since the ISO is attempting to set prices to maximize the effect on DSM
customers, of which there are none, but this is unknown to the utility. With no DSM, the mean
observed load is 332 kWh.
Next, we rerun the simulation setting κi = 0.5, ∀i . In the Goal 1 scenario in Figure 6b, we see that
the base load was reduced with a mean observed load of 269 kWh. In Figure 7b, the same simulation
40
Energies 2020, 13, 4331
is run with Goal 2. Here, the mean observed load was further reduced to 246 kWh, but large price
spikes occur at peak load times. Finally, we run simulations for κi = 0.99, ∀i to study the effects of high
penetration of DSM. In simulating both goals, shown in Figures 6c and 7c, the mean observed load is
reduced to 208 kWh, which is very close to the target load. However, in Goal 2, we see great resonating
feedback effects occur when prices spike very high. RTP increases as a response to large values in the
observed load. Then, when the load decreases to low levels, a price decrease causes the load to spike
more during the next time step, while under Goal 1, this is not observed. The higher prices set in Goal
2 would see a large cost to DSM participating customers.
• False Pricing Data Injection: The attacker can manipulate prices Pt received by each
compromised costumer i ∈ A, and the received price Pti can be different for various customers in
order to achieve the attacker’s desired effect:
a
Pt,i = Pt,i + at,i
P
, ∀i ∈ A, t ∈ Ta
This has the effect of compromising the demand response of a customer in the following way:
P P dsm
a
lt,i = (κi φt,i )( at,i ) + (1 − κi )φt,i
• Direct Load Manipulation: The attacker can manipulate the load of each compromised customer
lt,i , i ∈ D directly:
L
a
lt,i = lt,i + at,i
L
, ∀i ∈ A, t ∈ Ta
Under both attack modes, we would get a compromised aggregate load which may include one
or both attacks occurring simultaneously
N
P L
Lta = ∑ lt,ia + lt,i
a
.
i =1
These two modes of attack are equivalent as they both affect a customer’s load response as long
as part of the load is under cyber DSM control that is sensitive to price changes.
Theorem 1. Given a set of customers A compromised by the attacker, there always exists a direct load
manipulation attack such that all customers behave the same as a pricing data injection attack and vice versa for
κi > 0, ∀i , t ∈ Ta .
41
Energies 2020, 13, 4331
Proof. Setting both attacks to have the same load, then the attack load is set as follows:
P dsm
L
at,i = (κi φt,i )( at,i ) + (1 − κi )φt,i ,
If κi = 0, then the two attack modes are not equivalent since a price attack will have no effect on
customer load.
Since false pricing data injection and direct load manipulation attacks are equivalent, we focus
only on direct load manipulation attack analysis.
There are different goals an attacker can have to harm the power grid or exploit it. For example,
an attacker can cause chaotic metering by messing the metering data transmission, efficiency loss of
the energy provided by causing higher load volatility, or the energy system failure by overloading
the power lines or devices. The focus of this work is efficiency loss by increasing user loads through
direct load manipulation. In this scenario, we introduce three possible types of load attacks: ramp,
sudden, and point attacks. These types of attacks are visualized in Figure 8 wherein plot (a) an attacker
gradually increases a user’s load over time. In plot (b), an attacker suddenly ramps up the power
usage to a specified level, and, in plot (c), we demonstrate a point attack where the attacker increases
loads only for specific hours.
6. Attack Detection
Here, we outline sequential parametric, supervised learning-based, and nonparametric methods
for attack detection. The sequential detections methods are the one-sided CUmulative SUM (CUSUM)
test [30] and windowed GLRT [31]. Both these methods take as an input a residual time series that is
the output of applying a SARIMA filter to load observations. If enough past observations of load data
labeled as nominal or under attack are collected, then detection of attacks can be made by training
supervised learning classifiers. Supervised learning algorithms have been broadly adopted to the
42
Energies 2020, 13, 4331
smart grid literature to monitor and detect cyber attacks on power systems [32–34]. Here, we employ
several supervised learning methods for detecting attacks on cyber-DSM systems. It is unlikely that an
ISO can collect high-quality attack data due to the lack of such attacks occurring. However, such data
can be simulated. Using past load data, we simulate direct load manipulation attacks by creating
different types of attacks, as shown in Figure 8.
p P
(1− ∑ αi B)(1 − ∑ β i BS )(1 − B)d (1 − BS ) D Y (t + 1)
i =1 i =1
q Q
= (1 + ∑ ui B)(1 + ∑ γi BS ) x (t + 1)
i =1 i =1
where B is the back-shift operator of the model, αi , β i , ui , and γi are fixed coefficients determined by
the model after capturing the behavior of the series. Y (t + 1) is the price or load data at time t + 1
and x (t + 1) is the residual of the data. The exponents in the equation represent the times that the
back-shift operator is multiplied.
43
Energies 2020, 13, 4331
At > 0 is present for observation t, this will cause a definite shift in the mean of the residual series.
This detection problem can thus be stated as deciding if a null hypothesis H0 is true, where the residual
series has zero mean and known variance (invariant in time and estimated from a sample population),
or if the alternative hypothesis H1 is correct which states that the examined series has some mean not
equal to zero, thus being under attack. This can be modeled as a hypothesis test, and, for the GLRT
and CUSUM detectors, this translates to
H0 : xt ∼ N (0, σ2 ), t = 1, ..., N
H1 : xt ∼ N ( At , σ2 ), At > 0, t = 1, ..., N
For the GLRT detector, to simplify implementation, we model the attack as if it were constant
A > 0 but unknown. To produce a residual series, a SARIMA multi-step forecast for t = 1, ..., N is
made before the detection period. This forecast is conducted using a past window of training data that
was not under attack. The forecast is made for time t to t + k. These predictions are then subtracted
from the incoming observations to produce a residual time series, which is then fed as input into the
GLRT and CUSUM detectors.
gt = max(0, gt−1 + xt − k )
where k is called the reference value (sometimes also called drift) set priori to values such as 0, 0.5,
or A/2 if the size of A is known in advance. When gt = 0, then we define the change time as tc = t,
and, when gt > h > 0, we reset gt = 0 and flag an alarm at time t a = T. The alarm threshold is also set
priori to some value based on the sample population standard deviation such as h = 2σ, where σ is
estimated from past training data used to produce the SARIMA forecasts.
Building off the CUSUM, a more sophisticated nonparametric detector is the Mann–Whitney
method. Following the approach in [36], we consider a sequence of data samples sorted into descending
order. Each of them has k elements from set X and n − k elements from set Y. We label elements from
Y as a1 , X as a0 . Let Pk,n−k (U ) represent the probability of a sequence in which a1 precedes a0 U times.
We define the following statistic from the data
n k
Uk,n−k = ∑ ∑ ψ ( Xi − X j )
i = k +1 j =1
where ψ(t) is the unit step function (ψ(t) = 0, t < 0 and ψ(t) = 1, t ≥ 0). Consequently, we can write
k n−k
Pk,n−k (U ) = P (U − n + k ) + Pk,n−k−1 (U )
n k−1,n−k n
44
Energies 2020, 13, 4331
k(n−k)
|Uk,n−k − |
Tmax,k,n−k = max { 2
}
1≤ k ≤ n −1 k (n−k)(n+1)
12
T¯T = arg max ( Tmax,k,n−k )
1≤ k ≤ n −1
where Tk,n−k is the test statistic of the detector, Tmax,k,n−k is the maximum value of the Tk,n−k by
traversing k in 1 ≤ k ≤ n − 1 and T¯T is the corresponding k when the Tmax,k,n−k is located. With all
previous discussion, we can introduce a specified control limit hn and create a hypothesis test:
where yi = 1 if the i-th observation is under attack, and yi = 0 if there is no attack. A variety of
classifiers can be used for this learning problem. For detection of DSM attacks, we examine Logistic
Regression (LR), Random Forests (RF), Gaussian Naive Bayes (GNB), Gradient Boosting Classifier
(GBC), and an Artificial Neural Network (ANN). We chose these classifiers because they are all
compelling and have widespread use in both industry and academia; model descriptions of these
45
Energies 2020, 13, 4331
methods can be found in [39,40]. All these models were implemented using the Python package
scikit-learn [41] with each model using default parameters from the package.
GHWHFWLRQZLQGRZ
71 )1 73 )3
Figure 9. Confusion matrix imposed on a time axis of attack predictions vs. true observations.
We calculate accuracy as the ratio of correctly classified data points to total data points
TP + TN
Accuracy =
TP + TN + FP + FN
This measure provides the total classification success of the models. However, accuracy is not
enough to get a full picture of performance alone. Precision is calculated as the ratio of true positive
data points (attacks) to total points classified as attacks:
TP
Precision =
TP + FP
On the other hand, recall, also known as the True Positive Rate (TPR), refers to the portion of
attacks that were recognized correctly
TP
Recall =
TP + FN
Precision values give information about the prediction performance of the algorithms, whereas
recall values measure the degree of attack retrieval. For instance, a recall value equal to 1 signifies that
none of the attacked measurements were misclassified as nominal.
We use one more final measure of total performance of our detectors, the Receiver Operating
Characteristics (ROC) curve. The ROC curve is an assessment that enables visual analysis of the
trade-off between TPR and False Positive Rate (FPR). This can also be seen as the trade-off between the
probability of detection and the probability of false alarm. FPR is defined as follows:
FP
FPR = .
FP + TN
46
Energies 2020, 13, 4331
The ROC curve is constructed by plotting a two-dimensional graph with false positive rate on
the x-axis and true positive rate on the y-axis at various threshold settings. A detection algorithm
produces a (TPR, FPR) pair that corresponds to a single point in the ROC space. An ROC curve that’s
close to the upper axes boundary is reflective of a good detector. The best possible detection method
would produce a point in the upper left corner, coordinate (0,1) of the ROC space. A random prediction
would give a point along a diagonal line from (0,0) to (1,1). Points above this line are considered to
have performance, while points below are considered with performance worse than guessing.
47
Energies 2020, 13, 4331
Figure 10. ACF and PACF plots of the residual series between the SARIMA fit tp training data. From the
plots, we observe that the residuals are stationary.
Figure 11. Q-Q plot of the residual series between the SARIMA fit tp training data. From the plot, we
observe that, for extreme quantiles, the distribution is not Gaussian.
For training, the supervised learning methods for each of the six scenarios, we simulate more
data to ensure proper class learning. We keep the test sets the same, but we extend each training set
doubling its size. We keep the original training set, labeling it as nominal, and then make a copy of it.
In the copy, we split it into three parts; for each part, we add one of the types of attacks at random
levels. We label each observation of this set as under attack. We added the nominal and attacked
training sets to form a new training set with a total of 1344 observations. With our training and test
time series, we then create a set X features and Y labels that could be fed into the supervised learning
classifiers. Each training sample xi ∈ X is composed of 24 h of lagged data, each hour is one feature,
and each label yi ∈ Y corresponds to a class 0 (not under attack) and class 1 (under attack). Together,
we get N =1298 pairs of ( xi , yi ) training samples.
We run our experiments on a computer with an Intel i7 6700 2.6 GHz and 16 GB of RAM.
Implementation of the simulations, experiments, and sequential detectors were done in Python 3.6.
Implementation for SARIMA forecasts was done using the Python package Statsmodels [42],
the supervised learning methods, and confusion evaluation metrics were implemented using the
Scikit-Learn Python package [41], and ROC curves were created using the Matplotlib Python
package [43]. All classifiers used default hyperparameters from Scikit-Learn. We compiled all code
and data used in our work into a Python package titled LehighDSM, which is publicly available on
GitHub [44].
48
Energies 2020, 13, 4331
Experimental results from the six scenarios are reported in Table 2, in the form of accuracy,
recall, and precision metrics, and Figure 12 which showcases six ROC curves, one for each scenario.
All performance measures in Table 2 were calculated using the best thresholds found in the ROC
curves by searching for the shortest distance from each curve to the corner (0,1). As seen in Figure 12c,f
type 3 attacks, under any DSM level, had the lowest detection rates across thresholds with ANN being
the worst performer. In Figure 12a,d, LR yielded the best performance with GNB, and again ANN
being the worst. Type 2 attacks in Figure 12b,e resulted in the best detection with CUSUM, LR, GNB,
and GBC having perfect performance.
(a) Type 1 attack; κ = 0.1. (b) Type 2 attack; κ = 0.1. (c) Type 3 attack; κ = 0.1.
(d) Type 1 attack; κ = 0.9. (e) Type 2 attack; κ = 0.9. (f) Type 3 attack; κ = 0.9.
Figure 12. ROC curves for the six different experiment scenarios.
With the results from Table 2, we have the following conclusions: demonstrated findings imply
that detection of attacks has a higher accuracy with higher levels of DSM participation. This occurs
since higher DSM penetration results in more aggregated attacks. However, we note that an attacker,
if they have perfect knowledge of the grid, could decrease their intensity and make detection more
difficult. Furthermore, the supervised learning classifiers’ performance on average was on par or
better than sequential detection methods. From a detector comparison perspective, the parametric
and nonparametric detectors showcased higher performance on most of the attacks and varying
parameters. In the following section, we do a deeper dive on a comparison between the GLRT and
the nonparametric detectors to derive further inferences on operations paradigms for the attack
detection problem.
49
Energies 2020, 13, 4331
κ 0.1 0.9
Attack Type 1 2 3 1 2 3
LR 89.6 100.0 87.5 87.5 100.0 75.0
RF 77.1 97.9 64.6 77.1 93.8 60.4
Accuracy
50
Energies 2020, 13, 4331
Figure 13. Performance comparison between parametric and nonparametric detector under 200 users.
Figure 14. Performance comparison between parametric and nonparametric detectors with two
users only.
Figure 15. Performance comparison between nonparametric and GLRT detector, N = 2, PFA = 0.1,
P = 20%, k = 30%.
51
Energies 2020, 13, 4331
8. Conclusions
In this work, we study the exploitation of the hypothetical feedback between cyber-enabled
DSM programs, on the consumer side, and dynamic RTP on the utility side. An attacker with
exploitive economic or nefarious intentions, such as causing efficiency loss of energy provision,
can take advantage of the dependency between dynamic pricing and DSM load control. The utility
modifies prices in response to forecasted demand in order to push realized load up or down. This is
done to achieve some target load level to reduce peak load or achieve some other DSM objective.
On the user side, cyber DSM programs then autonomously respond to prices to adjust certain portions
of a user’s load up or down with some given elasticity.
We propose two modes of attacks, false price data injections, and direct load manipulation.
Under a false price data injection, an attacker modifies the RTP that users receive to alter their demand.
Through a direct load manipulation attack, attacker hacks and alters a user’s load profile directly.
In both of these attacks, aggregate load from the grid is modified, which can alter future prices or
demand. We showcase how these two modes of attacks are equivalent and introduce three ways an
attack can occur. The first type is a ramp attack, the second is a sudden attack, and the third is a point
attack. We simulate these types of attacks and review several methods to detect them.
We simulate and examine load-price data under different DSM participation levels with
three types of additive random attacks: ramp, sudden, and point attacks. We then conduct two
tutorial-style detection studies. First, we compare sequential change point and supervised learning
methods. Moreover, in a second study, we compare parametric and nonparametric detectors under
different conditions. These two studies are meant to provide a broad review of various possible
detection approaches. The proposed approaches are pure data driven approaches which can be
implemented without additional hardware assuming that the data are collected at a demand response
management system. The computational cost of the proposed detection methods would depend on
the specific methodologies adopted. Statistical tests can be implemented without incurring significant
computational cost whereas more complex detectors such as ANNs require hardware acceleration for
reasonable performance.
Results conclude that higher amounts of DSM participation can exacerbate attacks and lead
to better detection of such attacks, point attacks are the hardest to detect, and supervised learning
methods produce par or better results than sequential detectors. Amongst the GLRT based parametric
and the nonparametric detector, we show that, with growing window size, the performance of both
the detectors will increase. The performance of the GLRT detector increases faster when the detection
horizon increases due to the MLE mechanism. Thus, if the detection horizon is broad and the user
number is fixed, both the detectors become more potent due to the more information and larger size of
the attack data.
Due to the speculative nature of these types of attacks, we plan to look into many different issues
for future studies. All our detection approaches for attacks depend on predictions. When an accurate
detection is made, the ISO would take measures to mitigate that attack by adjusting price levels
appropriately. In the event of false negatives or false positives, for part of the future work, we envision
to study how an ISO could have a risk cost matrix that would define the amount of price adjustments
for which it would feel comfortable to mitigate attacks taking into account probability confidence levels
on the accuracy of the attacks. Additional new directions include expanding our model to incorporate
day ahead pricing and DSM scheduling, adding models of cost and utility on for a customer and ISO,
expanding our model taking into account grid topology, and modeling more advanced attack goals
such as system failure by power line overload. Additionally, in this work, we only looked at the DSM
goal of strategic load conservation. More complex strategies, such as load shifting, could result in
more complex types of cyber attacks; this also makes for compelling future work.
52
Energies 2020, 13, 4331
References
1. Faruqui, A.; Sergici, S. Household Response to Dynamic Pricing of Electricity-a Survey of The Empirical
Evidence. 2010. Available online: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1134132
(accessed on 5 April 2019).
2. Grochocki, D.; Huh, J.H.; Berthier, R.; Bobba, R.; Sanders, W.H.; Cárdenas, A.A.; Jetcheva, J.G. AMI threats,
intrusion detection requirements and deployment recommendations. In Proceedings of the 2012 IEEE
Third International Conference on Smart Grid Communications (SmartGridComm), Tainan, Taiwan, 5–8
November 2012; pp. 395–400.
3. Khan, A.R.; Mahmood, A.; Safdar, A.; Khan, Z.A.; Khan, N.A. Load forecasting, dynamic pricing and DSM
in smart grid: A review. Renew. Sustain. Energy Rev. 2016, 54, 1311–1322. [CrossRef]
4. Xie, L.; Mo, Y.; Sinopoli, B. Integrity data attacks in power market operations. IEEE Trans. Smart Grid 2011,
2, 659–666. [CrossRef]
5. Tan, R.; Badrinath Krishna, V.; Yau, D.K.; Kalbarczyk, Z. Impact of integrity attacks on real-time pricing
in smart grids. In Proceedings of the 2013 ACM SIGSAC Conference on Computer & Communications
Security, Berlin, Germany, 4–8 November 2013; ACM: New York, NY, USA, 2013; pp. 439–450.
6. Palensky, P.; Dietrich, D. Demand side management: Demand response, intelligent energy systems, and
smart loads. IEEE Trans. Ind. Inform. 2011, 7, 381–388. [CrossRef]
7. Logenthiran, T.; Srinivasan, D.; Shun, T.Z. Demand side management in smart grid using heuristic
optimization. IEEE Trans. Smart Grid 2012, 3, 1244–1252. [CrossRef]
8. Mohsenian-Rad, A.H.; Wong, V.W.; Jatskevich, J.; Schober, R.; Leon-Garcia, A. Autonomous demand-side
management based on game-theoretic energy consumption scheduling for the future smart grid. IEEE Trans.
Smart Grid 2010, 1, 320–331. [CrossRef]
9. Gelazanskas, L.; Gamage, K.A. Demand side management in smart grid: A review and proposals for future
direction. Sustain. Cities Soc. 2014, 11, 22–30. [CrossRef]
10. Batista, A.C.; Batista, L.S. Demand Side Management using a multi-criteria epsilon-constraint based exact
approach. Expert Syst. Appl. 2018, 99, 180–192. [CrossRef]
11. Dudek, G. Pattern-based local linear regression models for short-term load forecasting. Electr. Power
Syst. Res. 2016, 130, 139–147. [CrossRef]
12. Pappas, S.S.; Ekonomou, L.; Karampelas, P.; Karamousantas, D.; Katsikas, S.; Chatzarakis, G.; Skafidas, P.
Electricity demand load forecasting of the Hellenic power system using an ARMA model. Electr. Power
Syst. Res. 2010, 80, 256–264. [CrossRef]
13. Christiaanse, W. Short-term load forecasting using general exponential smoothing. IEEE Trans. Power
Appar. Syst. 1971, PAS-90, 900–911. [CrossRef]
14. Taylor, J.W. Triple seasonal methods for short-term electricity demand forecasting. Eur. J. Oper. Res. 2010,
204, 139–152. [CrossRef]
15. Kavousi-Fard, A.; Samet, H.; Marzbani, F. A new hybrid modified firefly algorithm and support vector
regression model for accurate short term load forecasting. Expert Syst. Appl. 2014, 41, 6047–6056. [CrossRef]
16. Baliyan, A.; Gaurav, K.; Mishra, S.K. A review of short term load forecasting using artificial neural network
models. Procedia Comput. Sci. 2015, 48, 121–125. [CrossRef]
17. Kong, W.; Dong, Z.Y.; Jia, Y.; Hill, D.J.; Xu, Y.; Zhang, Y. Short-term residential load forecasting based on
LSTM recurrent neural network. IEEE Trans. Smart Grid 2017, 10, 841–851. [CrossRef]
53
Energies 2020, 13, 4331
18. Dudek, G. Short-term load forecasting using random forests. In Intelligent Systems’2014, Proceedings of the
7th IEEE International Conference Intelligent Systems IS’2014, Warsaw, Poland, 24–26 September 2014; Springer:
Berlin/Heidelberg, Germany, 2015; pp. 821–828.
19. Li, S.; Goel, L.; Wang, P. An ensemble approach for short-term load forecasting by extreme learning machine.
Appl. Energy 2016, 170, 22–29. [CrossRef]
20. Chakraborty, S.; Ito, T.; Senjyu, T. Smart pricing scheme: A multi-layered scoring rule application.
Expert Syst. Appl. 2014, 41, 3726–3735. [CrossRef]
21. Nazar, N.; Abdullah, M.; Hassan, M.; Hussin, F. Time-based electricity pricing for Demand Response
implementation in monopolized electricity market. In Proceedings of the 2012 IEEE Student Conference on
Research and Development (SCOReD), Pulau Pinang, Malaysia, 5–6 December 2012; IEEE: Piscataway, NJ,
USA, 2012; pp. 178–181.
22. Zimmerman, R.D.; Murillo-Sánchez, C.E.; Gan, D. MATPOWER: A MATLAB Power System Simulation Package;
Manual, Power Systems Engineering Research Center: Ithaca, NY, USA, 1997; Volume 1.
23. Chassin, D.P.; Schneider, K.; Gerkensmeyer, C. GridLAB-D: An open-source power systems modeling and
simulation environment. In Proceedings of the 2008 IEEE/PES Transmission and Distribution Conference
and Exposition, Chicago, IL, USA, 21–24 April 2008; IEEE: Piscataway, NJ, USA, 2008; pp. 1–5.
24. Barker, S.; Mishra, A.; Irwin, D.; Cecchet, E.; Shenoy, P.; Albrecht, J. Smart*: An open data set and tools for
enabling research in sustainable homes. Sustkdd August 2012, 111, 108.
25. Brockwell, P.J.; Davis, R.A.; Calder, M.V. Introduction to Time Series and Forecasting; Springer:
Berlin/Heidelberg, Germany, 2002; Volume 2.
26. Gamerman, D.; Lopes, H.F. Markov Chain Monte Carlo: Stochastic Simulation for Bayesian Inference; Chapman
and Hall/CRC: Boca Raton, FL, USA, 2006.
27. Politis, D.N.; White, H. Automatic block-length selection for the dependent bootstrap. Econom. Rev. 2004,
23, 53–70. [CrossRef]
28. Thimmapuram, P.R.; Kim, J.; Botterud, A.; Nam, Y. Modeling and simulation of price elasticity of demand
using an agent-based model. In Proceedings of the 2010 Innovative Smart Grid Technologies (ISGT),
Gothenburg, Sweden, 19–21 January 2010; IEEE: Piscataway, NJ, USA, 2010; pp. 1–8.
29. Liang, G.; Zhao, J.; Luo, F.; Weller, S.R.; Dong, Z.Y. A review of false data injection attacks against modern
power systems. IEEE Trans. Smart Grid 2017, 8, 1630–1638. [CrossRef]
30. Tartakovsky, A.; Nikiforov, I.; Basseville, M. Sequential Analysis: Hypothesis Testing and Changepoint Detection;
Chapman and Hall/CRC: Boca Raton, FL, USA, 2014.
31. Key, S. Fundamentals of Statistical Signal Processing; Volume ii: Detection Theory; Prentice Hall PTR:
Upper Saddle River, NJ, USA, 1993.
32. Hink, R.C.B.; Beaver, J.M.; Buckner, M.A.; Morris, T.; Adhikari, U.; Pan, S. Machine learning for power
system disturbance and cyber-attack discrimination. In Proceedings of the 2014 7th International Symposium
on Resilient Control Systems (ISRCS), Denver, CO, USA, 19–21 August 2014; pp. 1–8.
33. Ozay, M.; Esnaola, I.; Vural, F.T.Y.; Kulkarni, S.R.; Poor, H.V. Machine learning methods for attack detection
in the smart grid. IEEE Trans. Neural Netw. Learn. Syst. 2016, 27, 1773–1786. [CrossRef]
34. Esmalifalak, M.; Liu, L.; Nguyen, N.; Zheng, R.; Han, Z. Detecting stealthy false data injection using machine
learning in smart grid. IEEE Syst. J. 2017, 11, 1644–1652. [CrossRef]
35. Apley, D.W.; Shi, J. The GLRT for statistical process control of autocorrelated processes. IIE Trans. 1999,
31, 1123–1134. [CrossRef]
36. Mann, H.B.; Whitney, D.R. On a test of whether one of two random variables is stochastically larger than the
other. Ann. Math. Stat. 1947, 18, 50–60. [CrossRef]
37. Yu, M. A nonparametric adaptive CUSUM method and its application in network anomaly detection. Int. J.
Adv. Comput. Technol. 2012, 4, 280–288.
38. Hawkins, D.M.; Deng, Q. A nonparametric change-point control chart. J. Qual. Technol. 2010, 42, 165–173.
[CrossRef]
39. Bishop, C. Pattern Recognition and Machine Learning. In Pattern Recognition and Machine Learning; Springer:
Berlin/Heidelberg, Germany, 2006.
40. James, G.; Witten, D.; Hastie, T.; Tibshirani, R. An Introduction to Statistical Learning; Springer:
Berlin/Heidelberg, Germany, 2013; Volume 112.
54
Energies 2020, 13, 4331
41. Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.;
Weiss, R.; Dubourg, V.; et al. Scikit-learn: Machine learning in Python. J. Mach. Learn. Res. 2011,
12, 2825–2830.
42. Seabold, S.; Perktold, J. Statsmodels: Econometric and statistical modeling with python. In Proceedings of
the 9th Python in Science Conference, SciPy Society Austin, Austin, TX, USA, 28 June–3 July 2010; Volume 57,
p. 61.
43. Hunter, J.D. Matplotlib: A 2D graphics environment. Comput. Sci. Eng. 2007, 9, 90–95. [CrossRef]
44. Hatalis, K. LehighDSM Python Package. 2019. Available online: https://github.com/hatalis/
DSM (accessed on 1 August 2020).
c 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
55
energies
Article
Simulation of Incidental Distributed Generation
Curtailment to Maximize the Integration of
Renewable Energy Generation in Power Systems
Ingo Liere-Netheler *, Frank Schuldt, Karsten von Maydell and Carsten Agert
DLR Institute of Networked Energy Systems, Carl-von-Ossietzky-Str. 15, 26129 Oldenburg, Germany;
[email protected] (F.S.); [email protected] (K.v.M.); [email protected] (C.A.)
* Correspondence: [email protected]; Tel.: +49-441-99906-457
Abstract: Power system security is increasingly endangered due to novel power flow situations
caused by the growing integration of distributed generation. Consequently, grid operators are forced
to request the curtailment of distributed generators to ensure the compliance with operational limits
more often. This research proposes a framework to simulate the incidental amount of renewable
energy curtailment based on load flow analysis of the network. Real data from a 110 kV distribution
network located in Germany are used to validate the proposed framework by implementing best
practice curtailment approaches. Furthermore, novel operational concepts are investigated to improve
the practical implementation of distributed generation curtailment. Specifically, smaller curtailment
level increments, coordinated selection methods, and an extension of the n-1 security criterion are
analyzed. Moreover, combinations of these concepts are considered to depict interdependencies
between several operational aspects. The results quantify the potential of the proposed concepts to
improve established grid operation practices by minimizing distributed generation curtailment and,
thus, maximizing power system integration of renewable energies. In particular, the extension of the
n-1 criterion offers significant potential to reduce curtailment by up to 94.8% through a more efficient
utilization of grid capacities.
Keywords: power system operation; power system security; renewable energy integration; load flow
analysis; congestion management; distributed generation curtailment
1. Introduction
The integration of distributed generators (DGs), such as wind turbines or photovoltaics, causes
formerly unknown load flow situations that can endanger the operational security of power systems.
In Germany, particularly in rural areas, the increasing installation of DG systems coincides with
low hosting capacities due to the historical dimensioning of the network [1]. Hence, violations of
operational security limits, such as overvoltage problems or thermal overloadings of grid components,
can occur more frequently [2]. To prevent this, grid operators are required to enhance existing
congestion management regimes. In the long term, grid expansion and reinforcement measures are
required to ensure larger DG penetrations without jeopardizing the operational security of power
systems. However, several studies have analyzed the grid expansion demand in Germany and
agreed that the provision of a system without any bottlenecks solely through grid expansion or
reinforcement is neither the most efficient solution at the distribution level (≤110 kV) [3] nor at the
transmission level (≥220 kV) [4]. Consequently, an efficient utilization of existing grid capacities
through operational congestion management approaches will become more relevant in the future in
times of high renewable feed-in [5].
58
Energies 2020, 13, 4173
practice approaches are first implemented into the framework and the results derived from load flow
simulations are compared to curtailment measures from real grid operation. Another main contribution
of this work is the proposal of three novel operational concepts to improve grid capacity utilization
in critical operating states. These concepts include increasing the grid operator’s ability to control
DG units by implementing smaller curtailment level increments, improving DG selection methods
with a special focus on increasing the coordination between different grid operators and adapting
the deterministic n-1 security criterion to maximize DG integration. Furthermore, combinations of
the proposed concepts are investigated to demonstrate possible interdependencies. The concepts are
implemented into the simulation framework and their potential to reduce the incidental amount of
renewable energy curtailment compared to the best practice is quantified. Hence, by combining the
proposal of different operational concepts and their implementation into the validated model of a real
distribution network, this work extends the existing research in the field by quantifying the influence of
several operational aspects on DG curtailment and by depicting potentials to maximize DG integration
into power systems by enhancing the utilization of existing transmission capacities.
The remainder of the paper is structured as follows. Section 2 introduces the simulation framework
and discusses different operational aspects influencing DG curtailment. The current best practice
regarding these aspects in grid operation and concepts to improve the practical implementation are
discussed. Section 3 includes a description of the 110 kV distribution grid model and the validation
based on the simulation of best practice curtailment as a reference scenario. Furthermore, scenarios
for the implementation of the proposed concepts are defined and simulation results for each of the
scenarios are presented, summarized, and discussed. Section 4 concludes the most relevant findings of
the paper.
2.1. DG Controllability
Grid operators can request the reduction of power generation of DG units but are not allowed to
directly control these. Instead, generation units receive curtailment requests and are required to realize
the requested power output reduction. The maximum permitted power injection Pinj,max of a DG unit
requested by a grid operator can be formulated as:
where Pnom describes the nominal capacity of the DG unit and lcurt the requested percentage level
included in the feasible set of all possible curtailment levels Ωcurt . Recently, the established granularity
of curtailment levels is four discrete output levels [1]:
Here, the curtailment level of 1 is used to cancel previously requested commands and allow the
maximum possible feed-in again. The ability of grid operators to request the most efficient amount
of DG curtailment is thus limited due to their limited control opportunities. As a consequence,
the practically curtailed power feed-in can exceed the theoretically necessary amount of power
reductions to solve congestions in certain situations, which can result in an inefficient utilization of
transmission capacities [1]. Therefore, implementing an extended feasible set of curtailment levels
59
Energies 2020, 13, 4173
with smaller increments can potentially enable the request of more efficient measures and reduce the
amount of incidental renewable energy curtailment compared to the best practice.
where ΔPbranch and ΔQbranch are the vectors of real and reactive power flow changes over all considered
grid components, such as lines or transformers, ΔPinj and ΔQinj are the vectors of changes in nodal
real and reactive power injections, and M is the sensitivity matrix of the system. Each value of M
approximates the effect of a change in power injection in a certain node in the network on a certain
grid component for one specific power flow solution. Accordingly, power flow sensitivities can be
used to determine the most sensitive DG unit to certain congested grid components.
However, the iterative selection scheme in combination with discrete curtailment level increments
minimizes the number of necessary curtailment requests but does not necessarily lead to the minimum
possible amount of curtailed power. This is due to the fact that the most sensitive unit can, e.g., be a
large wind farm and only be curtailed in relatively large discrete steps, whereas smaller units with
smaller possible curtailment levels with an equally large sensitivity are not considered.
To address this issue and enable the request of more efficient curtailment measures, the problem of
selecting DG units can also be implemented in the simulation framework as an optimization problem
aiming at minimizing the total amount of power reductions [20]:
nu
minf(x) = ΔPi (3)
x
i=1
where nu describes the total number of available DG units and ΔPi the curtailed power of unit i,
which amounts to:
ΔPi = Pi,init − Pinj,max (4)
60
Energies 2020, 13, 4173
where Pi,init describes the initial power infeed and Pinj,max the requested maximum power injection
from Equation (1). Accordingly, the decision variables are discrete due to the discrete curtailment level
increments which results in a mixed integer programming problem. Real power flow restrictions are
displayed by the addition of constraints for each grid component:
where Pcomp,init describes the real power flow over the considered grid component in the initial grid
state, ΔPcomp the change in the real power flow due to DG curtailment, and Pcomp,max the maximum
permitted power flow over the grid component. The power flow change can be written in terms of the
power flow sensitivities of each DG unit on the respective grid component:
nu
∂Pcomp
ΔPcomp = · ΔPi (6)
∂Pi
i=1
This optimization approach solely minimizes the amount of curtailed power, whereas the number
of curtailment requests is not directly considered in the objective function. On the contrary, the best
practice iterative approach curtails the least number of DGs and neglects the amount of reduced power
output. However, grid operators might practically want to reach a certain objective, such as minimizing
curtailment, with a limited number of control actions [21]. Therefore, a hybrid approach is also
proposed as a candidate selection method in this paper, which aims at reducing the resulting renewable
energy curtailment while limiting the complexity of the practical implementation by limiting the
number of curtailment requests. The three considered selection methods are compared schematically
regarding the amount of curtailment and the number of requests in Figure 1.
Figure 1. Schematic comparison of distributed generator (DG) selection methods regarding the resulting
amount of curtailed energy and number of curtailment requests.
The hybrid selection method takes the number of curtailment requests into account by
implementing a penalty term for each curtailment level leading to adjustments compared to the
initial grid state (ΔPi > 0) in the optimization objective (Equation (3)). This can be interpreted as
minimizing the total costs of curtailment consisting of a variable part depending on the amount of
curtailment and a fixed part for each measure. The variable costs describe the financial reimbursement
of curtailed DGs whereas the fixed proportion represents organizational expenses of grid operators
(e.g., costs of billing curtailment). The optimization objective is thus adapted as follows:
u
n
minf(x) = ΔPi · cvar · Δt + pc (ΔPi )
x i=1 (7)
0 if ΔPi = 0
with pc =
cfix else
61
Energies 2020, 13, 4173
The variable costs cvar are estimated based on the total amount of curtailed energy in Germany in
2017 and the respective amount of financial reimbursements, which amount to 610 M€ for 5518 GWh
of curtailed energy [11], resulting in variable costs of 110 €/MWh. The penalty term pc is set to 0 if
no curtailment is requested and to a fixed cost parameter for curtailment requests. This fixed cost
parameter cfix directly influences the solution of the optimization problem. The higher this parameter,
the higher the solution would prioritize the number of curtailment requests towards the amount of
curtailed energy and vice versa. Due to the available data set in this work, the time interval Δt is set to
15 min or 0.25 h, respectively.
Following the existing operative cascade (left side of Figure 2), in the first step, the TSO assesses
security criteria in its responsibility area (Step 1). If congestions, e.g., overloadings of transformers
linking the transmission and distribution system, are identified, the TSO is allowed to send curtailment
requests to the respective DSO. The DSO is then responsible for forwarding these signals to the respective
DG units (Step 2). Additionally, the DSO determines the security state in its own responsibility area
(Step 3), which can lead to additional curtailment requests (Step 4).
To avoid redundant or unnecessary curtailment requests, the proposed procedure aims at finding
the global optimal amount of DG curtailment (right side of Figure 2). Therefore, grid restrictions
of multiple system levels are considered simultaneously. Consequently, the TSO does not request
DG curtailment directly but rather transmits the necessary power adjustments on all congested grid
components ΔPcomp to the DSO (Step 2). These are added to the optimization problem as constraints
and complemented by additional security constraints due to distribution network restrictions (Step 3).
As a consequence, the optimization problem can be solved once considering all necessary security
constraints simultaneously (Step 4). The operative cascade is also extensible for multiple grid operators
on the distribution grid level by adding additional hierarchy levels. Because this work focuses on the
110 kV distribution level, subordinate grid levels are not represented in detail and the operational
cascade only consists of two system levels.
62
Energies 2020, 13, 4173
Due to the uncertain occurrence of contingency events, the increase of the maximum permissible
real power flow in n-1 events Pcomp,max,n-1 above the physical limit of the respective grid component is
not necessarily accompanied by actual overloadings.
63
Energies 2020, 13, 4173
Particularly in times of high wind occurrence accompanied by a high amount of DG feed-in, the demand
for congestion management measures can be reduced due to higher admissible currents on transmission
lines resulting from additional cooling effects [27].
The application of dynamic line rating is already well established in practice. Therefore, it is
not considered as a novel operational concept in this work but rather integrated into the simulation
environment to increase the accuracy of simulation results in comparison to real data from grid
operation practice because dynamic line rating has a significant impact on the amount of incidental DG
curtailment. Based on a meteorological dataset, the maximum transferable current of a transmission
line is calculated dynamically using the approach proposed by the Institute of Electrical and Electronics
Engineers (IEEE) [29]. Thus, the maximum real power transfer capabilities in Equations (5), (8), and (9)
change dynamically.
64
Energies 2020, 13, 4173
respective 110 kV transmission lines. For this, each transmission line is assigned to the substation it
has the shortest geographical distance to.
The balance of the total real and reactive power within power flow simulations in the region is
ensured by the reference node on the 220 kV level. The parameters of transmission lines on the 110 kV
level are provided by the respective grid operator (GO). Furthermore, the line lengths of the 220 kV
lines are derived from a publicly available GIS database and the electrical parameters are extracted
from a typical AL/ST 265/35 220 kV overhead line model [34]. Capacities of 220 kV/110 kV transformers
are also given through grid operator information, whereas the short-circuit voltage for power flow
calculations is specified to 13% based on [35]. A schematic graphical representation of the model is
given by Table 1 including an overview of all numbered model objects and the respective data sources.
Table 1. Schematic visualization of grid simulation model including the data sources of different
model parameters.
Furthermore, the iterative curtailment method is implemented and the curtailment requests by
TSO and DSO are derived independently. In addition, the maximum loading of grid components in
contingency situations is set to 100%. In the following, this scenario configuration is also considered as
the reference scenario to evaluate the effectiveness of the proposed concepts.
The published data set of historical curtailment measures includes starting time, level (1, 0.6,
0.3, 0), ending time, and the receiving substation of each request. Due to the frequent occurrence of
curtailment requests in the historical data set and the availability of comprehensive time series data,
February 2016 was selected as the investigated time period in this work.
Hence, the historical curtailment of DG units is determined using the respective time series and the
historical data set with a resolution of 15 min. For the same time horizon, the incidental curtailment using
the best practice approaches is simulated. Therefore, in each time step, a contingency analysis is executed
to assess the security state of the network. In the alert state, DG curtailment is requested as a preventive
65
Energies 2020, 13, 4173
congestion management measure. In case of multiple critical contingencies, grid operators usually
initiate countermeasures ordered by the severity of operational limit violations [22]. Subsequently,
the amount of curtailment is determined for the most severe contingency at first. This is iteratively
repeated until all congestions in each possible n-1 event are terminated.
Firstly, according to the cascading grid operation approach (Figure 2), congestions on the TSO
level are identified by performing a contingency analysis of all 220/110 kV transformers in the grid
area. For all insecure grid states, the resulting amount of DG curtailment is derived by the simulation
framework. Since most curtailment measures contained in the historical data set result from congestions
of transformers connecting the transmission and distribution system it is assumed that the TSO requests
FIM measures only to prevent overloadings of transformers. Thus, congestions of 220 kV transmission
lines are not considered in this work. After deriving curtailment requests by the TSO, another
contingency analysis is executed to determine the security state of the 110 kV system and derive
curtailment requests by the DSO if necessary. Although the curtailment demand is determined for
each installed DG unit individually, simulation results are validated for the total amount of curtailment
in the investigated region to assess the overall amount of incidental renewable energy curtailment.
Table 2. Depiction of different binary states to assess time steps resulting from combinations of the
presence or absence of simulated and historical curtailment measures.
Simulated Curtailment
Historical Curtailment True False
True a b
False c d
To increase the comparability of the simulation results achieved in this work, these are put
in relation to the study of [2] as a reference. There, the authors developed a congestion forecast
model based on autonomous neural networks for a German transmission network and validated it
based on historical data about congestion management measures. Although neither these measures
are quantified nor multiple system levels are considered in [2], the results can be compared to the
simulations carried out in this work regarding the ability to represent historical congestion situations.
The precision (p), recall (r), and f1-score (f) are used as similarity indices between simulation and
historical values in [2], which are defined as follows:
a
p= (10)
a+c
a
r= (11)
a+b
p·r
f = 2· (12)
p+r
66
Energies 2020, 13, 4173
For each of those indices, a value of 1 indicates total similarity. Additionally, in this work the
mean absolute error (MAE) and the root mean square error (RMSE) are used as error metrics to assess
the accuracy of curtailment simulations, which are defined as follows [38]:
1 1
n n
MAE = · |ei | = · Pcurt,historic − Pcurt,simulated (13)
n n
k=1 k=1
1 2 1 2
n n
RMSE = · |ei | = · Pcurt,historic − Pcurt,simulated (14)
n n
k=1 k=1
where n describes the number of simulated time steps, Pcurt,historic the amount of power reduction in
the historical data set, and Pcurt,simulated the simulated power curtailment per time step.
Due to the independent simulation of curtailment requests by the TSO and the DSO, the similarity
indices and error metrics are calculated separately for both grid operators. In addition, the aggregated
results, defined as the sum of TSO and DSO curtailment requests, are determined. Table 3 depicts the
resulting similarity indices from reference [2] compared to those achieved in this paper. Furthermore,
the resulting MAE and RMSE from this work are depicted. In order to visualize the significance of
the absolute values for the error metrics, the maximum amount of simulated power curtailment is
also depicted.
Table 3. Resulting similarity and error metrics between historical and simulated DG curtailment in
comparison to results published in [2].
This Work
[2]
DSO TSO Aggregated
Precision p 0.699 0.462 0.812 0.930
Recall r 0.636 0.263 0.748 0.760
f1-score f 0.648 0.335 0.779 0.837
Maximum curtailment [MW] / 162.844 389.393 389.393
MAE [MW] / 7.892 19.985 22.484
RMSE [MW] / 26.353 51.396 51.948
According to Table 3, regarding all similarity values, the simulation of TSO curtailment is more
accurate compared to [2], whereas DSO measures are depicted less accurately. The results also show
that the aggregated consideration has the highest values in all similarity indices. Both error metrics
have the lowest values for the DSO curtailment and the highest values for the aggregated consideration.
Consequently, the similarity metrics indicate a lack of accuracy regarding the simulation of
curtailment requests by the DSO. This is due to several reasons. In particular, the recall r of only 26.3%
indicates a poor model performance. Since this index is dependent on the inaccurate state b, it can
be concluded that a large number of historical DSO requests do not appear in the simulation results,
because the corresponding congestions are already resolved by the prior determined TSO curtailment
demand. In 86% of the times assigned to state b, TSO requests are simulated. Hence, congestions in
the distribution network are often already resolved and no additional DSO curtailment is necessary.
On the contrary, 58% of times assigned to state c are accompanied by an underestimation of TSO
curtailment requests. Consequently, the precision p is also influenced by this dependency but to a
smaller extent. Accordingly, this index indicates a larger similarity of 46.2%. The MAE values are
relatively small compared to the maximum occurring power reductions in all three cases. The RMSE
values are considerably larger than the MAE values. Hence, it can be concluded that the curtailment
simulation values occasionally include larger deviations from the historical values, since the RMSE
is more sensitive to large errors than the MAE [38]. However, both error metrics are relatively small
compared to the maximum occurring values. The differences between simulation results and historical
67
Energies 2020, 13, 4173
data are further visualized by the time series comparison of requests by the TSO, DSO and the
aggregated amount of DG curtailment in Figure 3. The time series confirm the high degree of temporal
coincidence between simulation and historical values. Occasionally, larger deviations are apparent,
which contribute to the larger RMSE values.
Figure 3. Time series comparison of simulated and historical curtailment measures requested by
the transmission system operators (TSO) (top), distribution system operators (DSO) (middle) and
aggregated (bottom).
Overall, the simulation of curtailment measures requested by the DSO is highly influenced by
the simulation of the TSO demand beforehand, which also contains errors, as can be seen in Table 3.
Nevertheless, the aggregated results gain the highest similarity values and only slightly higher values
in the error metrics. This confirms that some measures historically requested by the DSO are mistakenly
simulated as measures by the TSO and vice versa. The relatively low values of MAE and RMSE and
the graphical time series comparison in Figure 3 confirm the accuracy of simulated DG curtailment,
particularly considering the aggregated results.
Possible sources of errors decreasing the accuracy of curtailment simulations are the meteorological
data which affect both the wind power time series and the dynamic line rating approach. As stated
in [39], the quality of the dynamic line rating approaches could be improved by further increasing the
maximum geographical resolution of 2.8 km of established meteorological models, which are also used
in this work. This is especially relevant for the 110 kV lines and, thus, for the curtailment requests by
the DSO. Additionally, due to the available data basis, the simulation was carried out for 15-min time
intervals. However, grid security is typically assessed more frequently (e.g., in intervals of 2 [40] or
5 [41] min) by state estimation algorithms. Therefore, the practical security assessment fluctuates more
than in the simulations, which can also affect the derivation of curtailment measures.
68
Energies 2020, 13, 4173
Figure 4. Comparison of renewable energy curtailment requested by the TSO, DSO, and aggregated.
The curtailed energy due to requests by the TSO is slightly underestimated by 1.78 GWh (6.5%),
whereas the demand from the DSO is overestimated by 1.43 GWh (45.6%). Consequently, the combined
energy amount is underestimated by 0.35 GWh or 1.1%. Accordingly, the model is able to project the
historical curtailment measures regarding the total amount of curtailed energy very well. The share of
TSO requests in the aggregated amount is comparably large in the simulation (84.89%) and the historical
(89.74%) dataset. Consequently, most curtailment requests are due to congestions of transformers
linking the distribution and transmission system.
Overall, the similarity measures, time series analysis, and evaluation of the curtailed energy
indicate the capability of the simulation framework to represent the demand and number of curtailment
measures in the investigated region to a reasonable extent. Despite inaccuracies regarding the
allocation of curtailment measures to the requesting grid operator, the aggregated simulation results
are particularly promising. Therefore, the developed simulation framework is assumed to achieve
reasonable results regarding the simulation of DG curtailment. The resulting amount of renewable
energy curtailment of 30.27 GWh in the reference scenario is thus further used to assess the extent to
which the proposed operational concepts can increase grid capacity utilization and, thereby, decrease
the incidental amount of DG curtailment.
69
Energies 2020, 13, 4173
Table 4. Overview of investigated scenarios and adaptations regarding the specification of curtailment
level increments, DG selection method, and maximum permissible loading in contingency states
compared to best practice approaches.
Specification of Aspect
Scenario Description Feasible Set of Curtailment DG Selection Method and Maximum Permissible
Levels Ωcurt Grid Operator Coordination Loading in n-1 Events
Iterative selection method,
Best practice reference
[1,0.6,0.3,0] independent curtailment 100%
scenario
requests by TSO and DSO
Adaptations compared to best practice approaches
Scenario 1: Extended
X
feasible set
Scenario 2: Coordinated
X X
DG selection methods
Scenario 3: Increased
loadability in X X
contingency events
Firstly, the reference scenario includes four discrete curtailment levels, the iterative selection
scheme accompanied by independent curtailment request of TSO and DSO, and a maximum loadability
of grid components of 100% in contingency events. The first scenario quantifies the benefit of decreasing
curtailment level increments and thereby enabling the grid operator to control DG units to a larger
extent. The DG selection procedure and maximum permissible loading in contingency events remain
unmodified in comparison to the reference scenario. The second scenario supplements the larger
control opportunities by also implementing the optimization and hybrid approaches as candidate DG
selection methods and thereby improving the coordination between TSO and DSO. Finally, the last
scenario quantifies reduction potentials of DG curtailment dependent on increasing the maximum
loadability of grid components in n-1 events. This scenario also considers the different DG selection
methods. In the following, simulation results for each of the described scenarios are presented.
Table 5. Resulting renewable energy curtailment for different incremental sizes of curtailment levels.
Curtailment requested by the DSO can be reduced by 0.41 GWh (9%) for 10% and 0.58 GWh
(12.7%) for 1% increments. Moreover, TSO curtailment can be reduced by 0.51 GWh (2%) and 0.72 GWh
(2.8%). Accordingly, the aggregated savings amount to 0.92 GWh (3%) and 1.3 GWh (4.3%) for 10%
70
Energies 2020, 13, 4173
and 1% increments, respectively. The results show that the reduction of the total amount of curtailment
is comparable between DSO and TSO. Consequently, the relative savings achievable by the DSO are
distinctly higher. Due to the iterative selection scheme applied in this scenario, the benefit of smaller
curtailment level increments is limited to the least sensitive DG unit affected by the respective request,
because each of the other affected units is fully curtailed. Hence, the relative reduction of curtailed
energy is reduced significantly only in the case of less severe congestions that induce a small amount
of total power curtailment.
This effect is illustrated in Figure 5, which shows the relative reduction of the aggregated amount of
DG curtailment requested by TSO and DSO gained by 10% and 1% increments per time step depending
on the respective amount of total power reduction. The relative savings are related to the best practice
increments. It becomes clear that the highest relative savings are achieved at the less severe congestions.
Smaller increments tend to only have minor advantages in situations with large amounts of power
reductions. As depicted by Table 5, the relative reduction potential is thus considerably lower for TSO
requests, since these exhibit larger amounts of power adjustments compared to DSO curtailment.
Figure 5. Relative reduction of curtailed DG power feed-in achieved by 10% and 1% increments in
comparison to established curtailment levels.
Furthermore, a few measures show negative savings for 10% and 1% increments in Figure 5.
This is due to the fact that TSO curtailment also reduces congestions at the distribution level. Therefore,
the avoidance of TSO requests can inevitably lead to a higher amount of curtailment demand at the
distribution grid level. Consequently, the aggregated amount resulting from smaller increments can
even exceed the demand induced by larger increments in certain situations.
In total, the advantage of decreasing the curtailment level increments is its easy implementation
into existing grid control strategies, since modern communication standards and technologies already
enable a finer granularity in DG controllability compared to FIM [20]. However, the simulation results
in this scenario imply that the potential to reduce the curtailed energy is limited to a few percent due to
the frequent occurrence of severe congestions in real grid operation practice, especially in combination
with the iterative DG selection approach. Therefore, the benefit of this concept is also assessed in
combination with implementing the novel DG selection methods in the following scenario.
71
Energies 2020, 13, 4173
the number of necessary control signals is also considered to evaluate the communication demand
associated with the implementation of this concept.
To implement the hybrid approach (Equation (7)), firstly a suitable choice for the fixed cost
parameter cfix in the objective function has to be determined. Therefore, this parameter is varied
iteratively between 100 € and 500 € in steps of 100 € and analyzed regarding its influence on the curtailed
energy and the number of requested measures. Both variables are depicted in Table 6 dependent on
the respective fixed cost parameter. In addition, the results obtained from the reference scenario are
shown. As expected, the lowest fixed cost parameter (100 €) causes the lowest amount of curtailed
energy but the largest number of measures. In contrast, the highest fixed cost parameter of 500 €
prioritizes the number of measures over the amount of curtailed energy. Compared to the reference
case, an increase in the number of curtailment measures for all parameters is present, whereas the
amount of curtailed energy is lower. For the further analyses in this work, the parameter is set to a value
of 300 €, since this causes a considerable reduction of curtailed energy (2.79 GWh) while increasing the
number of curtailment measures acceptably (49.43%) compared to the iterative approach.
Table 6. Renewable energy curtailment and number of curtailment requests for different fix cost
parameter specifications.
The three different DG selection methods are compared regarding the incidental amount of
renewable energy curtailment and the number of resulting curtailment requests. In addition,
the approaches are combined with the concept of smaller curtailment level increments to show
possible interdependencies between both concepts. Therefore, nine different scenario specifications
are investigated. The resulting energy curtailment and the number of measures are visualized in
Figure 6. The different increments are represented by varying colors, and the number and the amount
of curtailment are reflected by different markers (‘o’,’x’), respectively.
Figure 6. Resulting renewable energy curtailment and number of curtailment requests for different DG
selection methods and incremental sizes of curtailment levels.
72
Energies 2020, 13, 4173
The iterative approach yields the least number of measures and the highest amount of curtailed
energy. Decreasing the curtailment increments mainly reduces the curtailed energy (Section 3.4.1.),
whereas the number of measures nearly remains constant. In contrast, the implementation of the
optimization approach reduces the amount of curtailed energy by up to 3 GWh (10%) at the expense of
1660 additional curtailment requests. In contrast to the iterative approach, the energy amount remains
constant with decreasing increments whereas the number of measures can be reduced considerably.
This is due to the fact that smaller increments result in a higher flexibility regarding the controllability
of DGs. Therefore, it is less probable that curtailment measures are firstly cancelled and, subsequently,
reactivated if congestions return. Smaller increments allow the grid operator to carefully return the
DGs feed-in to the initial value while monitoring the operational security limits of the network. Finally,
the hybrid approach nearly obtains the same amount of curtailed energy compared to the optimization
approach (dashed trend line) and a similar number of curtailment measures compared to the iterative
approach (dotted trend line). Moreover, the dependence of this approach on the incremental size is
relatively small.
Overall, the hybrid approach yields the best trade-off between the amount of curtailed energy
and the complexity of its practical implementation. Combining this strategy with smaller increments
only has a minor effect on both of these aspects. However, compared to the iterative approach,
the curtailed energy can be reduced by 2.78 GWh at the expanse of 216 additional curtailment measures.
Considering the average financial reimbursement of DG operators, reducing curtailment by 2.78 GWh
could save 305,800 € of compensation payments within the investigated time period of one month.
Concurrently, 1444 measures can be avoided compared to the optimization approach, which reduces
the communication demand and the complexity of the implementation into grid operation significantly.
This is realized at the expense of increasing the amount of curtailed energy by only 0.24 GWh.
73
Energies 2020, 13, 4173
Figure 7. Resulting renewable energy curtailment for different maximum permissible grid component
loadings in contingency events and different DG selection methods.
74
Energies 2020, 13, 4173
Table 7. Overview of resulting renewable energy curtailment of proposed concepts and challenges
regarding the practical implementation.
4. Conclusions
The on-going integration of DGs into the energy system causes an increase in curtailment measures
executed by grid operators due to limited transmission capacities. To address this problem, this paper
presents a simulation framework for optimizing the practical implementation of DG curtailment based
on load flow analysis. The framework proved to be sufficient for simulating the curtailment demand
in the investigated region in Germany, although the allocation of curtailment measures between DSO
and TSO shows some inaccuracies. The framework is further used to analyze the ability of three
operational concepts and combinations of these to reduce renewable energy curtailment. Firstly,
an enhancement of DG controllability is investigated. Furthermore, novel DG selection methods
are proposed that enable a simultaneous consideration of multiple voltage levels. Additionally,
an extension of the deterministic n-1 criterion is considered by enhancing the maximum thermal
loading of grid components in contingency events.
The simulation results obtained indicate that the amount of renewable energy curtailment can be
reduced both by implementing smaller curtailment level increments by up to 4.3% and by establishing
coordinated novel DG selection methods by up to 10%. To limit the complexity of the practical
implementation, the hybrid approach proves to be superior to solely minimizing the amount of power
adjustments. Both concepts could be a temporary solution to optimize the curtailment process since
the effort for their realization is rather small. However, the concept of increasing the utilization of
transmission capacities by enhancing operational limits in contingency situations has clearly proven
to be the most efficient concept to reduce preventive DG curtailment. Enhancing the maximum
permissible loadings of grid components to 130% reduces the amount of renewable energy curtailment
by up to 94.8% and, thus, nearly leads to a complete prevention of these congestion management
measures. Concurrently, this is also the most challenging concept regarding its implementation.
Therefore, future research should address possibilities to extend operational limits, such as considering
the thermal inertia of grid components or establishing fast-reacting, curative congestion management
measures, and compare the additional costs of these measures with their benefit regarding the reduction
of renewable energy curtailment.
Author Contributions: Conceptualization I.L.-N.; Writing-original draft preparation I.L.-N.; Writing-review and
editing I.L.-N., F.S. and K.v.M.; and Supervision C.A. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding.
Acknowledgments: The authors would like to thank all project partners of the research project “enera”.
In particular, the authors thank Avacon Netz GmbH for providing their grid model. Furthermore, the authors
thank EWE NETZ GmbH for the provision of the aggregated load time series and the German Meteorological
Service (DWD) for providing meteorological data through the COSMO-DE model. The project enera is funded by
the Federal Ministry for Economic Affairs and Energy (BMWi, grant no. 03SIN317).
75
Energies 2020, 13, 4173
References
1. Gross, T.; Reese, S.; Petters, B.; Cupelli, M.; Mildt, D.; Monti, A.A. Novel Approach to DG Curtailment in Rural
Distribution Networks—A Case Study of the Avacon Grid as Part of the InterFlex Field Trial. In Proceedings
of the 2018 IEEE 16th International Conference on Industrial Informatics (INDIN), Porto, Portugal, 18–20
July 2018; pp. 667–672.
2. Staudt, P.; Rausch, B.; Gärttner, J.; Weinhardt, C. Predicting Transmission Line Congestion in Energy Systems
with a High Share of Renewables. In Proceedings of the 2019 IEEE Milan PowerTech, Milan, Italy, 23–27 June
2019; pp. 1–6.
3. Büchner, J.; Katzfey, J.; Flörcken, O.; Moser, A.; Schuster, H.; Dierkes, S.; Leeuwen, T.V.; Verheggen, L.;
Amelsvoort, M.V.; Uslar, M. Smart Grids in Germany: How much costs do distribution grids cause at
planning time? In Proceedings of the 2015 International Symposium on Smart Electric Distribution Systems
and Technologies (EDST), Vienna, Austria, 8–11 September 2015; pp. 224–229.
4. Gunkel, D.; Möst, D. The German transmission grid expansion in long-term perspective—What is the impact
of renewable integration? In Proceedings of the 11th International Conference on the European Energy
Market (EEM14), Kracow, Poland, 28–30 May 2014; pp. 1–6.
5. Hoffrichter, A.; Kollenda, K.; Schneider, M.; Puffer, R. Simulation of Curative Congestion Management in
Large-Scale Transmission Grids. In Proceedings of the 2019 54th International Universities Power Engineering
Conference (UPEC), Bucharest, Romania, 3–6 September 2019; pp. 1–6.
6. Yusoff, N.I.; Zin, A.A.M.; Khairuddin, A.B. Congestion management in power system: A review.
In Proceedings of the 2017 3rd International Conference on Power Generation Systems and Renewable
Energy Technologies (PGSRET), Johor Bahru, Malaysia, 4–6 April 2017; pp. 22–27.
7. Huang, S.; Wu, Q. Real-Time Congestion Management in Distribution Networks by Flexible Demand Swap.
IEEE Trans. Smart Grid 2017, 9, 4346–4355. [CrossRef]
8. Berizzi, A.; Delfanti, M.; Marannino, P.; Pasquadibisceglie, M.; Silvestri, A. Enhanced Security-Constrained
OPF With FACTS Devices. IEEE Trans. Power Syst. 2005, 20, 1597–1605. [CrossRef]
9. Huang, S.; Wu, Q.; Liu, Z.; Nielsen, A.H. Review of congestion management methods for distribution
networks with high penetration of distributed energy resources. In Proceedings of the IEEE PES Innovative
Smart Grid Technologies, Europe, Istanbul, Turkey, 12–15 October 2014; pp. 1–6.
10. Bouloumpasis, I.; Steen, D.; Tuan, L.A. Congestion Management using Local Flexibility Markets: Recent
Development and Challenges. In Proceedings of the 2019 IEEE PES Innovative Smart Grid Technologies
Europe (ISGT-Europe), Bucharest, Romania, 29 September–2 October 2019; pp. 1–5.
11. Bundesnetzagentur für Elektrizität, Gas, Telekommunikation, Post und Eisenbahnen Monitoring Report 2018.
Berlin, Germany, 29 May 2019. Available online: https://bundesnetzagentur.de/SharedDocs/Downloads/EN/
Areas/ElectricityGas/CollectionCompanySpecificData/Monitoring/MonitoringReport2018.pdf?__blob=
publicationFile&v=3/ (accessed on 20 July 2020).
12. Liu, R.; Srivastava, A.K.; Bakken, D.E.; Askerman, A.; Panciatici, P. Decentralized State Estimation and
Remedial Control Action for Minimum Wind Curtailment Using Distributed Computing Platform. IEEE Trans.
Ind. Appl. 2017, 53, 5915–5926. [CrossRef]
13. Jupe, S.; Taylor, P.; Michiorri, A. Coordinated output control of multiple distributed generation schemes.
IET Renew. Power Gener. 2010, 4, 283–297. [CrossRef]
14. Müller, B.; Salzinger, M.; Lens, H.; Enzcnhöfer, R.; Gutekunst, F. Coordinated Congestion Management Across
Voltage Levels Using Load Flow Sensitivities. In Proceedings of the 2018 15th International Conference on
the European Energy Market (EEM), Lodz, Poland, 27–29 June 2018; pp. 1–6.
15. Tschampion, M.; Bucher, M.A.; Ulbig, A.; Andersson, G. N−1 security assessment incorporating the flexibility
offered by dynamic line rating. In Proceedings of the 2016 Power Systems Computation Conference (PSCC),
Genoa, Italy, 20–24 June 2016; pp. 1–7.
16. Meier, F.; Töbermann, C.; Braun, M. Retrospective Optimal Power Flow for Low Discriminating Active
Power Curtailment. In Proceedings of the 2019 IEEE Milan PowerTech, Milan, Italy, 23–27 June 2019; pp. 1–6.
17. Tonkoski, R.; Lopes, L.A.C.; El-Fouly, T.H.M. Coordinated Active Power Curtailment of Grid Connected PV
Inverters for Overvoltage Prevention. IEEE Trans. Sustain. Energy 2010, 2, 139–147. [CrossRef]
76
Energies 2020, 13, 4173
18. VDE-AR-N 4140 Kaskadierung von Maßnahmen für Die Sicherheit von Elektrischen Energieversorgungsnetzen;
VDE Verlag GmbH: Berlin, Germany, February 2017.
19. Stott, B.; Jardim, J.; Alsac, O. DC Power Flow Revisited. IEEE Trans. Power Syst. 2009, 24, 1290–1300.
[CrossRef]
20. Liere-Netheler, I.; Schuldt, F.; Von Maydell, K.; Agert, C. Optimised curtailment of distributed generators
for the provision of congestion management services considering discrete controllability. IET Gener.
Transm. Distrib. 2020, 14, 735–744. [CrossRef]
21. Capitanescu, F.; Wehenkel, L. Redispatching Active and Reactive Powers Using a Limited Number of Control
Actions. IEEE Trans. Power Syst. 2011, 26, 1221–1230. [CrossRef]
22. Stott, B.; Alsac, O.; Monticelli, A. Security analysis and optimization. Proc. IEEE 1987, 75, 1623–1644.
[CrossRef]
23. ENTSO-E Network Code on Operational Security; European Network of Transmission System Operators for
Electricity (ENTSO-E): Brussels, Belgium, February 2013.
24. Schlegel, S.; Westermann, D. Determination of remedial actions taking into account various operational rules.
In Proceedings of the 2017 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe),
Esslingen, Germany, 26–29 September 2017; pp. 1–6.
25. Westermann, D.; Schlegel, S.; Sass, F.; Schwerdfeger, R.; Wasserrab, A.; Haeger, U.; Dalhues, S.; Biele, C.;
Kubis, A.; Hachenberger, J. Curative actions in the power system operation to 2030. In Proceedings of the
International ETG-Congress 2019 ETG Symposium, Esslingen, Germany, 8–9 May 2019; pp. 1–6.
26. Kundur, P. AC Transmission. In Power System Stability and Control, 1st ed.; McGraw-Hill: New York, NY, USA,
1994; p. 226.
27. Alvarez, D.L.; Rosero, J.A.; Silva, F.F.; Bak, C.L.; Mombello, E.E. Dynamic line rating—Technologies and
challenges of PMU on overhead lines: A survey. In Proceedings of the 2016 51st International Universities
Power Engineering Conference (UPEC), Coimbra, Portugal, 6–9 September 2016; pp. 1–6.
28. DIN EN 50341-2-4 Freileitungen über AC 1kV; VDE Verlag GmbH: Berlin, Germany, September 2019.
29. 738-2012-IEEE Standard for Calculating the Current-Temperature Relationship of Bare Overhead Conductors;
IEEE SA: New York, NY, USA, 2013.
30. Memmel, E.; Peters, D.; Völker, R.; Schuldt, F.; Von Maydell, K.; Agert, C.; Voelker, R. Simulation of vertical
power flow at MV/HV transformers for quantification of curtailed renewable power. IET Renew. Power Gener.
2019, 13, 3071–3079. [CrossRef]
31. Hammer, A. Anwendungsspezifische Strahlungsinformationen aus Meteosat-Daten. In Dissertation,
Fachbereich Physik; Carl-von-Ossietzky-Universität: Oldenburg, Germany, 2000.
32. Baldauf, M.; Förstner, J.; Klink, S.; Reinhardt, T.; Schraff, C.; Seifert, A.; Stephan, K. Kurze Beschreibung
des Lokal-Modells Kürzestfrist COSMO-DE (LMK) und seiner Datenbanken auf dem Datenserver des
DWD. Offenbach, Germany, November 2016. Available online: https://dwd.de/SharedDocs/downloads/
DE/modelldokumentationen/nwv/cosmo_de/cosmo_de_dbbeschr_version_2_4_161124.pdf?__blob=
publicationFile&v=4 (accessed on 20 July 2020).
33. ENERCON. GmbH ENERCON Produktübersicht; ENERCON: Aurich, Germany, 2015.
34. Brakelmann, H. Netzverstärkungs-Trassen zur Übertragung von Windenergie: Freileitung oder Kabel?
Rheinberg, Germany, October 2004. Available online: https://www.wind-energie.de/fileadmin/redaktion/
dokumente/pressemitteilungen/2018/20180103-brak-studie-2004.pdf (accessed on 11 August 2020).
35. Heuck, K.; Dettmann, K.-D.; Schulz, D. Aufbau und Ersatzschaltbilder der Netzelemente. In Elektrische
Energieversorgung—Erzeugung, Übertragung und Verteilung für Studium und Praxis, 9th ed.; Springer Vieweg:
Wiesbaden, Germany, 2013; p. 156.
36. Neon Neue Energieökonomik, Technische Universität Berlin, ETH Zürich, DIW Berlin Open Power System
Data. March 2018. Available online: https://open-power-system-data.org/ (accessed on 20 July 2020).
37. Choi, S.-S.; Cha, S.-H.; Tappert, C.C. A Survey of Binary Similarity and Distance Measures. J. Syst.
Cybern. Inform. 2009, 8, 43–48.
38. Cort, J.W.; Kenji, M. Advantages of the mean absolute error (MAE) over the root mean square error (RMSE)
in assessing average model performance. Clim. Res. 2005, 30, 79–82.
39. Molinar, G.; Fan, L.T.; Stork, W. Ampacity forecasting: An approach using Quantile Regression Forests.
In Proceedings of the 2019 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference
(ISGT), Washington, DC, USA, 18–21 February 2019; pp. 1–5.
77
Energies 2020, 13, 4173
40. Primadianto, A.; Lu, C.-N. A Review on Distribution System State Estimation. IEEE Trans. Power Syst. 2016,
32, 1. [CrossRef]
41. Grigsby, L.L. State Estimation. In Power System Stability and Control, 3rd ed.; CRC Press: Boca Raton, FL,
USA, 2012.
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
78
energies
Article
Recloser-Based Decentralized Control of the Grid
with Distributed Generation in the Lahsh District of
the Rasht Grid in Tajikistan, Central Asia
Anvari Ghulomzoda 1 , Aminjon Gulakhmadov 2,3,4, *, Alexander Fishov 1 , Murodbek Safaraliev 5 ,
Xi Chen 2,3 , Khusrav Rasulzoda 6 , Kamol Gulyamov 7 and Javod Ahyoev 8
1 Department of Automated Electric Power Systems, Novosibirsk State Technical University,
Novosibirsk 630073, Russia; [email protected] (A.G.); fi[email protected] (A.F.)
2 Research Center for Ecology and Environment of Central Asia, Xinjiang Institute of Ecology and Geography,
Chinese Academy of Sciences, Urumqi 830011, China; [email protected]
3 State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography,
Chinese Academy of Sciences, Urumqi 830011, China
4 Ministry of Energy and Water Resources of the Republic of Tajikistan, Dushanbe 734064, Tajikistan
5 Department of Automated Electrical Systems, Ural Federal University, Ekaterinburg 620002, Russia;
[email protected]
6 Andritz Hydro GmbH, Vienna 1120, Austria; [email protected]
7 Department of Electric Drive and Electric Machines, Tajik Technical University named after
Academic M. S. Osimi, Dushanbe 734042, Tajikistan; [email protected]
8 Department of Electric Power Station, Tajik Technical University named after Academic M. S. Osimi,
Dushanbe 734042, Tajikistan; [email protected]
* Correspondence: [email protected]; Tel.: +86-173-9777-5315
Abstract: Small-scale power generation based on renewable energy sources is gaining popularity
in distribution grids, creating new challenges for power system control. At the same time, remote
consumers with their own small-scale generation still have low reliability of power supply and poor
power quality, due to the lack of proper technology for grid control when the main power supply is
lost. Today, there is a global trend in the transition from a power supply with centralized control to a
decentralized one, which has led to the Microgrid concept. A microgrid is an intelligent automated
system that can reconfigure by itself, maintain the power balance, and distribute power flows.
The main purpose of this paper is to study the method of control using reclosers in the Lahsh district of
the Rasht grid in Tajikistan with distributed small generation. Based on modified reclosers, a method
of decentralized synchronization and restoration of the grid normal operation after the loss of the
main power source was proposed. In order to assess the stable operation of small hydropower plants
under disturbances, the transients caused by proactive automatic islanding (PAI) and restoration of
the interconnection between the microgrid and the main grid are shown. Rustab software, as one of
the multifunctional software applications in the field of power systems transients study, was used for
simulation purposes. Based on the simulation results, it can be concluded that under disturbances,
the proposed method had a positive effect on the stability of small hydropower plants, which are
owned and dispatched by the Rasht grid. Moreover, the proposed method sufficiently ensures the
quality of the supplied power and improves the reliability of power supply in the Lahsh district
of Tajikistan.
Keywords: decentralized control; small hydropower plants; microgrid; emergency control; recloser;
synchronous coupler; power systems stability
1. Introduction
At present, many technologies provide high-quality and reliable power supply to consumers
locally, having the opportunity in emergency conditions to get power from backup sources. That is
why centralized electrification of power supply to remote rural areas is a hugely complicated problem
to be accomplished by grid companies.
The Tajikistan power system faces these difficulties for several reasons being (1) power supply is
centralized, and (2) there is power shortage during the autumn-winter period.
These two factors force the regulation of power generation, which is strongly reflected in the rural
areas making up about 70% of Tajikistan’s population [1].
Tajikistan is a country where that has a huge stock of hydro resources. Consequently, Tajikistan
ranks first among Central Asian countries, and eighth in the world, as noted by Zhan-Kristof et al. [2],
for hydropower research. The country possesses 4% of all world hydro-energy potential [1].
Fishov et al. [3] analyzed the condition and progress of Tajikistan’s hydro-energy potential, which
exceeds 527 billion kWh per year, and the country exploits only 6.6% of it [4]. Potential reserves of
small hydropower are 184.46 billion kWh per year [1].
In the field of small hydropower, the CASA-1000 (Central Asia South Asia) [5] project is of
paramount importance to Tajikistan. The role envisioned for the project is to limit the export of excess
power to neighboring countries during the summer. It is also important to recognize the effects not only
of large-scale hydropower (LHPP) plants but also small-scale hydropower plants (SHPPs) on power
generation. Implementation of the project will allow the efficient use of SHPPs and the development
of this energy sector in Tajikistan.
The 0.4–10 kV rural distribution grids are characterized by longline length and complex structure.
The key disadvantages of such grids are low reliability, inadmissible values of power quality indices, and
significant power losses. Cherkasova [6] has explored energy efficiency issues, including the reliability
and safety of rural electrical grids. The transmission outages can make up 40–90% of the total number
of outages. Moreover, due to the significant degradation of rural distribution grids, the reliability of
power supply is also deteriorating. For example, in the 6–20 kV grids, 30 blackouts/year × 100 km of
line length occurs on average, and in the 0.4 kV grids, 100 blackouts/year × 100 km [7]. It is possible
to reduce the number of blackouts using auto-reclosing in transmission lines. As Cherkasova [8]
shows in her research on the monitoring and analysis of failures, about 60% of power line damage is
unsustainable and eliminated by auto-reclosing. However, since rural distribution grids were built in
the 1960s and 1970s in Tajikistan, it is impossible to use auto-reclosing based on old oil circuit breakers
at the distribution center. A significant part of outages and interruptions can be reduced, and as a
consequence, the reliability of power supply will increase. The use of sectionalizing points when the
feeder is divided by switching devices is known to be an effective means of reliability increase. With
the development of engineering and technologies based on automatic sectionalizing points (ASP), it is
possible to significantly increase the reliability of power supply, especially in rural areas. The use of
reclosers can serve as a good opportunity for that purpose.
With the progress of power electronics and control theory, the Microgrid concept emerged, which
reflected in the research studies conducted in the USA, Europe, and Asia. Zhou et al. [9] have considered
key problems and prospects of microgrid progress. Barnes et al. [10] have considered a principle of
work versions of microgrid management. Zaidi et al. [11] discussed a new approach to modeling
intellectual microgrids with automatic load control consumers.
The microgrid is a small and independent system that combines small-scale generation (SSG),
consumers, energy storage systems, as well as control devices, forming an integrated controlled power
supply system. The microgrid is an intelligent automated system that can reconfigure itself, manage
power generation and load (maintain power balance), and distribute power flows.
In the present work, the problem of connecting the microgrid based on small generation to
an external power grid is solved using its direct connection to the grid. We applied the methods
in Zhdanov [12], namely the fixed-point iteration method, equal area criterion, and the method
80
Energies 2020, 13, 3673
of superposition for determining the stability of power plants operating in parallel with the grid.
To simulate the method used in the present article, the Rustab software based on RastrWin3 (2.3, JSC
“Scientific and Technical Center of Unified Power System”, Yekaterinburg, Russia) was used. It is a
multifunctional software devoted to the study of electromagnetic and electromechanical transients
during different load flow calculations [13]. The RastrWin3 software includes Eurostag and Mustang
libraries, which allow us to calculate, analyze, and optimize the load flow states; to monitor the power
generation/consumption; to calculate short circuit currents and to carry out asymmetrical load flow.
Currently, the transition from a centralized to a decentralized power supply system based on
the microgrid is relevant. It allows us to efficiently use to distribute power sources over the grid,
including renewable energy sources (RES). Therefore, decentralized synchronization and control of
the normal operation restoration during the parallel operation of the separated grid and generators is
considered in the present work. The control method using modified reclosers with distributed small
generation in the Lahsh district of the Rasht grid of Tajikistan is studied. Based on the methodology
above, the stability of all available SHPPs under disturbances in the Rasht grid was estimated.
Frequent emergency disturbances in the external network will lead to emergency shutdowns
of the SHPPs power units, which feed consumers located in the local network of the Lahsh district.
Consequently, consumers of this local network remain without power, and the available capacity of
SHPPs is underutilized. Therefore, the goal of this work is to study ways of decentralized management
using reclosers of a local network with SHPPs in the rural Lahsh district of Rasht electric networks
of Tajikistan.
The paper is structured as follows: Section 2 provides an overview of the literature; Section 3
describes information about the object of study and a problem description; Section 4 discusses methods
and tools for solving the problems under consideration; Section 5 presents the results obtained, as well
as plans for further research; Section 6 gives the conclusions.
2. Literature Review
A review of research works devoted to the decentralization of grids control is presented in several
publications. In Wang et al. [14] study, a decentralized power control system consisting of three
integrated microgrids and working both independently and in parallel with the main power grid is
considered. The microgrid may contain such power sources as a microturbine unit (under distribution
grid operator control), wind turbines, solar panels operating stochastically. When the microgrid is
in interconnected mode, the distribution grid operator and each microgrid connected to the grid
are considered as separate objects with individual goals, which are minimization of operating costs.
In an islanded mode, each microgrid’s goal is to provide a reliable power source for its consumers.
Tani et al. [15,16] study power control in decentralized power systems using wind turbines with
photovoltaic panels and energy storage systems. An energy storage system is needed for a quick
response to correct the imbalance between generation and load caused by sudden fluctuations in wind,
sun, or consumers. In this work, the energy consumption control method based on a frequency approach
using a fluctuating distribution of wind power/load is shown. The main advantage of the control
method in Tani et al. [16] is that the used frequency approach takes the dynamic characteristics of the
power sources into account and allows them to increase the operating time of the main power sources
and reduce the size of the energy storage system. Kargarian et al. [17] considered the decentralized
algorithms to determine the optimum power flow in electric power systems. The algorithms offered in
their work can be used for the accelerated process of deciding large-scale problems of optimization that
are difficult to solve centrally, as well as to control systems with several independent operating objects.
Dou et al. [18] proposed double-tier strategies for decentralized online optimization management for
an isolated microgrid without communication network channels to ensure frequency security and
minimize operating costs. Qi et al. [19] considered a decentralized approach to optimal performance,
based on analytical cascading of goals, to optimize the performance of AC/DC hybrid distribution
networks. Yin et al. [20] have offered real-time decentralized energy management for multi-source
81
Energies 2020, 13, 3673
hybrid energy systems that adapts to sudden changes in system configuration. Koukoula et al. [21] offer
a new fully decentralized method of controlling congestion in radial distribution networks. The authors
propose an algorithm for detecting and eliminating local restrictions on power flows without the need
for centralized control. Shi et al. [22] developed a distributed energy management strategy for optimal
microgrid performance, considering the distribution network and its associated limitations.
Reclosers are widely used in various countries around the world. Examples can be drawn from
several studies. Sazykin et al. [23] considered the decentralized control of 6–10 kV distribution grids
and planned to achieve this by installing reclosers that would act like sectionalizers. The placement of
reclosers for different versions of the grid design was determined analytically. Reliability indices of
the grid under consideration were calculated. Andrianova et al. [24] analyzed the use of reclosers in
6–35 kV rural distribution grids for decentralized control. Reliability indices such as SAIFI (System
Average Interruption Frequency Index) and SAIDI (System Average Interruption Duration Index),
as well as power supply quality indices, were calculated before and after installation of reclosers.
It was noted that the number of load shedding events reduced by 15 times, and the duration of load
loss was reduced by 26 times when reclosers were installed in distribution grids. Popovic et al. [25]
presented the methodology for optimizing and coordinating placement of reclosers and generators
in distribution grids. It was noted that optimal placement contributes to the improvement of system
reliability and voltage profile and the reduction of power losses. The locations of reclosers and
generators were determined using a genetic algorithm, which helped to improve the overall reliability
of the grid. The study [26] was devoted to finding the optimal placement of reclosers in the grid.
The reliability index SAIFI was calculated for various grid designs using the developed method.
De Bruyn et al. [27] studied the behavior of recloser in the case of distributed wind energy presence in
the grid. Misoperations and dead zones of relay protection were found and appropriately addressed
using the developed improved version of the method. Sazykin et al. [28] examined the criteria of
optimal recloser placement in 6–10 kV distribution grids. The results indicated a reduction in the
number and duration of load loss events, with the total energy not served value being reduced from
69.7% to 83.5%.
A review of research on divisive protection and emergency automation can be found in some
papers. Onisova [29] has reviewed the basic functions and principles of divisive protections in
electric power systems, formulated new requirements for divisive protections, and has proposed
improvements in devisive voltage automation. Fishov et al. [30] considered the problem of integrating
small generation into electric networks. They offer new technical solutions aimed at ensuring the
reliability and cost-effectiveness of systems with distributed small generation through highly automated
emergency control in line with the modern SMART GRID concept. Ilyushin et al. [31] presented the
features of separation of power plants from the power system with their own needs and load at a
voltage of 6–20 kV. Formulated requirements for the speed of unloading during the forced separation of
the power plant from the power system. Ilyushin [32] formulated possible limitations in the application
of multi-parameter dividing automation related to the peculiarities of active power regulation and
the settings of relay protection devices of generating installations of distributed generation facilities.
Shabad [33] considered schemes and calculations of the settings of divisive protection, designed to
prevent various emergencies in distribution electric networks and power plants. Kalentionok et al. [34]
examined the problem of creating divisive automation in industrial enterprises that have their own
generating units. The authors’ proposed algorithms for the operation of dividing automation, which
ensure the minimum possible power imbalances in the allocation of generating units for stand-alone
operation, as well as possible parameters for its operation.
82
Energies 2020, 13, 3673
reflect this in their analysis of the development of small hydropower in Tajikistan. The construction
dynamics of these SHPPs is shown in Figure 1.
4XDQWLW\
<HDU
Figure 1. Construction dynamics of the small-scale hydropower plants (SHPPs) from 1990–2013.
The country continues to build SHPPs under the “Long-term program for the construction of
SHPPs during 2009–2020.” Construction of 189 SHPPs is part of the program, with a total installed
capacity of 103 MW (Table 1).
The “Barqi Tojik” Open Joint-Stock Holding Company (OJSHC) energy company is responsible
for power supply in the majority of the country’s regions, including the Lahsh district of the Rasht
grids. The Rasht region has 6 SHPPs [3], and the generation retrospective is shown in Figure 2.
As can be seen from the figures above, the plant utilization factor (which means the ratio of the
total actual energy produced or supplied by SHPPs in the Rasht grid over a definite period, to the
energy that would have been produced if the plant had operated continuously at the maximum rating)
differs from the assigned one, since the capacity of SHPPs is not used correctly. The reasons for the
decrease in power generation are mainly frequent disturbances in the grid leading to the shutdown of
power units, the operation of the generation units only in stand-alone mode, the impact of external
factors, and others.
83
Energies 2020, 13, 3673
Frequent disturbances in the Lahsh grid strongly affect consumers, since they lose power both
from the main grid and from the local SHPPs. Consequently, it is crucial to solve the problem of
customer outage, reducing energy not served, ensuring the quality of supplied power, and improving
the reliability of the power system in the Lahsh district of the Rasht grid in Tajikistan. Figure 3 shows a
geographical map of the grid under study in the Lahsh district.
(a) (b)
Figure 2. (a) The power generation and (b) the plant utilization factor of small-scale hydropower plants
(SHPPs) in the Rasht grids.
NP
Figure 3. The geographical map of the grid under study of the Lahsh district in Tajikistan.
84
Energies 2020, 13, 3673
4. Methodology
([WHUQDO
HOHFWULFDOQHWZRUN
%LRPDVV
(OHFWULFLW\ :LQGWXUELQHV 39FHOOV
*HQHUDWRU
FRQVXPHUV
(OHFWULFLW\
FRQVXPHUV 7KHUPDOSRZHU +\GUR3RZHU
)XHOFHOOV (QHUJ\VWRUDJH
SODQWV 3ODQWV
+HDWFRQVXPHUV +HDWFRQVXPHUV
The microgrid can be either in interconnected mode, where the microgrid exchanges power with
the main grid through the distribution substation transformer, or in the islanded mode, where the
microgrid is disconnected from the main grid, and operates autonomously, serving its own local
demand using DG and other distributed energy resources. During contingency Microgrid can be
independently disconnected from the main grid and operate autonomously. After eliminating the
contingency, microgrid can again reconnect to the main grid. Thus, microgrid is an intelligent automated
system that is able to reconfigure by itself, maintain power balance and distribute power flows.
This study presents the solution of microgrid (on the base of small generation) connection problem
by means of synchronous coupling or the use of microgrid in the grid without centralized control.
A prerequisite is the integration of SHPPs into the grid, the integration of several local power supply
systems (on the base of SHPPs) into islanded or interconnected power systems.
85
Energies 2020, 13, 3673
connection to the main network of a local power supply system based on small generation. This type
of emergency control has proven its efficiency, and it can be applied both to future SSG projects and to
existing facilities, which were not initially considered for use in that way. In order to ensure reliable
parallel operation of the microgrid in terms of emergency control, a unique method of emergency
control was suggested and implemented. Fishov et al. [39] developed a method for emergency control
of the mode of parallel operation of synchronous generators in electric networks. The concept behind
it is the proactive islanding of the microgrid at selected points when a contingency occurs. After that,
the transition to the island mode is made with the synchronization and transition to normal operating
conditions with a proper equipment loading schedule. This type of emergency control has already
proved its efficiency in test conditions. It was tested in the Berezovoe district, Pervomayskiy area of
Novosibirsk, in the Russian Federation.
Another key element of decentralized operation and control of Microgrids is an automatic
recloser. Fishov et al. [40] considered a decentralized reconfiguration of the electrical network with
a microgrid using advanced reclosers. Reclosers perform the functions of isolation and redundancy
in 6–10 kV distribution grids of radial and network types with long transmission lines. The recloser
includes a vacuum circuit-breaker, power conversion systems, an auxiliary DC control power system,
a microprocessor-based protection and control system, and SCADA communications ports and software.
Previously Buzin et al. [41] investigated modern relay protection and automation of 6–10 kV overhead
power lines.
The reclosing is that the lines connecting the microgrid to the power system are restricted for grids
with generation facilities (SSG in this case). The synchronization conditions should be met in order for
these lines to be switched on. If not, out-of-step switching may result in significant dynamic torque,
which represents a possible threat to generators and their auxiliaries. Hachaturov [42] presented the
principles of synchronization and the consequences of the non-synchronous inclusion of power plant
generators in the network. It should be considered when reclosers are to be installed in grids with
generation facilities. This work examines whether it is possible to use reclosers for the recovery of
power generators parallel operation.
Modern reclosers do not possess synchronization properties needed in grids with SSG penetration.
Since reclosers are located away from generators, the synchronization is not carried out using the
generator circuit breakers but the reclosers. The following modification was made: the synchronization
package was added to the control cabinet of a vacuum circuit-breaker. The functional control diagram
is shown in Figure 5.
In Figure 5, the high-voltage equipment includes the Q—vacuum circuit-breaker; ST1,
ST2—auxiliary transformers; TV1, TV2—voltage transformers, and a TA—current transformer.
The control cabinet is the power supply unit; the control system unit; the relay protection and
the control unit (RP and A); the measurement unit; the battery; the GSM modem (in case if SCADA
connection is not provided); the monitor and the synchronization package.
The recloser operation modification allows extending the range of control capabilities by adding
the synchronization package to the cabinet. The synchronization package receives such measurement
signals as voltage module and phase measured by voltage transformers (TV1, TV2) and frequency on
both sides of a vacuum circuit-breaker.
The following synchronization conditions are to be met:
⎫
U i − U j < dUph. ⎪⎪
⎪
⎬
δi − δ j < δperm. ⎪
⎪ (1)
⎪
fi − f j < sperm. ⎭
where Ui —root-mean-square voltage value in the “from” bus of the transmission line, kV;
U j —root-mean-square voltage value in the “to” bus of the transmission line, kV; dUph. —permissible
voltage drop of the transmission line, kV; δi —voltage phase angle in the “from” bus of the transmission
line, el.; δ j —voltage phase angle in the “from” bus of the transmission line, el.; δperm. —permissible
86
Energies 2020, 13, 3673
phase angle difference of the transmission line, el.; fi —frequency in the “from” bus of the transmission
line, Hz; f j —frequency in the “to” bus of the transmission line, Hz; sperm. —permissible slip of the
transmission line, Hz.
If based on values of the mentioned parameters, the synchronization conditions are met, then the
synchronizer transmits a signal to the control unit passing through the relay protection and control
unit in order to switch on the vacuum circuit-breaker. The process of synchronization can be seen on
the monitor.
67
3RZHUVXSSO\
3RZHU EDWWHU\
EULFN
79
*60PRGHP
6&$'$
&RQWURO
4 PRGXOH 'LVSOD\
67
3RZHUVXSSO\
53DQG$
79 XQLW 6\QFKURQL
]LQJPRGXOH
7$ 0HDVXUH
PHQWXQLW
+LJKYROWDJHHTXLSPHQW &RQWUROFDELQHW
PRGXOH
Such microprocessor-based devices as «AC-М» [43], «Sprint-М» [44] and «SYNCHROTACT» [45]
are used as synchronizers. These devices provide compliance with all synchronization conditions.
They carry out the functions of measurement, analysis, and, if necessary, synchronization parameters
adjustment. When all conditions are met, the synchronizer sends a signal to switch on the recloser.
Additionally, the synchronizers are capable of self-testing and displaying state data, which can be
considered as an advantage.
87
Energies 2020, 13, 3673
Swinging of synchronous generators might be caused by sudden sharp load drop or rise, outage
of a transmission line, a transformer, or a generator, with a fault being the most dangerous one.
Generally, transmission outages are followed by sudden changes in the output of power plants. Rotors
of generators and turbines respond to disturbances with some time delay due to significant inertia.
As a result, the power balance between generator and turbine shifts, resulting in excess torque of the
generator, in turn causing rotor speed increase.
According to the fixed-point iteration method [12], the rotor acceleration is:
ΔM ΔP
α= = (2)
Tj Tj
where T j —inertia constant, kg·m2 /s2 ; ΔM = Mmech − Mel —excess torque of turbine, determined
by the difference between mechanical torque and electrical torque, Nm; ΔP—excess power of the
generator, MW.
Taking into account that angular acceleration of the rotor is the second derivative of rotor angle
with respect to time α = d2 δ/dt2 ; and the difference between turbine power and generator power
output ΔP = P0 − Pm · sin δ, the swing equation of rotor is:
d2 δ
Tj = P0 − Pm sin δ (3)
dt2
Solving the Equation (3) shows how to load angle changes over time and allows the estimation of
generator stability.
The stable parallel operation of two power plants can be estimated using the fixed-point iteration
method and equal area criterion. The power output of each power plant is:
⎫
P1 = E21 y11 sin α11 + E1 E2 y12 sin(δ12 − α12 ) ⎪
⎪
⎬
⎪
⎪ (4)
P2 = E22 y22 sin α22 − E1 E2 y12 sin(δ12 + α12 ⎭
where P1 , P2 —the power output of the first and the second power plant, respectively; E1 , E2 —EMF
of the first and the second power plant, respectively; y11 , y22 , y12 —self-admittances of both power
plants and mutual admittance, respectively; δ12 —angle difference between power plants; α11 , α22 ,
α12 —absolute acceleration of the first power plant generator rotor, the second power plant generator
rotor, and their relative acceleration, respectively.
Absolute rotor acceleration of both power plants, and also their relative acceleration is:
⎫
ΔP1 ⎪
⎪
360 f
α1 = T j1 ⎪
⎬
⎪
⎪ (5)
ΔP2 ⎪
360 f
α2 = T j2 ⎭
ΔP1 ΔP2
α12 = α1 − α2 = 360 f − (6)
T j1 T j2
where f —utility frequency;
ΔP1 , ΔP1 —excess power of generators:
ΔP1 = P0 − P1 · sin δ
(7)
ΔP2 = P0 − P2 · sin δ
Hence, the stability of power plants working in parallel can be estimated based on the equal area
criterion (equality of deceleration and acceleration areas).
88
Energies 2020, 13, 3673
The stability of systems with complex structures can be analyzed using the superposition technique.
The power equation for a system with n—generators is:
⎫
P1 = E21 y11 sin α11 + E1 E2 y12 sin(δ12 − α12 )+E1 E3 y13 sin(δ13 − α13 )+ . . . + E1 En y1n sin(δ1n − α1n ); ⎪
⎪
⎪
⎪
P2 = E2 E1 y21 sin(δ21 − α21 )+E22 y22 sin α22 + E2 E3 y23 sin(δ23 − α23 )+ . . . + E2 En y2n sin(δ2n − α2n ); ⎪
⎪
⎪
⎪
⎪
⎪
... ⎪
⎪
⎪
⎪
P(n−1) = E(n−1) E1 y(n−1)1 sin(δ(n−1)1 − α(n−1)1 )+E(n−1) E2 y(n−1)2 sin(δ(n−1)2 − α(n−1)2 )+. . . ⎬
⎪
⎪ (8)
⎪
⎪
. . .+E2(n−1) y(n−1)(n−1) sin α(n−1)(n−1) + E(n−1) En y(n−1)n sin(δ(n−1)n − α(n−1)n ); ⎪
⎪
⎪
⎪
⎪
⎪
Pn = En E1 yn1 sin(δn1 − αn1 )+En E2 yn2 sin(δn2 − αn2 )+. . . ⎪
⎪
⎪
⎪
. . .+En E(n−1) yn(n−1) sin(δn(n−1) − αn(n−1) )+E2n ynn sin αnn ⎭
Consequently, the excess power of each generator can be found using the methods mentioned
above, despite the complexity of the system structure. Expressions (5)–(7), and angle-time curves form
the basis of generators stability analysis.
89
Energies 2020, 13, 3673
Figure 6. Model of the rural area grid of the Lahsh region of Tajikistan, Central Asia.
90
Energies 2020, 13, 3673
$QJOH
į
D WVHF
$QJOH
į
WVHF
E
Figure 7. The angle-time curve in case of disturbance: (a) ±180◦ curve; (b) incremental angle curve.
į*>DQJOH@ WVHF
Figure 8. The angle-time curve in disturbance conditions while using the suggested method.
91
Energies 2020, 13, 3673
,5>$@ ,5>$@ 3*>0:@ 3*>0:@ 3*>0:@ 3*>0:@ I>+]@ WVHF
Figure 9. The microgrid transient response with 20% of the maximum load.
Figure 10. The microgrid transient response with 60% of the maximum load.
92
Energies 2020, 13, 3673
,5>$@ ,5>$@ 3*>0:@ 3*>0:@ 3*>0:@ 3*>0:@ I>+]@ WVHF
Figure 11. The microgrid transient response with 100% of the maximum load.
Figure 8 shows a curve of the change in the angle of the generator’s rotor, and Figures 9–11 show
the curves of the transition process on the microgrid network when an emergency disturbance occurs.
On the listed curves, the vertical bars indicated by the letters A–E are action indicators. Curves 9–11
show the following dependencies: changes in currents through reclosers R1, R2—I (R1), I (R2), A;
power of generators G1–G4—P (G1), P (G2), P (G3), P (G4), MW; Microgrid frequency—f, Hz. The PAI
acted according to the algorithms aimed at forcing generator outages depending on the value of the
microgrid load. The simulation scenario was as follows: when a fault occurs (Figure 6), PAI acts ahead
of current relay protection and sends signals to reclosers R1, R2 to switch off and to recloser R3 to
switch on (point A, at t = 1.5 s, Figures 9–11) and generators to turn off (point C, at t = 2.27 s, G1, G2
on Figure 9 and G3, G4 in Figure 10). It eliminates excess power in the microgrid being separated.
In case of power shortage in the Microgrid UFLS acts and sends a signal to recloser R4 to switch off
(point C, at t = 4.85 s, Figure 11); then current relay protection acts (point B at t = 2 s, Figures 9–11)
and sends signals to circuit-breaker Q6 to switch off, causing a fault to self-clear and circuit-breaker
Q6 to reconnect automatically (point D, at t = 5 s, Figures 9–11). When voltage appears on both
sides of reclosers R1, R2, the mechanism of auto-reclosing with a synchronism check is triggered. If
synchronization conditions are met, then reclosers R1, R2 switch on, and recloser R3 switches off (point
E, at t = 6, 5 s, Figure 9, Figure 10, and Figure 11). The power supply of customers is restored. No more
load losses occur due to the available generation in the grid.
The simulation results prove the efficiency of the suggested method, which leads to the
improvement of the small hydro plant stability. This, in its turn, provides the required power
quality and power supply reliability for customers in the Lahsh district. The suggested method of
power system control is planned to be applied in other regions of Tajikistan in Central Asia with
distributed generation. In addition, at the next stage of the study, this method will be applied for
studying voltage control capabilities, reliability analysis, and the influence of reclosers placement on
these parameters.
6. Conclusions
The suggested method of distributed synchronization of the microgrid connection to the external
grid was studied. The obtained data of the object under study was analyzed and used in the simulation.
93
Energies 2020, 13, 3673
The wide range of Rustab package functions was used to simulate disturbances and study transient
stability of the small hydro plants in the Lahsh Microgrid in Tajikistan.
The suggested modified recloser was used in the study. Using such reclosers makes it possible to
decentralize emergency control, provide independent operation of the sectionalizing point, minimize
human error, ensure fast recovery of supply in areas untouched by disturbance, and, as a consequence,
keep the values of energy not served to a minimum. Automation and control in overhead power lines
based on reclosers were considered by Vorotnickij et al. [46,47].
According to the results of the simulation, the suggested method has a positive impact on the
ability of small hydropower plants in the Rasht grid to withstand disturbances. It provides the required
power quality and improves power supply reliability in remote areas of Tajikistan’s Lahsh district.
Author Contributions: All authors have made valuable contributions to this work. A.G. (Anvari Ghulomzoda)
and A.F. engaged in the development, research of the main part of the work, and wrote the manuscript; M.S.
assisted in modeling and software; A.G. (Aminjon Gulakhmadov) and X.C. helped revise the manuscript structure
and funding; K.R., K.G., and J.A. helped in the calculation and organization of data. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was funded by the Strategic Priority Research Program of the Chinese Academy of
Sciences, Pan-Third Pole Environment Study for a Green Silk Road (Grant No. XDA20060303), the International
Cooperation Project of the National Natural Science Foundation of China (Grant No. 41761144079), the Xinjiang
Tianchi Hundred Talents Program (Grant No. Y848041), and the project of the Research Center of Ecology and
Environment in Central Asia (Grant No. Y934031).
Acknowledgments: The authors are thankful to the Ministry of Energy and Water Resources of the Republic of
Tajikistan and the Rasht electric networks OJSHC “Barqi Tojik” for providing the data for this research work.
Conflicts of Interest: The authors declare no conflicts of interest.
References
1. Tajikistan. Rapid Assessment and Gap Analysis. Available online: https://www.undp.org/content/dam/
tajikistan/docs/library/UNDP_TJK_SE4ALL_Rapid_Assessment_and_gap_analysis_Rus.pdf (accessed on 10
April 2020).
2. Zhan-Kristof, F.; Varasteh, E.; Jako, R.; Chobanova, B.; Bilolov, F. Tajikistan: In-Depth Energy Efficiency Review;
Spotinov Print LLC: Brussels, Belgium, 2013; pp. 76–79. ISBN 978-905948-138-1.
3. Fishov, A.G.; Gulomzoda, A.H.; Kasobov, L.S. Analysis of the state and direction of development of small
hydropower of Tajikistan. Polytech. Bull. Ser. Eng. Stud. 2019, 1, 13–20.
4. Barki Tojik OJSHC. Available online: http://www.barqitojik.tj (accessed on 15 April 2020).
5. CASA-1000. Available online: http://www.casa-1000.org/indexr.php (accessed on 17 April 2020).
6. Cherkasova, N.I. Fundamentals of Managing Technological Risks and the Efficient Functioning of Power
Supply Systems for Agricultural Consumers. Ph.D. Thesis, Altai State Technical University, Barnaul, Russia,
15 December 2017.
7. Standing Order of PJSC “ROSSETI” on a Unified Technical Policy in the Electric Grid Complex; Protocol No.138.;
PJSC Rosseti: Moscow, Russia, 2013.
8. Cherkasova, N.I. Analysis of the state of rural electric networks of 10 kV in the light of failure monitoring.
Polzunovsky Bull. 2012, 4, 49–54.
9. Zhou, X.; Guo, T.; Ma, Y. An Overview on Microgrid Technology. In Proceedings of the 2015 IEEE International
Conference on Mechatronics and Automation (ICMA), Beijing, China, 2–5 August 2015; pp. 76–81.
10. Barnes, M.; Kondoh, J.; Asano, H.; Oyarzabal, J.; Ventakaramanan, G.; Lasseter, R.; Hatziargyriou, N.;
Green, T. Real-World MicroGrids-An Overview. In Proceedings of the 2007 IEEE International Conference
on System of Systems Engineering, San Antonio, TX, USA, 16–18 April 2007; pp. 1–8.
11. Zaidi, A.A.; Kupzog, F. Microgrid Automation—A Self-Configuring Approach. In Proceedings of the 2008
IEEE International Multitopic Conference, Karachi, Pakistan, 23–24 December 2008; pp. 565–570.
12. Zhdanov, P.S. Stability Issues of Electrical Systems; Energy: Moscow, Russia, 1979; pp. 77–105.
ISBN 31.27-01-6112.11.
13. RastrWin. Available online: http://www.rastrwin.ru (accessed on 7 May 2020).
94
Energies 2020, 13, 3673
14. Wang, Z.; Chen, B.; Wang, J.; Kim, J. Decentralized Energy Management System for Networked Microgrids
in Grid-Connected and Islanded Modes. IEEE Trans. Smart Grid 2016, 7, 1097–1105. [CrossRef]
15. Tani, A.; Camara, M.B.; Dakyo, B. Energy Management in the Decentralized Generation Systems Based
on Renewable Energy—Ultracapacitors and Battery to Compensate the Wind/Load Power Fluctuations.
IEEE Trans. Ind. Appl. 2015, 51, 1817–1827. [CrossRef]
16. Tani, A.; Camara, M.B.; Dakyo, B. Energy Management in the Decentralized Generation Systems Based
on Renewable Energy Sources. In Proceedings of the 2012 International Conference on Renewable Energy
Research and Applications (ICRERA), Nagasaki, Japan, 11–14 November 2012; pp. 1–6.
17. Kargarian, A.; Mohammadi, J.; Guo, J.; Chakrabarti, S.; Barati, M.; Hug, G.; Kar, S.; Baldick, R. Toward
Distributed/Decentralized DC Optimal Power Flow Implementation in Future Electric Power Systems.
IEEE Trans. Smart Grid 2018, 9, 2574–2594. [CrossRef]
18. Dou, C.; An, X.; Yue, D.; Li, F. Two-Level Decentralized Optimization Power Dispatch Control Strategies for
an Islanded Microgrid without Communication Network. Int. Trans. Electr. Energy Syst. 2017, 27, e2244.
[CrossRef]
19. Qi, C.; Wang, K.; Fu, Y.; Li, G.; Han, B.; Huang, R.; Pu, T. A Decentralized Optimal Operation of AC/DC
Hybrid Distribution Grids. IEEE Trans. Smart Grid 2018, 9, 6095–6105. [CrossRef]
20. Yin, H.; Zhao, C.; Ma, C. Decentralized Real-Time Energy Management for a Reconfigurable Multiple-Source
Energy System. IEEE Trans. Ind. Inform. 2018, 14, 4128–4137. [CrossRef]
21. Koukoula, D.I.; Hatziargyriou, N.D. Gossip Algorithms for Decentralized Congestion Management of
Distribution Grids. IEEE Trans. Sustain. Energy 2016, 7, 1071–1080. [CrossRef]
22. Shi, W.; Xie, X.; Chu, C.-C.; Gadh, R. Distributed Optimal Energy Management in Microgrids. IEEE Trans.
Smart Grid 2015, 6, 1137–1146. [CrossRef]
23. Sazykin, V.G.; Kudrjakov, A.G. Actual Issues of Technical Sciences: Theoretical and Practical Aspects. Collective
Monograph; Aeterna: Ufa, Russia, 2017; pp. 64–87.
24. Andrianova, L.P.; Kabashov, V.J.; Hajrislamov, D.S. Reliable indicators of SAIFI and SAIDI rural electric
networks with intelligent sectional reclosers. Bull. Bashkir State Agrar. Univ. 2019, 3, 83–92. [CrossRef]
25. Popović, D.H.; Greatbanks, J.A.; Begović, M.; Pregelj, A. Placement of Distributed Generators and Reclosers
for Distribution Network Security and Reliability. Int. J. Electr. Power Energy Syst. 2005, 27, 398–408.
[CrossRef]
26. Karpov, A.I.; Akimov, D.A. Integral Indicators Improvement (SAIFI) of Power Supply Reliability in Electric
Distribution Systems Based on Reclosers Placement Optimization. In Proceedings of the 2018 IEEE
Conference of Russian Young Researchers in Electrical and Electronic Engineering (EIConRus), Moscow,
Russia, 29 January–1 February 2018; pp. 663–666.
27. De Bruyn, S.; Fadiran, J.; Chowdhury, S.; Chowdhury, S.P.; Kolhe, P. The Impact of Wind Power Penetration
on Recloser Operation in Distribution Networks. In Proceedings of the 2012 47th International Universities
Power Engineering Conference (UPEC), London, UK, 4–7 September 2012; pp. 1–6.
28. Sazykin, V.; Kudryakov, A.; Bagmetov, A. Criteria for Optimizing the Place of Reklouser Installation in the
6-10 KV Distribution Network. Electrotech. Syst. Complexes 2018, 1, 33–39. [CrossRef]
29. Onisova, O.A. Subdivision Protection in Electric Power Systems Containing Low-Power Electric Power
Plants. Power Technol. Eng. 2014, 48, 322–330. [CrossRef]
30. Fishov, A.G.; Landman, A.K.; Serdyukov, O.V. SMART Technologies for Connecting to Electric Networks
and Control Small Generation Modes. In Proceedings of the VIII International Scientific and Technical
Conference “Electricity through the Eyes of Youth—2017”, Samara, Russia, 2–6 October 2017; pp. 27–34.
31. Ilyushin, P.V.; Pazderin, A.V. Requirements for dividing automation of distributed generation facilities,
taking into account the influence of adjacent network parameters and load. Electr. Transm. Distrib. 2018, 4,
42–47.
32. Ilyushin, P.V. Features of the implementation of multi-parameter dividing automation in energy areas with
distributed generation facilities. Relay Prot. Autom. 2018, 2, 12–24.
33. Shabad, M.A. Fission Protections—Automation of Division During Failures, 7th ed.; Dyakov, A.F., Ed.;
Energoprogress: Moscow, Russia, 2006; pp. 5–37.
34. Kalentionok, E.V.; Filipchik, Y.D. Methodological approaches to the creation of dividing automation in
industrial enterprises with generating power plants. Energy Proc. Higher Educ. Establ. Energy Assoc. CIS 2010,
6, 14–19.
95
Energies 2020, 13, 3673
35. Gulomzoda, A.; Fishov, A.G.; Nikroshkina, S.V. Development of Small-Scale Hydropower Generation in
Tajikistan. In Proceedings of the 2019 8th International Academic and Research Conference of Graduate and
Postgraduate Students, Novosibirsk, Russia, 28 March 2019; pp. 123–126.
36. Gulomzoda, A.; Fishov, A.G.; Nikroshkina, S.V. Technology of Managing the Modes of Local Energy Supply
Systems. In Proceedings of the 2018 2nd All Russia Academic and Research Conference of Graduate and
Postgraduate Students, Novosibirsk, Russia, 20 December 2018; pp. 70–72.
37. Ismoilov, S.T.; Gulomzoda, A.H.; Rahimov, F.M. Putting distributed generation into the network to provide
system services. In Proceedings of the 11th International Conference on Mining, Construction and Energy,
Tula, Russia, 5–6 November 2015; pp. 362–367.
38. Gezha, E.N.; Glazyrin, V.E.; Glazyrin, G.V.; Ivkin, E.S.; Marchenko, A.I.; Semendjaev, R.J.; Serdjukov, O.V.;
Fishov, A.G. System automation for the integration of local power supply systems with synchronous low
generation into electrical networks. Prot. Eng. 75 Years Dep. RP APS NRU MPEI 2018, 2, 24–31.
39. Fishov, A.G.; Mukatov, B.B.; Marchenko, A.I. Method for Emergency Control of the Mode of Parallel
Operation of Synchronous Generators in Electrical Networks. RU Patent 2662728, 30 July 2018.
40. Fishov, A.G.; Ghulomzoda, A.H.; Kasobov, L.S. Decentralised reconfiguration of a Microgrid electrical
network using reclosers. Proc. Irkutsk State Tech. Univ. 2020, 24, 382–395. [CrossRef]
41. Buzin, S.; Vorotnickij, V. Modern Relay Protection and Automation for the Automation of 6–10 Kv Overhead Electrical
Networks; Technical Report; RK Tavrida Elektrik Ltd.: Krasnodar, Russia, 2010; pp. 1–4.
42. Hachaturov, A.A. Non-Synchronous Switching and Resynchronization in Power Systems, 2nd ed.; Energy: Moscow,
Russia, 1977; pp. 11–25. ISBN 621-311-018-5.
43. Beljaev, N.A. Synthesis of Adaptive Synchronization Systems for Generators with an Electric Network Based
on Automatic Control Methods with a Reference Model. Master’s Thesis, Tomsk Polytechnic University,
Tomsk, Russia, 7 October 2015. (In Russian).
44. Sprint-M Accurate Automatic Synchronization Device. Available online: https://www.rza.ru/catalog/
zahita-i-avtomatika-stancionnogo-oborudovaniya/sprint-m.php?sphrase_id=29938&click=130 (accessed on
7 April 2020).
45. Synchrotact. Sync Devices and Systems. Available online: https://new.abb.com/power-electronics (accessed
on 25 March 2020).
46. Reliability of Distribution Electric Networks 6 (10) Kv Automation Using Reclosers. Available online:
http://www.news.elteh.ru/arh/2002/17/08.php (accessed on 23 April 2020).
47. Vorotnickij, V.; Buzin, S. The Recloser is a New Level of Automation and Control of Overhead Lines of
6 (10) kV. Available online: http://www.news.elteh.ru/arh/2005/33/11.php (accessed on 27 April 2020).
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
96
energies
Article
Location and Sizing of Micro-Grids to Improve
Continuity of Supply in Radial
Distribution Networks
Fernando Postigo Marcos *, Carlos Mateo Domingo, Tomás Gómez San Román
and Rafael Cossent Arín
Institute for Research in Technology (IIT), Comillas Pontifical University, 28015 Madrid, Spain;
[email protected] or [email protected] (C.M.D.); [email protected] or
[email protected] (T.G.S.R.); [email protected] or [email protected] (R.C.A.)
* Correspondence: [email protected] or [email protected]; Tel.: +34-91-542-2800 (ext. 2729)
Abstract: The steady decline in the prices of distributed energy resources (DERs), such as distributed
renewable generation and storage systems, together with more sophisticated monitoring and control
strategies allow power distribution companies to enhance the performance of the distribution network,
for instance improving voltage control, congestion management, or reliability. The latter will be
the subject of this paper. This paper addresses the improvement of continuity of supply in radial
distribution grids in rural areas, where traditional reinforcements cannot be carried out because they
are located in secluded areas or in naturally protected zones, where the permits to build new lines
are difficult to obtain. When a contingency occurs in such a feeder, protection systems isolate it,
and all downstream users suffer an interruption until the service is restored. This paper proposes a
novel methodology to determine the optimal location and size of micro-grid systems (MGs) used
to reduce non-served energy, considering reliability and investment costs. The proposed model
additionally determines the most suitable combination of DER technologies. The resulting set of MGs
would be used to supply consumers located in the isolated area while the upstream fault is being
repaired. The proposed methodology is validated through its application to a case study of an actual
rural feeder which suffers from reliability issues due to the difficulties in obtaining the necessary
permissions to undertake conventional grid reinforcements.
1. Introduction
The digitalization of the energy industry represents a turning point in the sector development [1].
In the case of electricity network planning [2], until the end of the 20th century, networks were mainly
composed of a set of passive assets governed by electromechanical protection elements. With the
sector digitalization, a wide range of possibilities enabling an increase in global social welfare has
been opened. Regarding system planning and operation, these possibilities are translated into new
management strategies that allow optimizing the performance of the system.
Nowadays, the levels of monitoring and control of transmission grids are fairly high; however,
transmission only accounts for 3% of the total length of the European electricity network [3].
The remaining 97% corresponds to the distribution system [4], whose monitoring levels are significantly
lower, especially in the case of rural networks [5]. These grids are mostly radial. Consequently, a failure
in any of the network elements usually implies the loss of supply for all customers located downstream
of the nearest circuit-breaking device [6]. Moreover, in many cases, orography hampers fast fault
detection and repair, leading to long restoration times and deterioration in quality of service.
It is worth stressing that in more than half of the European countries [4], Distribution System
Operators (DSOs) are exposed to incentive-penalty schemes associated with the power quality and
continuity of supply indexes in their networks [7]. Increasing grid reliability has conventionally
been achieved through traditional network reinforcements, such as the installation of switching and
control equipment throughout feeders, the construction of new parallel power feeders, or meshing
existing feeders to allow network reconfiguration [8]. However, these solutions are not always possible,
since rural networks are sometimes placed in secluded and/or protected areas where permits to build
this type of infrastructures are very hard to obtain. To overcome these difficulties, DSOs may consider
alternative solutions based on micro-grids (MGs) in order to improve quality of supply without causing
a major environmental impact.
MGs can be defined as network subsets with a high degree of automation and distributed energy
resources (DERs) that can be operated in an islanded manner [9]. As several international experiences
show [10], MGs are not a futuristic solution, but current solutions for which new policies and regulations
are already being developed [11]. This is mainly due to the sharp reduction in the costs of DERs in
recent years [12].
According to the existing literature, MGs can be classified according to different criteria. In [13],
two types of MGs are defined from a regulatory perspective: customer microgrids (or true microgrids)
which are located downstream of an end-user meter, and utility microgrids (or milligrids) that involve
a segment of the common distribution network. While milligrids are not technologically different
from customer microgrids, they are radically different from a regulatory viewpoint, largely because
milligrids integrate the conventional utility network as in this case. However, other categorizations
can be made as indicated in [14], where a taxonomy is presented based on their use, with the main
categories being: remote, utility distribution, community, institutional, military, and commercial.
Islanded operation of MGs also imposes new technological requirements on the operation and control
of DER. If only inverter-connected DER is available, at least one of the DER units must operate in
grid-following or voltage-source-based grid-supporting mode [15].
Regarding distribution networks, MGs are able to increase the number of control variables
(e.g., islanded operation), and enable new management strategies [16] that can be applied in order
to achieve diverse objectives depending on the network needs [17]. Previous research has already
addressed several of these applications, including: decreasing technical losses [18], helping voltage
control [19], increasing PV hosting capacity [20], rural electrification [21], increasing global efficiency [22],
or market services [23]. Nevertheless, one of the most relevant advantages of these solutions comes
from the opportunity of improving continuity of supply in the case of network failures. To this end,
some publications have developed methodologies for locating generators [24,25] and reclosers [26] in
distribution networks.
Nonetheless, as far as we know, to date, no reference in the literature has addressed the same
problem tackled in this paper. This paper proposes a novel methodology that allows obtaining optimal
MG designs that enable supplying all those consumers that, after a contingency, could not be supplied
from the original source (normally the upstream substation). In this paper, the location, the type
(storage, PV systems, and diesel groups) and the size of the DER technologies in charge of feeding the
MG are chosen according to a multicriteria optimization, in which both the investment in these new
technologies and the increase of reliability obtained with them are evaluated.
The remainder of this paper is structured as follows. Section 2 defines the problem addressed
and poses the type of solutions that can be obtained. Section 3 presents the nomenclature used in this
paper. Section 4 digs into the methodology applied to optimize the location and size of MGs. Section 5
applies the previously developed methodology to a synthetic realistic rural network. Finally, Section 6
outlines the main conclusions.
98
Energies 2020, 13, 3495
2. Problem Statement
As presented in the introduction, this paper addresses the problem of improving the reliability
of radial distribution networks in rural and secluded areas. Currently, when a contingency occurs
in such a feeder, protection systems isolate it, and all downstream users suffer an interruption until
the fault is repaired and service is restored. This paper proposes a methodology for determining the
optimal size and location of a set of DERs (a combination of generators and storage), considering both
reliability improvements and investment minimization. In the event of a contingency, and once the
fault is isolated, the installed DERs allow to keep supplying the affected consumers while the fault is
being repaired, thus minimizing the duration of the interruption times of those customers. This process
is shown in Figure 1. The proposed reliability improvement application can be combined with other
potential services that DERs can provide, such as voltage control or congestion management, since the
islanded operation will only be used sporadically when network outages occur.
&ĂŝůƵƌĞ
EĞƚǁŽƌŬ ZƐ
/ŶŝƚŝĂů ƌĞĐŽŶĨŝŐƵƌĂƚŝŽŶ ƐƵƉƉůLJ /ŶŝƚŝĂů
ƐŝƚƵĂƚŝŽŶ ĂŶĚŝƐůĂŶĚŝŶŐ ŝƐůĂŶĚƐ ƐŝƚƵĂƚŝŽŶ
For illustrative purposes, this section provides an overview the kind of MG-based solutions
that the proposed model would yield, and the main concepts used throughout the paper. However,
these concepts are further detailed in Section 3. In this line, Figure 2 shows an example of a network
divided into four zones (z1 , z2 , z3 , z4 ). These zones are delimited by three smart switches (ss1 , ss2 , ss3 )
that enable the reconfiguration of the network when an outage occurs. This figure shows a possible
solution to improve the reliability of this network. Two sets of DERs (derd1 and derd2 ) are installed
in different zones of the system. The colors used denote that derd1 is designed to supply three zones
(z1 , z2 , and z3 ), while derd2 is dimensioned to supply just z4 .
The Network Solution (nsg1 ) for the previous network example is described in Figure 3.
The distribution network has four zones (z1 , z2 , z3, and z4 ), and two MGs (mgf1 and mgf2 ) designed to
99
Energies 2020, 13, 3495
improve the reliability of the network. These MGs are dimensioned considering the load within two
groups of zones (gzc1 and gzc2 ) and comprise two sets of DERs (derd1 and derd2 ), respectively.
ŶƐŐϭ
ŵŐĨϭ ŵŐĨϮ
ĚĞƌĚϭ ĚĞƌĚϮ
ŐnjĐϭ ŐnjĐϮ
njϭ njϮ njϯ njϰ
Figure 3. Nomenclature diagram for a selected Network Solution.
Considering the same network, under a conventional scenario, when a failure takes place in z2,
the nearest upstream switching element (ss1 ) would isolate it from the upstream grid, leaving z3 and
z4 without supply, while z1 would continue being supplied from the upstream substation. However,
with the Network Solution exemplified in Figure 2, the same fault in z2, would only lead to the loss of
supply of the affected zone, since derd1 would supply z3 , and derd2 would supply z4, while z1 would
continue being supplied from the upstream substation. It should be noted that DERs would only
be operated for this purpose while the failed zone is out of service, and only supplying those zones
that cannot be supplied from the upstream substation due to the fault location. The rest of the time,
these DERs may be used for other purposes like voltage control, energy production, or energy arbitrage
in the case of batteries. It is important to note that if the contingency takes place in the same zone
where the DERs are located, it is assumed that the corresponding micro-grid would lose its ability to
be operated in an islanding manner. For instance, in the example of Figure 2, if the fault occurs in zone
z3 , the load of this zone would not be supplied, counting as non-served energy, whereas zones z1 and
z2 will be supplied by the upstream substation, and derd2 would supply z4.
The methodology presented in this paper is observed from a DSO point of view. The utility can
use this tool to plan the resources of its property used in case of failure, and which traditionally have
consisted of diesel units. These DERs could be located in the point located further upstream of the
indicated zone. In this way, the network will operate in normal conditions and congestion or voltage
problems will be avoided. However, different strategies can be applied according to the wishes of the
operator as described in the paper.
• Smart Switches (ssa ): Tele-controlled switching devices able to isolate damaged parts of the
network when a failure or contingency takes place. It is assumed that their actuation—status
change to open and network reconfiguration—is fast enough to avoid worsening reliability
indexes commonly counted for interruption durations longer than 3 min [27]. The set that includes
all the smart switches located in the network is defined as SS, while A is the total number of
elements—number of smart switches—as shown in Equation (1)
SS = {ssa }aa=
=1
A
(1)
100
Energies 2020, 13, 3495
• Zone (zb ): Network subset bounded by smart switches. The set that includes all the network
zones is defined as Z, and B is the total number of zones. B can be obtained as the number of
smart switches A plus one. The proposed notation is shown in Equation (2).
Z = {zb }bb=
=1
B=A+1
(2)
• Group of Zones (gzc ): Set of adjacent network zones potentially supplied by the same generation
and storage installations under islanded operation. The set that includes all the possible groups of
zones is defined as GZ, and the total number is C. This value depends on the network topology
and the number/location of smart switches, and it is calculated as the number of different sets of
connected zones as Equation (3) shows.
C
GZ = gzc gzc ⊆ P(Z) AND Zones( gzc ) are connected cc=
=1 (3)
where:
“P (input element)” is a function that provides the power-set of an input set, in this case,
the Zones set. The power-set, in mathematics, is defined as the set of all the subsets of a set.
“Zones (input element)” is a function that provides the set of zones of which the input
element is composed.
• Distributed Energy Resources (dersd ): Tuple of generation facilities and storage installations
as shown in Equation (4). In this paper, different solar PV installations (pvi ), diesel units (duj ),
and batteries (bessk ) are considered. The indexes i, j, and k determines which elements of the
equipment catalog are selected for each DERs design derd . The PV installations and the diesel
units are defined by their rated power (kW). The storage installations are defined by their rate
capacity (kWh) and their ratio power/capacity (kW/kWh). In addition, the annualized CAPEX
(Capital Expenditure) and OPEX (Operating Expenditures) are considered for all the installations.
The set that includes all the possible distributed energy resources combinations, according to the
equipment catalog, is defined as DER, and the total number is equal to D.
D
DER = derd | d = (i, j, k) dd=
=1 (4)
Moreover, the set that includes all the optimal DERs for the group of zones gzc in terms of
investment and non-served energy (reliability index) is defined as DER_OPTc as shown in Equation
(5). The total number of optimal DERs for the group of zones gzc is equal to E. Thus, DER_OPTc is just
the optimal subset of DER for the group of zones gzc .
E
DER_OPTc = der_optce | der_optce ⊆ DER) ee=
=1 (5)
• Micro-Grid (mgf ): Tuple composed by a group of zones gzc , a tuple of optimal DERs der_optc e
feeding that group of zones, and the zone zb , where the DERs are located as shown in
Equation (6).The set that includes all the possible MG designs is defined as MG, and the
total number is equal to F.
f =F
MG = mg f mg f = ( gzc , derce , za ) AND za Ie Zones mgf (6)
f =1
• Network Solution (nsg ): Subset of the micro-grids (MG) set, where all the zones are included
but only once, as Equation (7) details. In other words, all the network is included in a network
101
Energies 2020, 13, 3495
solution like the combination of different MGs. The set that includes all the possible network
solutions is defined as NS, and the total number is equal to G.
g=G
NS = ns g ns g ⊆ MG AND Z ⊆ Zones ns g AND Zones ns g \Z = o (7)
g=1
Section 4 presents the developed methodology for obtaining the set of optimal Network Solutions
to improve the continuity of supply under network contingencies while minimizing the associated
investment of DER installations.
4. Methodology
This section details the methodology developed to obtain the aforementioned Network Solutions,
understood as a combination of micro-grid system designs and locations. As shown in Figure 4,
the construction of the set of Network Solutions starts with a network partition based on the location of
the existing smart switches. Next, based on the grid connectivity layout, all possible Groups of Zones
are identified. Subsequently, the DER installations for supplying each of the Groups of Zones—and
therefore the MGs—are sized and located, finding the optimal combinations of technologies for each
case. This optimization is carried out attending to a double perspective: minimization of investment
costs and non-supplied energy. Since this is a multicriteria optimization, the sizing step makes it
possible to find several optimal DER combinations for each MG, and therefore, different Network
Solutions. Finally, network reliability is assessed for each Network Solution, selecting those ones that
are optimal (non-dominated solutions in the Pareto front) in terms of reliability and investment.
Broadly speaking, multicriteria optimization problems such as the one addressed herein can be
tackled through two main approaches [28]. On the one hand, weights can be assigned to each of the
functions to be optimized in order to obtain a single objective function which, once solved, yields a
single optimal solution. This solution strongly depends on the weighting factors selected, which vary
according to the planner preferences. On the other hand, a second alternative consists of evaluating all
the individual objective functions separately and obtain a set of optimal, non-dominated, solutions that
belong to the Pareto front. In this case, there is no longer a single solution, but a set of non-dominated
ones, i.e., there is no other possible solution that performs better than these in all the separate objective
functions at the same time. The methodology proposed in this paper follows this last approach.
Distribution network planning, and specifically the installation of DERs, entails important
investments. Thus, it is really convenient for distribution companies to have a wide range of
well-performing solutions so that they can choose the one that fits better their needs. The proposed
methodology offers the advantage of obtaining all the non-dominated solutions without exploring
all possible combinations. This is achieved thanks to a two-step optimization following a bottom-up
102
Energies 2020, 13, 3495
approach. In the first stage, MGs are constructed sizing the DERs that minimize the MG non-served
energy and investment cost. In a second step, an optimization that builds the Network Solutions
with minimal non-served energy and investment is performed. Therefore, the total complexity of the
whole design problem is significantly reduced. The computational burden of solving this problem is
essentially determined by the number of smart switches in the feeder, as they increase the network
reconfiguration possibilities. Nevertheless, the type of practical applications of the proposed planning
solutions would be mostly installed in rural areas. In rural areas, the level of distribution network
automation is typically low (normally between three and eight smart switches per feeder); consequently,
the methodology described in this section is totally valid as demonstrated by the case study in Section 4.
In the following sub-sections, the different methodology steps, according to Figure 3, are described
in depth, including the assumptions made and the developed mathematical formulation.
Z = {z1 , z2 , z3 , z4 } (9)
Before determining the location of DER, it is important to determine the zones they would be
able to supply. For this purpose, all the possible Groups of Zones are determined. According to
the definition in Equation (3), the set of all the Group of Zones GZ is determined by every group of
interconnected zones that belongs to the power-set of Z. In the example, Equation (10) shows the
resulting power-set of Z.
P(Z)
= {{ }, {z1 }, {z2 }, {z3 }, {z4 }, {z1 , z2 }, {z1 , z3 }, {z1 , z4 }, {z2 , z3 }, {z2 , z4 }, {z3 , z4 }, {z1 , z2 , z3 }, {z1 , z2 , z4 }, {z2 , z3 , z4 }, {z1 , z3 , z4 }, {z1 , z2 , z3 , z4 }}
(10)
The pseudocode shown in Table 1 summarizes the process used to obtain GZ from the power-set
of Z, and its total number of elements C.
Pseudocode
GZ = ø; c = 0
for each “element” in P(Z)
if the “element” is connected
Add “element” to GZ
c += 1
Accordingly, Equation (11) presents the GZ obtained for the example network through the
pseudocode in Table 1.
GZ = {{z1 }, {z2 }, {z3 }, {z4 }, {z1 , z2 }, {z2 , z3 }, {z2 , z4 }, {z1 , z2 , z3 }, {z1 , z2 , z4 }, {z2 , z3 , z4 }, {z1 , z2 , z3 , z4 }} (11)
103
Energies 2020, 13, 3495
gzc , a tuple of optimal DERs der_optc e , and finally the zone za where the DER installations are located.
These last two elements will be the object of study in this sub-section.
The process of obtaining the MGs is carried out in two stages. In the first one, for each Group
of Zone gzc , the optimal elements of the DER set “der_optc e ” are determined. These are those that
minimize investment costs and non-served energy. Secondly, the obtained der_optc e is located in
each one of all the zones za , which belongs to the considered Group of Zones gzc . This process is
described as pseudocode in Table 2 however, the sizing and siting procedures will be further explained
in Sections 4.2.1 and 4.2.2 respectively.
Pseudocode
MG = ø; DER_OPT = ø c
DER sizing
nsecd= Non-Served Energy in gzc supplied by derd
invd = Investment of derd
solutions.add (nsecd, invd)
for each “element” in solutions
if ”element” belongs to Pareto front
DER_OPTc.add (derd)
for each der_optce in DER_OPTc
DER siting
for each za in gzc
mgf= (gzc ,der_optce, za)
MG.add(mgf)
It should be noted that all selected DER technologies (solar PV, diesel units, and batteries),
specified in the input catalog—DER set—are considered in each MG. Thus, those DER combinations
that minimize the investment and the MG non-served energy are considered optimal and part of the
DER_OPTc set, while the rest of other DER combinations are discarded. Figure 5 shows this process.
104
Energies 2020, 13, 3495
The red triangles represent the optimal subset of DER for the analyzed MG—DER_OPTc —while the
grey circles represent all the rest DER combinations included in the catalog—DER set.
Figure 5. Pareto optimization for DER combinations for each analyzed MG.
To calculate the MGs’ non-served energy, two algorithms are used. The first one is based on the
installed DER capacity (kW), while the second one is based on the energy demanded by the affected
customers during the contingency. The latter is needed due to the energy-constrained operation of
batteries. The capacity (kWh) and the power (kW) are the two design parameters of batteries.
It should be mentioned that only those hours with a generation deficit are counted according
to Equation (14). In the deficit case, the difference (Di fP (h)) between the micro-grid demand (D(h))
and the DER availability at that hour is computed. As previously mentioned, DERs are composed
of photovoltaic generation, where (PPV (h)) is the estimated PV production at hour h, diesel units,
where (Pdie) is the rated capacity (kW), and batteries, where (Pbat) is the maximum power that can be
delivered by the battery, as shown in Equation (15). This calculation is made for all of the hours of the
year, since the demand and solar generation profiles vary hour by hour, and the hour when the fault
occurs is not a priori known.
⎧
⎪
⎪
⎨Di fP (h) Di fP (h) ≥ 0
NSE1 (h) = ⎪
⎪ (14)
⎩0 Di fP (h) < 0
105
Energies 2020, 13, 3495
the previous power-based calculation, this micro-grid non-supplied energy approach (NSEE (h)),
only accounts for those hours with a generation deficit, see Equation (16).
⎧
⎪
⎪
⎨Di fE (h) Di fE (h) ≥ 0
NSEE (h) = ⎪
⎪ (16)
⎩0 Di fE (h) < 0
h1=h
+hrep
The difference (Di fE (h)) between the energy to be supplied to the demand ( D ( h1 ) )
h1=h
during the repair time of the fault, and the generation that can be provided by the photovoltaic
h1 =h+hrep
panels ( PPV (h1 )), the batteries (Ebat ) and the diesel units (Pdie ∗ hrep ) is shown in Equation (17).
h1 =h
It should be noted that the available energy storage in the battery is calculated considering the battery
capacity (kWh) and the state of charge of the battery when the fault occurs (S.C.pu ).
h1=
h+hrep h1=
h+hrep
Di fE(h) = D(h1) − PPV (h1) − EBAT ∗ S.C.pu − PDIE ∗ hrep (17)
h1=h h1=h
The micro-grid energy demand (E(h)) during the contingency time is obtained through the sum
of the hourly demands in that period according to Equation (19).
h1 =h+hrep
E(h) = D(h1 ) (19)
h1 = h
To express the micro-grid non-served energy in unitary terms, the MG non-served energy
coefficient is computed as the ratio between the micro-grid non-served energy (NSE(h)) and the total
demanded energy (E(h)) as shown in Equation (20), assuming that the fault may equiprobably happen
in any of the hours of the year.
h NSE(h)
NSEpu = (20)
h E(h)
As mentioned above, those non-dominated combinations of the DER set that minimize the
non-served energy and the investment will be selected as optimal to be part of the DER_OPTc set.
Only the set of optimal DER combinations will be considered in the algorithms presented in the
following sections, while the rest of DER combinations will be discarded.
106
Energies 2020, 13, 3495
considered Group of Zones. Thus, for each Group of Zones, gzc , and optimal set of DER, der_optc e ,
there will be as many different micro-grids, mge , as zones are included in the corresponding Group of
Zones. This algorithm is clarified in the second part of the pseudocode presented in Table 1.
1. The Group of Zones are combined to compose the whole network. Table 3 shows all the possible
network combinations for the example network shown in Figure 2.
2. These Groups of Zones that compose the network combinations are substituted with all the MGs
that contain them to create the Network Solutions set, NS. It should be emphasized that more
than one MG may contain the same group of zones.
3. In the case of our example network (see Figures 2 and 3), the nsg1 is formed following the network
combination number 5, where the Group of Zones {z1 , z2 , z3 } and {z4 } correspond to mgf1 and
mgf2 , respectively.
107
Energies 2020, 13, 3495
DERs, the MG non-served energy coefficient will decrease proportionally to the installed power in the
non-served MG zones.
It should be noted that under normal conditions (without contingencies) the operation of the
entire network is radial. Only in the contingency case is the failure zone isolated, and the rest of
the zones that cannot be supplied from the upstream substation (due to the fault location) would be
supplied by the DERs allocated within their associated MGs.
For illustrative purposes, the example network used in the previous sections is analyzed.
This network solution is composed of two MGs, the first one “mgf1 ” is designed to supply z1 ,
z2, and z3 where the derd1 are located, the second one “mgf2 ” is only formed by z4 where derd2 are
placed. Assuming the installed power in each zone is the same, and mgf1 and mgf2 have a 0.46 pu and
0.38 pu MG non-served energy coefficients, respectively (result of Equation (20)), the obtained zone
interaction matrix for the nsg1 is the one shown in Equation (21).
⎛ ⎞
⎜⎜ 1 0 0 0 ⎟⎟
⎜⎜ ⎟
⎜⎜ 0.19 1 0 0 ⎟⎟⎟⎟
ZInsg = ⎜⎜⎜ ⎟ (21)
⎜⎜ 0.19
⎜⎝ 0 1 0 ⎟⎟⎟⎟
⎠
0.38 0.38 0 1
Once the Zone Interaction matrix ZInsg is calculated for each Network Solution, the reliability
assessment for each Network Solution is carried out until complete the whole NS set. For the reliability
assessment, it is necessary to consider some additional parameters, such as the number of supply
points in each analyzed zone (Nb ), the average repair time of the fault in the considered zone (h repb ),
and the annual failure rate of the network included in the considered zone ( fb ) obtained as the sum of
the product of the failure rate of overhead lines, expressed per km, times the overhead line lengths
in the zone plus the product of the failure rate of underground lines, expressed per km, times the
underground line lengths in the zone. Equation (22) shows the SAIDI calculation for a generic Network
Solution nsg .
b1=B b2=B ns g
b2=1 b2=1 fb2 ∗ h repb2 ∗ Nb1 ∗ ZIb1,b2
SAIDI =ns g
b1=B (22)
b1=1 Nb1
108
Energies 2020, 13, 3495
Once the Network Solutions have been selected, a specific location will be assigned for the
DERs of each zone according to the established criteria, and a powerflow analysis will be carried out.
Those Network Solutions that do not cause operational problems will be validated to be implemented.
5. Case Study
5.1. Description
In this section, the proposed methodology is applied to an actual distribution network.
The analyzed feeder presents reliability problems. In addition, the feeder is located in a protected
natural area where traditional network reinforcements are not an option due to the difficulty of
obtaining permits. For this reason, the DSO in charge of supplying this area is exploring the use of
non-conventional network solutions that minimize the impact on the environment.
The MV feeder analyzed presents three smart-switches and follows an identical structure as the
example used for describing the proposed methodology, as shown Figure 7.
Table 4 shows the number of supply points located in each of the feeder zones, as well as the
overhead and underground lengths of the power lines included in each zone.
Zone Number of Supply Points Overhead Length (km) Underground Length (km)
z1 0 10.99 0.04
z2 585 9.54 0.82
z3 1386 9.94 2.07
z4 771 6.04 0.15
Regarding failure rates, 12 outages/100 km-year was set for overhead lines and 6 outages/
100 km-year for underground lines. Concerning the fault repair time, an average time of 6 h was
considered. This data is consistent with the values posed in reference [29].
Regarding the catalog of DERs, Appendix A details the rated capacities, as well as investment
and operation and maintenance costs of PV, batteries, and diesel units. It should be mentioned that
the irradiation profile used to obtain PV generation is based on the geographical coordinates of the
analyzed zone and obtained from [30].
109
Energies 2020, 13, 3495
5.2. Results
Applying the abovementioned methodology on the proposed network, 1,531 network solutions
were obtained. Figure 8 shows a Pareto diagram in which the obtained reliability index (SAIDI) and
the associated annualized investment in each network solution are displayed. Each of the dominated
network solutions is represented by a circle, while the non-dominated solutions are represented by
(red) triangles. Furthermore, in order to facilitate the selection of the solution that best suits the system
needs, an additional parameter has been calculated. This parameter is, for each network solution,
the standard deviation of the zonal SAIDIs with respect to the overall network SAIDI, measured
in percentage. This parameter is represented by a color code in which the warmer the color the
higher disparity between zonal reliability indicators. Therefore, at the same level of overall SAIDI and
investment, solutions with similar reliability indices between zones (cooler colors) would be preferred
ensuring more homogeneous solutions in terms of reliability.
Figure 8. Network solutions for the case study network (colors represent the standard deviation of
zonal SAIDI in percentage).
110
Energies 2020, 13, 3495
Another important issue to analyze is, for each non-dominated network solution, in which network
zone DERs are located and which DER technologies are selected. Table 6 presents this information.
For each network solution, it identifies in which zone the DERs are located (DERs Zone) and the zones
they supply (z1 , z2 , z3 , and z4 ). The size of the PV systems (PPV ), batteries (PBAT and EBAT ) and diesel
units (PDIE ) are also included. It can be seen that all the optimal solutions include DERs in zone number
3. This is an expected result since this is the area with the highest number of connected supply points
and located farthest from the supply point. The same happens with zone number 4 —the zone with
the second-highest number of supply points—for similar reasons.
Regarding the DER combinations selected, it is observed that all network solutions except
one include diesel units, in some cases supported by batteries. These batteries have a high power
(kW)–capacity (kWh) ratio, which leads to the conclusion that batteries are mainly used for peak
shaving, avoiding the installation of an additional diesel unit with a low use rate. This effect can
be observed by comparing the evolution of selected DERs starting from the Base Case and going
through network solutions 18, 17, 16 and 15. It can be observed that as the reliability index improves,
the selection of batteries (solutions 18 and 16) is alternated with that of diesel units (solutions 17 and
15). Therefore, a higher granularity in diesel unit sizes presumably would eliminate the need for
batteries in the optimal network solutions.
On the other hand, it is observed that there is no optimal solution selecting PV facilities. This is
mainly explained by two reasons. The first is the comparatively high PV investment cost. Nowadays,
PV systems can be cost-effective when the produced energy is self-consumed and/or sold to the market
during the expected life of the installation. However, in this paper, we are assuming that PV is only
used to obtain benefits of its production during the small number of hours per year in which network
outages take place. The second reason is that, unlike storage or diesel units, the PV production is limited
to the sunlight hours. This effect is modeled through solar radiation hourly profiles for the specific
111
Energies 2020, 13, 3495
network location. The use of annual profiles is recommended to capture seasonality. For example, if the
network failure occurs at night or at the end of the day, PV will not be able to provide the necessary
energy to supply the isolated customers and to improve the continuity of supply indices.
Therefore, it can be concluded that for the network studied, and with current investment costs,
the use of diesel units by themselves or in combination with batteries is the most cost-effective network
solution to increase the reliability of the system. During network faults, the diesel units would cover
the base demand and the batteries would cover the peaks of the isolated supply points.
A sensitivity analysis to the investment cost of the batteries was performed. To isolate the effect of
this parameter, a new case was run for which battery systems were the only DER technology available.
There are four reasons for this. Firstly, batteries are less environmentally pollutant than diesel units
(at least locally). Additionally, batteries are already being used in the optimal solutions of 6. Secondly,
112
Energies 2020, 13, 3495
PV systems do not seem to be a good option to improve system reliability because their use is restricted
to the sunlight hours and, therefore, they cannot be used when outages occur at night. Moreover,
unlike batteries, in none of the 18 optimal network solutions the PV systems are chosen as a design
option, being more cost-effective to invest in batteries. Third, diesel units are a mature technology
from which no major price variations are expected. In this case, the only factor that could increase
their cost would be a sharp rise in the cost of diesel (variable cost included in the O&M cost defined in
the Annex). However, these are units that operate very few hours per year and are therefore not very
sensitive to this factor. Finally, the cost of batteries nowadays presents a steady downwards trend.
Therefore, the break-even point for batteries to become the preferred design option was evaluated.
Figure 9 shows the Pareto front for different percentages of battery cost reductions. It can be
observed that to obtain a Pareto front similar to the one presented with current investment prices for
PV and batteries and including diesel units, reductions in the cost of batteries by close to 80% would
be needed.
Analyzing the results obtained, it can be concluded that diesel generators and batteries (as long as
their initial state of charge is sufficient) are the preferred option, since, unlike PV systems, they are
available at any time of the day and under any atmospheric conditions. This does not mean that PV
systems cannot be used for this purpose, but they do have some features that make them difficult to
apply. However, each case must be studied individually and new clean and hybrid solutions such as
PV systems with batteries should be considered in future low-cost scenarios.
6. Conclusions
This paper proposes a novel methodology for determining the optimal location and sizing of
MGs based on a multicriteria optimization in which both the DER investment cost and the network
reliability levels are considered. Unlike other publications, this methodology allows the mix of DERs
to be selected that best fits the needs of the network. Additionally, the result is a set of non-dominated
113
Energies 2020, 13, 3495
solutions, allowing the DSO to select the one that best suits other possible non-modeled requirements
(e.g., local emissions). This methodology was tested in a real case study, from which some relevant
findings have been extracted, as discussed below.
The results show that the combination of diesel units and batteries seems to be the most
cost-effective option to increase the reliability of the network through islanded operation. During the
failure repair time, diesel units would be in charge of covering the base demand of islanded zones,
whereas batteries would operate as peaking units. Moreover, the results show that selecting solar PV
installations for this purpose is not cost-effective. The reason for this is twofold. On the one hand,
PV presents a comparatively high investment cost. On the other hand, PV can only supply the demand
during sunlight hours, thus being useless to tackle outages occurring at night. However, each case
must be studied individually and adjusted to the specific climatic conditions of the network location.
Finally, it has been shown that if battery investment costs were to drop by close to 80% with respect to
current values, MG solutions based mainly on batteries instead of more highly polluting solutions
based on diesel units would be the preferred option.
As observed in the case study, emission-free DERs are expensive and are only used in less than 1%
of the hours to enhance the reliability of the distribution grid. For this reason, DSOs may explore the
procurement of a service through which third parties, such as DER owners, would provide network
support when a network outage takes place. The remuneration of this service would be based on
the location of the DER in the network and its commitment to offering the service when requested.
In this way, the DSO would reduce investments in network infrastructure and the owners of DERs
could obtain additional benefits to the regular ones derived from their participation in the market or
cost-savings thanks to behind the meter generation.
This work opens up future research lines to be explored. The main limitation of this study
is the degree of automation of the network. The greater the degree of automation of the network,
the greater the number of smart-switches, and the possibilities of reconfiguring the network would
grow exponentially. For this reason, the proposed solution perfectly fits with the problems of rural
networks, which are characterized by low automation levels. However, urban areas with higher
automation levels may sharply increase the computational load. For this reason, this limitation opens,
for example, a future research line in applying metaheuristic algorithms (e.g., genetics algorithms) to
overcome this issue.
Author Contributions: Conceptualization, F.P.M., C.M.D., T.G.S.R. and R.C.A.; methodology, F.P.M., C.M.D. and
T.G.S.R.; software, F.P.M. and C.M.D.; validation, F.P.M., C.M.D., T.G.S.R. and R.C.A.; writing—review and editing,
F.P.M., C.M.D., T.G.S.R. and R.C.A. All authors have read and agreed to the published version of the manuscript.
Funding: This research has received funding from the Fundación Iberdrola España through the ‘Ayudas a la
investigación 2019 program.
Acknowledgments: The authors would like to thank i-DE for partially supporting this research. In particular,
comments and suggestions for improving the practical application of the developed algorithms are acknowledged
to Diego Haro Ginés, Roberto García Gómez, Fernando Palacios González, and Cristina Vilá Castro.
Conflicts of Interest: The authors declare no conflict of interest.
114
Energies 2020, 13, 3495
The price of the batteries is based on [33]. Other additional costs like de O&M costs have been
considered according to the breakdown given in [34]. For the power-capacity ratio, the costs indicated
in [32] have been taken into account. To obtain the total investment cost of batteries for different
power-capacity ratios, the power and energy values are averaged. The useful life considers that very
few charge/discharge cycles are expected per year in this application. Table A2 shows the data used.
Rated Power (kW) Power (kW)—Capacity (kWh) Ratio Investment Cost (€/kW)
0 3.33 314.26
100 1 Investment cost (€/kWh)
200 0.4 346.21
400 0.25 O&M (€/kWh)
800 0.356
Useful life (years)
15
The price of diesel units depends mainly on their rated capacity [35]. Apart from the price of
the diesel unit itself, an installation cost of 60% was considered. For the fuel, a cost of 1.3 €/L and an
expense of 0.3l/kWh has been assumed [36] to obtain the total O&M cost. A useful life of 20 years
is considered, assuming that the unit will be used only a few hours per year. Table A3 shows the
data used.
References
1. European Commission. Digitalisation of the Energy Sector; SETIS Magazine; European Commission: Brussels,
Belgium, 2018.
2. Makris, P.; Efthymiopoulos, N.; Nikolopoulos, V.; Pomazanskyi, A.; Irmscher, B.; Stefanov, K.; Pancheva, K.;
Varvarigos, E. Digitization Era for Electric Utilities: A Novel Business Model Through an Inter-Disciplinary
S/W Platform and Open Research Challenges. IEEE Access 2018, 6, 22452–22463. [CrossRef]
3. Staschus, K. ENTSO-E Memo 2011; Secretariat of ENTSO-E: Brussels, Belgium, 2011.
4. Pavla Mandatova; Gunnar Lorenz. Power Distribution in Europe: Facts & Figures; Eurelectric: Brussels,
Belgium, 2013.
5. Short, T.A. Electric Power Distribution Handbook; CRC Press: Boca Raton, FL, USA, 2014.
6. Electric Power Distribution Engineering. Available online: https://www.crcpress.com/Electric-Power-
Distribution-Engineering-Third-Edition/Gonen/p/book/9781482207002 (accessed on 20 February 2019).
115
Energies 2020, 13, 3495
7. Pérez-Arriaga, I.J. Regulation of the Power Sector; Springer: New York, NY, USA, 2013; ISBN 978-1-4471-5033-6.
8. Chen, T.-H.; Huang, W.-T.; Gu, J.-C.; Pu, G.C.; Hsu, Y.F.; Guo, T.Y. Feasibility study of upgrading primary
feeders from radial and open-loop to normally closed-loop arrangement. IEEE Trans. Power Syst. 2004,
19, 1308–1316. [CrossRef]
9. Lasseter, R.H.; Paigi, P. Microgrid: A conceptual solution. In Proceedings of the 2004 IEEE 35th Annual
Power Electronics Specialists Conference (IEEE Cat. No.04CH37551), Aachen, Germany, 20–25 June 2004;
Volume 6, pp. 4285–4290.
10. Hatziargyriou, N.; Asano, H.; Iravani, R.; Marnay, C. Microgrids. IEEE Power Energy Mag. 2007, 5, 78–94.
[CrossRef]
11. Marnay, C.; Asano, H.; Papathanassiou, S.; Strbac, G. Policymaking for microgrids. IEEE Power Energy Mag.
2008, 6, 66–77. [CrossRef]
12. Pérez-Arriaga, J.I.; Batlle, C.; Gómez, T.; Chaves, J.P.; Rodilla, P.; Herrero, I.; Dueñas, P.; Ramírez, C.R.V.;
Bharatkumar, A.; Burger, S.; et al. Utility of the Future: An MIT Energy Initiative Response to an Industry in
Transition; Massachusetts Institute of Technology: Cambridge, MA, USA, 2016.
13. Types of Microgrids|Building Microgrid. Available online: https://building-microgrid.lbl.gov/types-
microgrids (accessed on 10 April 2020).
14. Asmus, P.; Lawrence, M. Emerging Microgrid Business Models. Available online: /paper/Emerging-
Microgrid-Business-Models-Asmus-Lawrence/67052852b5b104d9e7f61daaa1b7220d74cdb39d (accessed on
9 June 2020).
15. Rocabert, J.; Luna, A.; Blaabjerg, F.; Rodríguez, P. Control of Power Converters in AC Microgrids. IEEE Trans.
Power Electron. 2012, 27, 4734–4749. [CrossRef]
16. Lopes, J.A.P.; Moreira, C.L.; Madureira, A.G. Defining control strategies for MicroGrids islanded operation.
IEEE Trans. Power Syst. 2006, 21, 916–924. [CrossRef]
17. Lasseter, R.H. Smart Distribution: Coupled Microgrids. Proc. IEEE 2011, 99, 1074–1082. [CrossRef]
18. Patel, Y.P.; Patel, A.G. Placement of DG in Distribution System for loss reduction. In Proceedings of the 2012
IEEE Fifth Power India Conference, Delhi, India, 19–22 December 2012; pp. 1–6.
19. Anand, S.; Fernandes, B.G.; Guerrero, J. Distributed Control to Ensure Proportional Load Sharing and
Improve Voltage Regulation in Low-Voltage DC Microgrids. IEEE Trans. Power Electron. 2013, 28, 1900–1913.
[CrossRef]
20. Armendáriz, M.; Heleno, M.; Cardoso, G.; Mashayekh, S.; Stadler, M.; Nordström, L. Coordinated microgrid
investment and planning process considering the system operator. Appl. Energy 2017, 200, 132–140. [CrossRef]
21. Ciller, P.; Ellman, D.; Vergara, C.; González-García, A.; Lee, S.J.; Drouin, C.; Brusnahan, M.; Borofsky, Y.;
Mateo, C.; Amatya, R.; et al. Optimal Electrification Planning Incorporating On- and Off-Grid Technologies:
The Reference Electrification Model (REM). Proc. IEEE 2019, 107, 1872–1905. [CrossRef]
22. Park, W.Y.; Shah, N.; Phadke, A. Enabling access to household refrigeration services through cost reductions
from energy efficiency improvements. Energy Effic. 2019, 12, 1795–1819. [CrossRef]
23. DeForest, N.; MacDonald, J.S.; Black, D.R. Day ahead optimization of an electric vehicle fleet providing
ancillary services in the Los Angeles Air Force Base vehicle-to-grid demonstration. Appl. Energy 2018,
210, 987–1001. [CrossRef]
24. Celli, G.; Ghiani, E.; Mocci, S.; Pilo, F. A multi-objective formulation for the optimal sizing and siting of
embedded generation in distribution networks. In Proceedings of the 2003 IEEE Bologna Power Tech
Conference Proceedings, Bologna, Italy, 23–26 June 2003; Volume 1, p. 8.
25. Ippolito, M.G.; Morana, G.; Sanseverino, E.R.; Vuinovich, F. Risk based optimization for strategical planning
of electrical distribution systems with dispersed generation. In Proceedings of the 2003 IEEE Bologna Power
Tech Conference Proceedings, Bologna, Italy, 23–26 June 2003; Volume 1, p. 7.
26. Popović, D.H.; Greatbanks, J.A.; Begović, M.; Pregelj, A. Placement of distributed generators and reclosers for
distribution network security and reliability. Int. J. Electr. Power Energy Syst. 2005, 27, 398–408. [CrossRef]
27. Tabatabaei, N.M.; Ravadanegh, S.N.; Bizon, N. (Eds.) Power Systems Resilience: Modeling, Analysis and Practice;
Springer International Publishing: Cham, Switzerland, 2019; ISBN 978-3-319-94441-8.
28. Ehrgott, M. Multicriteria Optimization, 2nd ed.; Springer-Verlag: Berlin/Heidelberg, Germany, 2005;
ISBN 978-3-540-21398-7.
29. Kjolle, G.; Sand, K. RELRAD-an analytical approach for distribution system reliability assessment. IEEE Trans.
Power Deliv. 1992, 7, 809–814. [CrossRef]
116
Energies 2020, 13, 3495
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
117
energies
Article
Towards a Sustainability Assessment Model for
Urban Public Space Renewable Energy Infrastructure
Kaan Ozgun
School of Design, Queensland University of Technology, Brisbane 4000, Australia; [email protected]
Abstract: As cities develop new interventions for climate change mitigation, incorporating renewable
energy in urban public spaces becomes a common norm to address sustainability objectives. However,
current built projects and assessment practices mainly uses a “techno-fixes” approach focusing on
strategies that are related to the environmental benefits and neglecting other potential strategies
instigating social and economic benefits of renewable energy. The purpose of this study is to present
a potential sustainability assessment model introducing new strategies for public space renewable
energy use where social and economic benefits of renewables become evident. Supplemented with
theories and principles from ecology, the model’s economic strategies refer whether the project
considers meaningful part of the produced electricity for generating a local economy; environmental
strategy comprises embodied energy, energy storage and self-maintenance; social strategy includes
whether the project considers generating active and passive interaction using on-site electricity.
Ballast Point Park in Sydney was used as a test bed to examine the model and sustainability of park’s
renewable energy use. The findings showed that environmental strategies were evident in the park,
social strategies remained average and economic strategies with renewable energy were lacking.
Interactions with on-site produced electricity was further claimed to be an imperative feature of
any public space. Recommendations were made specific to operational and planning impacts of the
integrated model.
Keywords: renewable electricity distribution for public space; sustainability assessment model;
integrated assessment for public space; tripartite altruism; urban renewable energy; ecological
infrastructures
1. Introduction
Energy autonomy is an emerging trend at city, neighborhood, community and building scales.
Renewable energy, distributed generation, transition from alternative circuit to direct circuit infrastructure
and resilient micro and smart grids, all indicate a transition to new energy environments [1]. Using renewable
energy in built environment becomes a common solution for climate change mitigation. However,
its current use into urban environments is problematic from a sustainability perspective. Despite some
exceptions, most built interventions concentrate on improving cities’ environmental sustainability by
retrofitting urban environments with technological additions such as green walls and photovoltaic
panels [2]. Rather than perceiving renewable energy applications as crude technological addition that
simply address the environmental sustainability objective of a sustainable top-down developmental
approach, scholars and practitioners need to treat renewable energy as “ecological infrastructures”.
For example, positioning the location of “energy–ecology–society relations in a ‘commons’ space”
and for a focus on techniques and social arrangements can serve the aims of sustainability and
equity [3]. Public space showcases for the renewable energy commons approach, a bridge that connects
mainstream energy with the emerging alternative decentralized energy movements. Public space can
also be designed as a real living demonstration of what can be achieved in renewable energy.
The success of sustainable development depends on several effective energy programs and
their management strategies. This includes energy innovations; raising awareness through public
and professional organizations; informing users about energy use and renewable types; increasing
environmental knowledge through training and education; encouraging renewable energy use through
finance, tax initiatives, and/or policies; and integrating assessment and auditing tools [4]. Because the
concept of sustainability has evolved, scholars have recognized the importance of a multidisciplinary
approach to the study and practice of sustainability. “The disciplines are essential for quality control” [5].
Each discipline creates its own quality control tools, one of which is the assessment of sustainability.
Tools that have been highly effective in pushing architectural industry to a better sustainability
standard and have, by association, also influenced sustainable Landscape Architecture. The most
appropriate ones concomitant to the scope of this study comprise ‘LEED’ (Leadership in Energy
and Environmental Design) and ‘SITES’ (Sustainable Sites Initiative—a tool specifically for public
open space assessment) in the USA, ‘BREEAM’ (Building Research Establishment Environmental
Assessment Method) in the UK and ‘Greenstar’ in Australia.
SITES—which was jointly developed by the Lady Bird Johnson Wildflower Center at The
University of Texas (Austin), the United States Botanic Gardens and the American Society of
Landscape Architects—incorporates ecosystems services into its agenda aiming to foster sustainable
land development and management practices [6]. While SITES aims to “create guidelines and
performance benchmarks for sustainable design, construction and maintenance in Landscape
Architecture projects” [6], it has not yet gained acceptance by the Landscape Architecture profession in
Australia as a sustainability assessment tool for public spaces.
While renewable energy assessment is an important aspect of the SITES, their point rating
scheme assesses public spaces based on a set of criteria focusing on environmental objectives such
as making clean electricity available, promoting energy efficiency and reducing carbon footprint.
For example, in their recently updated assessment guidelines (SITES v2), renewable energy is credited
as “use renewable energy sources for landscape electricity needs” [7]. More specifically, if a project
addresses fifty percent of annual outdoor site electricity, three points are scored. If a project generates
one hundred percent of annual outdoor site electricity, four points are scored. Their description clearly
states that renewable energy in a public space context currently does not count as a potential social
sustainability initiator, but mainly considers the economic and environmental aspect of renewable
energy and electricity production. While the relevant information in the guideline booklet includes a
section about community renewable energy systems, its focus is on the produced electricity leasehold
and management [7]. The current SITES v2 guidelines neglect the economic and social sustainability of
on-site produced electricity and overlook renewable energy interactions from a public space context,
thus misses the potential for enriching public space programs with new ways of public engagement.
Both theory and practice currently lack a sound knowledge base. Thus, the study formulates
the following research question—which theories, principles and strategies can better contribute to
the renewable energy-embedded public space and to what extent these principles can be used to
form a potential assessment tool? With the advancement of renewable energy technologies and their
increasing production capacities, renewable energy within a public space context can offer invaluable
outcomes beyond the accepted environmental benefits. The purpose of this study is thus, to present a
potential sustainability assessment model that realizes novel strategies instigating social and economic
benefits of on-site produced electricity on top of the commonly accepted environmental benefits.
The optimal electricity distribution (OED) framework and its strategies were adapted and tested
by assessing renewable energy projects designed for the Freshkills Park site in New York City [8].
A selection committee and multidisciplinary experts from the Land Art Generator Initiative (LAGI)
design competition assessed and shortlisted twenty-five speculative projects [8]. This study considers
LAGI as an authority in this emerging topic because LAGI organizes design competitions since 2010
and has gathered an archive of speculative projects over four different countries attracting experts,
designers, and artists who work on the verge of science, art and technology. Therefore, LAGI’s design
120
Energies 2020, 13, 3428
proposals are likely to be indicative of some dominant contemporary approaches to the design of
renewable energy-embedded public spaces. Part of the strategies presented in this study have been
regularly used by the organizers in the last decade as criteria for judging schemes submitted to the
LAGI design competitions. Both author’s previous studies—as well as LAGI’s frequently tested judging
criteria—lay the groundwork for presented strategies in this study [8,9].
Because the SITES assessment framework is still in its development phase, the study’s another
objective is to enrich the scope of current assessment tools advocating creative processes and strategies
that augment the sustainability of public space renewable energy. Contrary to the focus on the LAGI’s
speculative projects [8], this study adapts the OED framework and its strategies for assessing a built
project site. Ballast Point Park (BPP) in Sydney is used as a test bed to demonstrate the model’s
performance and compatibility of its strategies.
To find out how each strategy is addressed in the design and how BPP design addresses
sustainability from a public space renewable energy distribution perspective, the study acquires
data from site observation, user survey and semi-structured interviews with consulted experts and
designers. The following strategies are formulated coming out of author’s earlier works and the
LAGI’s speculative project assessment criteria [8,9].
• Strategies under economic benefits involve selling 1/3 or meaningful portion of the produced
on-site electricity to city grid for generating income to the public space and its local community;
• Strategies under environmental benefits include using 1/3 or meaningful portion of produced
on-site electricity in the life time of renewable energy infrastructure for embodied energy,
energy storage and maintenance of the renewable energy infrastructure and/or public space;
• Strategies under social benefits focus on the remaining portion of the onsite electricity to be used
in creating activities whether the design considers social active and/or passive engagement with
public space renewable energy.
Strategies are further substantiated with theories from ecology and described in the result section.
The findings enable a blueprint for other sustainability practitioners who consider using renewable
energy in their projects elsewhere. The study discusses the operations and planning impact of the
potential assessment model and concludes with the limitations and future research. Recommendations
are made how renewable energy can be effectively incorporated into public space. Local energy
trading [10] and interaction with on-site produced electricity are further claimed to be an imperative
feature for future public space designs. The need for an integrated approach when using renewables
in public spaces is also discussed.
Next section first introduces the case then presents the methods of the study.
121
Energies 2020, 13, 3428
intelligent verbatim [14]. The interview questions addressed the following topics; the essence of
sustainability, triple-bottom-line (TBL), how renewable energy was integrated into the park design,
and the original goals of overall design concerning social, economic, environmental and esthetic benefits.
(b)
(a) (c)
Figure 1. (a) Ballast Point Park (BPP) location map illustrates the Sydney Harbor context, map extracted
from Google Earth; (b) 1943 BPP site; (c) 2010 BPP site, maps extracted from maps.six.ns.gov.au.
Site observation was used to understand the extend of activities concerning renewable energy
infrastructure in the park. These included how renewable energy was incorporated into the design
and how people interacted. The number of people using the specific location of the renewable energy
infrastructure was also questioned against the use of other park functions. The observation was
conducted during a two-week period in January 2014 including, weekdays, weekends and a public
holiday when the summer started in the southern hemisphere. The site was divided into six zones
and seven exposed spots were investigated, respectively to conduct the observations. By moving
every twenty minutes from one spot to another “with rotating shifts of early morning 7:30–10:00 a.m.;
morning 10:00 a.m.–12:00 p.m.; midday 12:00 p.m.–2:00 p.m., early afternoon 2 p.m.–4 p.m.; and late
afternoon 4 p.m.–7 p.m.”, park program was recorded and documented at each observation spot [14].
The type of renewable energy used in the design was vertical micro wind turbine, eight of which
were integrated into a recycled steel construction in Figure 2 and erected at the park center as a
renewable energy sculpture where the former tank 101 was standing.
Since wind was the primary renewable source in the project, wind intensity was measured using
an anemometer from seven observations zones to understand if turbine locations associated with the
site wind potentials [14].
To comprehend the social aspect of the model, thirty-four users at various locations were
approached randomly during the site observation. A small survey was conducted asking two questions
addressing if they reside near the vicinity and noticed the park’s capacity for renewable electricity
production. The outcome of the survey informed a better understanding of park’s local or regional
user group against their awareness about the park’s renewable energy aspect [14].
Micro wind turbines technical specification notes and one interviewee’s energy calculations
were used to inform the turbines’ life cycle and energy payback time which helped to recognize
environmental strategies of the model.
122
Energies 2020, 13, 3428
Figure 2. Renewable energy sculpture incorporated eight micro wind turbines was made from the
former tank 101 recycled steel structures. Image by the author.
3. Results
Copious amount of sustainability assessment tools are used for different human environments,
ones specific to landscape architecture profession are inadequate. While the effectiveness and
leadership of assessment tools are key to energy future, current tools including the ‘SITE’, disregards a
comprehensive sustainability assessment of public space on-site renewable electricity.
Next section presents theories and principals from ecology substantiating the strategies of potential
assessment model, its components—and finally, its performance and outcomes after tested against the
case study.
3.1. Theories behind the Potential Assessment Model: The Link between Energy, Information and Public Space
To provide food, water and shelter and to improve the quality of our life, energy is required.
Energy comes in many forms, and the science of thermodynamics assesses energy transfers and
transformations in processes, systems and devices. Two laws in principle rules the nature of energy.
According to the first law of thermodynamics, energy cannot be extinguished or created and only its
state can be altered and kept. The second law of thermodynamics can be defined within these states.
Respectively, when the state of useful energy, the work capacity, a.k.a., ‘exergy’ dissipates, the state of
disorder, a.k.a., ‘entropy’ arises [15].
Howard T. Odum, a renowned 20th century ecologist, introduced the ecosystem concepts of energy
hierarchy and energy flow to the human environment. His early ideas on energy and nature stem from
a provisional hypothesis—‘tripartite altruism’ which was later developed to ‘emergy’ concept, a.k.a.,
‘the quality of energy’ [16]. Emergy refers to the history of the system and the time and processes that
have contributed to its current state [17]. The letters ‘em’ of emergy represents ‘energy memory’ and
relate directly to the now more common term ‘embodied energy’.
Odum’s energy-nature equation so-called ‘tripartite altruism’ postulates that organisms in nature
uses their energy based on 1/3 ratio comprising higher order, lower order and, maintenance and storage
purposes. He presented this theory using rabbits as an example illustrated in Figure 3. According to
this theory, “They eat grass to live, grow and reproduce. Their manure fertilizes the grass that feed[s]
them and they ‘sacrifice’ weak rabbits to predators to keep the population fit and in balance” [18].
123
Energies 2020, 13, 3428
Such formula can be useful when the public space is purposed to be designed toward an ecological
sophistication. Particularly, Tripartite altruism’s higher order, lower order and storage and maintenance
strategy is simply about distributing the available energy on the system equally between three types
of needs.
This study claims that active and passive interaction with renewable energy is an imperative
feature of public spaces to establish a give-and-take connection between society and energy and to
reach an evocative and assessable mode of sustainability. Similarly, the maximum empower principle
also known as the Odum’s fourth law states that during the self-organization process, systems advance
interactions, processes and parts that amplify efficiency and production [19–21]. Odum’s fifth law
further asserts that “system processes maximize power by interacting abundant energy forms with
ones of small quantity, but a larger amplification ability” [19]. Thus, design processes and outcomes
targeting an ecological sophistication in public space can be used as a tool to educate the public about
a sustainable energy lifestyle and to increase general environmental awareness, so maximizes energy
efficiency and production in the broader community, leading to long-term benefits.
Focusing on the social aspect of public space where people mingle through communication,
interaction and learning from their surroundings and from each other, Miller [22] argues “[p]ublic
spaces do not exist as static physical entities, but are constellations of ideas, actions and environments”.
Therefore, public space can shift society into a sustainable energy lifestyle when used as an educational
and information platform. When conceptualized as a self-organizing state, public space can become, in
Odum’s terms, “a platform to create useful information for community”, in doing so, endorse a larger
commitment to sustainable energy across multiple social realms.
• Economic strategy refers to using one third or meaningful part of the on-site produced electricity
to be sold to city grid for generating income to the public space and its local community;
• Environmental strategy suggests whether the design considers embodied energy, energy storage
and maintenance of the renewable energy infrastructure and/or public space. That is, one third or
meaningful part of the produced total electricity throughout the life time of renewable energy
infrastructure needs to pay back its maintenance cost and embodied energy considering its
depreciation value;
124
Energies 2020, 13, 3428
• The model’s final one third or meaningful part of the produced electricity focuses on creating
social strategies using on-site renewable energy. That is whether the design considers any active
and/or passive engagement using public space renewable energy.
More precisely, the economic strategy is quantified by using a scale from zero to four points.
Since each case study site would have different energy distribution needs, appropriateness of the
distribution strategy would score the highest point. The project scores zero point if no generated
electricity within public space is used for local energy trading. The project scores one point if designers
are aware of and have the intention to implement the model’s economic strategy however project
fails to do so for several reasons as some of them are exemplified and presented in this case study.
If the strategy is implemented, but unmeasured and unmonitored the project scores one and a half
points. A project scores two points if total production capacity is used for local energy trading [10].
A project scores three points if environmental or social strategy is considered in addition to economic
engagement. A project scores three and a half points both environmental and social strategies are
considered in addition to the economic strategy. A project scores four points if full generated electricity
amount is equally and/or meaningfully distributed between three assessment strategies.
To comply with the environmental strategy, the project shall consider embodied energy,
energy storage and contributes to the maintenance of the infrastructure and/or the public space.
A project scores zero point, either designers are not aware or consider any above factors. A project
scores one point if an intention for addressing all three factors of environmental strategy is clear
however project fails to achieve this. If an intention for addressing two of the three factors is obvious,
the score would be 2/3 point. If an intention for addressing at least one of the three factors is apparent,
the score would be 1/3 point. A project scores two points, if successfully and fully addresses one of the
three factors. A project scores three points, if any two factors are fully addressed. A project scores four
points, if all three factors are considered simultaneously.
The model’s social strategy assesses whether public space renewable energy generates both active
and passive engagement with public. Educational, informative and place making activities can involve
active and/or passive engagement with on-site renewable energy. Here active and passive engagement
is twofold.
The first one is about using on-site renewable electricity through either production or consumption
activity within public space. An example of active renewable electricity production is Piezoelectric
generator. Active renewable electricity consumption activities comprise mobile phone charge, car battery
charge, using electricity for street performance while passive renewable energy consumption may
include digital information screens where users consume electricity in an indirect manner.
The second one refers to ‘energy quality’ concept introduced earlier in the theory section which is
literally about ‘quality design’ or quality of the renewable energy infrastructure. More specifically,
how the design communicates with public, what diverse uses intentionally and unintentionally
present, observable or hidden. Active engagement with renewable energy may include a meta
sustainability narrative the design is trying to communicate regardless of functioning as a renewable
electricity generator. Interpretative signage of renewable energy may be incorporated into the design
to communicate with public at a symbolic level. Whereas, passive engagement involves spontaneous
public space activities, for example the design is used as a backdrop for market stall installation,
photography and performances. Or the design triggers questions in public about its function,
evokes several other meanings and promote multifunctionality.
Respectively, a project scores zero point, if renewable energy design does not generate any
engagement with public whatsoever. A project scores one point if any intention creating either active
or passive interaction is apparent, but the project is failed. A project scores two points if any intention
creating both active and passive interaction is obvious, but the project is failed. A project scores three
points, if either active or passive engagement is considered. A project scores four points, if both active
and passive engagements using on-site renewable energy are considered. Points in each segment
125
Energies 2020, 13, 3428
would be divided into two–one covers using on-site renewable electricity and other one concerns with
parameters related to design quality of the renewable energy infrastructure.
Table 1 in the following section presents the assessment scoring in detail while incorporating
information from this prescribed text.
Table 1. Ballast Point Park renewable energy infrastructure sustainability assessment scores.
3.3.1. Ballast Point Park Renewable Energy Infrastructure Economic Sustainability Assessment
• Renewable energy’s economic benefit was an incentive to include the concept in the park; however,
the integrated wind turbines did not function and did not generate any electricity [14,23–25].
126
Energies 2020, 13, 3428
While BPP was a well-regarded project that fulfilled many sustainability aspects, specific to local
economy of the park, little information was found. Minor activities in the park such as wedding
ceremonies support park’s local economy and a recent study has revealed a value increase over fifty
percent on surrounding properties since the park opened to public in 2009 [14,26].
Economic benefit from on-site generated electricity was initially planned to balance out the operational
cost of the park, however micro wind turbines installed did not operate as anticipated [14,23–25].
Annual wind data were not incorporated into the process of decision making since no calculations
were done by renewable energy consultant. Wind turbines’ technical specification notes were used as
the only primary source to predict possible yield. However, to anticipate the wind turbines’ possible
location, local wind measurements were needed which were never realized [23]. Following upon this
gap, also to better understand reasons behind the dysfunctional wind turbines, wind measurements
were taken by using an anemometer at seven different observation locations during site trips as
presented in Figure 4a.
(a) (b)
Figure 4. (a) BPP location map showing observation spots; (b) graph showing the daily averaged wind
values for each spot.
During regular visits for two weeks in January 2014, no spinning movement was monitored on
any single wind turbine despite seeing the park at some extremely windy days.
Figure 4b presents a graph showing daily averaged wind values for each spot, respectively while
Table 2 demonstrates each zone’s average wind values during one week period. Each value in Table 2
represents daily average wind speed calculated from 7:30 a.m. to 7 p.m. by moving every twenty
minutes from one observation location to another. Black thick line in Figure 4b represents the turbines’
average wind speed at their exact location where wind values repeatedly stayed below the other
measured average wind speeds. While local wind measurements are crucial to determine the turbines’
exact location, the economy of generated electricity requires yearly wind data since every wind turbines
including very successful ones show strong seasonality in its output. Therefore, data presented in
Table 2 is indicative for a general understanding of site wind potentials and so possible ideal locations
of the turbines.
According to the model’s economic sustainability assessment, BPP scored one from four points.
In this study, local electricity production’s economic value was an essential strategy for sustainable
renewable energy-embedded public spaces and BPP has failed to do so. BPP has accomplished critical
success in its own right and has deservedly won many awards for addressing numerous environmental
sustainability criteria, but in the context and for the purpose of this study,—when assessed against
the more thorough renewable energy economy potential which the proposed model shows here—it
fell short.
More specifically, the model indicated an equation using 1/3 or meaningful amount of the produced
electricity for local energy trading which could bring income to the community. However, at the time,
the technology was new and underdeveloped. Generated electricity was not efficient for providing
127
Energies 2020, 13, 3428
enough economic incentive [23–25]. Such shortcoming was also discussed later as one limitation of the
model. Presumably, past technologies would not able to address this strategy fully, therefore for the
time being, designer intentions are invaluable to understand whether this strategy originally included
in the project’s agenda.
Date O—1 O—2 O—3 O—4 O—5 O—6 O—Wind Turbines Average of Average
Day 1 19.7 7.1 5.6 8.5 9.6 9.8 7.1 10.9
Day 2 6 7.6 6.9 4.8 7.4 6.5 6.4 7.1
Day 3 19.3 8.4 10.8 9.7 12.7 11.2 11 8.2
Day 4 7.4 8 8.6 4.9 3 7.1 5.2 6.6
Day 5 3.9 4.9 3.5 3.4 3.9 3.8 2.1 7.0
Day 6 9.5 6.7 7.8 6.7 6.3 6.8 2.1 7.9
Day 7 10.4 6.7 14.1 8.1 5.9 10.4 4.3 5.5
Average 10.9 7.1 8.2 6.6 7.0 7.9 5.5 7.6
O—N: observation location.
3.3.2. Ballast Point Park Renewable Energy Infrastructure Environmental Sustainability Assessment
According to the model’s environmental assessment, BPP shall consider embodied energy,
energy storage and the maintenance of the infrastructure and/or park successfully.
• The park’s ‘cradle-to-cradle’ (CTC) design philosophy favorites embodied energy consideration
of the renewable energy infrastructure.
• Wind turbines and batteries are part of the renewable energy infrastructure, however none of
them functioned and delivered the anticipated energy yield.
• Public space maintenance using on-site renewable electricity was discussed by interviewees
however never become a reality.
What was interesting about embodied energy aspect of the BPP renewable energy infrastructure
was the 101 oil tank and its recycled steel frames repurposed as a renewable energy sculpture on
the basis of park’s post-industrial history. However, no life cycle analysis done; designers applied
the ‘cradle-to-cradle’ (CTC) philosophy to the park’s overall design at multiple scales indicating a
focus on environmental sustainability for design decisions. According to the model, embodied energy
prevails important implications for renewable energy devices. Embodied energy of a renewable energy
infrastructure refers to ‘energy pay back times’ (EPT); in other words, the energy and revenue spent on
a renewable energy device until constructed and functioning [27–29].
Commenting on EPT, one interviewee presented his preliminary calculations referring to
twenty-nine years pay-back time when eight micro wind turbines spin at 10 m/s (36 km/h–Force 5
fresh breeze). The wind turbines catalog used in the project specified each turbine generating 800 W at
14 m/s [23]. Considering the latest developments in renewable energy technology, EPT has significantly
dropped nowadays in comparison to eight years ago. Thus, despite the limitations of past technological
shortcomings, designers were aware of the topic and considered embodied energy. In the case of BPP,
the CTC design philosophy was applied to the design of energy sculpture. When assessed by the
proposed model, the CTC approach taken by designers addressed embodied energy, which scored two
from four points and partly fulfilled the model’s environmental strategy. Designers used pre-existing
structures and materials from the site during the construction phase, thus reduced the art piece’s
embodied energy cost.
Concerning wind turbines and energy storage, data from site observation and interviews revealed
that neither batteries nor turbines functioned. Turbine moving segments have rusted over time because
the sprayed sea salt prevented turbines spinning [25]. As originally calculated by the distributer,
wind turbines’ power would have sufficed park’s energy need. However, to acquire full performance
128
Energies 2020, 13, 3428
from the installed turbines, several criteria needed to be addressed. For example, determining the most
productive wind harvesting location was a crucial strategy to be fulfilled which was not implemented.
Therefore, the project scored 1/3 point partially addressing this strategy.
What stood out from an environmental assessment of the relation between maintenance and
renewable energy infrastructure, designers intended to use generated electricity from the renewables for
reducing park maintenance costs. However, because any technical and expert knowledge was lacking,
the renewable energy devices did not function as intended, and did not meet the goal of supplying
electricity to the park for daily use [23–25]. Thus, the park did not generate electricity for itself or
the community, and thereby neglected both economic and partially environmental (self-maintenance)
aspect of on-site electricity production. From this finding, the project achieved a score of 1/3 point.
The environmental sum total was then calculated as 2.7 from 4 points.
3.3.3. Ballast Point Park Renewable Energy Infrastructure Social Sustainability Assessment
• No active and/or passive engagement with on-site renewable electricity was recorded on the site.
This has also never become the designers’ intention [23,25].
• Active engagement was generated through ‘quality design’; powerful symbolic meaning from a
former fossil fuel tank to a renewable energy sculpture; interpretive signage.
• Passive engagement was generated with renewable infrastructure through spontaneous use.
• Designer’s esthetical concerns, community will and site wind exposure influenced the decision
for using wind turbines in the design.
Throughout regular site visits over two weeks in January 2014, no spinning movement was
monitored on any single wind turbine, owing to this, no active or passive engagement with on-site
renewable energy was recorded. Moreover, designers did not intend creating activities in public
space using on-site generated electricity [23,25]. However, transforming a former fossil fuel tank to
a renewable energy sculpture conveyed a powerful symbolic meaning which was the main design
intention. The principal designer emphasized his view on the societal impact of the designed renewable
energy sculpture with his own words “the big fossil fuel tank turned into the biggest wind turbine
thing on Sydney Harbor. There is poetry there” [14,25]. Such narrative was also reflected in using
interpretive signages across the park as seen in Figure 5. In doing so, the project scored one point
through addressing the ‘quality design’ strategy–active engagement side of renewable energy.
(a) (b)
Figure 5. (a) The interpretive signage of renewable energy sculpture; (b) The renewable energy
sculpture. Images by the author.
129
Energies 2020, 13, 3428
(a) (b)
Figure 6. (a) Renewable energy sculpture used as a background for wedding photography; (b) geo-tags
were pinned to the renewable energy sculpture. Images by the author.
During site visits, the survey presented in Figure 7 was completed with thirty-four participants.
Only 24 percent of the park users were aware of the park’s capacity to produce electricity, and these
park users were living in the vicinity.
Figure 7. Survey results revealing user’s response to the park’s renewable energy.
This finding was further elaborated with interview data for example using on-site produced
electricity to encourage public space activities was never brought up in any community consultation
process. However, during these meetings some initial ideas were seeded for example using the wind
as a choice for renewable source and retaining the former oil tank on the site. As discussed with the
principal designer, the idea of generating on-site electricity originated from the community and its
stories about flying kites on the site because of strong wind exposure. The idea of keeping the former
tank was originally suggested by community to be used as a landmark for captain’s sense of direction
while roaming in the harbor. These ideas together with designers’ esthetical concerns further formed
the final design which was developed in a design charette and the former tank turned into a renewable
energy sculpture [25].
130
Energies 2020, 13, 3428
According to the assessment model, simply relying on the renewable energy’s environmental
and economic aspect was not sufficient to create genuinely sustainable renewable energy embedded
public space. An important factor in accomplishing social outcomes of renewable energy use was its
public acceptance [30,31]. For example, photovoltaic (PV) panels, despite their controversial esthetic
qualities, needed to be visible to users to increase their social acceptance [32]. These esthetic concerns
were reiterated in the case study that designers were hesitant to integrate PV panels into the BPP
design [23]. Moreover, because most park users did not know the park had the capacity to produce
electricity, the socially sustainable knowledge and acceptance of renewable energy was limited [14].
Therefore, some issues emerging from these findings were related specifically to the designers’ priority
of focusing on the esthetic and technological aspect of renewable energy, rather than on the potential
socioeconomic relations stemming from its local production.
From a social sustainability viewpoint, BPP’s renewable energy scored two points through active
and passive engagement. Several aspects of which referred to ‘quality design’—by definition design
quality of the energy sculpture and its well-thought narrative, but not from on-site produced electricity
as clearly presented in this section.
4. Discussion
Current sustainability assessment tools specific to the Landscape Architecture profession devalue
public space renewable energy use by neglecting the potential social, economic and environmental
engagements of local electricity production that could otherwise increase the sustainability of the
public space. One common issue that emerges from the BPP case study findings as well as the reviewed
literature is that although a broad range of sustainability assessment tools have been used for different
human environments, ones specific to the Landscape Architecture profession are limited.
Therefore, for the purpose of this study, the SITES initiative is selected as a typical example
which is a legitimate well-regarded sustainability assessment tool in the profession. As discussed,
the sustainable SITES initiative assesses the sustainability of public spaces with a point-based system [7].
However, SITES’ assessment approach to renewable energy-embedded public open spaces is one-sided,
often misses the social and economic relations around electricity production and, thus, only ticking
the environmental sustainability box. This may be because SITES’ focus is only on the renewable
energy devices, rather than on the activity of electricity production. The SITES’ limitation to assess the
renewable energy side of a public space led the development of a potential future assessment tool as
presented through the BPP case in Sydney.
Having a device-based focus on renewable energy is symptomatic of a top-down sustainability
development approach. A possible explanation for this may be ingrained in the definition of
‘techno-fix’, which holds that present market conditions generally satisfy demand with quick solutions
to increase consumerism [33]. Respectively, urban renewable energy is widely framed as a technological
addendum, an afterthought, by definition a retrofit to public space designs. Currently renewable
energy infrastructures embedded in urban environments have a relatively small electricity production
capacity. In addition, energy storage utilities are still expensive, and distributed energy infrastructure
systems in cities around the world are not widespread. For the time being, existing industry practices
may be reluctant to make use of producing clean electricity in cities at maximum potential. However,
such parameters indicate only present day constraints and it is foreseeable that these constraints will
be overcome in the very near future. Therefore, a balanced quantity between the proposed strategies
may gain more credit if not now, but later.
The presented model promotes social, economic and environmental engagement using on-site
renewable energy. For this reason, it requires negotiation among many stakeholders, including policy
makers, local government, neighborhood residents, public space users, designers and experts involved
in the project. It makes the design process more transparent and gives each stakeholder equal
opportunity for management and/or involvement. It is likely, therefore, that designers can be more
cognizant of the sustainability of renewable energy. This in turn, can reduce the propensity for
131
Energies 2020, 13, 3428
superficial ‘green wash’ outcomes and increase the importance of social and economic factors involved
in electricity production.
The model can be scaled for household and neighborhood uses. For example, researchers can
identify how much clean electricity a household generates; sells to the city grid; uses for its own needs
including maintenance, storage, ordering, installation, transportation, as well as its own embodied
energy. The electricity’s specific consumption also becomes the focus of some recent studies that
investigates integrated solutions for households’ peak energy demand [34]. The proposed model
can guide a multidisciplinary research team in specifying the social, economic and environmental
engagement strategies according to a household’s energy needs. Simply put, a balance is needed
between these three aspects. The model can also be used for long-term sustainable development to
evaluate the efficacy of renewable energy investments for impoverished communities. It may also
benefit energy independence studies for energy poverty alleviation in remote settlements [35].
The model’s economic engagement strategy not only fosters a surplus to community but can
indicate a green rating for the neighborhood. For example, some emerging distributed energy
neighborhoods have energy information centers. Their main roles are marketing and advertising the
engagement with renewable energy and promote sustainable energy education. These neighborhoods
are widespread in some European countries such as Austria, Germany, Denmark, Sweden and the
Netherlands [36–40]. Public spaces designs in line with the thinking promoted through the proposed
model here can showcase its neighborhood electricity use as a public space amenity. A distributed
energy market economy is reliant on smart grid systems, which require active interaction with
its stakeholders [41]. Therefore, each public space design can support such interaction addressing
site-specific social, environmental and economic data while integrating renewable energy technology
to support a local energy economy, monitoring electricity production and consumption, and informing
residents about their energy use, with a view to helping them to be more energy responsive.
Urban administrative policies can then use such data to channel subsidies for energy responsive
neighborhoods and encourage the transition to sustainable energy.
Another local energy trading application is so-called transactive energy which uses the existing
infrastructure for renewable energy distribution. A real world example was tested through Monash
Micro grid case study in Melbourne, Australia where a complete hardware and software infrastructure
are staged for micro-grid energy management [10]. Furthermore, some alternative administrative
tools are evolving to provide on-site energy services. For example, the sustainable energy utility
(SEU) offers on-site energy services to empower private energy markets and facilitate self-finance
for communities [42,43]. However, the presented model sheds new light on the community energy
distribution using public space as a point of departure, it requires side information, policies and
administrative innovation similar to the SEU for better integration within human environments.
The proposed model’s environmental assessment strategy formed a significant connection between
embodied energy, energy storage and maintenance of the infrastructure and/or park. The BPP designers
seemed to be more aware of such connection, because environmental, social and economic strategies
using renewable energy scored highest to lowest, respectively. In the previous study of the author,
twenty-five speculative designs submitted to the LAGI design competition were assessed against
very similar strategies whereas the findings revealed an entirely different outcome—the shortlisted
projects‘ renewable energy based economic, environmental, and social strategies scored highest to
lowest average, respectively [8]. This outcome perhaps suggests environmental interventions in
general prevail dominant trend for the mainstream urban sustainability practices which is clearly
the case as presented in this study. However, LAGI’s philosophy on renewable energy and public
space is highly responsive, their approach to the sustainability esthetics, as noted by Elizabeth Meyer,
needs to be revisited. Meyer states that the esthetics of sustainable design need to be considered
within two contexts—‘appearance’ and ‘function’—to have a significant cultural and societal impact.
She argues that ‘function’, also referred to as ‘fitness’ or ‘performance’, cannot be experienced through
representation, but through direct engagement [44].
132
Energies 2020, 13, 3428
5. Conclusions
This study investigated the link between renewable energy and public space and developed a
potential integrated model to assess the sustainability of public space renewable energy infrastructure.
The model was tested through assessing a renowned public space awarded for its environmental
sustainability innovations. Since incorporating renewable energy use in urban public space is an
emerging field, this study is built upon author’s previous work on assessing sustainability of speculative
designs in which findings revealed an imbalance for distributing onsite generated electricity from a
public space context.
The devised model, strategies and related ecological theories are the key contribution of this study.
Sustainability practitioners have not developed a full understanding of public space renewable energy
use beyond it being a mere design retrofit. The presented strategies in Section 3.2 would be useful
for designers to better incorporate renewable energy into their designs. Such strategies would also
benefit public space assessment bodies such as the SITES initiative. Importantly, the link between three
strategies in this study was argued to become indispensable for an ecologically sophisticated public
space design.
Another crucial contribution is the acknowledgement of the importance of using generated
electricity immediately within public spaces through active and passive interactions. An ecologically
sophisticated public space design should be energy responsive in the sense that it increases the social,
economic and environmental sustainability of its locale. At the same time, such a public space may
generate useful information for the community, which can then be used as a basis for sustainable
energy transition and social change. This claim stems from the “maximum power principle”, which is
considered as Odum’s fourth law of thermodynamics. Odum’s fifth law also posits that active and
133
Energies 2020, 13, 3428
passive interactions are imperative for on-site renewable energy, and, respectively, the more these
interactions occur in public spaces, the greater the chances the public space would help society
transitioning to a sustainable energy life style.
Finally, the study presented the need for an integrated approach to renewable energy and public
space, to achieve not only meaningful and measurable sustainability, but also to communicate the
reciprocal relations between society and energy.
5.1. Limitations
The model’s major limitation is the devised strategies and their appropriateness with the renewable
energy project. While these strategies explicitly engage with renewable electricity at multi-level;
site-specific features, designer’s esthetical concerns, decisions on renewable energy technology types
and many other unforeseen factors may influence full exploitation of renewable energy as presented
through the BPP. From an overall sustainability perspective, BPP is successful on its own right.
However, when the renewable aspect of the project is scrutinized, a different outcome becomes evident.
Another limitation of the study is time factor. Because the project examined was built in 2009,
at the time renewable energy technologies were less efficient and economically infeasible considering
its highly expensive manufacturing cost against lower production capacity. Therefore, in the past a
design fully responsive to the proposed model’ strategies in an urban condition was not practical and
was perhaps impossible. This is why the project used the BPP as a test bed, merely to demonstrate the
model’s function rather than focusing on the project flaws.
134
Energies 2020, 13, 3428
T. Odum’s early theories. To generate local energy economies and create smart learning spaces in
addition to the social and ecological function of public space, this study conceptualized public space
as an innovation and transformation hub for social and environmental change while considering
advancements in renewable energy technologies and their increasing production rates. A transition
from fossil fuel to renewable energy is a significant start on the road to future energy autonomy but
requires systemic bottom-up and top-down policies and interventions to create a truly sustainable
society. The outcome of this study is one such intervention that can inform and advance this critical
transition process.
Funding: This research did receive a fee of waive from Energies Journal, many ideas presented here was seeded
in my PhD which was funded by Australian Postgraduate Award (APA).
Acknowledgments: I would like to thank Ian Weir, Debra Cushing and Laurie Buys for their support during my
PhD process. I would like to express my deep gratitude to the interviewees and Mcgregor + Coxall team for
their time and support. I would also like to thank three reviewers for their constructive feedback helping me to
improve this manuscript.
Conflicts of Interest: The author declares no conflict of interest.
References
1. Droege, P. 100 Per Cent Renewable: Energy Autonomy in Action; Earthscan: London, UK, 2009.
2. Moldan, B.; Janoušková, S.; Hák, T. How to understand and measure environmental sustainability:
Indicators and targets. Ecol. Indic. 2012, 17, 4–13. [CrossRef]
3. Byrne, J.; Martinez, C.; Ruggero, C. Relocating Energy in the Social Commons: Ideas for a Sustainable Energy
Utility. Bull. Sci. Technol. Soc. 2009, 29, 81–94. [CrossRef]
4. Dincer, I. Renewable energy and sustainable development: A crucial review. Renew. Sustain. Energy Rev.
2000, 4, 157–175. [CrossRef]
5. Cairns, J. Sustainability and specialisation. Ethics Sci. Environ. Politics ESEP 2004, 4, 33–38. [CrossRef]
6. SITES, T.S.S.I. Guidelines and Performance Benchmarks; The American Society of Landscape Architects:
Washington, DC, USA; The United States Botanic Garden: Washington, DC, USA; The Lady Bird Johnson
Wildflower Center at the University of Texas: Austin, TX, USA, 2009.
7. SITES, T.S.S.I. SITES v2 Rating System For Sustainable Land Design and Development; The American Society of
Landscape Architects: Washington, DC, USA; The United States Botanic Garden: Washington, DC, USA;
The Lady Bird Johnson Wildflower Center at the University of Texas: Austin, TX, USA, 2014.
8. Ozgun, K.; Weir, I.; Cushing, D. Optimal Electricity Distribution Framework for Public Space:
Assessing Renewable Energy Proposals for Freshkills Park, New York City. Sustainability 2015, 7, 3753–3773.
[CrossRef]
9. Monoian, E.; Ferry, R. LAGI 2012 Design Guidelines. In Regenerative Infrastructures: Freshkills Park, NYC:
Land Art Generator Initiative; Klein, C., Ed.; Prestel: Munich, Germany; London, UK; New York, NY, USA,
2013; p. 30.
10. Khorasany, M.; Azuatalam, D.; Glasgow, R.; Liebman, A.; Razzaghi, R. Transactive Energy Market for Energy
Management in Microgrids: The Monash Microgrid Case Study. Energies 2020, 13, 2010. [CrossRef]
11. Silverman, D. Doing Qualitative Research, 4th ed.; SAGE: London, UK; Thousand Oaks, CA, USA, 2013; p. 470.
12. AILA. Ballast Point Park, Birchgrove, Sydney. Available online: http://www.aila.org.au/projects/nsw/ballast-
point/overview.htm (accessed on 1 December 2013).
13. O’Neill, R.; Consultant, Sydney, Australia; Ozgun, K.; Researcher, Sydney, Australia. Personal
Communication, 2014.
14. Ozgun, K.; Cushing, D.; Buys, L. Renewable energy distribution in public spaces: Analyzing the case of
Ballast Point Park in Sydney, using a triple bottom line approach. Jola J. Landsc. Archit. 2015, 10, 18–31.
[CrossRef]
15. Rosen, M.A.; Dincer, I. EXERGY: Energy, Environment and Sustainable Development; Elsevier: Amsterdam,
The Netherlands, 2007.
16. Odum, H.T. Self-Organization, Transformity, and Information. Science 1988, 242, 1132. [CrossRef]
135
Energies 2020, 13, 3428
17. Bastianoni, S.; Marchettini, N. Emergy/exergy ratio as a measure of the level of organization of systems.
Ecol. Model. 1997, 99, 33–40. [CrossRef]
18. Holmgren, D. Permaculture: Principles & Pathways beyond Sustainability; Holmgren Design Services: Hepburn,
Australia, 2002; p. 286.
19. Tilley, D.R.; Howard, T. Odum’s contribution to the laws of energy. Ecol. Model. 2004, 178, 121–125. [CrossRef]
20. Odum, H.T. Environmental Accounting; Wiley: Hoboken, NJ, USA, 1996.
21. Odum, H.T.; Odum, E.C. A Prosperous Way Down: Principles and Policies; Univ Pr of Colorado: Boulder, CO,
USA, 2008.
22. Miller, K.F. Designs on the Public: The Private Lives of New York’s Public Spaces; University of Minnesota Press:
Minneapolis, MN, USA, 2007.
23. McDermott, L.; Landscape Architect, Brisbane, Australia; Ozgun, K.; Researcher, Brisbane, Australia.
Personal Communication, 2014.
24. Kennedy, T.; Project Manager, Sydney, Australia; Ozgun, K.; Researcher, Sydney, Australia. Personal
Communication, 2014.
25. Coxall, P.; Principal Landscape Architect, Sydney, Australia; Ozgun, K.; Researcher, Sydney, Australia.
Personal Communication, 2014.
26. Kilbane, S.; Toland, A.; Pham, K. Ballast Point Park; Landscape Architecture Foundation: Washington, DC,
USA, 2017. [CrossRef]
27. Alsema, E.A.; Fthenakis, V.M. Photovoltaics Energy Payback Times, Greenhouse Gas Emissions and External
Costs: 2004–early 2005 Status. Prog. Photovolt. 2006, 14, 275–280. [CrossRef]
28. Hammond, G.P. Energy and sustainability in a complex world: Reflections on the ideas of Howard, T. Odum.
Int. J. Energy Res. 2007, 31, 1105–1130. [CrossRef]
29. Roberts, F. Energy accounting of alternative energy sources. Appl. Energy 1980, 6, 1–20. [CrossRef]
30. Assefa, G.; Frostell, B. Social sustainability and social acceptance in technology assessment: A case study of
energy technologies. Technol. Soc. 2007, 29, 63–78. [CrossRef]
31. Rogers, J.; Convery, I.; Simmons, E.; Weatherall, A. Local renewables for Local Places? Attitudes to Renewable
Energy and the Role of Communities in Place-based Renewable Energy Development. In Making Sense of
Place: Multidisciplinary Perspectives; Convery, I., Corsane, G., Davis, P., Eds.; Boydell & Brewer: Woodbridge,
UK, 2012.
32. Thayer, R.L. Gray World, Green Heart (1994). In Theory in Landscape Architecture: A Reader; Swaffield, S.R., Ed.;
University of Pennsylvania Press: Philadelphia, PA, USA, 2002.
33. Huesemann, M.; Huesemann, J. Techno-Fix: Why Technology Won’t Save Us Or the Environment; New Society
Publishers: New York, NY, USA, 2011.
34. Buys, L.; Vine, D.; Ledwich, G.; Bell, J.; Mengersen, K.; Morris, P.; Lewis, J. A Framework for Understanding
and Generating Integrated Solutions for Residential Peak Energy Demand. PLoS ONE 2015, 10, e0121195.
[CrossRef]
35. González-Eguino, M. Energy poverty: An overview. Renew. Sustain. Energy Rev. 2015, 47, 377–385. [CrossRef]
36. Dóci, G.; Vasileiadou, E.; Petersen, A.C. Exploring the transition potential of renewable energy communities.
Futures 2015, 66, 85–95. [CrossRef]
37. Hauber, J.; Ruppert-Winkel, C. Moving towards Energy Self-Sufficiency Based on Renewables: Comparative
Case Studies on the Emergence of Regional Processes of Socio-Technical Change in Germany. Sustainability
2012, 4, 491–530. [CrossRef]
38. Scheer, H. Energy Autonomy: The Economic, Social and Technological Case for Renewable Energy; EARTHSCAN:
London, UK; Sterling, VA, USA, 2007.
39. Van Timmeren, A.; Zwetsloot, J.; Brezet, H.; Silvester, S. Sustainable Urban Regeneration Based on Energy
Balance. Sustainability 2012, 4, 1488–1509. [CrossRef]
40. Wächter, P.; Ornetzeder, M.; Rohracher, H.; Schreuer, A.; Knoflacher, M. Towards a Sustainable Spatial
Organization of the Energy System: Backcasting Experiences from Austria. Sustainability 2012, 4, 193–209.
[CrossRef]
41. Ilic, D.; Da Silva, P.G.; Karnouskos, S.; Griesemer, M. An energy market for trading electricity in smart grid
neighbourhoods. In Proceedings of the Digital Ecosystems Technologies (DEST), 2012 6th IEEE International
Conference on, Campione d’Italia, Italy, 18–20 June 2012; pp. 1–6.
136
Energies 2020, 13, 3428
42. Byrne, J.; Taminiau, J. A review of sustainable energy utility and energy service utility concepts and
applications: Realizing ecological and social sustainability with a community utility. Wiley Interdiscip. Rev.
Energy Environ. 2015. [CrossRef]
43. Houck, J.; Rickerson, W. The sustainable energy utility (SEU) model for energy service delivery. Bull. Sci.
Technol. Soc. 2009, 29, 95–107. [CrossRef]
44. Meyer, E.K. Sustaining beauty. The performance of appearance: A manifesto in three parts. J. Landsc. Archit.
2008, 3, 6–23. [CrossRef]
© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
137
energies
Article
Adaptable Source-Grid Planning for High Penetration
of Renewable Energy Integrated System
Ming Tang 1 , Jian Wang 2, * and Xiaohua Wang 2
1 College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China; [email protected]
2 Tsinghua Sichuan Energy Internet Research Institute, Chengdu 610213, China;
[email protected]
* Correspondence: [email protected]
Abstract: To adapt to the growing scale of renewable energy and improve the consume ability of
the power system, it is necessary to design a highly adaptable planning scheme for high penetration
of the renewable energy integrated system. Thus, this paper firstly gives the conception of system
adaptability and designs an adaptability index system, which considers the supply and demand
balance, operation state, and network structure of the high penetrated renewable energy integrated
system. It can help to comprehensively evaluate the system ability towards uncertain shocks. Then, a
two-stage source-grid coordinative expansion planning model is presented. The adaptability indexes
of supply and demand balance are used as objection of the source planning stage, the adaptability
indexed of the operation state and network structure are used to guide the grid planning stage.
The model is further solved based on the coordination between the source and grid planning stage.
Finally, the case study verifies that the obtained optimal plan has good adaptability to the impact of
renewable energy on the power supply capacity and security operation.
1. Introduction
1.1. Motivation
The pursuit of a sustainable low-carbon society has made renewable energy, represented by wind
power and photovoltaics, popularized and applied on a large scale. With the rapid development of
related technologies and the breakthrough of key issues, the penetration of renewable energy increases
each year [1]. The newly installed capacity of renewable energy units in the world reached 181 GW
by 2018, in which wind and photovoltaic power accounted for 83%, while the renewable energy
unit supplied more than 26% of load demand [2]. The obvious influence of natural resources and
seasonal climate cause the renewable energy unit to have strong volatility and randomness [3]. Variable
renewable energy sources, such as solar photovoltaic and wind power, not only change the patterns
of electricity utilization on the demand side, but also bring challenges for the balancing and security
of power supply on the source side [4]. Considering the scale effect, the uncertain characteristics of
multi-spatiotemporal differential distribution of renewable energy are coupled in the operation features
of the power system. That seriously impacts the safety and stability of the power system and brings
great challenges to the planning and operation. The limitation of network structure and shortage of
flexibility resources will further restrict the adaptability of the power system for renewable energy
integration. Thus, a source-grid coordinative planning scheme adaptable for the high penetration
system is crucial to improve the system adaptability to renewable energy [5].
140
Energies 2020, 13, 3304
proposed, which combines the source planning and the gird planning. While preserving the power
balance and regulation capabilities, the multiple attributes of the operation state and network structure
are optimized. The matching and adaptability between the source planning scheme and grid planning
scheme are enhanced, improving consumption ability of the system to high penetration of renewable
energy. The simulation and analysis carried out in the Garver-18 bus test system verify the practicability
and effectiveness of the proposed method.
1.4. Structure
The rest of this paper is organized as follows. Section 2 presents the system adaptability index
system. Section 3 introduces the adaptive source-grid coordinative expansion planning model and its
solution method. The case study is illustrated in Sections 4 and 5 concludes this paper.
Figure 1. Typical characteristics of high penetration of the renewable energy integrated system.
141
Energies 2020, 13, 3304
uncertain characteristics and then effectively characterizes the adaptability to the uncertain impact of
renewable energy from multiple perspectives [17]. At the same time, the design of the index system
should be oriented to be a guidance criterion for source-grid planning. Here, the proposed index
system for the adaptability of high penetration of the renewable energy integrated system is given in
Figure 2.
Figure 2. Evaluation index system for system adaptability to high penetration of renewable energy.
The regulation demand of traditional power systems mainly comes from the power shortage
caused by faults. Specifically, when the power system lacks power supply, spare capacity is required to
fill the power gap in time to ensure the supply and demand balance. So, the fault regulation demand
has a single directionality, i.e., up-regulation. However, under the high penetration of renewable
energy integration, renewable energy sources have become the main source of the power supply and
will bear most of the load demand. Due to the strong fluctuation and intermittency of renewable
energy, its output power will continuously change with time and the direction of change is uncertain.
In this situation, the system needs to be able to quickly adjust the resources to match it, respond to
changes in the renewable energy, form a complementary relationship of upward and down-regulation,
and ultimately ensure reliable power supply and efficient consumption of renewable energy. Therefore,
the high penetration of the renewable energy integrated system has the regulation demand coming
from the renewable energy power change and its directionality has the upward and downward
bidirectional characteristics.
142
Energies 2020, 13, 3304
When there is a generator and transmission line failure, it will cause up-regulation demand.
However, with the improvement of the safety and reliability of the power system, the probability of its
occurrence is very low and the frequency of the up-regulation demand is the same as the frequency of
the failure. At the same time, the response time of its regulation is usually required to be within 10–30
min, which is a medium-long regulation time scale. However, because the output power of renewable
energy varies with natural resources, the fluctuation is relatively frequent and there are extreme cases of
steep drop and steep increase. Therefore, the frequency of upward and down-regulation requirements
resulting from changes in the output of renewable energy is relatively higher. Similarly, the response
time of the regulation resources is shorter, which belongs to the short-term regulation time scale.
Comparing with the temporal characteristics of traditional reserve regulation, the regulation demand
caused by renewable energy has a higher frequency and shorter time scale.
The actual regulation capability of the power system is determined by the capacity of the system
configuration regulation resource and network structure, the capacity of the regulation resource
configuration determines the upper limit of the system regulation capability. In the operation of the
system, due to the limitation of network transmission, the actual regulation capability of the power
system is often less than the regulation capacity. Therefore, in the analysis of the power supply and
demand balance relationship between source and demand, it is not possible to simply use the sum
relationship to determine whether the available regulation resources can match the demand. On the
one hand, the quantitative relationship can theoretically reflect whether the upper limit of regulation
capability can meet demand; on the other hand, when further analyzing the balance between supply
and demand, power flow constraint should be added to consider the support of the network.
In high penetration of the renewable energy integrated system, the fluctuation of renewable energy
output power is the main reason for the regulation demand. Compared with changes in load power, the
fluctuation of renewable energy power is larger and more frequent and there is no corresponding rule
to follow. This paper mainly considers the regulation demand caused by the fluctuation of renewable
energy, the relevant regulation demand is formulated as [18]:
⎧
Pi,REG (t + τ) − Pi,REG (t) Pi,REG (t) > Pi,REG (t + τ)
up
⎪
⎪ PD (t, τ) =
⎪
⎪
⎨ Pi,REG ∈ΩREG
⎪
Pi,REG (t + τ) − Pi,REG (t) Pi,REG (t) < Pi,REG (t + τ)
⎪ Pdown (t, τ) = (1)
⎪
⎪
⎩ D
P ∈Ω
i,REG REG
where PD up (t, τ) and PD down (t, τ) represent the up- and down-regulation demand respectively; Pi,REG (t)
and Pi,REG (t + τ) represent the output power of the renewable energy unit i at the time t and t + τ;
ΩREG is the set for the renewable energy unit.
It can be seen from (1) that when the output power of renewable energy decreases at t + τ,
up-regulation demand will be generated. If the fast regulation resources can match this demand,
the power supply reliability can be maintained continuously. Otherwise, there will be a load
shedding. Similarly, when the output power of renewable energy increases at t + τ, there will be
a down-regulation demand. If the fast regulation of resources cannot be matched, there will be a
curtailment of renewable energy.
143
Energies 2020, 13, 3304
The regulation resources include various types of generators with rapid regulation capabilities,
such as rapid regulation units, energy storage devices, etc., and various regulation methods, such
as demand-side response, rapid load shedding. However, as far as the technical level and practical
application of the above-mentioned various measures are concerned, energy storage devices are
expensive to build and have limited capacity and are mostly used in distribution networks and below.
Demand-side responses are similar. Two types of regulation measures, such as rapid load shedding
and curtailment of wind and photovoltaic, are contrary to the original intention of reliable power
supply and efficient consumption of renewable energy and are often used as the last level of security
control. Therefore, at the transmission grid level, large-capacity fast regulating units are currently the
most important regulation resource. For such resources, its ability to regulate can be expressed as [18]:
⎧ up up
!
⎪
⎪ PS (t, τ) = min Ri,G ·τ, Pmax − Pi,G (t)
⎪
⎪ i,G
⎨ Pi,G ∈ΩG
! ,
⎪
⎪ (2)
⎪
⎪
down (t, τ) = min Rdown ·τ, Pi,G (t) − Pmin
⎩ PS P ∈Ω
i,G i,G
i,G G
where PS up (t, τ) and PS down (t, τ) represent the up- and down-regulation capability and the subscript S
refers to “Supply”; ΩG is the set of fast regulating units; Pi,G (t) is the output power of the fast regulating
units i at time t; Pi,G max and Pi,G min are the upper and lower output of the fast regulating generator
units respectively; Ri,g up and Ri,g down respectively represent the maximum up and down ramp rate of
the fast regulation units.
As can be seen from (2), that the regulation capability of the fast regulation unit is determined by
the current operating state and regulation capability of the unit. Taking the up-regulation capability as
an example, the actual regulation capability is the smaller of the maximum climbing regulation rate,
the upper limit of the output power, and the difference between the actual output power.
⎧ up
⎪
⎪
up up
⎨ PM (t, τ) = PS (t, τ) − PD (t, τ)
⎪
⎪ , (3)
⎩ PM (t, τ) = PS
down down (t, τ) − Pdown
D
(t, τ)
where PM τ) and PM
up (t, τ) represent the upper and lower regulation resource margins and
down (t,
the subscript M refers to “Margin”; B is bus admittance matrix; θ, Pg , PG , PREG , PL are voltage angel
vector, the output vector of the conventional generator unit integrated bus, the output vector of the fast
regulation generator unit integrated bus, the output vector of the renewable energy unit integrated bus
and the power vector of the load bus.
The presented state evaluation model of the supply and demand balance consists of two parts. (3)
can be used to judge whether the upper limit of system regulation resources is sufficient or insufficient,
so as to indicate whether the regulation resources allocated by the system can meet the demand of
regulation. (4) requires that the operation calls of each unit meet the network transmission constraints.
Therefore, (3) and (4) can be used to comprehensively reflect the satisfaction relationship between the
actual available regulation capacity and the demand. To sum up, when PM up (t, τ) or PM down (t, τ) are
greater than 0, it indicates that the system can maintain the balance of power supply and demand at
this time; on the contrary, there is a lack of fast regulation in the system, which will lead to a certain
degree of forced load shedding or curtailment of renewable energy [18].
144
Energies 2020, 13, 3304
IRSD is introduced to characterize the ratio of the time when the power supply and demand
balance is insufficient to the total operation time, i.e., the probability of the insufficient supply and
demand balance [18]:
T
Rins = ins , Tins = tPM (t,τ)<0 , (5)
Ttotal
t∈Ttotal
where Tins and Ttotal represent the time of insufficient supply and demand balance and the total
running time; tPM < 0 represents the moment when the supply and demand balance is insufficient.
IRUR is introduced to characterize the ratio of the time between insufficient supply and demand
balance due to lack of up-regulation resources and the total time of up-regulation demand, i.e., the
probability of the supply and demand imbalance and load shedding due to lack of up-regulation
resources [18]:
up
Tins
up up
Rins = up , Tins = tPup (t,τ)<0 , (6)
Ttotal up M
t∈Ttotal
up up
where Tins and Ttotal are the time when the up-regulation resources are insufficient and the total
operation time when the up-regulation demand occurs respectively, tPM up < 0 is the time when the
up-regulation is insufficient.
IRDR is introduced to characterize the ratio of the time when the supply and demand balance is
insufficient due to the lack of down-regulation resources and the total time when the down-regulation
demand occurs, i.e., the probability of the supply and demand imbalance and renewable energy
curtailment due to the lack of down-regulation resources [18]:
down
Tins
Rdown = down
, Tins = tPdown (t,τ)<0 , (7)
ins
Tdown M
total t∈Tdown
total
where Tins down and Ttotal down represent the time when the down-regulation resource is insufficient
and the total operation time when the down-regulation demand occurs; tPM down < 0 represents the
time when the down-regulation is insufficient.
AIUR is introduced to characterize the average shortage of up-regulation resources per minute,
due to the mismatch of up-regulation resources and requirements [18]:
up
PM (t, τ)
up
up t∈Tins
Eins = up , (8)
Tins
AIDR is introduced to characterize the average shortage of down-regulation resources per minute,
due to the mismatch of down-regulation resources and requirements [18]:
145
Energies 2020, 13, 3304
down
PM (t, τ)
t∈Tdown
ins
Edown
ins
= down
, (9)
Tins
where the numerator represents the total insufficiency of down-regulation resource.
System operation security refers to the safety degree of the overall state, which can be quantitatively
characterized by measuring the safety distance between the current operating state and the fault
state. Self-organized critical theory is a typical evaluation method based on the above ideas [19].
The self-organized critical state represents the marginal state of the system operation. In this state,
any disturbance may cause system failure or even large-scale cascading failure. Since the traditional
operation mode of the power system is relatively fixed, the self-organized critical theory mainly starts
from the spatial dimension and evaluates the safety of the system by analyzing the concentration
and order of the load levels between lines under a certain operation mode. However, due to the
superposition of the fluctuations of the renewable energy output, the load level of lines shows
spatiotemporal differences. In addition, the traditional self-organized critical theory emphasizes
that the lower the load rate, the safer it is and ignores the economics of operation. Therefore, in
high penetration of the renewable energy integrated system, when the operating state of each line is
concentrated and orderly at a reasonable load level recognized by the operator, the actual operating
state of the power system is safer.
The operational efficiency refers to the economic utilization level of power system transmission
equipment and reflects the reasonable degree of the power system investment, planning, and operation.
If the utilization efficiency is too low, it will cause a large number of transmission equipment to be
idle and redundant, indicating that the grid investment is too large and there is a lot of waste of
resources. The high utilization efficiency of the power system improves the economics of power
system investment and operation, but this situation often has a limited safety margin and there is a
risk of safe operation. Existing operational efficiency indexes just reflect the average utilization level
of transmission equipment and cannot reflect the penetration of lines with high or low utilization
efficiency and the impact on overall efficiency. In addition, for the operation state with reasonable
utilization efficiency, the operational efficiency of the system cannot be simply characterized by the
load level. Therefore, in an uncertain operating environment where high penetration of renewable
energy is integrated, operation state with good efficiency and adaptability means that each line is
always operating within a reasonable range as determined by the planner. The utilization level of each
line is similar, so that the overall operation state is efficient and reasonable.
146
Energies 2020, 13, 3304
The stability of the operating state can be understood as the adaptability of the operation state to
power fluctuations such as renewable energy or the ability to maintain the operation state in a dynamic
environment. The uncertainty of the integrated renewable energy unit is coupled with the overall
operation state and there will be a phenomenon where the power flow distribution is continuously
shifted and the system operating state changes rapidly. At this time, if the ability to maintain the
operation state is poor, it will result in a large difference in the safety and efficiency characteristics
of the system state under the flow section at different times. This will not only lead to poor safety,
low utilization efficiency, or heavy overloading, but also the evaluation index cannot cover the above
extreme scenarios, which ultimately leads to the ambiguity of the index and reduces the accuracy of
the overall assessment. In the face of various power fluctuations and shocks coupled in the operation,
the operating state of the various lines has a low degree of change and the fluctuation range of the
transmission power is small. The power disturbance is suppressed to the greatest extent and eventually
it will not cause a huge change in the overall operating state. Therefore, when the load level of all
lines is concentrated and orderly within the reasonable operating range recognized by the operating
personnel, it indicates that the safety and security of the current operating state is at a good level. In
other words, showing good adaptability to the uncertainty of the complex spatiotemporal distribution
of renewable energy.
The high frequency and large range of output power changes of renewable energy, the
superimposition of spatiotemporal characteristic lead to the continuous transfer of the state. For each
line, the transmission power fluctuates and changes, which can be understood as a time series. The
line undertakes the corresponding power transmission task at different times according to the power
flow distribution. Therefore, this paper first defines the expected line load rate to describe the average
load level of the transmission power in each scenario [18]:
" +∞
−∞
Pi ϕ(Pi )dPi
Fi = , (10)
Si,max
where Pi and ϕ(Pi ) represent the actual transmission power of line i and the probability density function
respectively; the numerator represents the expected value of the transmission power; Si,max represents
the rated transmission power capacity of the line.
147
Energies 2020, 13, 3304
where E(·) is the expected value of the relevant parameters; the numerator is the fluctuation value of
the transmission power; the denominator is the allowable power fluctuation range, i.e., the power
fluctuation capacity.
The operating state is regarded as the system as a whole and each load rate interval is regarded as
a system event. The ratio of the number of lines in each interval to the total number of lines is regarded
as the probability of occurrence of events in this interval. The weights of each interval are subjectively
determined according to planners, and set the optimal operating interval weight to the minimum.
The weighted entropy theory can be used to establish WEELLR to evaluate the concentration and
rationality of the operating state distribution.
The load rate interval is expressed as a set M = [M1 , M2 , . . . , M5 ], where interval M1 indicates
that the load rate interval is (0, 20%] of the rated capacity and so on, M5 indicates (80%, 100%]. NMi is
the number of lines whose expected load rate belongs to the interval. Then, the probability that the
line is in the ith expected load rate interval can be expressed as [20]:
NMi
p ( FM i ) = . (14)
5
NMi
i=1
The weight of the expected load rate interval of each line ω(FMi ) can be determined according to
the planner’s comprehensive consideration of the grid efficiency and safety margin. For example, the
rational ranking of the load rate interval is: M3 > M2 > M1 > M4 > M5 , where the “>” symbol indicates
that the left side of the symbol is more reasonable than the right-side object. The above sorting method
shows that the planner thinks that when the line load rate in (80%, 100%] of the rated capacity is the
most reasonable and its weight is the smallest. Therefore, WEELLR is expressed as [20]:
5
HF = − ω(FMi )p(FMi ) ln p(FMi ). (15)
i=1
This index reflects the concentration and rationality of the distribution of the average load levels
between lines from the spatial dimension. The smaller the index value, the more the expected load
rate of the line is concentrated in the range that the operator believes is more reasonable. The overall
average state has both safety and efficiency and has good adaptability.
Due to fluctuations of renewable energy output, the operating state will shift from the average
state. When the offset is large, it will cause the safety and efficiency of the running state to not maintain
the average state. Therefore, WELPFR is introduced to quantify the degree of this deviation and to
evaluate the stability of the system operating state.
The operating state is still regarded as the system as a whole and each power fluctuation rate
interval is regarded as a system event. The ratio of the number of lines existing in each interval to the
total number of lines is regarded as the probability of occurrence of events in this interval. Because
148
Energies 2020, 13, 3304
the smaller the line power fluctuation, the more stable the overall system state, so the weight of the
operating interval with the smallest power fluctuation rate is set to the minimum. The weighted
entropy theory can be used to establish WELPFR to evaluate the stability of the grid operating state
distribution. The power fluctuation rate interval is expressed as a set V = [V1 , V2 , . . . , V10 ], where
interval V1 represents the power fluctuation rate value interval is (0, 10%] and so on, V10 indicates
(90%, 100%]. NVj is the number of lines whose line power fluctuation rate belongs to the interval Vj .
Then, the probability that the line is in the jth power fluctuation rate interval can be expressed as [20]:
NV j
p(BV j ) = . (16)
10
NV j
j=1
The smaller the power fluctuation of each line is, the more stable the system is. Therefore, the
smaller the weight of the power fluctuation rate interval, the higher the rationality and the average
value of the line power fluctuation rate of each interval can be directly used as the weight of the interval.
Let k denote the line whose line power fluctuation rate is in the interval Vj , the weights is calculated as
follows:
1
ω(BV j ) = Bk . (17)
NV j
k∈V j
10
HB = − ω(BV j )p(BV j ) ln p(BV j ). (18)
j=1
This index considers the fluctuation of the transmission power of all lines and can reflect the
stability operating state from the time and space dimensions. The smaller the index value, the more
the line power fluctuation range is smaller, power system has better adaptability and smoothing ability
to the renewable energy power fluctuation, and shows high inertia operating characteristics.
WEELLR evaluates the safety and efficiency operating state, WELPFR evaluates the stability of
the operating state. The above two indexes and three attributes jointly determine the adaptability and
rationality of the system operating state. However, in practical applications, it is often more desirable
to use an index to directly assess the adaptability and rationality of the operation state. Therefore, this
paper combines the two indexes and further propose JWEOS to comprehensively evaluate the safety,
efficiency, and stability of operation state.
The operating state of high penetration of the renewable energy integrated system can be
characterized by the expected line load rate and line power fluctuation rate. If the grid operating state
is regarded as the whole system, the expected line load rate and line power fluctuation rate can be
regarded as the two types of events that constitute the overall system, the concentration and rationality
jointly determine the adaptability operating state. Because, under the considered operation scenario
and corresponding operation mode, the expected load rate and power fluctuation rate distribution are
obtained through simulation. Therefore, JWEOS is formulated as [21]:
5
10 !
HFB = − ω(FMi )ω(BV j )p(FMi )p(BV j ) ln p(FMi )p(BV j ) . (19)
i=1 j=1
The smaller the index value, the more lines in the system are operating within the load level
interval that the planner thinks reasonable and the more line with smaller power fluctuation range.
Combined with the definition of adaptability of operation state, it can be seen that the system considers
149
Energies 2020, 13, 3304
both safety and efficiency in this operating state, with strong robustness and a comprehensive high
level of adaptability.
On the one hand, bus robustness is determined by the load level and the integrated generator
capacity, on the other hand, by the spatial position of bus in the power system topology. Bus that
bears important loads or bus with important generators often has a high importance level and needs to
pay special attention to ensure reliable power supply of the critical load. Bus with more complicated
contact relationships in the system also tends to have a higher importance level. Once this type of bus
withdraws due to various uncertain shocks or reduces its contribution to the operation, it has a wide
range of coverage and is often easy to cause larger failures. Therefore, bus with higher criticality tends
to become weak links, it is of great significance to improve the robustness of the network structure to
identify the key bus and reduce its weakness through reasonable planning.
The transmission line assumes the key role of connecting various types of generator and load
buses, which is the basic platform to ensure the reliable power supply. On the one hand, branch
robustness is determined by the importance of buses connected by the transmission corridor, on the
other hand, by the transmission capacity undertaken by the transmission corridor in the system power
supply and demand balance. The transmission lines with more complicated contact relationships and
contact buses belonging to important buses often have higher importance levels. Once such lines exit
the system due to various uncertain impacts, it is easy to cause large-scale chain failures and large-scale
power outages. Therefore, identifying the weak links of the power transmission line and on this basis,
reducing the number of weak links and reducing the degree of weakness is the key to improving the
robustness of network structure.
150
Energies 2020, 13, 3304
buses and branches and use graph theory, statistical theory, etc. The classic index of the method to
quantify the network structure characteristics of complex systems. Among them, buses of the complex
network represent the basic units that make up the system and branches reflect the association between
the basic units. In the analysis of complex networks in power systems, the electric betweenness is the
most commonly used. This index is based on the network intermediary index and combined with the
improvement of the actual characteristics of the power flow distribution. It is divided into electric
betweenness of buses and branches. Then, clarify the network’s weak links and their weaknesses.
It means that any power generation and load bus pair in the power system is selected, the power
generation and load power of the bus pair are used as power weights and then the unit current is added
between bus pairs and Kirchhoff’s law solves the current in the entire network After the traversal
considers all bus pairs, the current through a bus or branch is weighted and accumulated and the
electric betweenness of bus or branch can be obtained. The index can reflect the power capacity of the
different power generation and load bus pairs by adjusting the weights and can accurately characterize
the occupancy of each bus and branch element based on Kirchhoff’s theorem.
(1) Electric Betweenness of Bus
Electric betweenness of bus characterizes the occupancy of all bus components by each power
transmission path of the grid, i.e., the contribution of different bus in the power transmission. The
larger the electric betweenness of bus, the greater the contribution and importance of bus in the
power transmission of the whole network and the more easily and disturbed the system power is. By
obtaining the electric betweenness values of all buses in the power system, the weak ranking of buses
can be obtained. The electric betweenness of bus i can be calculated as [22]:
√
Di = ωa ωb Be,ab (i), (20)
a∈Ge,b∈Lo
where a, b respectively represent the number of generator and load bus; Ge and Lo represent the set
of generator and load bus; ωa , ωb represent the weights of generator and load bus, their values are
denoted as the actual power; Be,ab (i) is the current distribution of bus i, after add a unit current element
between pairs of generator load bus (a, b), which can be calculated as [22]:
⎧
⎪
⎪ 1
I (i, j) i a, b
⎪
⎨ 2 j ab
Be,ab (i) = ⎪
⎪ , (21)
⎪
⎩ 1 i = a, b
where Iab (i, j) is the current of branches (i, j) after adding the unit current element between the generator
load bus pair (a, b) and bus j represents the remaining buses directly connected to bus i. When bus
i is one of the generator load bus pairs, the current value passing through bus is regarded as the
unit current.
(2) Electric Betweenness of Branch
Electric betweenness of branch characterizes the occupancy of all power system components
by each power transmission line, i.e., the contribution of each branch in the power transmission of
the entire network. The greater the electric betweenness of branch, the greater the contribution and
importance of branch in the power transmission. It also indicates that branch is more susceptible to
system power disturbances and has less robustness. The easier it is to become the weak link in the
overall grid. The electric betweenness of branch m can be calculated as [22]:
√
Lm = ωa ωb Iab (m), (22)
a∈Ge,b∈Lo
where Iab (m) is the current of branch m after the unit current element is added between the generator
load bus pair (a, b).
151
Energies 2020, 13, 3304
Although the electric betweenness of bus can clarify the order of the weakness of each bus within
the power system, it cannot compare the robustness of different network structures from the perspective
of the overall bus. Therefore, this paper proposes WEEBBus to clarify the difference in robustness of
different network structures. The network topology is regarded as the system, the electric betweenness
interval of each bus is regarded as event. The ratio of the number of buses in each interval to the
total number of buses is regarded as the probability of occurrence of events in this interval. Since the
smaller the electric betweenness of bus, the more robust the component and the overall structure, the
interval weight with the smallest electric betweenness is set to the minimum. Thus, the weighted
entropy theory can be used to establish WEEBBus and the robustness of network structure can be
evaluated from the perspective of bus.
Represent the interval for electric betweenness of bus as a set O = [O1 , O2 , . . . , O5 ], NOi is the
number of bus whose electric betweenness is in the interval Oi . Then, the probability that the electric
betweenness of bus belongs to the ith interval Oi can be expressed as [20]:
NOi
p(DOi ) = . (23)
5
NOi
i=1
The smaller the electric betweenness of bus, the less the weak links of the network, the better the
robustness of the structure. Therefore, the smaller the electric betweenness value is, the smaller the
weight value of interval is. Here, let k denote bus whose electric betweenness belongs the interval
Oi , the average value of the electric betweenness for each interval is used as the weight value of the
interval [20]:
1
ω(DOi ) = Dk . (24)
NOi
k∈Oi
5
HD = − ω(DOi )p(DOi ) ln p(DOi ). (25)
i=1
The smaller the index value, the more buses in the power system are concentrated in a smaller
value interval. This shows that from the perspective of buses, the overall structure is balanced, robust,
and has few weak links and is not easily affected by power disturbances. Network structure has higher
adaptability and stronger ability to withstand uncertain shocks.
When the electric betweenness of all branches is concentrated in a relatively small range, it
indicates that there is no weak link in the branch of the network structure. The topological structure is
regarded as the system as a whole, the electric betweenness interval of each branch is regarded as a
system event and the ratio of the number of branches in each interval to the total number of branches
is regarded as the probability of occurrence of events in this interval. The interval weight with the
smallest electric betweenness is set to the minimum and so on. In this way, the weighted entropy
theory can be used to establish WEEBBra and the robustness of the network structure can be evaluated
from the perspective of the branch.
152
Energies 2020, 13, 3304
Represent the interval for electric betweenness of branch as a set Z = [Z1 , Z2 , . . . , Z5 ], NZj is the
number of branches whose electric betweenness belongs to interval Zj . Then, the probability that the
electric betweenness of bus belongs to the jth interval Zj can be expressed as [20]:
NZ j
p(LZ j ) = . (26)
5
NZ j
j=1
On this basis, determine the weight ω(LZi ) of the interval for electric betweenness of each branch.
Similarly, the smaller the electric betweenness is, the smaller the weight setting. Let k denote the branch
whose electric betweenness belong the interval Zj , the average value of electric betweenness of the
branch for each interval is used as the weight of the interval [20]:
1
ω ( LZ j ) = Lk . (27)
NZ j
k∈Z j
5
HL = − ω(LZ j )p(LZ j ) ln p(LZ j ). (28)
j=1
The smaller the index, the more branches are concentrated in the smaller value interval of electric
betweenness. This shows that from the perspective of branch circuit, the overall structure is balanced,
robust, and not easily affected by power disturbances. Network structure has higher adaptability and
stronger ability to withstand uncertain shocks.
From the above analysis, we can see that buses and branches are the most basic components of the
power network and their respective robustness will directly affect the ability of the overall structure
to withstand uncertain shocks. Although the above two weighted entropy indexes can describe the
robustness of all buses or all branches separately, they cannot describe the robustness of the overall
structure. Therefore, this paper further proposes JWENS. Based on the overall consideration of all the
basic components, the robustness topology is comprehensively evaluated and then the adaptability of
network structure to uncertain shocks is reflected.
The most basic components topology includes two types of buses and branches. If the power
system topology is regarded as the system as a whole, then buses and branches can be regarded as
two types of events that constitute the system as a whole. The rationality together determines the
adaptability of the network structure. For a certain network structure, the connection relationship
between its internal buses and branches is fixed and the distribution of electric betweenness of buses
and branches is also independent and determined. Therefore, the joint weighted entropy model of the
decoupling calculation is used to establish JWENS as follows [21]:
5
5 !
HDL = − ω(DOi )ω(LZ j )p(DOi )p(LZ j ) ln p(DOi )p(LZ j ) . (29)
i=1 j=1
The smaller the index value, the more buses and branches have smaller electric betweenness. In
other words, the fewer weak links in system, the smaller the impact of power disturbance, the better
the overall robustness, and the higher the adaptability to uncertain shocks.
153
Energies 2020, 13, 3304
r(1 + r)n
k= , (32)
(1 + r)n − 1
CWR = D p(s) cREG ·Pdown
M (t, τ)·τ, (33)
s∈S t∈Tdown
ins
cLOAD ·PM (t, τ)·τ,
up
CWL = D p(s) (34)
s∈S up
t∈Tins
154
Energies 2020, 13, 3304
where CG is the cost of regulation capacity transformation, CWR is the annual cost of renewable energy
generation, CWL is the annual load shedding cost; k1 , k2 are the fund recovery factor and the project
fixed operating rate, which are used to allocate the one-time investment cost to each year of life span, r
is the discount rate, n is the economically applicable year of project; ΩG is set of fast regulating units; zi
is the sign for transformation, xi is the transformation capacity of unit, yi is the transformation quantity
of regulation rate; c1,i is the unit cost for capacity transformation, c2,i is unit cost for transformation
of regulation rate; D is the number of days of operation per year, S is the set for typical scenarios of
renewable energy outputs, p(s) is the probability of typical scenarios s, cREG is the unit cost of renewable
energy curtailment, and Tins down is the time of renewable energy curtailment; |PM down (t,τ)·τ| is the
curtailment power of the renewable energy, cLOAD is the cost of load shedding, Tins up is the time of
load shedding, |PM up (t,τ)·τ| is the power of load shedding.
5
10 !
HFB = − ω(FMi )ω(BV j )p(FMi )p(BV j ) ln p(FMi )p(BV j ) , (37)
i=1 j=1
5
5 !
HDL = − ω(DOi )ω(LZ j )p(DOi )p(LZ j ) ln p(DOi )p(LZ j ) , (38)
i=1 j=1
where CN is the cost of grid expansion; HFB is JWEOS; HDL is JWENS; Ωl is the set of expanded lines,
ci is the unit investment of reconstructed line, ni is the circuit number of expanded lines, li is the total
length of the reconstructed line.
3.2. Constraints
Pmin
i,G ≤ Pi,G (s, t) ≤ Pmax
i,G , i ∈ ΩG , (40)
Rins ≤ α, (44)
155
Energies 2020, 13, 3304
where ΣPL (s,t) is the total load at time slot t in scenario s; Pi,g (s, t), Pi,G (s,t), and Pi,REG (s,t) are the actual
output power of the conventional unit, the fast regulating unit, and the renewable energy unit at time
slot t in scenario s; Ωg , ΩG , ΩREG are the sets of the corresponding units; Pi,G max and Pi,G min are the
upper and lower limits for the output power of fast regulating unit; Pi,g max and Pi,g min are the upper
and lower limits for the output power of the conventional unit; Ri,G up and Ri,G down are the maximum
ramp up and down rate of the fast-regulation unit under the time scale τ; B is the admittance matrix; θ
is the voltage phase angle vector; Pi is the transmission power of line i; Si,max is the rated transmission
capacity of line i; Rins is the insufficient rate index of supply and demand balance; α is the set threshold
value.
nmin
i ≤ ni ≤ nmax
i , i ∈ Ωl ni ∈ Z, (45)
⎧
⎪
⎪ −Bθ + Pg + PG + PREG = PL
⎪
⎪
⎪
⎪
⎨ |Pi | ≤ Si,max , i ∈ ΩL
⎪
⎪ Pmin ≤ Pi,g ≤ Pmax , i ∈ Ωg
, (46)
⎪
⎪
⎪
⎪
i,g i,g
⎩ Pmin ≤ P ≤ Pmax , i ∈ Ω
i,G i,G i,G G
⎧
⎪ −B N−1 N−1
+ PN−1 + PN−1 + PN−1 = PN−1
⎪
⎪
⎪
⎪ θ g G REG L
⎪
⎪ Pi ≤ Si,max , i ∈ ΩL
N−1
⎨
⎪
⎪ Pmin ≤ PN−1 ≤ Pmax , i ∈ Ωg
, (47)
⎪
⎪
⎪
⎪ i,g i,g i,g
⎪
⎩ Pmin ≤ PN−1 ≤ Pmax ,i ∈ ΩG
i,G i,G i,G
⎧
⎪
⎪ −Bθ
(s, t) + Pg (s, t) + PG (s, t) + PREG (s, t) = PL (s, t)
⎪
⎪
⎪
⎪
⎪ Pi (s, t) ≤ Si,max , i ∈ ΩL
⎪
⎪
⎨ Pmin ≤ Pi,g (s, t) ≤ Pmax , i ∈ Ωg
⎪
⎪ i,g i,g , (48)
⎪
⎪
⎪
⎪ Pi,G ≤ Pi,G (s, t) ≤ Pmax
min , i ∈ Ω
⎪
⎪ i,G G
⎪
⎩ up
−Rdown
i,G
·τ ≤ Pi,G (s, t) − Pi,G (s, t − τ) ≤ Ri,G ·τ, i ∈ ΩG
where ni min , ni max are the upper and lower limits of the circuit number of lines; the superscript “N−1”
represents the variable in the case of N−1 condition.
156
Energies 2020, 13, 3304
157
Energies 2020, 13, 3304
158
Energies 2020, 13, 3304
resources of the system are relatively abundant and the sum of the theoretical output of conventional
power generation resources and renewable energy resources far exceeds the load power supply
requirements of the system. At the same time, even if there is an extreme situation where the output of
renewable energy is zero, the upper limit of the capacity allocation of conventional power generation
resources can meet the load power. The main contradiction that causes the supply and demand balance
to remain unmaintained is that the system cannot compensate for the regulation demand brought
about by changes in renewable energy power in a relatively short time scale. Therefore, the planning
model optimizes the regulation capability parameters of the conventional generator set, so that the
regulation capability of the system can be improved. It can follow the magnitude and rate of the
response to the regulation demand and truly releases the capacity advantage of the system. Therefore,
under the new background of renewable energy grid connection, the main demand for power system
regulation has changed from capacity demand to demand for rapid regulation capability.
The objective value of optimal grid planning scheme in Table 1 is given in Table 2. As can be
seen from the results that although the economically optimal scheme can guarantee various safety
constraints operation and meet the power supply requirements of the load, its power system status
and structural adaptability indexes are weaker than the comprehensive optimal scheme. This can
be understood as that although the comprehensive optimal scheme sacrifices part of the economy,
it effectively considers the operating state and adaptability of the structure in the planning process
and has a more reasonable operating state and a more robust grid structure. It can better adapt to
the development trend of large-scale grid connection of renewable energy. At the same time, the
comprehensive optimal plan and the power supply construction and operation plan have a better
matching degree. The following is a further analysis of the specific differences between the operating
state and the network structure.
159
Energies 2020, 13, 3304
Table 5 shows the specific distribution of the power fluctuation rate of each line of the two schemes.
Lines whose actual maximum load rate is less than 10% are not considered, because the absolute load
level of such lines is too low and small changes in power will cause a large power fluctuation rate
index, but the absolute magnitude of the change in load level is still small. It can be seen from the table
that the fluctuation rate of 86% line in the comprehensive optimal scheme is below 20% of its allowable
fluctuation interval, of which the line power fluctuation rate is less than 10% and the penetration
of lines is 71%. The two data corresponding to the optimal economic plan correspond to 71% and
21%, respectively. In summary, the operating state of the optimal scheme is more stable, which can
better resist the impact of renewable energy power fluctuations on the system state under uncertain
160
Energies 2020, 13, 3304
operating environments and always maintain the safety and efficiency of the operating state during
actual operation. At the same time, it can be found that the two schemes still have a small amount of
line power fluctuation rate, which is because the line mainly undertakes the task of renewable energy
transmission. Large fluctuations in the output of renewable energy have led to large fluctuations in the
power of these lines, but they have not had a significant impact on the overall operating state.
161
Energies 2020, 13, 3304
5. Conclusions
From the perspective of adapting to the security integration of renewable energy, this paper firstly
proposes the adaptability indexes system and applies it to the two-stage adaptable expansion planning
of the transmission grid. The following conclusions can be draw:
(1) The adaptability of supply and demand balance not only evaluates the power balance of the
system, but also focuses on the adequacy of the system regulation capacity and rate, which can
reflect the level of renewable energy consumption and the risk of being urgently loaded; the
adaptability of operation state consider the characteristics of power flow distribution change
caused by renewable energy output fluctuation and the security, efficiency and stability of the
actual operation of system; the adaptability of network structure evaluation the balance and
robustness of system based on the order and severity of the electric betweennesses.
(2) The weighted entropy index can evaluate the adaptability state and network structure based
on the given objective criteria. Not only can it achieve the ranking of relative advantages and
disadvantages between the programs, but also it can refine the specific differences of program
states and structures based on the objective standards. The objective evaluation criteria can be
set by planner according to different considerations in different development periods, which
improves the flexibility and applicability of the indexes. The application of joint weighted
entropy can combine the physical meaning of each index to achieve dimensionality reduction and
integration of indexes, so as to comprehensively measure the security of the state and structure
with a single value. It is more suitable for the analysis safety adaptability under multi-dimensional
influence factors and the multi-objective planning.
(3) The expansion planning of the high renewable energy penetrated system is vital and decisive for
the actual consumption capacity in the operation stage, both the planning of the power plant in
the source side and transmission line in the grid side are interrelated. For the improvement of
safe integration of renewable energy in the planning stage, it is suggested to comprehensively
consider the adaptability of supply and demand balance, operation state, and network structure
in source-grid planning issues.
Author Contributions: Conceptualization, M.T.; methodology, J.W.; validation, J.W. and X.W.; writing—original
draft preparation, J.W.; writing—review and editing, M.T. and X.W. All authors have read and agreed to the
published version of the manuscript.
Funding: This work is founded by the National Key R&D Program of China (2018YFB0905200)—Research and
application demonstration on complementary combined power generation technology for distributed photovoltaic
and cascade hydropower.
Conflicts of Interest: The authors declare no conflict of interest.
162
Energies 2020, 13, 3304
Nomenclature
PD up , PD down up- and down-regulation demand (MW)
Pi,REG output power of the renewable energy unit i (MW)
ΩREG the set for renewable energy units
PS up , PS down up- and down-regulation capability (MW)
ΩG the set of fast regulating units
Pi,G output power of the fast regulating units i at time t (MW)
Pi,G max , Pi,G min upper and lower output of the fast regulating generator units (MW)
Ri,g up , Ri,g down maximum up and down ramp rate of the fast regulation units (MW/15min)
PM up , PM down upper and lower regulation resource margins (MW)
Tins up the time when the up-regulation resources are insufficient
Ttotal up the total operation time when the up-regulation demand occurs
Tins down the time when the down-regulation resource is insufficient
Ttotal down the total operation time when the down-regulation demand occurs
Pi the actual transmission power of line i
Si,max the rated transmission power capacity of the line
ωa , ω b the weights of generator and load bus
Be,ab (i) the current distribution of bus i
Iab (i, j) the current of branches (i, j)
CG the cost of regulation capacity transformation
CWR the annual cost of renewable energy generation
CWL the annual load shedding cost
CN the cost of grid expansion
ΣPL (s,t) the total load at time slot t in scenario s
IRSD insufficient rate of supply and demand balance
IRUR insufficient rate of up-regulation
IRDR insufficient rate of down-regulation
AIUR average insufficiency of up-regulation
AIDR average insufficiency of down-regulation
WEELLR weighted entropy of expected line load rate
WELPFR weighted entropy of line power fluctuation rate
JWEOS joint weighted entropy of operation state
WEEBBus weighted entropy for electric betweenness of bus
WEEBBra weighted entropy for electric betweenness of branch
JWENS Joint weighted entropy of network structure
References
1. Eitan, A.; Rosen, G.; Herman, L.; Fishhendler, I. Renewable energy entrepreneurs: A conceptual framework.
Energies 2020, 13, 2554. [CrossRef]
2. Murdock, H.E.; Gibb, D.; André, T. Renewables 2019 Global Status Report; REN21: Paris, France, 2019.
3. Kiviluoma, J.; Holttinen, H.; Weir, D.; Scharff, R.; Soder, L.; Menemenlis, N.; Cutululis, N.A.; Danti Lopez, I.;
Lannoye, E.; Estanqueiro, A.; et al. Variability in large-scale wind power generation. Wind Energy 2016, 19,
1649–1665. [CrossRef]
4. Zsiborács, H.; Baranyai, N.H.; Vincze, A.; Zentkó, L.; Birkner, Z.; Máté, K.; Pintér, G. Intermittent Renewable
Energy Sources: The Role of Energy Storage in the European Power System of 2040. Electronics 2019, 8, 729.
[CrossRef]
5. Lu, Z.; Huang, H.; Shan, B.; Wang, Y.; Du, S.; Li, J. Morphological evolution model and power forecasting
prospect of future electric power systems with high proportion of renewable energy. Autom. Electr. Power
Syst. 2017, 41, 12–18.
6. Lannoye, E.; Flynn, D.; O’Malley, M. Transmission, variable generation, and power system flexibility. IEEE
Trans. Power Syst. 2015, 30, 57–66. [CrossRef]
7. Wheelan, S.A. Harnessing Variable Renewables; International Energy Agency: Paris, France, 2011.
163
Energies 2020, 13, 3304
8. Lannoye, E.; Flynn, D.; O’Malley, M. Evaluation flexibility. IEEE Trans. Power Syst. 2012, 27, 922–931.
[CrossRef]
9. Nosair, H.; Bouffard, F. Flexibility envelopes for power system operational planning. IEEE Trans. Sustain.
Energ. 2015, 6, 800–809. [CrossRef]
10. Ma, J.; Silva, V.; Belhomme, R.; Kirschen, D.S.; Ochoa, L.F. Evaluating and planning flexibility in sustainable
power systems. IEEE Trans. Sustain. Energ. 2013, 4, 200–209. [CrossRef]
11. Zhuo, Z.; Du, E.; Zhang, N.; Kang, C.; Xia, Q.; Wang, Z. Incorporating Massive Scenarios in Transmission
Expansion Planning with High Renewable Energy Penetration. IEEE Trans. Power Syst. 2020, 35, 1061–1074.
[CrossRef]
12. Li, X.; Fan, J.; Cao, L.; Wang, L.; Hu, T.; Wu, T.; Zhou, Z. Analysis on the adaptability of large-scale
grid-connected PV station. Power Sys. Prot. Control 2018, 46, 164–169.
13. Liu, W.; Cai, W.; Yangqing, D.; Liang, C.; Wang, J.; Wang, N.; Ma, Y. The evolution of grid’s self-organized
criticality state based on entropy theory under the circumstance of large-scale wind power centralized grid.
Power Syst. Technol. 2013, 37, 3392–3398.
14. Zhang, C.; Cheng, H.; Liu, L.; Wang, Z.; Lu, J.; Zhang, X. Adaptability index and evaluation method for
power transmission network structure with integration of high penetration of renewable energy. Autom.
Electr. Power Syst. 2017, 41, 55–61.
15. Dai, W.; Yang, Z.; Yu, J.; Zhao, K.; Wen, S.; Lin, W.; Li, W. Security region of renewable energy integration:
Characterization and flexibility. Energy 2019, 187, 115975. [CrossRef]
16. Papavasiliou, A.; Oren, S.S. Large-Scale integration of deferrable demand and renewable energy sources.
IEEE Trans. Power Syst. 2014, 29, 489–499. [CrossRef]
17. Khan, S.U.; Mehmood, K.K.; Haider, Z.M.; Rafique, M.K.; Kim, C. A bi-level EV aggregator coordination
scheme for load variance minimization with renewable energy penetration adaptability. Energies 2018, 11,
2809. [CrossRef]
18. Sun, W.; Wang, C.; Zeng, P.; Han, J. Review on evaluation method and index of power system homogeneity.
Power Syst. Technol. 2015, 39, 1205–1212.
19. Liu, W.; Cai, W.; Zhang, N.; Dan, Y.; Liang, C.; Wang, W.; Liang, C. Evolution of self-organizing of grid critical
state based on united weighted entropy theory. Proc. CSEE 2015, 35, 1363–1370.
20. Sheng, S.; Gu, Q. A day-ahead and day-in decision model considering the uncertainty of multiple kinds of
demand response. Energies 2019, 12, 1711. [CrossRef]
21. Ji, W.; Wu, J.; Zhang, M.; Liu, Z.; Shi, G.; Xie, X. Blind image quality assessment with joint entropy degradation.
IEEE Access 2019, 7, 30925–30936. [CrossRef]
22. Bai, H.; Miao, S. Hybrid flow betweenness approach for identification of vulnerable line in power system.
IET Gener. Transm. Distrib. 2015, 9, 1324–1331. [CrossRef]
23. Teimourzadeh, S.; Aminifar, F. MILP Formulation for Transmission Expansion Planning with Short-Circuit
Level Constraints. IEEE Trans. Power Syst. 2016, 31, 3109–3118. [CrossRef]
24. Cheng, Y.; Zhang, N.; Wang, J.; Li, H.; Wang, Z.; Xie, L.; Kang, C. Comprehensive Evaluation of Transmission
Network Planning for Integration of High-penetration Renewable Energy. Autom. Electr. Power Syst. 2019,
43, 33–42; 57.
25. Huang, Y.; Li, Y.; Gao, C. Multi-objective transmission network planning based on non-dominated sorting
differential evolution. Dianwang Jishu/Power Syst. Technol. 2011, 35, 85–89.
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
164
energies
Article
Optimising a Microgrid System by Deep
Reinforcement Learning Techniques
David Domínguez-Barbero *, Javier García-González, Miguel A. Sanz-Bobi and
Eugenio F. Sánchez-Úbeda
Institute for Research in Technology (IIT), ICAI School of Engineering, Comillas Pontifical University,
28015 Madrid, Spain; [email protected] (J.G.-G.); [email protected] (M.A.S.-B.);
[email protected] (E.F.S.-Ú.)
* Correspondence: [email protected]
Abstract: The deployment of microgrids could be fostered by control systems that do not require very
complex modelling, calibration, prediction and/or optimisation processes. This paper explores
the application of Reinforcement Learning (RL) techniques for the operation of a microgrid.
The implemented Deep Q-Network (DQN) can learn an optimal policy for the operation of the
elements of an isolated microgrid, based on the interaction agent-environment when particular
operation actions are taken in the microgrid components. In order to facilitate the scaling-up of this
solution, the algorithm relies exclusively on historical data from past events, and therefore it does not
require forecasts of the demand or the renewable generation. The objective is to minimise the cost of
operating the microgrid, including the penalty of non-served power. This paper analyses the effect
of considering different definitions for the state of the system by expanding the set of variables that
define it. The obtained results are very satisfactory as it can be concluded by their comparison with
the perfect-information optimal operation computed with a traditional optimisation model, and with
a Naive model.
Keywords: machine learning; microgrids; optimisation methods; power systems; reinforcement learning
1. Introduction
The transformation of the electric power industry to reduce carbon emissions is changing the
roles of both generators and consumers. On the one hand, the generation system is hosting larger
amounts of renewable energy sources (RES), with an increasing share of distributed generation, i.e.,
small generators placed at the distribution level. On the other hand, end consumers are no longer
passive agents of the system, but rather active ones that can modulate their consumption during the
day, and reduce their net energy balance by means of more efficient and smarter operation of the loads.
In this increasingly decentralised power system, involved agents must make decisions continuously
in a limited information environment. This decentralisation can be materialised by the creation of
micro-networks (microgrids) that are groupings of distributed generation resources, consumption and
storage systems that can be partially controlled and that can work both connected to the conventional
network or in an isolated manner. The development of smart grids suggests that despite the existing
barriers, the penetration of these microgrids will increase considerably soon [1].
Finding the optimal operation of the different elements of the microgrid is not a simple task,
and there are several possible approaches in the literature such as robust optimisation [2], stochastic
optimisation [3], Mixed Integer-Linear Programming (MILP) combined with a Particle Swarm
Optimisation [4] or Reinforcement Learning (RL) [5]. It is important to highlight that robust or
stochastic optimisation models require an explicit representation of the microgrid, a mathematical
formulation that includes the objective function to be optimised and the set of all the constraints
that define the feasible domain. In addition, the input parameters required to forecast the random
variables of interest, such as loads and RES generation, are also a practical need for those optimisation
approaches. Therefore, the development of an optimisation-based controller requires the precise
selection and estimation of all the parameters and models used to represent each component. Obtaining
such information could be extremely difficult in the case of a massive deployment of microgrids.
For instance, in a hypothetical future scenario of the power system hosting millions of microgrids (from
residential prosumers to larger microgrids in industrial areas), the diversity of consumption patterns
and the intermittent nature of renewable generation might represent a serious barrier, as each specific
microgrid may require different parameters and even different models [6] leading to unacceptable
implementation costs.
In this context, the main advantage of RL is that it allows making decisions in a limited information
framework based on the actions taken in the past, and on the observed effects of such actions
by automatically adapting the control strategy over time. Unlike standard model-based control
approaches, RL techniques do not require the identification of the system in the form of a model.
In this way, RL is able to learn from the experience similarly to human behaviour. For example, in [7],
it is shown through numerical examples that the proposed learning algorithms can work effectively
despite not having prior information on the dynamics of the system. Another example can be seen
in [8] where an RL algorithm is proposed to coordinate the loading and unloading of several storage
systems with different characteristics. The authors in [9] present a comprehensive review of the main
research works found in the literature where RL is applied to a variety of typical problems of power
systems, and it presents an RL framework to operate real systems meanwhile a simulation model is
used to learn.
Over the last years, a new version of RL methods has emerged. It is called Deep Reinforcement
Learning (DRL) because it combines RL with Deep Learning (DL). These methods have an important
impact on the academic field, but the industrial applications are currently at the early stages of
development. As this is a very active research field, there is an increasing trend of papers that study
the application of DRL to power systems. For instance, in [10] several research works that apply RL to
power systems are discussed, and some of them are focused on DRL methods. Among them, Deep
Q-Network is the most popular technique given its simplicity in contrast to other more complex DRL
methods. It should be noted that despite being a recent publication, none of the papers reviewed
in [10] address the problem of finding the optimal steady state operation of a microgrid by using DRL.
A more updated and very recent survey paper is [11], where the authors present a broad review
of RL applications to power systems, although the categorisation differs from [10]. In particular,
one of the categories is the energy supply side, which is very related to the topic studied in this paper,
i.e., the microgrid optimal scheduling problem. Under this research area, reference [12] should be
highlighted as it addresses the optimisation of the operation of the microgrid using a Deep Q-Network
(DQN). The Artificial Neural Network (ANN) used in [12] to implement the DQN is a Multi-Layer
Perceptron (MLP). Nevertheless, there are more advanced ANN architectures such as Convolutional
Neural Networks (CNN), with one-dimension convolution layers, and Recurrent Neural Networks
(RNN), with memory cell layers. These architectures are very suitable for improving the train phase
when the input data has a time series structure. For instance, the authors in [5] developed a CNN to
find the optimal operation of a microgrid that could be operated both isolated or connected to grid.
The microgrid can be modelled as a Markov Decision Process (MDP). Therefore, beyond the
selection of the ANN architecture, the application of DRL (and also RL) requires a definition of a state
space of the environment so that the agent can select the optimal actions based on which state the
system is in. However, deciding what defines the state of the system may not be trivial. For example,
in chess, the position of all the pieces on the chessboard at any given time defines the state of the
system, and it is perfectly observable by the player. However, in a microgrid, the decision of what
defines the state of the system is not so direct, since the number of variables to be measured can be as
large as desired and the evolution over time is subject to uncertainty. The work in [13] presents the
166
Energies 2020, 13, 2830
theoretical consequences of varying the configuration of the state and discusses the bias-overfitting
tradeoff of DRL. This paper delves also on the effect of the state configuration, and following the
suggested research guidelines of [5], the impact of different state definitions of the microgrid are
analysed. Therefore this paper takes [5] as a starting point using a DQN based on CNN. In addition, as
the time window used to define the state can have different lengths, the main contribution of this paper
is the comparative analysis of the impact of the duration of such time window on the performance
of the model. In addition, the considered microgrid is more complex than the one studied in [5] as it
includes a diesel generator that can act as backup power. Finally, another contribution of this paper is
that the economic assessment of the operation of the batteries does not require fixing any price, as the
opportunity cost is properly modelled in the sense that the charge/discharge decisions have a direct
impact on the usage of the diesel generator. Therefore, this approach is more realistic and can make its
application easier to any particular system.
The paper is organised as follows. Section 2 introduces the microgrid elements, and also the
modelling fundamentals. The DRL algorithm is described in Section 3, and Section 4 presents its
application to a study case where the microgrid structure is described in detail. Section 5 analyses
the RL behaviour, comparing the results among the different state configurations. Finally, the main
conclusions are summarised in Section 6.
167
Energies 2020, 13, 2830
2.2.1. Demand
The load is modelled as a time series with hourly intervals that represent the average consumption
of each hour, and it is assumed that historical data is available. Furthermore, this demand is not
dependent on the behaviour of all the other elements of the microgrid. In case there is not enough
generation to supply the microgrid demand, the non-served power will take the required positive
value to ensure the demand balance equation. This non-served power will be severely penalised.
However, in case there is an excess of RES generation, it will be assumed that such extra production
can be curtailed without any cost (RES spillages).
2.2.2. Generation
The microgrid generation system consists of a solar panel and a diesel generator. Regarding the
solar panel, the maximum PV generation profile can be represented by an hourly time series, and its
variable operation cost can be neglected. On the other hand, the diesel generator can provide backup
energy in case of a lack of RES production. This diesel generator can be dispatched by the control
system that establishes the power to be produced in a deterministic manner, i.e., possible failures of the
equipment is out of the scope of this paper. Two different operation modes will be considered. In the
first one, the output power is allowed to take any continuous value between the minimum and the
maximum power. In the second one, only discrete values are possible. For simplicity, the input-output
cost curve is modelled as a quadratic function where the independent term, i.e., the no-load cost,
is only incurred when the generator is committed. Start-up and shutdown costs are neglected, and
further details can be seen in Section 4.
2.2.3. Storage
With respect to the storage, a Li-ion battery is used as short-term storage supporting the microgrid
in periods when RES cannot supply the demand. It is characterised by the maximum capacity RtB
[kWh], the maximum power PtB [kW] for charge and discharge and the charge and discharge efficiencies
μtB , ζ tB . It is important to highlight, under the proposed RL approach, that it would be easy to take into
account a more detailed model of the battery to capture the non-linear relationship among the state of
charge (SoC), the maximum power, the efficiencies and other characteristics. However, as the classical
optimisation model used to compare the results cannot include all those non-linear relationships,
a simple model has been preferred. The other storage device, i.e., a hydrogen system, is used as
168
Energies 2020, 13, 2830
long-term storage, supporting large periods of high demand and low RES. The characterisation of the
hydrogen storage is analogous to the Li-ion battery: maximum capacity RtH [kWh], maximum power
PtH [kW] for charge and discharge and the charging and discharging efficiencies μtH and ζ tB .
3. Reinforcement Learning
RL belongs to the area of Machine Learning benefiting from ideas of optimal control of
incompletely-known Markov decision process, and it has been applied to many fields [14]. An RL
problem is modelled as an agent and an environment that interact with each other as in Figure 2.
The agent makes decisions following a policy, exploring the environment dynamics and obtaining
rewards. The agent updates its policy π guided by these rewards rt over a discrete space of time
divided in time steps t, improving the cumulative reward.
Figure 2. General interactions in Reinforcement Learning (RL).
In the RL context, a state space S and an action space A has to be defined. Given an initial
state s0 , a policy π ∈ Π performs a trajectory τ = (s0 , a0 , s1 , a1 , ..., sn ) where sn is a terminal state.
A reward function R(st , at , st+1 ) and a return function G (τ ) have to be defined also according to the
RL problem. The optimal policy π ∗ is the one that maximises the expected return. An intrinsic feature
of an RL algorithm is the policy-based and the value-based characteristic, which affects the method
of computing the next action and the learning behaviour. Some methods have both characteristics,
but the proposed method in this paper has only the value-based characteristic.
169
Energies 2020, 13, 2830
and with this function, the algorithm could perform the best believed action:
with convergence to the optimal Q∗ (s, a). s and a are the next state and the action to be performed
in s respectively.
The curse of dimensionality phenomena makes the Q-Learning algorithm unsuitable to solve
high dimensional problems, and approximations of the Q(·) function have to be introduced to deal
with this drawback [14]. Since ANNs are universal approximators, DQN can be implemented to model
the Qθ (·) function, being θ the parameters of the ANN. The updating process of the Q-values has to be
computed to update each parameter θ of the ANN. Additionally, DQN uses a replay memory to help to
efficiently update the parameters and get a good trade-off between exploration and exploitation [22].
Regarding the variables to be monitored, in this paper it is proposed that each temporal slice is
defined by the next variables:
pv
slicet = ( Pt−1 , Dt−1 , Stb , Sth2 ). (6)
Notice that the state of charge (SoC) corresponds to the value of energy stored at the end of
each hour, and that the generated or consumed power represents its average value during each
hourly period.
In order to compute the SoC of both storage devices, the following energy balance equations must
be taken into account: ⎧
⎨Sb − | Pb | /ζ b ⇔ Pb ≥ 0
t −1 t t
St =
b
, (7)
⎩Sb + | Pb | η b ⇔ t <0
P b
t −1 t
⎧
⎨Sh2 − | Ph2 | /ζ h2 ⇔ Ph2 ≥ 0
t −1 t t
Sth2 = . (8)
⎩Sh2 + | Ph2 | η h2 ⇔ Ph2 < 0
t −1 t t
Notice that both storage devices must be operated within their feasible limits, and this is ensured
by a continuous monitoring at each decision step.
170
Energies 2020, 13, 2830
where
C d ( x ) = δ2 x2 + δ1 x + δ0 (10)
pns
is the diesel cost function [e] with quadratic coefficients δ2 , δ1 and δ0 . Pb , Pt and Ptcurt are calculated
using the demand balance-constraint equation given by
pns pv
( Dt − Pt ) = ( Pt − Ptcurt ) + Ptb + Pth2 + Ptd , (11)
prioritising the usage of the battery to reduce the not-supplied power and the curtailment.
4. Case Study
This section describes the case study used in this paper, as well as the different configurations
of the proposed Deep Reinforcement Learning (DRL) that have been used to perform the
comparative analysis.
171
Energies 2020, 13, 2830
Table 1. Cont.
4.1.2. Load
The hourly data for the load profile has also been obtained from [23]. This load profile represents
the typical residential consumption of a consumer with an average daily demand of 18.33 kWh. In case
the microgrid is unable to supply all the load, it is assumed a cost 1e per not supplied kWh.
172
Energies 2020, 13, 2830
of the batteries, the possible generation of the diesel group, the not supplied power and the possible
spillages. This benchmark model was implemented in GAMS [27], where the detailed equations
were omitted here for the sake of simplicity, but added in Appendix A. The structure of the resulting
optimisation problem is the following one:
min c( x )
x
, (12)
subject to g( x ) ≤ 0
where the objective function c( x ) is the sum of the diesel generation cost along the entire time horizon
plus the penalty term related to the possible non-supplied energy. The cost of the diesel generator is
assumed to be a quadratic function, where the independent term is only incurred when the generator
is on. This fact requires the use of unit-commitment binary variables.
The constraints g( x ) are the energy balance at the microgrid, and at the storage devices taking into
account their charging and discharging efficiencies, their maximum storage levels and the maximum
rated power.
As the objective is a quadratic function and some binary variables are needed, the resulting model
is a Mixed Integer Quadratic Programming model (MIQP).
173
Energies 2020, 13, 2830
"
!" !" #$% #$% $
#$% #$% ( ( (
&""' &""'
Figure 3. Convolutional Neural Network architecture used in the proposed Deep Q-Network (DQN).
5. Results
The CNN architecture proposed is similar but not equal to the one proposed in [5]. It takes
a whole tensor as the input of the first layer instead of scatter time-series input in 1D-convolution
layers and being merged with the remain inputs after the convolution computation. This configuration
brings more simplicity and scalability to the model without losing performance in the results despite
the information redundancy. This CNN has two 1-dimension convolution layer. The initialisation
procedures of these layers are the ones from Glorot [29] for Convolutional layers and He [30] for Dense
layers. All the parameter values used are shown in Table 2.
Parameter Value
Batch size 20
During the training phase, to deal with the lack of data, a regularisation technique called early
stop is applied. In this technique, the training set is split into a smaller training set and the development
174
Energies 2020, 13, 2830
set, and after a fixed number of training steps, the model is evaluated with the development set saving
the best snapshot of the model during the training.
This paper carries out the analysis of the window size of the state throughout different
configurations of the parameter k in Equation (5). The window size ranges from k = 3 to k = 24
in steps of 3 h. Following the state definition of Equation (5), some parameters in t < 0 are required
(state s0 contains some temporal slices in t < 0), and these are set to zero.
This section presents the results obtained with the DQN method applied to the microgrid
described previously. The DQN has been implemented in Python 3.7.7 using Tensorflow 2.1.0.
The MIQP model has been coded in GAMS using the solver CPLEX 12.9.0.0. The computer system was
an Intel Core i7-8550U (1.80 GHz–4.00 GHz) with 16 GB RAM running under Ubuntu 18.04 LTS x64.
The results gather several options for the definition of the state of the microgrid varying the
window size parameter k in Equation (5). As stated before, the motivation of this analysis is to find
how much recent information should be included in the state definition in order to obtain the best
performance of the microgrid. To put the results obtained with the DQN in perspective, other models
have been included in this analysis: the perfect-information MIQP model, the Naive strategy and a
Random Policy. The first one represents a lower bound of the operation cost that can be achieved
with such microgrid. The third one can be seen as an upper bound, where the average value of ten
samples (runs of the model) are included in the assessment. Based on the historical three-year data,
the return G (τ ) of the trajectory τ followed by the policy of the proposed DQN is used to measure
its goodness. The return match the cost of the microgrid in e and the trajectory τ is the operation
of the microgrid. Note that the first two years of available data are used to train the DQN model,
separated in the training set of the first year and the development set of the second year, whereas the
last year is used for testing. Concerning the reference provided by the MIQP, the model operates the
microgrid without any discretisation of the decision variables. The same applies for the Naive and the
Random strategies.
The MIQP model generates the lower bound given by a three-year accumulated cost of 2677.43 e
with a relative GAP of 6.06 as GAMS defines ( BestFeasible − BestPossible)/BestFeasible, taking 24 h
to find this solution. The accumulated cost for every year is shown in Table 3. The Naive strategy
achieves an expenses of 11,138.60 e, and the Random strategy 14,066.59 e, 316.02% and 425.38% more
than MIQP, respectively. Notice that the Naive strategy is not much better than the Random strategy,
and this fact calls for smarter policies to operate the microgrid, as the one proposed in this paper.
Table 3 shows the accumulated cost obtained after three years of operation of the microgrid using
the policy learned, for the different configurations of the proposed DQN. These results are significantly
better than the other models, achieving results between 3653.59 e for the best case with a window
size of 9, and 462,722 e for the worst-case with a window size of 21. Even the worst-case is much
better than a Naive or Random strategies. Relative to the results of the MIQP model, the worst-case
performance is two times worse than the best case. These results show that the choice of the window
size parameter has significant consequences in the performance of the DQN. Larger window size
involves learning difficulties in the policy for the DQN.
Another observation is the comparative between the best case DQN and the MIQP
results. The DQN achieves a good performance, being a method easy to implement and with
near-optimal results.
Analysing the agent behaviour, Figure 4 shows an example of the hourly microgrid operation
carried out by the proposed RL method. It consists of six days of the first year (three consecutive days
of winter and three of summer).
175
Energies 2020, 13, 2830
Cost [e] Year 1 Year 2 Year 3 Total Relative ( X − BEST )/BEST · 100
Figure 4. DQN operation of the microgrid, for summer (a) and winter (b). Hourly steps in a
3-day window.
Since the DQN only works with discrete action spaces, the continuous action space defined in
the MDP model of the microgrid in Equation (4) has to be discretised. Each hour t, the action at
consists of two variables (Ptd and Pth2 ) corresponding to the power generated by the diesel generator
and the power generated (consumed if negative) by the hydrogen fuel cell respectively. The required
discretisation of the diesel output power Ptd leads to the following cases:
176
Energies 2020, 13, 2830
• Ptd = Pmax
d , i.e., the agent configures the diesel generator to dispatch the maximum power (1 kW
power it can (0.5 kW in the example, and “Low” label in Figure 4).
• Ptd = 0, i.e., the agent configures the diesel generator in off. (0 kW in the example, and “Off”
label in Figure 4).
The agent considers also three different configurations for the values that Pth2 can take:
• Pth2 = − Pmaxh2 , i.e., the agent configures the hydrogen electrolyser to charge at maximum rate
from the system (load at 1 kW in the example, and “Load” label in Figure 4).
• Pth2 = 0, i.e., the agent configures the hydrogen components in an idle state (load/discharge at
0 kW, and “Idle” label in Figure 4).
• Pth2 = Pmax
h2 , i.e., the agent configures the hydrogen fuel cell to inject the maximum power into
the system (discharge at 1 kW in the example, and “Disx” label in Figure 4).
To summarise, in each hour the agent can take nine different configurations. The battery SoC
takes values between 0 kWh (0% in the figure) and 2.9 kWh (Full in the figure that correspond to 100%
of the maximum SoC).
According to Figure 4, there are no significant differences between the summer and winter load
profiles. In contrast, the available solar irradiation makes the difference between these opposite
seasons, being the main responsible for the change in the operation policy decided automatically by
the RL method.
There is a clear daily pattern operation of the Li-ion battery, but it differs with the season.
During summer the daily energy surplus provided by the solar panels is used to fill this battery during
daylight hours, being systematically discharged during the rest of the hours. However, the lack of solar
irradiation in winter is compensated by the continuous generation of the diesel to cover the demand in
the main hours, using the diesel surplus of the off-peak hours to charge the battery for it usage during
non-sunlight hours. On the contrary, during summer the use of the diesel is limited to some hours
where there is not enough solar irradiation. Concerning the hydrogen fuel cell, it is used in a similar
diesel pattern, to cover the demand when there is not enough solar irradiation.
In order to illustrate the behaviour of the model, a numerical example is presented hereafter.
Assuming that the state st considers a tensor of shape (k, 4) with k = 9, from the output results shown
in Figure 4a, the detailed values that correspond to hour t = 4380 are the next ones:
st = slicet−k , slicet−k+1 , ..., slicet−1 , slicet
⎛ pv pv pv pv
⎞
P , Pt−k , ..., Pt−2 , Pt−1
⎜ t − k −1 ⎟
⎜ Dt−k−1 , Dt−k , ..., Dt−2 , Dt−1 ⎟
= ⎜
⎜ b
⎟
⎟
⎝ St − k , Stb−k+1 , ..., Stb−1 , Stb ⎠
Sth2
−k , Sth2
− k +1 , ..., Sth2
−1 , Sth2
⎛ ⎞
0.002, 0.151, 0.461, 1.122, 1.973, 3.301, 4.295, 4.767, 4.891
⎜ ⎟
⎜ 0.061, 0.186, 0.447, 0.837, 1.222, 1.398, 1.270, 0.963, 0.711 ⎟
= ⎜
⎜
⎟.
⎟
⎝ 1.028, 0.989, 1.001, 1.257, 1.932, 2.260, 2.900, 2.900, 2.900 ⎠
36.000, 36.000, 36.000, 36.000, 36.000, 36.650, 37.300, 37.950, 38.600
Then, the action taken by the DQN using Equation (2) is:
at = Ptd , Pth2 =
0.000, −1.000
177
Energies 2020, 13, 2830
As a consequence of this action, the new state in hour t + 1 = 4381 is reached, with its
corresponding new slice:
slicet+1 = pv
Pt , Dt , Stb+1 , Sth2
+1
= 4.899, 0.672, 2.900, 39.250 ,
in which Stb+1 , that is calculated using the Equation (A3), takes the value 0.0 kW because it is full and
the curtailment Ptcurt , that is calculated using the balance Equation (11), takes the value of 3.217 kW.
Another interesting result that can be highlighted is the management of the hydrogen storage
along the year. One could expect that since there is less solar generation in winter, it would be better to
store energy during summer in order to have it available when necessary. Figure 5 shows the hydrogen
storage SoC for the whole 3-year period since this profile is barely appreciable in the 3-day window
of Figure 4. It can be seen that the proposed DQN is capable of finding an optimal yearly pattern,
charging the hydrogen tank in summer and discharging it in winter, and this optimal behaviour is
obtained just with the most recent information at each step. A summary of the usage of the hydrogen
storage is shown in Table 4.
6. Conclusions
This paper presents the application of DRL techniques to manage the elements of an isolated
microgrid. The advantage of the proposed approach is the ability of the algorithm to learn from its
own experience, and this can be an interesting feature to foster the up-scaling of these type of solutions
that do not require a tailor-made specific optimisation model adapted to the particularities of each
microgrid. However, as the application of RL and DRL to power systems is an emerging research field,
the selection of the most appropriate ANN architecture, or the definition of what variables should
be included in the the so-called system state, are still open questions. With respect to the first one,
the proposed CNN architecture provides simplicity and a good performance. Regarding the second
question, this paper provides a sensible set of variables to configure the system state. In addition,
178
Energies 2020, 13, 2830
the effect of increasing the time window of such state has been analysed, and the numerical results
show that 9 h is the optimal size of such time window for the studied microgrid. In order to measure
the quality of the DRL results, this paper presents a traditional Mixed Integer Quadratic Programming
optimisation model to compute the theoretical best operation possible, i.e., the hypothetical case of
having complete information for the whole time period. Using such model as benchmark, whereas
a Naive algorithm similar to the ones used in practical implementation results in an operation 316.02%
worse, the developed DRL model reaches a value of 36.46%. Future work will be oriented to the
implementation of Transfer Learning and Robust Learning to address multi-agent microgrid problems.
Author Contributions: Conceptualisation, J.G.-G. and M.A.S.-B.; methodology, D.D.-B., J.G.-G. and M.A.S.-B.;
software, D.D.-B.; validation, D.D.-B., J.G.-G., M.A.S.-B. and E.F.S.-Ú.; formal analysis, D.D.-B., J.G.-G. and
M.A.S.-B.; data curation, D.D.-B. and E.F.S.-Ú.; writing—original draft preparation, D.D.-B.; writing—review
and editing, D.D.-B., J.G.-G., M.A.S.-B. and E.F.S.-Ú.; visualisation, D.D.-B.; supervision, J.G.-G. and M.A.S.-B.
All authors have read and agreed to the published version of the manuscript.
Funding: This research has been funded by the Strategic Research Grants program of Comillas University, and by
“Programa microrredes inteligentes, Comunidad de Madrid (P2018/EMT4366)”.
Acknowledgments: The authors want to express their gratitude to Francisco Martín Martínez and Miguel M.
Martín-Lopo for their valuable help during the development of this research.
Conflicts of Interest: The authors declare no conflict of interest.
Notation
To facilitate the mathematical representation of the elements of the microgrid, the following
nomenclature is used throughout the paper.
Sets and indexes:
t∈T time periods from 1 to | T |
Parameters:
pv
Pt Maximum available PV power generation at time period t, [kW]
pv
Pmax Nominal rate of the solar panel, [kW]
Dt Load of the microgrid at time period t, [kW]
Dmax Maximum power the microgrid can consume, [kW]
ζb, Discharge efficiency of the Li-ion battery
ηb, Charge efficiency of the Li-ion battery
ζ h2 Discharge efficiency constants of the hydrogen storage
η h2 Charge efficiency hydrogen storage
δ2 Quadratic term of the diesel generator cost function, [e/kW2 ]
δ1 Linear term of the diesel generator cost function, [e/kW]
δ0 No-load term of the diesel generator cost function when committed, [e]
cpns Cost of not supplied energy, [e/kW]
d
Pmax Maximum output power of the diesel generator, [kW]
b
Pmax Maximum power the Li-ion battery can dispatch, [kW]
b
Smax Maximum SoC of the battery, [kWh]
S0b Initial SoC of the battery, [kWh]
h2
Pmax Maximum power the hydrogen fuel cell can dispatch, [kW]
h2
Smax Maximum SoC of the hydrogen storage, [kWh]
S0h2 initial SoC of the hydrogen storage, [kWh]
179
Energies 2020, 13, 2830
Variables:
Ptd Output power of the diesel generator at time period t, [kW]
Ptb Generated (+) and consumed (-) power from the Li-ion battery at time period t, [kW]
Pth2 Generated (+) and consumed (-) power from the hydrogen storage at time period t, [kW]
Stb State of Charge (SoC) of the Li-ion battery at time period t, [kWh]
Sth2 State of Charge (SoC) of the hydrogen storage at time period t, [kWh]
pns
Pt Not-supplied power at time period t, [kW]
Ptcurt Non-negative variable that represents the PV curtailment at time period t, [kW]
Abbreviations
The following abbreviations are used in this manuscript:
ANN Artificial Neural Network
CNN Convolutional Neural Network
DQN Deep Q-Network
DRL Deep Reinforcement Learning
GAMS General Algebraic Model System
LTS Long-Term Support
MCTS Monte Carlo Tree Search
MIQP Mixed Integer Quadratic Programming
MSE Mean Squared Error
PCC Point of Common Coupling
PV Photovoltaic Panel
RES Renewable energy sources
RL Reinforcement Learning
SoC State of Charge
pns power not supplied
180
Energies 2020, 13, 2830
where Equation (A1) is the objective function, Equation (A2) is the demand balance constraint,
Equations (A3) and (A4) model the dynamics of the battery and the hydrogen storage, respectively,
Equation (A5) establishes a minimum hydrogen storage level in the last hour, and Equation (A6)
indicates that unit commitment of the diesel generator is a binary variable. The remaining equations
impose the upper and lower bounds to the decision variables.
References
1. Ali, A.; Li, W.; Hussain, R.; He, X.; Williams, B.; Memon, A. Overview of Current Microgrid Policies, Incentives
and Barriers in the European Union, United States and China. Sustainability 2017, 9, 1146. [CrossRef]
2. Gupta, R.A.; Gupta, N.K. A robust optimization based approach for microgrid operation in deregulated
environment. Energy Convers. Manag. 2015, 93, 121–131. [CrossRef]
3. Reddy, S.; Vuddanti, S.; Jung, C.M. Review of stochastic optimization methods for smart grid. Front. Energy
2017, 11, 197–209. [CrossRef]
4. Kim, R.K.; Glick, M.B.; Olson, K.R.; Kim, Y.S. MILP-PSO Combined Optimization Algorithm for an Islanded
Microgrid Scheduling with Detailed Battery ESS Efficiency Model and Policy Considerations. Energies
2020, 13, 1898. [CrossRef]
5. François-Lavet, V.; Taralla, D.; Ernst, D.; Fonteneau, R. Deep Reinforcement Learning Solutions for
Energy Microgrids Management. In Proceedings of the European Workshop on Reinforcement Learning
(EWRL 2016), Barcelona, Spain, 3–4 December 2016.
6. Mbuwir, B.V.; Ruelens, F.; Spiessens, F.; Deconinck, G. Battery Energy Management in a Microgrid Using
Batch Reinforcement Learning. Energies 2017, 10, 1846. [CrossRef]
7. Kim, B.; Zhang, Y.; van der Schaar, M.; Lee, J. Dynamic Pricing and Energy Consumption Scheduling with
Reinforcement Learning. IEEE Trans. Smart Grid 2016, 7, 2187–2198. [CrossRef]
8. Qiu, X.; Nguyen, T.A.; Crow, M.L. Heterogeneous Energy Storage Optimization for Microgrids. IEEE Trans.
Smart Grid 2016, 7, 1453–1461. [CrossRef]
9. Glavic, M.; Fonteneau, R.; Ernst, D. Reinforcement Learning for Electric Power System Decision and Control:
Past Considerations and Perspectives. IFAC-PapersOnLine 2017, 50, 6918–6927. [CrossRef]
10. Glavic, M. (Deep) Reinforcement learning for electric power system control and related problems: A short
review and perspectives. Annu. Rev. Control 2019, 48, 22–35. [CrossRef]
181
Energies 2020, 13, 2830
11. Yang, T.; Zhao, L.; Li, W.; Zomaya, A.Y. Reinforcement learning in sustainable energy and electric systems:
A survey. Annu. Rev. Control 2020. [CrossRef]
12. Ji, Y.; Wang, J.; Xu, J.; Fang, X.; Zhang, H. Real-Time Energy Management of a Microgrid Using Deep
Reinforcement Learning. Energies 2019, 12, 2291. [CrossRef]
13. Francois-Lavet, V.; Rabusseau, G.; Pineau, J.; Ernst, D.; Fonteneau, R. On Overfitting and Asymptotic Bias in
Batch Reinforcement Learning with Partial Observability. J. Artif. Intell. Res. 2019, 65, 1–30. [CrossRef]
14. Sutton, R.S.; Barto, A.G. Reinforcement Learning: An Introduction, 2nd ed.; The MIT Press: Cambridge,
UK, 2018.
15. Arulkumaran, K.; Deisenroth, M.P.; Brundage, M.; Bharath, A.A. A Brief Survey of Deep Reinforcement
Learning. IEEE Signal Process. Mag. 2017, 34, 26–38. [CrossRef]
16. Li, Y. Deep Reinforcement Learning: An Overview. arXiv 2017, arXiv:1701.07274.
17. Francois-Lavet, V.; Henderson, P.; Islam, R.; Bellemare, M.G.; Pineau, J. An Introduction to Deep
Reinforcement Learning. Found. Trends® Mach. Learn. 2018, 11, 219–354. [CrossRef]
18. Watkins, C.J.C.H.; Dayan, P. Q-learning. Mach. Learn. 1992, 8, 279–292. [CrossRef]
19. Mnih, V.; Kavukcuoglu, K.; Silver, D.; Graves, A.; Antonoglou, I.; Wierstra, D.; Riedmiller, M. Playing Atari
with Deep Reinforcement Learning. arXiv 2013, arXiv:1312.5602.
20. Sutton, R.S. Integrated Architectures for Learning, Planning, and Reacting Based on Approximating
Dynamic Programming. In Proceedings of the Seventh International Conference on Machine Learning,
Morgan Kaufmann, Austin, TX, USA, 21–23 June 1990; pp. 216–224.
21. Chaslot, G.; Bakkes, S.; Szita, I.; Spronck, P. Monte-Carlo Tree Search: A New Framework for Game
AI. In Proceedings of the Fourth Artificial Intelligence and Interactive Digital Entertainment Conference,
Palo Alto, CA, USA, 22–24 October 2008; Mateas, M., Darken, C., Eds.; AAAI Press: Menlo Park, CA, USA,
2008; pp. 216–217.
22. Mnih, V.; Kavukcuoglu, K.; Silver, D.; Rusu, A.A.; Veness, J.; Bellemare, M.G.; Graves, A.; Riedmiller, M.;
Fidjeland, A.K.; Ostrovski, G.; et al. Human-level control through deep reinforcement learning. Nature
2015, 518, 529–533. [CrossRef] [PubMed]
23. François-Lavet, V.; Taralla, D. DeeR. 2016. Available online: http://deer.readthedocs.io (accessed on 16
April 2020).
24. François-Lavet, V.; Gemine, Q.; Ernst, D.; Fonteneau, R. Towards the Minimization of the Levelized Energy Costs
of Microgrids using Both Long-Term and Short-Term Storage Devices; CRC Press: Boca Raton, FL, USA, 2016.
25. EnergySage. EnergySage: 6 kW Solar System. 2019. Available online: http://news.energysage.com/6kw-
solar-system-compare-prices-installers (accessed on 16 April 2020).
26. Jasmin, E.A.; Imthias Ahamed, T.P.; Jagathy Raj, V.P. Reinforcement Learning approaches to Economic
Dispatch problem. Int. J. Electr. Power Energy Syst. 2011, 33, 836–845. [CrossRef]
27. Corporation, G.D. General Algebraic Modeling System (GAMS) Release 24.2.1. 2013. Available online:
http://www.gams.com/ (accessed on 16 April 2020).
28. Lin, L.J. Reinforcement Learning for Robots Using Neural Networks; Technical Report CMU-CS-93-103; School of
Computer Science, Carnegie Mellon University: Pittsburgh, PA, USA, 1993.
29. Glorot, X.; Bengio, Y. Understanding the difficulty of training deep feedforward neural networks.
In Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics,
Chia Laguna Resort, Sardinia, Italy, 13–15 May 2010; Teh, Y.W., Titterington, M., Eds.; Proceedings
of the Machine Learning Research (PMLR); 2010; Volume 9, pp. 249–256. Available online: http:
//proceedings.mlr.press/v9/glorot10a.html (accesed on 16 April 2020).
30. He, K.; Zhang, X.; Ren, S.; Sun, J. Delving Deep into Rectifiers: Surpassing Human-Level Performance on
ImageNet Classification. In Proceedings of the IEEE International Conference on Computer Vision (ICCV),
Santiago, Chile, 7–13 December 2015.
31. Dozat, T. Incorporating Nesterov Momentum into Adam. In Proceedings of the ICLR 2016 Workshop,
San Juan, Puerto Rico, 2–4 May 2016.
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
182
energies
Review
Review of Positive and Negative Impacts of Electric
Vehicles Charging on Electric Power Systems
Morsy Nour 1,2, *, José Pablo Chaves-Ávila 1 , Gaber Magdy 2 and Álvaro Sánchez-Miralles 1
1 Institute for Research in Technology (IIT), ICAI School of Engineering, Comillas Pontifical University,
28015 Madrid, Spain; [email protected] (J.P.C.-Á.); [email protected] (Á.S.-M.)
2 Department of Electrical Engineering, Faculty of Energy Engineering, Aswan University,
Aswan 81528, Egypt; [email protected]
* Correspondence: [email protected]
Abstract: There is a continuous and fast increase in electric vehicles (EVs) adoption in many countries
due to the reduction of EVs prices, governments’ incentives and subsidies on EVs, the need for energy
independence, and environmental issues. It is expected that EVs will dominate the private cars
market in the coming years. These EVs charge their batteries from the power grid and may cause
severe effects if not managed properly. On the other hand, they can provide many benefits to the
power grid and get revenues for EV owners if managed properly. The main contribution of the article
is to provide a review of potential negative impacts of EVs charging on electric power systems mainly
due to uncontrolled charging and how through controlled charging and discharging those impacts
can be reduced and become even positive impacts. The impacts of uncontrolled EVs charging on
the increase of peak demand, voltage deviation from the acceptable limits, phase unbalance due
to the single-phase chargers, harmonics distortion, overloading of the power system equipment,
and increase of power losses are presented. Furthermore, a review of the positive impacts of controlled
EVs charging and discharging, and the electrical services that it can provide like frequency regulation,
voltage regulation and reactive power compensation, congestion management, and improving power
quality are presented. Moreover, a few promising research topics that need more investigation in
future research are briefly discussed. Furthermore, the concepts and general background of EVs,
EVs market, EV charging technology, the charging methods are presented.
Keywords: electric vehicles; uncontrolled charging; delayed charging; controlled charging; V2G; V2B;
V2H; peak shaving; valley filling; congestion management; renewable energy sources
1. Introduction
Greenhouse gases (GHGs) emissions, global warming, and climate change are getting significant
attention worldwide [1]. Countries aim to diminish the use of fossil fuel which is the main reason
behind these issues. Most of fossil fuel consumption is in electricity generation and transportation
sectors [2]. In 2014, 35% of the total energy consumption was by the transportation sector [3].
Electricity generation can cut fossil fuel usage by moving to renewable energy sources (RESs) instead
of traditional nonrenewable generation and transportation sector can cut fossil fuel usage by moving to
electrified transportation. With this transformation in the transportation sector, carbon dioxide (CO2)
emissions can be reduced [4,5]. The advancement in EV technology can lower the reliance on fossil
fuels and leads to emission reduction [5]. Charging the EVs from a power grid with high share of RESs
generation contributes more to decreasing pollutant emissions [6].
Electrified transportation like metros, electric trains, trolleybuses, trams, etc., are widely used in
many countries public transportation and is considered a mature technology, but private electrified
transportation like private EVs have not had large adoption in the past due to limitations in battery
technology in terms of heavy weight, high price, short life, and long charging duration. Due to the
breakthroughs and recent developments in battery technology, EVs are proposed as a competitor to
traditional vehicles driven by gasoline due to the fast decline of EV prices [7]. Moreover, EVs produce
no pollution (locally) and have a very high energy conversion efficiency compared to traditional
vehicles because the internal combustion engine (ICE) is less efficient than the electric motor (EM).
Furthermore, several countries already initiated governmental programs to increase EVs market share
by financial subsidies and tax exemptions to motivate people to buy EVs. Additionally, many car
manufacturers produce various types of EVs that have different specifications and prices. This gives
the buyer a lot of options to choose from depending on his budget and car usage (i.e., long distance or
short distance).
These EVs can be charged from large charging stations, street chargers, workplace chargers,
and private home chargers [8], which will be supplied from the distribution network. It is expected
that this new load which draws a large amount of electric energy from the power system in a short
time will have harmful effects on the distribution network if not managed optimally and will foster
the need for large infrastructure upgrades, which are an economic burden for utility companies.
Therefore, comprehensive studies should be executed to understand and assess the characteristics of
EVs load and probable effects they may have on the electric power system, especially the distribution
networks [9], and consider this new load in the design, operation, and planning processes. Additionally,
techniques to minimize and alleviate these negative effects and for optimal integration of EVs into the
power grid should be developed [10].
A large number of research studies have investigated and assessed the impacts of uncontrolled
EV charging on distribution networks due to uncertainties related to these studies. Many uncertainties
should be considered like charging start time, the charger power rating, the charging location,
EVs battery capacity, EVs battery state of charge (SoC) [11] when started charging, penetration level,
and distribution network status. The studies assessed the impacts of uncontrolled EVs charging on
distribution networks total power demand [12], transformer loading and life [13], cable loading [14],
voltage profile [15], power losses [16], voltage and current unbalance [17], and harmonics distortion [18].
Moreover, various studies proposed solutions for optimal integration of EVs by delayed charging,
smart charging, vehicle to grid (V2G) technology, vehicle to building (V2B) technology, and vehicle
to home (V2H) technology, and how it can provide electrical services like frequency regulation,
voltage regulation and reactive power compensation, peak shaving and valley filling, integration of
RESs, spinning reserves, and improving power quality [19,20].
This paper is organized as follows: Section 2 presents a general background of EVs, EVs market,
EVs types, the benefits and challenges of EVs, and EV charging technology. Section 3 presents
the charging and discharging methods; for instance, uncontrolled charging, delayed charging,
controlled charging, V2G technology, V2B technology, and V2H technology. Section 4 presents
the negative impacts of uncontrolled EVs charging on electric power systems. Section 5 presents the
positive impacts of controlled charging and discharging of EVs on electric power systems. Section 6
presents a few promising research topics that need more investigation in future research. Finally,
conclusions are presented in Section 7.
184
Energies 2020, 13, 4675
showed better performance at that time because the energy density of fuel was better than lead–acid
batteries. Usually, EV terminology is widely used to refer to private electric cars and not to other
electrified transportation methods like trams, metros, etc.
185
Energies 2020, 13, 4675
Figure 1. Basic structure of different electric vehicles (EVs) types. (a) Hybrid Electric Vehicle (HEV);
(b) Plug-in Hybrid Electric Vehicle (PHEV); (c) Battery Electric Vehicle (BEV); and (d) Fuel Cell Electric
Vehicle (FCEV).
186
Energies 2020, 13, 4675
this type [29] depending on different companies’ designs but the basic configuration and principal
components are shown in Figure 1c. The battery is charged from the distribution network by different
kinds of EV chargers. Moreover, during deceleration and braking of EV the motor operates as a
generator and produces electric energy that charges the battery. This is called regenerative braking
which is used in other public electrified transportation like metros and trams too. To enable regenerative
braking bidirectional DC/AC converter is used. It allows electric energy to flow from battery to AC
motor in the driving mode and from the AC motor to the battery in braking mode [29,30]. The main
challenge for BEV is the long charging period and limitation of public charging infrastructure. This issue
will be tackled in the near future with the advancement in battery technology which will extend the
electric range that the EV can drive before the need to be recharged. Furthermore, the large-scale
deployment of public fast charging stations will reduce EV owners range anxiety. Moreover, in the
future, BEV manufacturers may enable battery swapping. This means replacing the empty battery
with a fully charged one in battery swapping station which needs a very short time. More details
about battery swapping are given in Section 2.4.3. This type of EVs has the largest battery capacity
and its charging can result in a severe negative impact on the power system if not managed properly.
Additionally, its ability to provide electrical services is higher than other EV types.
Advantages Disadvantages
Long range Low efficiency (15 to 35%)
Fast refueling No regenerative braking
Low cost per unit of power Pollution due to exhaust gases
Noise
EVs show many benefits compared with ICEVs but few challenges must be faced to increase EVs
adoption. Table 2 shows both the benefits and challenges of EVs [29,31]. Because both traditional
vehicles and EVs have their strengths and weaknesses, hybrid vehicles (i.e., HEVs and PHEVs) with
several configurations appeared, which try to combine the strengths of both types and avoid weaknesses.
187
Energies 2020, 13, 4675
Benefits Challenges
Emissions are localized at power stations and not in cities
Limited electric range
which improves citizens’ health
Limited number of charging stations and long charging
High efficiency because of high electric motor efficiency
(refueling) time
Negative impacts on the power grid if uncontrolled charging
Low (for PHEV) or no emissions (for BEV)
is used
Fuel or energy independence from countries which
Consumer acceptance
produce fuel
Simple transmission, fewer moving parts and low maintenance High prices
Less noise pollution Battery low energy density
Provide electrical services to the power system and helps the
Expensive battery and needs replacement after few years
RESs integration.
188
Energies 2020, 13, 4675
bidirectional chargers [34]. Unidirectional charging has simple charging hardware and allows power
flow from grid to EV only. Bidirectional charging allows power flow from the grid to EV and can inject
power from the EV battery to the grid, building, or home.
Availability and advancement of EV charging infrastructure can decrease the required onboard
energy storage and decrease the range anxiety of EV owners. According to the Society of Automotive
Engineers (SAE) Standard J1772 [35], there are three charging levels. Most of the EV owners are
expected to charge at home overnight according to the Electric Power Research Institute (EPRI) so
Level 1 and Level 2 chargers will be the primary option [36], while level 3 will be used for commercial
charging stations.
• Level 1 Charging
This is the slowest charging way and the simplest because no additional infrastructure is needed,
and any wall outlet can be used. In the US, a standard 120 V/15 A wall outlet is used for Level 1. It is
available only as an on-board charger. Although its cost is less than other charging levels, the EV needs
a long time to be fully charged. Due to its low power rating, this charging level has the lowest impacts
on distribution systems.
• Level 2 Charging
Level 2 charging uses 208 V or 240 V at currents up to 80 A, and 19.2 kW charging power.
EV owners prefer Level 2 compared with Level 1 because of its shorter charging time. It may need
dedicated electric vehicle supply equipment (EVSE) installation for public or home charging. Some EVs
like Nissan Leaf have an on-board charger of this charging level.
• Level 3 Charging
Level 3 is for fast charging and it operates as a commercial refueling station (i.e., less than 1-h
charging time) similar to the conventional gas station which can be installed at city main roads and
highways. It is supplied from a three phase circuit with 480V or higher voltages. It is available only
as an off-board charger because the charging power is high and may exceed 100 kW. It is clear that
level 3 charging is not suitable for home charging. It has a high installation cost, and this represents
a potential issue. It is expected that public chargers will use Level 2 or Level 3 for fast charging in
shopping centers, parking lots, restaurants, hotels, theaters, etc. High charging power represents an
advantage from charging time point of view, but it may generate a peak demand and overload the
distribution network equipment in addition high installation cost.
189
Energies 2020, 13, 4675
Current WC is designed for unidirectional power flow from grid to vehicle, but future development
of this technology is to enable EVs to discharge power to the grid wirelessly to provide electrical
services. The advantages of this technology are electrical safety, no cables needed, and user convenience.
The challenges of this technology are the high infrastructure cost compared to conductive charging,
and low power transfer efficiency between coils [40].
190
Energies 2020, 13, 4675
Figure 6. Total power demand profile with different charging and discharging methods. (a) Uncontrolled
charging; (b) Delayed Charging; (c) Controlled Charging; and (d) V2G, V2B, and V2H.
191
Energies 2020, 13, 4675
of charge (SoC) is 100%), similar to any electric device with a battery (i.e., laptop, smart phone, etc.).
Several studies concluded that uncontrolled charging of EVs may result in severe negative impacts on
distribution networks such as increase in peak load demand, overloading of transformers and cables
and shorten their life, increase voltage drop, increase system unbalance due to single phase chargers,
increase power losses, and increase harmonic distortion [49,50]. Moreover, this type of charging limits
the EVs acceptable penetration level, because EV owners charge their vehicles when arriving home
from work, which usually coincides with peak hours as shown in Figure 6a. The figure shows the
total power demand of residential consumers during the day [51] before connecting EVs (in grey) and
how the total power demand will change when EVs are connected to the distribution network for
charging. Furthermore, it will accelerate the need for infrastructure upgrade. This charging method is
also known as dumb charging, uncoordinated charging, and unregulated charging.
192
Energies 2020, 13, 4675
In decentralized charging control, the control entity (i.e., DSO or aggregator) sends signals to
EVs like electricity prices to incentivize them to take a specific action such as reducing their charging
power or stop charging at peak hours. In this case no need to send EVs private information to the
control entity. Although decentralized control methods need a cheaper processing and communication
infrastructure compared with centralized control methods, they have few disadvantages. The optimal
utilization of the system is not ensured, their ability to participate in ancillary service markets is limited
and they are vulnerable to variations in customers behavior [56].
In autonomous charging control, the charging power is regulated based on local inputs such
as voltage and battery SoC without any communication between EVs and control entity [57,58].
Usually, autonomous charging control is classified as decentralized control with no communication.
Although this control architecture is the single option for distribution systems with no communication
facilities and it can enable the integration of a larger number of EVs to the power grid in its current status,
it has few disadvantages. Their ability to participate in ancillary service markets is limited, they are
vulnerable to variations in customers behavior and the system cannot be operated optimally [56].
193
Energies 2020, 13, 4675
• EV battery charger
Depending on the charging infrastructure and the EV model single phase or three phase chargers
may be used for charging. Moreover, there are many power ratings for EVs chargers operating at
different voltage levels. So, the fast chargers are expected to increase the peak demand in a larger value
compared with slow chargers.
• Time
The time of connecting the EVs for charging is not certain so it is unknown how many EVs
may be charging at the same time and the probability that the EV charging time interferes with the
peak demand time of the distribution networks. Some studies assume the charging start time of EVs,
and other studies make surveys in a geographical area or a city to know the home arriving time of
vehicles and model it as probability distribution.
• Location
The place at which an EV charge is random. EV owner may charge at home, friend’s home,
street charger, or charging station. It may be different from distribution network to another where are
the charging locations and what are the EV charger type (i.e., private or commercial chargers).
• Battery capacity
There are a wide range of EV batteries capacity. PHEV usually contains a small battery capacity
while BEV contains higher battery capacity. EVs with high battery capacity will draw a larger amount
of energy from the power grid.
• Battery SoC
The battery SoC at the plug-in instant is stochastic. Many studies assume the SoC and others
consider it as a probability distribution.
194
Energies 2020, 13, 4675
• Penetration level
EVs still have a very small market share which will increase quickly for few countries like Norway
and slowly for many other countries. Therefore, many studies considered different EVs penetration
levels ranging from 10% to 100%.
There are many differences between distribution systems in terms of structure (i.e., radial or
ring), equipment loading condition before connecting EVs, voltage level (i.e., medium voltage or low
voltage), voltage profile, daily load profile of loads, etc.
Therefore, numerous studies were executed to assess the impacts of EVs charging on the power
system. Most of the studies focus on the distribution network at which EVs charging occurs, because the
most severe impacts are expected to occur at the distribution level. Classification of positive and
negative impacts of EVs charging/discharging on electric power system can be seen in Figure 7 and
they are explained in the following sections.
195
Energies 2020, 13, 4675
In [72], the impact of uncontrolled charging of EVs on total power demand for three distribution
networks (i.e., urban, suburban, and rural) with different capacities were evaluated. The study
considered penetration levels up to 25%. For the highest penetration level, the peak demand increased
by 9% for both urban and rural networks and 11% for suburban network. This increase made the rural
network exceed the maximum capacity limit while urban and suburban did not exceed the capacity
limit. In [73], it was found that uncontrolled charging increased the peak demand. The study proposed
a demand response method to control EVs demand and other flexible home devices. This method
enabled the integration of EVs without increase in peak demand from the base case (i.e., without EVs).
In a recent study [74], the impacts of uncontrolled EVs charging on Great Britain power system
was investigated. The study also considered how controlled charging can mitigate these impacts. For a
100% penetration level, it was found that the peak demand will increase by 8 GW at generation and
transmission level. Moreover, the results showed that controlled charging can significantly reduce the
increase in peak demand and eliminate the need for new generation infrastructure. At the distribution
level 100% penetration level will require 28% upgrades at distribution network. Controlled charging
can reduce the distribution network upgrade requirements to only 9%. Another interesting conclusion
in this study is that load profile at distribution level and load profile at transmission level cannot be
flattened at the same time using controlled charging. For instance, flattening the load at transmission
level (i.e., national demand) will require 19% upgrades at the distribution level. Moreover, flattening the
load at distribution level will result in 6 GW increase in national demand.
In [75,76], it was concluded that the increase in peak demand can be crucially mitigated by the
use of optimized charging and time of use (ToU) charging. A study on the Estonian grid [77] with
30% penetration level of the passenger’ cars available was conducted and the results showed an
increase in the peak load with 5% for uncontrolled charging and 4% increase for controlled charging.
Section 5.3 gives more details on how controlled charging can flatten the load profile and how controlled
charging/discharging can cut the peak demand (i.e., peak shaving).
From the previous studies, it is clear that large integration of EVs charging in uncontrolled manner
may lead to a significant increase in peak demand. This increase in peak demand will result in higher
operation costs because expensive generation must be operated at peak hours for short duration to
supply peak load. In addition, infrastructure upgrade is required if the peak demand is higher than
components capacity at generation, transmission, or distribution levels. Moreover, the use of delayed
charging and controlled charging is effective in minimizing peak demand increase due to EVs demand
and the accompanied negative impacts; this can enable higher EVs penetration.
196
Energies 2020, 13, 4675
charging of EVs on voltage drop for three distribution networks (i.e., urban, suburban, and rural) was
estimated. The study considered penetration levels up to 25%. For rural network which has long
feeders, higher voltage drop was recorded due to EVs charging which fosters the need for voltage
regulation devices. The impact of uncontrolled charging of EVs on voltage deviation in primary and
secondary distribution systems was examined in [81]. The study considered different penetration levels
(i.e., 30% and 50%), different EVs types (i.e., BEV and PHEV), and different charging levels (i.e., level 1
and level 2). Results showed no voltage limits violation at any node of primary distribution system.
However, voltage limits violations were recorded in the secondary system. Moreover, it showed that
BEV caused more voltage drop than PHEV due to its higher battery capacity and level 2 charging
caused more voltage drop than level 1. In another work [82], the effect of large-scale integration
of EVs with high power charging in IEEE 39 bus distribution system and how the distribution
network reliability was improved by optimal charging of EVs was investigated. Another method for
reducing the voltage instability was proposed in [83]; the method is based on voltage control by the
tap-changing transformer.
In [84], the impact on voltage unbalance due to uneven EVs distribution on the three phases was
investigated. The study considered two scenarios for EVs distribution on phases. For scenario A,
50% of EVs are connected to phase a, 30% to phase b, and 20% to phase c. For scenario B, 80% of EVs
were connected to phase a, 20% to phase b, and 0% to phase c. For scenario A, the voltage unbalance
factor (VUF) did not violate the limits until 50% penetration level. However, for scenario B, the VUF
reached the limits at 25% penetration level. Another study [85] investigated the impact of single phase
EVs charging (i.e., G2V) and discharging (i.e., V2G) on voltage unbalance of low voltage distribution
network. The study considered many cases and results showed that in some cases VUF exceeded the
limit for both charging and discharging modes. In [86] a significant phase unbalance occurs due to
the EVs charging with single phase chargers at level 1 at the residential network which results from
the unequal distribution of EVs chargers in the three phases. However, a small impact on voltage
and current unbalance was observed due to EVs charging in [87]. It should be mentioned that by
using controlled charging and discharging of EVs with considering VUF minimization, the voltage
unbalance at distribution network can be minimized. More details are provided in Section 5.4.
Previous studies showed that usually low EVs penetration levels can have a small impact on
voltage values even if uncontrolled charging was used. However, high EVs penetration levels can
cause high voltage drop, and voltage value may exceed the acceptable limits especially at the end
of long feeders (i.e., rural network), which will require the installation of voltage regulation devices.
Studies showed that secondary distribution networks have a higher probability of exceeding voltage
limits than primary distribution networks. Moreover, phase unbalance represents a challenging issue
since most EVs are expected to be charged by single phase private chargers. Very small EVs penetration
levels can cause small phase unbalance. However, high EVs penetration levels and high uneven
distribution of chargers on the three phases may result in VUF higher than acceptable limits.
197
Energies 2020, 13, 4675
In contrast, many studies concluded that EVs charging had a large impact on the distribution
system. In [90], the THDv increased to 11.4% due to uncontrolled rapid charging and this value
exceeded the limit of 8% based on EN 50,160 standard and a solution to the harmonic problem was
proposed by using the control of PV inverter as an active filter. In [91], the impact of EVs charging with
slow and fast charger on total harmonic distortion (THD) was assessed for different EVs models. A high
total harmonic distortion of current (THDi) was recorded for fast charging between 12% to 24%. In [92],
the impact of traditional EV charger on the quality of system voltage and drawn current was evaluated.
Traditional charger caused very high THD. The authors proposed a smart charger that draws sinusoidal
current and has a unity power factor. The smart charger significantly reduced THD compared with
traditional charger. The negative effects of EV home chargers on distribution network power quality
and transformer life were studied in [93]. The study found a quadratic relation between THDi and
life consumption of the transformer. Moreover, the study concluded that THDi should not be more
than 25 to 30% to have acceptable increase in life consumption of the transformer. In [94], the impact
of EV chargers’ current harmonics on distribution system capacity was studied. Results showed
that the 10 kV cable was overloaded at 27.25% penetration level when current harmonics due to
EV chargers was considered. However, the cable was overloaded at 30.74% penetration level when
current harmonics due to EV chargers was neglected. For harmonics reduction from EVs integration,
filters must be added to the EV chargers.
Only a few studies found that EV chargers harmonics result in a minor impact in the distribution
network. Most of the studies showed that traditional EV chargers can cause unacceptable harmonics
values. These high harmonics will result in decreasing the life cycle of distribution network components
(i.e., transformers and cables). However, by proper design of EV charger circuits, control strategy,
and filters integrated into the charger circuit, the charger harmonics can be alleviated significantly.
More details can be found in Section 5.4.
198
Energies 2020, 13, 4675
when controlled charging was used. In [98], it was found that the presence of EVs will lead to reducing
life duration of distribution transformer. In [99], it was concluded that the uncontrolled charging of
EVs will lead to aging of 25 kVA distribution transformer. Moreover, it was found that transformer
aging can be reduced using controlled charging strategies.
Another study [100], concluded that uncontrolled level 1 EVs charging has insignificant impact
on transformer life but the massive penetration of EVs may have a severe effect on transformer
lifespan. The transformer lifespan can be enhanced by using off-peak EV charging and load
management [95,100,101]. In [102] the cable loading was examined for EVs peak charging hours.
The results demonstrate that the cable can handle up to 25% penetration level for slow charging and
up to 15% penetration level for fast charging and cannot handle massive EVs penetration easily.
The high energy demand of EVs will increase the loading at different parts of the power system
(i.e., generation, transmission, and distribution). Distribution level is highly affected by EVs charging
compared to transmission and generation levels, and most of the studies focused on the impacts of EVs
charging on distribution network components. The studies showed that the acceptable EVs penetration
level before overloading of network components varies depending on the network components’
capacity and their loading condition before connecting EVs. Studies concluded that the uncontrolled
charging of EVs result in overloading of many transformers and cables at the distribution network and
can result in reducing their lifespan and requires components upgrade which represents economical
challenge to electric utilities. Results showed that many distribution networks can only allow 10%
penetration level before overloading if level 2 charging was used. The network can handle higher
penetration levels if level 1 slow charging was used. Using proper charging and discharging methods
(i.e., delayed charging, controlled charging, V2G, V2B, and V2H), distribution networks can integrate
higher EVs penetration levels before they reach their capacity limits. More details can be found in
Section 5.3.
199
Energies 2020, 13, 4675
Previous studies focused on assessing the impact of uncontrolled EVs charging on the power
losses at distribution networks. The studies concluded that uncontrolled EVs charging resulted in a
high increase in distribution network power losses. Using proper charging and discharging methods
(i.e., delayed charging, controlled charging, V2G, V2B, and V2H), power losses due to EVs charging
can be reduced significantly. Additionally, the installation of distributed generation (DG) near the
charging location can decrease the energy supplied from the grid, and hence reduce the power losses.
200
Energies 2020, 13, 4675
In [121], the authors studied two modes of EVs charging control for frequency regulation while
considering EV owner diving behavior. The first mode only controls the charging power and the
second mode controls charging and discharging power. Results showed that EVs were effective in
reducing frequency fluctuation. A coordinated control strategy between EV operating in V2G mode
and traditional generation for load frequency control (LFC) was proposed in [122]. The proposed
control strategy was tested on the Great Britain power system. The results showed the effectiveness of
the proposed strategy in improving frequency regulation and reduction of power mismatch. Moreover,
due to EV participation, the traditional power generation output variations were reduced. The EVs
ability to participate in LFC in a microgrid operating in isolated mode was investigated in [123].
The microgrid contains both renewable and nonrenewable generation. The studies proved that EVs
can enhance frequency stability in addition to reducing emissions from nonrenewable generation and
increase microgrid operator profit. In [124], EVs operating at V2G mode were used to provide LFC in a
multi-area power system containing traditional nonrenewable generation (i.e., thermal, hydro, and gas
turbines).
Previous studies results proved the feasibility of using EVs to provide frequency regulation due to
its fast response. Results showed that the use of EVs can achieve rapid control action in balancing the
generated power and the power demand during load and source variations and the system frequency
perturbation controlled by EVs is much lower than the other generation units. EVs can provide frequency
control either by regulating charging power with no V2G capability or by regulating charging and
discharging power (i.e., V2G capability). In the first approach, frequency regulation can be achieved
by only change the EVs charging power. This approach is simple, requires simple infrastructure,
and have a limited effect on battery degradation. In the second approach, frequency regulation is
provided by controlling both charging and discharging power. This approach is more effective than
the first approach. However, this approach is more complex, requires infrastructure upgrade (i.e.,
ICT and bidirectional chargers), and affect battery life cycle due to continuous charging and discharging.
The studies showed that EVs are effective in providing frequency regulation at traditional power
systems containing only nonrenewable generation. It resulted in reducing frequency fluctuation and
reducing the variation of generation units output power. Furthermore, it is effective when there
is a renewable generation installed in traditional power systems. It can reduce fluctuations due to
intermittent renewable generation, enable the integration of more renewable generation, and reduce
curtailment of renewable generation production. Moreover, EVs proved their effectiveness in providing
frequency regulation in microgrids with a high share of renewable generation. It can reduce frequency
fluctuation, increase microgrid profit, reduce curtailment from renewable generation, and enable
integrating high share of intermittent renewable generation while operating in acceptable limits.
201
Energies 2020, 13, 4675
202
Energies 2020, 13, 4675
28% compared to uncontrolled charging. The impact of uncontrolled EVs charging on total power
demand was investigated in [134]. Results showed that uncontrolled charging will increase peak
demand. Delayed charging and controlled charging were proposed for alleviating the impact of
uncontrolled charging. The results showed that the proposed methods were very efficient in shifting
the EVs charging from peak period and postponing any network upgrade.
Another study showed that the German transmission system cannot handle large scale uncontrolled
EV charging even with future expansion plans [135]. It also showed that V2G technology is effective
in reducing transmission system congestion and enhance grid stability. The effectiveness of V2G in
providing peak shaving and valley filling was proved in [136]. By allowing some of the EVs to inject
power back to the grid at peak period resulted in reducing the original peak demand (i.e., peak shaving).
By allowing EVs to charge at off peak period only resulted in valley filling. In [137], the integration
of EVs in a weak isolated grid in a Spanish island was studied. The study proposed a strategy for
coordinated charging and discharging of EVs. Results proved the effectiveness of EVs in the efficient
management of the grid, in addition to filling the valley and shaving the peak demand.
In [138], V2B was used in a building-integrated microgrid which contains EVs, battery storage,
and PV. The objective was to reduce the peak load, which will decrease subscribed power rating
and reducing the building electricity bill. The study showed that V2B can enable peak load shaving.
By encouraging more buildings to have a similar management system it can reduce the peak load
demand in the distribution system or even the whole grid. The effectiveness of V2H in reducing home
peak demand was tested in [64]. Results showed that V2H was very effective in minimizing home
peak demand and home electricity bills.
Previous studies’ results showed that uncontrolled charging of EVs will result in an increase of
peak demand at different levels of the power system (i.e., generation, transmission, and distribution)
and can overload many network components and will require components upgrade which represents
high investment costs for system operators. A more cost-effective solution to these issues is using
proper charging methods. The studies proved that simple charging method like delayed charging,
which only depends on variable electricity prices during the day, can shift most of the EVs load
to off-peak hours by motivating EV owners to decrease the charging costs. Delayed charging
can lead to reducing the increase in peak demand, reduce the number of overloaded components,
postpone infrastructure upgrades, and valley filling. A more advanced charging method is controlled
charging, where EVs can change their charging time and power depending on many variables such as
electricity prices, load demand, components loading, grid constraints, etc., depending on the proposed
control strategy. Controlled charging can lead to reduced charging costs, reduced increase in peak
demand, reduction in the number of overloaded components, delay in network components upgrade,
and valley filling. Moreover, the use of V2G, V2B, and V2B can provide more benefits to the grid and
EV owners. By injecting stored energy in EV batteries to the grid, the peak demand can be shaved
(i.e., peak shaving) and reduce power system operation costs by shutting down generation units with
high operating costs. Furthermore, they can defer the need for infrastructure upgrades, gain revenue
for EV owners by charging when the electricity price is low and discharge when the electricity price is
high (i.e., energy arbitrage), and reduce home or building electricity bills.
203
Energies 2020, 13, 4675
of EVs to minimize VUF. An unbalanced distribution network with 1.93% VUF was used to test the
proposed strategy. Uncontrolled charging of EVs increased the VUF significantly to 7.7%. The controlled
charging of EVs declined the VUF to 0.71%. Controlled charging and discharging of EVs dropped
the VUF to 0.5%. To control the voltage unbalance, the number of EVs connected at each phase
was optimized using a genetic algorithm in [142]. The authors of [143] proposed a hybrid method
combining centralized controlled charging of EVs and decentralized controlled discharging of reactive
power for voltage unbalance minimization. The proposed strategy was effective in reducing VUF
compared to uncontrolled charging case.
It was presented in Section 4.3 that EV chargers can result in significant harmonic distortion.
However, the proper design of charger power electronics circuits and control methods can minimize
or eliminate this issue. Few studies proposed EV chargers that cause no harmonic distortion or can
provide harmonic filtering. A novel control strategy applied to onboard bidirectional three phase EV
charger was proposed in [144]. The proposed method enabled the charger to draw or inject sinusoidal
current with no harmonics regardless of the power quality of the grid. Another study [144] proposed
EV charger that can operate as active filter and eliminate the harmonics caused by other loads at the
charger connection point in addition to providing reactive power compensation. In [145], the EV
charging station was used to eliminate the harmonics in distribution network by acting as active filter.
The high penetration of PV generation at distribution network can cause power quality issues due
to its intermittent nature and dependence on weather conditions. The variation of weather conditions
(i.e., clouds) results in fluctuation of PV power output which can cause voltage fluctuation and light
flicker. Therefore, DSOs need to develop mechanisms to mitigate power quality issues caused by the
fluctuation of PV output. Several studies proposed controlled charging of EVs as a possible solution to
alleviate the fluctuations caused by PV generation [146–148]. For instance, reference [148] assessed the
impacts of rapid variation of PV output on voltage fluctuation at low voltage distribution network
for three scenarios. The study proposed a controlled charging of EVs to mitigate voltage fluctuations.
The results proved the effectiveness of the proposed strategy in reducing voltage fluctuations and
accompanied light flicker.
The previous sections demonstrated that EVs can provide many benefits and services to electric
power systems. A very important secondary benefit that EVs can provide is helping the integration of
RESs. It is well known that to face environmental challenges and depletion of fossil fuel, large capacities
of RESs are being installed worldwide. Most of RESs installations are PV and wind due to the
technology advancement and fast drop in their prices. Electricity generation from RESs like PV
and wind are variable and intermittent depending on environmental conditions (i.e., sun and wind)
and hard to predict or forecast. This increases fluctuation and uncertainty in power system besides
load fluctuation [149]. Moreover, they are non-dispatchable generation and they cannot change the
generated power following demand variations like traditional power plants, which increase the need
for more flexibility in the power system which cannot be provided by central power plants only.
More flexibility in the power system can be achieved from distribution systems by demand response
(i.e., controllable or dispatchable loads) and energy storage.
Till now there is no economical utility-scale energy storage technology. EVs can act as controllable
loads by controlled charging and as energy storage by V2G, V2B, and V2H. Several studies investigated
how EVs can enable integrating more RESs [150]. EVs can absorb excess RESs generation and deliver
it to the grid when needed. In [151], it was concluded that EVs and heat pumps can reduce the cost
of integrating RESs, such as required balancing cost and required back-up generation cost. In [152],
based on the German 2030 scenario of RESs penetration, it was found that by controlling EVs charging
more RESs can be integrated into the power system. In [153], two case studies were considered for
Germany and California 2030 scenario with a high share of EVs and RESs. It concluded that the smart
charging of EVs can mitigate RESs fluctuation.
204
Energies 2020, 13, 4675
Reference [154], investigated the impact of V2G on two power systems and with different
penetration of wind generation, from 0% to 100%. It was found that V2G can enable a higher share of
wind power generation in the power system. An optimization strategy was tested in [155] to schedule
the charging of EVs which were connected to microgrid with PV and wind generation. The proposed
strategy achieved load leveling in addition to reducing microgrid operating costs and EVs charging
costs. The impact of EVs operating at controlled charging mode or V2G mode on the operation of a
power system with high share of RESs was examined in [156]. Results showed that controlled EVs
charging and discharging reduced deployed reserves usage, better usage of renewable generation by
reducing wind spillage, and reduced operation cost. In [157], it was found that controlled charging
and discharging of EVs can reduce operation costs and reduce emissions, besides maximizing RESs
utilization. In [158], an optimization algorithm was used to manage EVs charging and discharging to
increase the penetration of RESs. The proposed strategy was able to increase PV penetration up to 50%.
Figure 8. Uncontrolled charging negative impacts on the power system and how it can be mitigated
and become positive impact using the proper charging/discharging method.
As described in previous sections, there is a lot of ongoing research on the integration of EVs
to electric power system, focusing on accurate evaluation of the negative impacts of uncontrolled
EV charging or the services and benefits EVs may provide if other smart charging and discharging
methods are used. Other research areas that are being investigated by researchers are:
Most of the studies that investigate the benefits that smart charging and discharging methods
can provide focus on the technical feasibility and charging cost. However, there is a need to assess
205
Energies 2020, 13, 4675
the cost of providing these services on EV batteries degradation, because EV battery is a very
expensive component in EVs. After that, the economic feasibility of different charging and discharging
methods can be accurately estimated. Reference [159] executed a cost–benefit analysis of controlled
charging and V2G implementation considering EV batteries degradation cost. The results showed
that controlled charging is economical. Moreover, the study concluded that V2G implementation is
not economical without wind generation presence, while it is economical in the presence of wind
generation. This area requires more studies to evaluate the economic feasibility of different EV smart
charging and discharging methods on different case studies and different scenarios. The different
scenarios can consider the economic feasibility of using EVs to provide various services (i.e., frequency
regulation, voltage regulation, congestion management, etc.; or consider different power systems with
different characteristics (i.e., presence of different types of RESs, different penetration levels of RESs,
types of power plants, etc.).
• Coordination between transmission system operator (TSO) and distribution system operator
(DSO) for providing EV services
As explained in previous sections, EVs can provide many local and system-wide power and energy
services. Figure 9 summarizes the services provided to different power system parties (i.e., transmission
system operator (TSO), DSO, and loads such as buildings or homes) [160]. It is worth mentioning that
the provision of system-wide services by EVs may result in issues at distribution system at which EVs
are connected which may result in conflict of interests between TSO and DSO. For instance, the use
of EVs to provide a TSO service such as frequency regulation which requires continuous change at
charging and discharging power may cause negative impacts at the distribution system managed by
DSO such as overloading of distribution network components, phase unbalance, etc. Therefore, there is
a need for coordination between TSO and DSO to guarantee reliable and cost-efficient EVs based
services [161,162]. This topic is rarely investigated in the literature and there many open questions that
require study. Therefore, it requires more attention in future research.
With the large-scale adoption of EVs in the near future, many public EV charging stations will be
deployed in streets, highways, workplaces, shopping centers, etc. One of the main challenges for EVs
acceptance is its limited range compared to ICEVs and driver range anxiety. Therefore, the EV chargers
deployment must be planned accurately and efficiently to achieve both transportation and power system
206
Energies 2020, 13, 4675
objectives and needs. Many studies investigated the planning of future EV charging infrastructure.
The studies considered the optimal location and capacity of EV chargers [163]. Some studies considered
the transportation network only without considering power system conditions. Other studies
considered the power system only without considering transportation networks. These studies
consider power system economic and operation constraints while reducing investments required for
power system infrastructure upgrade. It is worth mentioning that EV charging stations couple both
transportation network and power system; therefore, both must be taken into account at EV charging
infrastructure planning. Few studies considered both transportation network and power system at the
planning of EV chargers [164], because EV infrastructure planning studies require real data for both
the transportation sector and power system, which varies between countries. There is a room for more
research in this area considering different case studies.
7. Conclusions
The paper presented EVs technology and the current need for it as well as its benefits compared
to traditional vehicles in addition to challenges it must tackle to achieve high adoption and social
acceptance. Furthermore, it presented the current EVs market and future predictions. Different charging
technologies were presented such as conductive charging, which is the current charging method and
other charging methods such as wireless charging and battery swapping which may have future
potential. A review of the negative impacts that EV may cause on electric power systems if uncontrolled
EV charging is used was presented. Conclusions of many studies that assessed these impacts were
discussed. All the studies showed that uncontrolled EV charging will result in unwanted negative
impacts on the power system especially the distribution networks and it will foster the need for
infrastructure upgrade. The severity of these impacts varies between different studies due to many
uncertainties in EV charging impacts studies (e.g., distribution system status, EV battery capacity,
EV battery SoC, time and location of charging, EV charger power rating, and EV penetration level).
The paper also showed that these impacts can be mitigated using delayed charging and controlled
charging methods, which can benefit both the power system and EV owner. More benefits and electrical
services can be obtained from EVs with advanced charging control methods like V2G, V2B, and V2H.
Different electrical services that can be provided with controlled charging/discharging of EVs were
presented and discussed. This review article briefly discussed the main research areas that are being
investigated for the integration of EVs to sustainable future power system, which can be helpful for
engineers and researchers. It further discussed a few interesting research topics that need more study
in future research.
Author Contributions: M.N. wrote the manuscript first draft. J.P.C.-Á., G.M., and Á.S.-M. reviewed, edited,
and suggested modifications to the manuscript. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.
Abbreviations
List of abbreviations used in this paper.
AC Alternating Current
BEV Battery Electric Vehicles
BSS Battery Swapping Stations
CO2 Carbon Dioxide
DC Direct Current
DG Distributed Generation
DSO Distribution System Operator
207
Energies 2020, 13, 4675
EM Electric Motor
EV Electric Vehicle
EVSE Electric Vehicle Supply Equipment
FCEV Fuel Cell Electric Vehicle
G2V Grid to Vehicle
GHGs Greenhouse gases
h Hour
HEV Hybrid Electric Vehicles
ICE Internal Combustion Engine
ICEV Internal Combustion Engine Vehicle
IEA International Energy Agency
IEEE Institute of Electrical and Electronics Engineers
IWC Inductive Wireless Charging
km Kilometer
kVA Kilo volt ampere
kW Kilowatt
kWh Kilowatt hour
LFC Load Frequency Control
Li-ion Lithium-ion
OLTC On-Load Tap Changer
PFC Primary Frequency Control
PHEV Plug-in Hybrid Electric Vehicles
PLDV passenger light-duty vehicles
PV Photovoltaic
RES Renewable Energy Source
SoC State of Charge
THD Total Harmonic Distortion
THDi Total Harmonic Distortion of current
THDv Total Harmonic Distortion of Voltage
ToU Time of Use
TSO Transmission System Operator
V2B Vehicle to Building
V2G Vehicle to Grid
V2H Vehicle to Home
VUF Voltage Unbalance Factor
WC Wireless Charging
References
1. UNFCCC; Conference of the Parties (COP). Adoption of the Paris Agreement Proposal by the President.
In Proceedings of the Paris Climate Change Conference, Paris, France, 30 November–12 December 2015.
2. Labatt, S.; White, R.R. Carbon Finance: The Financial Implications of Climate Change; John Wiley & Sons:
Hoboken, NJ, USA, 2007.
3. OECD. World Energy Statistics 2017; OECD: Paris, France, 2017.
4. Shaukat, N.; Khan, B.; Ali, S.; Mehmood, C.; Khan, J.; Farid, U.; Majid, M.; Anwar, S.M.; Jawad, M.; Ullah, Z.
A survey on electric vehicle transportation within smart grid system. Renew. Sustain. Energy Rev. 2018, 81,
1329–1349. [CrossRef]
5. Hedegaard, K.; Ravn, H.; Juul, N.; Meibom, P. Effects of electric vehicles on power systems in Northern
Europe. Energy 2012, 48, 356–368. [CrossRef]
6. Nanaki, E.A.; Koroneos, C.J. Comparative economic and environmental analysis of conventional, hybrid and
electric vehicles—The case study of Greece. J. Clean. Prod. 2013, 53, 261–266. [CrossRef]
7. Agency, I.E. Global EV Outlook 2019 to electric mobility. IEA Publ. 2019.
8. Knez, M.; Zevnik, G.K.; Obrecht, M. A review of available chargers for electric vehicles: United States of
America, European Union, and Asia. Renew. Sustain. Energy Rev. 2019, 109, 284–293. [CrossRef]
208
Energies 2020, 13, 4675
9. Dubey, A.; Santoso, S. Electric Vehicle Charging on Residential Distribution Systems: Impacts and Mitigations.
IEEE Access 2015, 3, 1871–1893. [CrossRef]
10. García-Villalobos, J.; Zamora, I.; Martin, J.S.; Asensio, F.J.; Aperribay, V. Plug-in electric vehicles in electric
distribution networks: A review of smart charging approaches. Renew. Sustain. Energy Rev. 2014, 38, 717–731.
[CrossRef]
11. Haus, B.; Mercorelli, P. Polynomial Augmented Extended Kalman Filter to Estimate the State of Charge of
Lithium-Ion Batteries. IEEE Trans. Veh. Technol. 2020, 69, 1452–1463. [CrossRef]
12. Nour, M.; Said, S.M.; Ali, A.; Farkas, C. Smart Charging of Electric Vehicles According to Electricity Price.
In Proceedings of the 2019 International Conference on Innovative Trends in Computer Engineering ITCE,
Aswan, Egypt, 2–4 February 2019; pp. 432–437. [CrossRef]
13. Nafisi, H. Investigation on distribution transformer loss-of-life due to plug-in hybrid electric vehicles
charging. Int. J. Ambient. Energy 2019, 1–7. [CrossRef]
14. Nour, M.; Ramadan, H.; Ali, A.; Farkas, C. Impacts of plug-in electric vehicles charging on low voltage
distribution network. In Proceedings of the 2018 International Conference on Innovative Trends in Computer
Engineering (ITCE), Aswan, Egypt, 19–21 February 2018; pp. 357–362.
15. Tong, X.; Ma, Q.; Tang, K.; Liu, H.; Li, C. Influence of electric vehicle access mode on the static voltage stability
margin and accommodated capacity of the distribution network. J. Eng. 2019, 2019, 2658–2662. [CrossRef]
16. Nour, M.; Ali, A.; Farkas, C. Mitigation of Electric Vehicles Charging Impacts on Distribution Network
with Photovoltaic Generation. In Proceedings of the 2019 International Conference on Innovative Trends in
Computer Engineering (ITCE), Aswan, Egypt, 19–21 February 2018; pp. 384–388.
17. Ul-Haq, A.; Azhar, M.; Mahmoud, Y.; Perwaiz, A.; Al-Ammar, E.A. Probabilistic Modeling of Electric Vehicle
Charging Pattern Associated with Residential Load for Voltage Unbalance Assessment. Energies 2017, 10,
1351. [CrossRef]
18. Leou, R.-C.; Teng, J.-H.; Lu, H.-J.; Lan, B.-R.; Chen, H.-T.; Hsieh, T.-Y.; Su, C.-L. Stochastic analysis of electric
transportation charging impacts on power quality of distribution systems. IET Gener. Transm. Distrib. 2018,
12, 2725–2734. [CrossRef]
19. Yong, J.Y.; Ramachandaramurthy, V.K.; Tan, K.M.; Mithulananthan, N. A review on the state-of-the-art
technologies of electric vehicle, its impacts and prospects. Renew. Sustain. Energy Rev. 2015, 49, 365–385.
[CrossRef]
20. Hota, A.R.; Juvvanapudi, M.; Bajpai, P. Issues and solution approaches in PHEV integration to the smart
grid. Renew. Sustain. Energy Rev. 2014, 30, 217–229. [CrossRef]
21. Richardson, D.B. Electric vehicles and the electric grid: A review of modeling approaches, Impacts, and
renewable energy integration. Renew. Sustain. Energy Rev. 2013, 19, 247–254. [CrossRef]
22. Talebizadeh, E.; Rashidinejad, M.; Abdollahi, A. Evaluation of plug-in electric vehicles impact on cost-based
unit commitment. J. Power Sources 2014, 248, 545–552. [CrossRef]
23. Lyon, T.P.; Michelin, M.; Jongejan, A.; Leahy, T. Is “smart charging” policy for electric vehicles worthwhile?
Energy Policy 2012, 41, 259–268. [CrossRef]
24. Weis, A.; Jaramillo, P.; Michalek, J.J. Estimating the potential of controlled plug-in hybrid electric vehicle
charging to reduce operational and capacity expansion costs for electric power systems with high wind
penetration. Appl. Energy 2014, 115, 190–204. [CrossRef]
25. Windecker, A.; Ruder, A. Fuel economy, cost, and greenhouse gas results for alternative fuel vehicles in 2011.
Transp. Res. Part. D Transp. Environ. 2013, 23, 34–40. [CrossRef]
26. Karabasoglu, O.; Michalek, J.J. Influence of driving patterns on life cycle cost and emissions of hybrid and
plug-in electric vehicle powertrains. Energy Policy 2013, 60, 445–461. [CrossRef]
27. Lunz, B.; Yan, Z.; Gerschler, J.B.; Sauer, D.U. Influence of plug-in hybrid electric vehicle charging strategies
on charging and battery degradation costs. Energy Policy 2012, 46, 511–519. [CrossRef]
28. Gass, V.; Schmidt, J.; Schmid, E. Analysis of alternative policy instruments to promote electric vehicles in
Austria. Renew. Energy 2014, 61, 96–101. [CrossRef]
29. Tie, S.F.; Tan, C.W. A review of energy sources and energy management system in electric vehicles.
Renew. Sustain. Energy Rev. 2013, 20, 82–102. [CrossRef]
30. Un-Noor, F.; Padmanaban, S.; Mihet-Popa, L.; Mollah, M.N.; Hossain, E. A Comprehensive Study of Key
Electric Vehicle (EV) Components, Technologies, Challenges, Impacts, and Future Direction of Development.
Energies 2017, 10, 1217. [CrossRef]
209
Energies 2020, 13, 4675
31. Andwari, A.M.; Pesiridis, A.; Rajoo, S.; Martinez-Botas, R.; Esfahanian, V. A review of Battery Electric Vehicle
technology and readiness levels. Renew. Sustain. Energy Rev. 2017, 78, 414–430. [CrossRef]
32. Singh, B.; Chandra, A.; Al-Haddad, K.; Pandey, A.; Kothari, D.P. A review of single-phase improved power
quality ac~dc converters. IEEE Trans. Ind. Electron. 2003, 50, 962–981. [CrossRef]
33. Singh, B.; Singh, B.; Chandra, A.; Al-Haddad, K.; Pandey, A.; Kothari, D. A Review of Three-Phase Improved
Power Quality AC–DC Converters. IEEE Trans. Ind. Electron. 2004, 51, 641–660. [CrossRef]
34. Yılmaz, M.; Krein, P.T.; Yilmaz, M. Review of charging power levels and infrastructure for plug-in electric
and hybrid vehicles. In Proceedings of the 2012 IEEE International Electric Vehicle Conference (IEVC),
Greenville, SC, USA, 4–8 March 2012; pp. 1–8. [CrossRef]
35. Young, K.; Wang, C.; Wang, L.Y.; Strunz, K. Electric Vehicle Integration into Modern Power Networks; Springer:
Berlin/Heidelberg, Germany, 2013.
36. Botsford, C.; Szczepanek, A. Fast Charging vs. Slow Charging: Pros and cons for the New Age of Electric
Vehicles. Int. Battery Hybrid Fuel Cell Electr. Veh. Symp. 2009.
37. Yang, Y.; El Baghdadi, M.; Lan, Y.; Benomar, Y.; Van Mierlo, J.; Hegazy, O. Design Methodology, Modeling,
and Comparative Study of Wireless Power Transfer Systems for Electric Vehicles. Energies 2018, 11, 1716.
[CrossRef]
38. Cao, Y.; Ahmad, N.; Kaiwartya, O.; Puturs, G.; Khalid, M. Intelligent Transportation Systems Enabled ICT
Framework for Electric Vehicle Charging in Smart City. In Handbook of Smart Cities; Springer Science and
Business Media LLC: Berlin, Germany, 2018; pp. 311–330.
39. Wang, Z.; Wei, X.; Dai, H. Design and Control of a 3 kW Wireless Power Transfer System for Electric Vehicles.
Energies 2015, 9, 10. [CrossRef]
40. Musavi, F.; Eberle, W. Overview of wireless power transfer technologies for electric vehicle battery charging.
IET Power Electron. 2014, 7, 60–66. [CrossRef]
41. Li, S.; Mi, C.C. Wireless power transfer for electric vehicle applications. IEEE J. Emerg. Sel. Top. Power Electron.
2015, 3, 4–17.
42. Ahmad, A.; Khan, Z.A.; Alam, M.S.; Khateeb, S. A Review of the Electric Vehicle Charging Techniques,
Standards, Progression and Evolution of EV Technologies in Germany. Smart Sci. 2017, 6, 36–53. [CrossRef]
43. Sarker, M.; Pandžić, H.; Ortega-Vazquez, M. Optimal operation and services scheduling for an electric vehicle
battery swapping station. In Proceedings of the 2015 IEEE Power & Energy Society General Meeting, Denver,
CO, USA, 26–30 July 2015.
44. Zheng, Y.; Dong, Z.; Xu, Y.; Meng, K.; Zhao, J.H.; Qiu, J. Electric Vehicle Battery Charging/Swap Stations in
Distribution Systems: Comparison Study and Optimal Planning. IEEE Trans. Power Syst. 2013, 29, 221–229.
[CrossRef]
45. Sarker, M.R.; Pandžić, H.; Ortega-Vazquez, M. Electric vehicle battery swapping station: Business case and
optimization model. In Proceedings of the 2013 International Conference on Connected Vehicles and Expo
(ICCVE), Las Vegas, NV, USA, 2–6 December 2013; pp. 289–294. [CrossRef]
46. Rao, R.; Zhang, X.; Xie, J.; Ju, L. Optimizing electric vehicle users’ charging behavior in battery swapping
mode. Appl. Energy 2015, 155, 547–559. [CrossRef]
47. Yang, S.; Yao, J.; Kang, T.; Zhu, X. Dynamic operation model of the battery swapping station for EV (electric
vehicle) in electricity market. Energy 2014, 65, 544–549. [CrossRef]
48. Electric Car Maker Tesla Unveils 90-s Battery Pack Swap—Reuters. Available online: https://www.reuters.com/
article/us-tesla-swap/electric-car-maker-tesla-unveils-90-s-battery-pack-swap-idUSBRE95K07H20130621
(accessed on 31 December 2019).
49. Shareef, H.; Islam, M.; Mohamed, A. A review of the stage-of-the-art charging technologies, placement
methodologies, and impacts of electric vehicles. Renew. Sustain. Energy Rev. 2016, 64, 403–420. [CrossRef]
50. Alshahrani, S.; Khalid, M.; AlMuhaini, M.M. Electric Vehicles Beyond Energy Storage and Modern Power
Networks: Challenges and Applications. IEEE Access 2019, 7, 99031–99064. [CrossRef]
51. Papathanassiou, S.; Nikos, H.; Kai, S. A benchmark low voltage microgrid network. In Proceedings of the
CIGRE Symposium: Power Systems with Dispersed Generation, Athens, Greece, 13–16 April 2005; pp. 1–8.
52. Shao, S.; Zhang, T.; Pipattanasomporn, M.; Rahman, S. Impact of TOU rates on distribution load shapes in a
smart grid with PHEV penetration. IEEE PES T&D 2010 2010, 1–6. [CrossRef]
210
Energies 2020, 13, 4675
53. Gao, Y.; Wang, C.; Wang, Z.; Liang, H. Research on time-of-use price applying to electric vehicles charging.
In Proceedings of the IEEE PES Innovative Smart Grid Technologies, Berlin, Germany, 14–17 October 2012;
pp. 1–6.
54. Tan, K.M.; Ramachandaramurthy, V.K.; Yong, J.Y. Integration of electric vehicles in smart grid: A review on
vehicle to grid technologies and optimization techniques. Renew. Sustain. Energy Rev. 2016, 53, 720–732.
[CrossRef]
55. Rahman, I.; Vasant, P.M.; Singh, M.S.B.; Abdullah-Al-Wadud, M.; Adnan, N. Review of recent trends in
optimization techniques for plug-in hybrid, and electric vehicle charging infrastructures. Renew. Sustain.
Energy Rev. 2016, 58, 1039–1047. [CrossRef]
56. Faddel, S.; Al-Awami, A.T.; Mohammed, O.A. Charge Control and Operation of Electric Vehicles in Power
Grids: A Review. Energies 2018, 11, 701. [CrossRef]
57. Faddel, S.; Mohamed, A.A.S.; Mohammed, O.A. Fuzzy logic-based autonomous controller for electric
vehicles charging under different conditions in residential distribution systems. Electr. Power Syst. Res. 2017,
148, 48–58. [CrossRef]
58. Al-Awami, A.T.; Sortomme, E.; Akhtar, G.M.A.; Faddel, S. A Voltage-Based Controller for an Electric-Vehicle
Charger. IEEE Trans. Veh. Technol. 2015, 65, 4185–4196. [CrossRef]
59. Saldaña, G.; Martin, J.S.; Zamora, I.; Asensio, F.J.; Oñederra, O. Electric Vehicle into the Grid: Charging
Methodologies Aimed at Providing Ancillary Services Considering Battery Degradation. Energies 2019, 12,
2443. [CrossRef]
60. Habib, S.; Kamran, M.; Rashid, U. Impact analysis of vehicle-to-grid technology and charging strategies of
electric vehicles on distribution networks—A review. J. Power Sources 2015, 277, 205–214. [CrossRef]
61. Thompson, A. Economic implications of lithium ion battery degradation for Vehicle-to-Grid (V2X) services.
J. Power Sources 2018, 396, 691–709. [CrossRef]
62. Pang, C.; Dutta, P.; Kezunovic, M. BEVs/PHEVs as Dispersed Energy Storage for V2B Uses in the Smart Grid.
IEEE Trans. Smart Grid 2011, 3, 473–482. [CrossRef]
63. Shin, H.; Baldick, R. Plug-In Electric Vehicle to Home (V2H) Operation Under a Grid Outage. IEEE Trans.
Smart Grid 2017, 8, 2032–2041. [CrossRef]
64. Alahyari, A.; Fotuhi-Firuzabad, M.; Rastegar, M. Incorporating Customer Reliability Cost in PEV Charge
Scheduling Schemes Considering Vehicle to Home Capability. IEEE Trans. Veh. Technol. 2014, 64, 1. [CrossRef]
65. Habib, S.; Khan, M.M.; Abbas, F.; Sang, L.; Shahid, M.U.; Tang, H. A Comprehensive Study of Implemented
International Standards, Technical Challenges, Impacts and Prospects for Electric Vehicles. IEEE Access 2018,
6, 13866–13890. [CrossRef]
66. Bohn, S.; Feustel, R.; Agsten, M. MC-based Risk Analysis on the Capacity of Distribution Grids to Charge
PEVs on 3-ph 0.4-kV Distribution Grids Considering Time and Location Uncertainties. SAE Int. J. Passeng.
Cars Electron. Electr. Syst. 2015, 8, 394–400. [CrossRef]
67. Ahmadian, A.; Mohammadi-Ivatloo, B.; Elkamel, A. A Review on Plug-in Electric Vehicles: Introduction,
Current Status, and Load Modeling Techniques. J. Mod. Power Syst. Clean Energy 2020, 8, 412–425. [CrossRef]
68. Wang, Z.; Paranjape, R. An Evaluation of Electric Vehicle Penetration under Demand Response in a
Multi-Agent Based Simulation. In Proceedings of the 2014 IEEE Electrical Power and Energy Conference,
Calgary, AB, Canada, 12–14 November 2014; pp. 220–225.
69. Putrus, G.A.; Suwanapingkarl, P.; Johnston, D.; Bentley, E.C.; Narayana, M. Impact of electric vehicles on
power distribution networks. In Proceedings of the 2009 IEEE Vehicle Power and Propulsion Conference,
Dearborn, MI, USA, 7–10 September 2009; pp. 827–831.
70. McCarthy, D.; Wolfs, P. The HV system impacts of large scale electric vehicle deployments in a metropolitan
area. In Proceedings of the AUPEC 2010—20th Australasian Universities Power Engineering Conference
2010, Christchurch, New Zealand, 5 December 2010.
71. Schneider, K.P.; Gerkensmeyer, C.; Kintner-Meyer, M.; Fletcher, R. Impact assessment of plug-in hybrid
vehicles on pacific northwest distribution systems. In Proceedings of the 2008 IEEE Power and Energy
Society General Meeting—Conversion and Delivery of Electrical Energy in the 21st Century, Pittsburgh, PA,
USA, 20–24 July 2008; pp. 1–6. [CrossRef]
72. Kelly, L.; Rowe, A.; Wild, P. Analyzing the impacts of plug-in electric vehicles on distribution networks in
British Columbia. In Proceedings of the 2011 IEEE Electrical Power and Energy Conference, Montreal, QC,
Canada, 22–23 October 2009; Institute of Electrical and Electronics Engineers (IEEE); pp. 1–6.
211
Energies 2020, 13, 4675
73. Shao, S.; Pipattanasomporn, M.; Rahman, S. Grid Integration of Electric Vehicles and Demand Response
with Customer Choice. IEEE Trans. Smart Grid 2012, 3, 543–550. [CrossRef]
74. Crozier, C.; Morstyn, T.; McCulloch, M. The opportunity for smart charging to mitigate the impact of electric
vehicles on transmission and distribution systems. Appl. Energy 2020, 268, 114973. [CrossRef]
75. Kristoffersen, T.K.; Capion, K.; Meibom, P. Optimal charging of electric drive vehicles in a market environment.
Appl. Energy 2011, 88, 1940–1948. [CrossRef]
76. Denholm, P.; Short, W. Evaluation of Utility System Impacts and Benefits of Optimally Dispatched Plug-In Hybrid
Electric Vehicles; NREL: Golden, CO, USA, 2006.
77. Drovtar, I.; Rosin, A.; Landsberg, M.; Kilter, J. Large scale electric vehicle integration and its impact on the
Estonian power system. In Proceedings of the 2013 IEEE Grenoble Conference, Grenoble, France, 16–20 June
2013; Institute of Electrical and Electronics Engineers (IEEE); pp. 1–6.
78. Van Cutsem, T. Voltage instability: Phenomena, countermeasures, and analysis methods. Proc. IEEE 2000,
88, 208–227. [CrossRef]
79. Dharmakeerthi, C.; Mithulananthan, N.; Saha, T.K. Impact of electric vehicle fast charging on power system
voltage stability. Int. J. Electr. Power Energy Syst. 2014, 57, 241–249. [CrossRef]
80. Clement-Nyns, K.; Haesen, E.; Driesen, J. The Impact of Charging Plug-In Hybrid Electric Vehicles on a
Residential Distribution Grid. IEEE Trans. Power Syst. 2009, 25, 371–380. [CrossRef]
81. Gray, M.K.; Morsi, W.G. Power Quality Assessment in Distribution Systems Embedded With Plug-In Hybrid
and Battery Electric Vehicles. IEEE Trans. Power Syst. 2014, 30, 663–671. [CrossRef]
82. Xiong, J.; Zhang, K.; Liu, X.; Su, W. Investigating the impact of plug-in electric vehicle charging on power
distribution systems with the integrated modeling and simulation of transportation network. In Proceedings
of the ITEC Asia-Pacific 2014—IEEE Transportation Electrification Conference & Expo 2014, Beijing, China,
31 August–3 September 2014; pp. 1–5. [CrossRef]
83. Soroudi, A.; Keane, A. Plug in Electric Vehicles in Smart Grids; Springer: Berlin/Heidelberg, Germany, 2015;
p. 89.
84. Ul-Haq, A.; Cecati, C.; Strunz, K.; Abbasi, E. Impact of Electric Vehicle Charging on Voltage Unbalance in an
Urban Distribution Network. Intell. Ind. Syst. 2015, 1, 51–60. [CrossRef]
85. Shahnia, F.; Ghosh, A.; Ledwich, G.; Zare, F. Predicting Voltage Unbalance Impacts of Plug-in Electric
Vehicles Penetration in Residential Low-voltage Distribution Networks. Electr. Power Compon. Syst. 2013, 41,
1594–1616. [CrossRef]
86. Richardson, P.; Flynn, D.; Keane, A. Impact assessment of varying penetrations of electric vehicles
on low voltage distribution systems. In Proceedings of the 2008 IEEE Power and Energy Society
General Meeting—Conversion and Delivery of Electrical Energy in the 21st Century, Providence, RI,
USA, 25–29 July 2010; pp. 1–6. [CrossRef]
87. Liu, R.; Dow, L.; Liu, E. A survey of PEV impacts on electric utilities. In Proceedings of the Innovative Smart
Grid Technologies (ISGT), 2011 IEEE PES, Perth, Australia, 13–16 November 2011.
88. Jiang, C.; Torquato, R.; Salles, D.; Xu, W. Method to assess the power quality impact of plug-in electric
vehicles. In Proceedings of the 2014 16th International Conference on Harmonics and Quality of Power,
Bucharest, Romania, 25–28 May 2014; pp. 177–180. [CrossRef]
89. Bass, R.; Harley, R.; Lambert, F.; Rajasekaran, V.; Pierce, J. Residential harmonic loads and EV charging.
In Proceedings of the 2001 IEEE Power Engineering Society Winter Meeting Conference Proceedings (Cat No
01CH37194), Columbus, OH, USA, 28 January–1 February 2001; pp. 803–808. [CrossRef]
90. Nguyen, V.-L.; Tran-Quoc, T.; Bacha, S. Harmonic distortion mitigation for electric vehicle fast charging
systems. In Proceedings of the 2013 IEEE Grenoble Conference, Grenoble, France, 16–20 June 2013; pp. 1–6.
91. Melo, N.; Mira, F.; De Almeida, A.; Delgado, J. Integration of PEV in Portuguese distribution grid: Analysis
of harmonic current emissions in charging points. In Proceedings of the 11th International Conference on
Electrical Power Quality and Utilisation, Lisbon, Portugal, 17–19 October 2011; pp. 1–6. [CrossRef]
92. Monteiro, V.; Gonçalves, H.; Afonso, J.L. Impact of Electric Vehicles on power quality in a Smart Grid context.
In Proceedings of the 11th International Conference on Electrical Power Quality and Utilisation, Lisbon,
Portugal, 17–19 October 2011; pp. 1–6. [CrossRef]
93. Gomez, J.; Morcos, M. Impact of EV battery chargers on the power quality of distribution systems. IEEE Trans.
Power Deliv. 2003, 18, 975–981. [CrossRef]
212
Energies 2020, 13, 4675
94. Zhou, C.; Wang, H.; Zhou, W.; Qian, K.; Meng, S. Determination of maximum level of EV penetration with
consideration of EV charging load and harmonic currents. IOP Conf. Ser. Earth Environ. Sci. 2019, 342,
012010. [CrossRef]
95. Hilshey, A.D.; Rezaei, P.; Hines, P.D.H.; Frolik, J. Electric vehicle charging: Transformer impacts and smart,
decentralized solutions. In Proceedings of the 2012 IEEE Power Energy Society General Meeting, San Diego,
CA, USA, 22–26 July 2012; pp. 1–8. [CrossRef]
96. Papadopoulos, P.; Skarvelis-Kazakos, S.; Grau, I.; Awad, B.; Cipcigan, L.M.; Jenkins, N. Impact of residential
charging of electric vehicles on distribution networks, a probabilistic approach. In Proceedings of the 45th
International Universities Power Engineering Conference, Cardiff, UK, 31 August–3 September 2010; pp. 1–5.
97. Verzijlbergh, R.A.; Lukszo, Z.; Slootweg, J.G.; Ilic, M.D. The impact of controlled electric vehicle charging
on residential low voltage networks. In Proceedings of the 2011 International Conference on Networking,
Sensing and Control, Delft, The Netherlands, 11–13 April 2011; pp. 14–19. [CrossRef]
98. Turker, H.; Chatroux, D.; Hably, A.; Bacha, S. Low-Voltage Transformer Loss-of-Life Assessments for a High
Penetration of Plug-In Hybrid Electric Vehicles (PHEVs). IEEE Trans. Power Deliv. 2012, 27, 1323–1331.
[CrossRef]
99. Hilshey, A.; Hines, P.; Rezai, P.; Dowds, J. Estimating the impact of electric vehicle smart charging on
distribution transformer aging. In Proceedings of the 2013 IEEE Power Energy Society General Meeting,
Vancouver, BC, Canada, 21–25 July 2013; p. 1. [CrossRef]
100. Razeghi, G.; Zhang, L.; Brown, T.; Samuelsen, S. Impacts of plug-in hybrid electric vehicles on a residential
transformer using stochastic and empirical analysis. J. Power Sources 2014, 252, 277–285. [CrossRef]
101. Yan, Q.; Kezunovic, M. Impact analysis of Electric Vehicle charging on distribution system. In Proceedings of
the 2012 North American Power Symposium (NAPS), Champaign, IL, USA, 9-11 September 2012; pp. 1–6.
102. Akhavan-Rezai, E.; Shaaban, M.F.; El-Saadany, E.F.; Zidan, A. Uncoordinated charging impacts of electric
vehicles on electric distribution grids: Normal and fast charging comparison. In Proceedings of the 2012
IEEE Power and Energy Society General Meeting, San Diego, CA, USA, 22–26 July 2012; pp. 1–7.
103. Fernandez, L.P.; GomezSanRoman, T.; Cossent, R.; Domingo, C.M.; Frias, P. Assessment of the Impact of
Plug-in Electric Vehicles on Distribution Networks. IEEE Trans. Power Syst. 2010, 26, 206–213. [CrossRef]
104. Pillai, J.R.; Bak-Jensen, B. Impacts of electric vehicle loads on power distribution systems. In Proceedings of
the 2010 IEEE Vehicle Power and Propulsion Conference, Lille, France, 1–3 September 2010; pp. 1–6.
105. Masoum, M.A.; Moses, P.S.; Smedley, K.M. Distribution transformer losses and performance in smart grids with
residential Plug-In Electric Vehicles. In Proceedings of the ISGT 2011, Anaheim, CA, USA, 17–19 January 2011.
106. Deb, S.; Tammi, K.; Kalita, K.; Mahanta, P. Impact of Electric Vehicle Charging Station Load on Distribution
Network. Energies 2018, 11, 178. [CrossRef]
107. Chang, M.; Bae, S.; Yoon, G.-G.; Park, S.-H.; Choy, Y. Impact of Electric Vehicle Charging Demand on a Jeju
Island Radial Distribution Network. In Proceedings of the 2019 IEEE Power & Energy Society Innovative
Smart Grid Technologies Conference (ISGT), Bucharest, Romania, 29 September–2 October 2019; pp. 1–5.
108. Tomić, J.; Kempton, W. Using fleets of electric-drive vehicles for grid support. J. Power Sources 2007, 168,
459–468. [CrossRef]
109. Pavić, I.; Capuder, T.; Kuzle, I. Value of flexible electric vehicles in providing spinning reserve services.
Appl. Energy 2015, 157, 60–74. [CrossRef]
110. Izadkhast, S.D. Aggregation of Plug-in Electric Vehicles in Power Systems for Primary Frequency Control.
Delft Univ. Technol. Comillas Pontif. Univ. KTH R. Inst. 2017, 240. [CrossRef]
111. Yang, Z.; Li, K.; Foley, A.; Guo, Y. Computational scheduling methods for integrating plug-in electric vehicles
with power systems: A review. Renew. Sustain. Energy Rev. 2015, 51, 396–416. [CrossRef]
112. Prabha, K. Power System Stability and Control; McGraw-Hill Education: New York, NY, USA, 1994; p. 1176.
113. Babahajiani, P.; Shafiee, Q.; Bevrani, H. Intelligent Demand Response Contribution in Frequency Control of
Multi-Area Power Systems. IEEE Trans. Smart Grid 2016, 9, 1282–1291. [CrossRef]
114. Ulbig, A.; Borsche, T.; Andersson, G. Impact of Low Rotational Inertia on Power System Stability and Operation;
Elsevier: Amsterdam, The Netherlands, 2014; Volume 47, pp. 7290–7297.
115. Rodrigues, E.M.G.; Osório, G.; Godina, R.; Bizuayehu, A.; Lujano-Rojas, J.; Catalão, J. Grid code reinforcements
for deeper renewable generation in insular energy systems. Renew. Sustain. Energy Rev. 2016, 53, 163–177.
[CrossRef]
213
Energies 2020, 13, 4675
116. Cabrera-Tobar, A.; Bullich-Massagué, E.; Aragues-Penalba, M.; Gomis-Bellmunt, O. Review of advanced
grid requirements for the integration of large scale photovoltaic power plants in the transmission system.
Renew. Sustain. Energy Rev. 2016, 62, 971–987. [CrossRef]
117. Marinelli, M.; Martinenas, S.; Knezović, K.; Andersen, P.B. Validating a centralized approach to primary
frequency control with series-produced electric vehicles. J. Energy Storage 2016, 7, 63–73. [CrossRef]
118. Mu, Y.; Wu, J.; Ekanayake, J.; Jenkins, N.; Jia, H. Primary Frequency Response From Electric Vehicles in the
Great Britain Power System. IEEE Trans. Smart Grid 2012, 4, 1142–1150. [CrossRef]
119. Wang, L.; Chen, B. Dual-level consensus-based frequency regulation using vehicle-to-grid service. Electr. Power
Syst. Res. 2019, 167, 261–276. [CrossRef]
120. Almeida, P.R.; Soares, F.J.; Lopes, J.P. Electric vehicles contribution for frequency control with inertial
emulation. Electr. Power Syst. Res. 2015, 127, 141–150. [CrossRef]
121. Meng, J.; Mu, Y.; Jia, H.; Wu, J.; Yu, X.; Qu, B. Dynamic frequency response from electric vehicles considering
travelling behavior in the Great Britain power system. Appl. Energy 2016, 162, 966–979. [CrossRef]
122. Jia, H.; Li, X.; Mu, Y.; Xu, C.; Jiang, Y.; Yu, X.; Wu, J.; Dong, C. Coordinated control for EV aggregators and
power plants in frequency regulation considering time-varying delays. Appl. Energy 2018, 210, 1363–1376.
[CrossRef]
123. Vahedipour-Dahraie, M.; Rashidizadeh-Kermani, H.; Ghamsari-Yazdel, M.; Khaloie, H.; Guerrero, J.M.
Coordination of EVs Participation for Load Frequency Control in Isolated Microgrids. Appl. Sci. 2017, 7, 539.
[CrossRef]
124. Debbarma, S.; Dutta, A. Utilizing Electric Vehicles for LFC in Restructured Power Systems Using Fractional
Order Controller. IEEE Trans. Smart Grid 2016, 8, 2554–2564. [CrossRef]
125. Hajforoosh, S.; Masoum, M.A.; Islam, S. Online optimal variable charge-rate coordination of plug-in electric
vehicles to maximize customer satisfaction and improve grid performance. Electr. Power Syst. Res. 2016, 141,
407–420. [CrossRef]
126. Nour, M.; Said, S.M.; Ramadan, H.; Ali, A.; Farkas, C. Control of Electric Vehicles Charging Without
Communication Infrastructure. In Proceedings of the 2018 Twentieth International Middle East Power
Systems Conference (MEPCON), Cairo, Egypt, 18–20 December 2018; pp. 773–778.
127. Ramadan, H.; Ali, A.; Nour, M.; Farkas, C. Smart Charging and Discharging of Plug-in Electric Vehicles
for Peak Shaving and Valley Filling of the Grid Power. In Proceedings of the 2018 Twentieth International
Middle East Power Systems Conference (MEPCON), Cairo, Egypt, 18–20 December 2018; pp. 735–739.
128. Marra, F.; Yang, G.Y.; Fawzy, Y.T.; Traeholt, C.; Larsen, E.; Garcia-Valle, R.; Jensen, M.M. Improvement
of Local Voltage in Feeders with Photovoltaic Using Electric Vehicles. IEEE Trans. Power Syst. 2013, 28,
3515–3516. [CrossRef]
129. Kisacikoglu, M.C.; Ozpineci, B.; Tolbert, L.M. Examination of a PHEV bidirectional charger system for
V2G reactive power compensation. In Proceedings of the 2010 Twenty-Fifth Annual IEEE Applied Power
Electronics Conference and Exposition (APEC) 2010, Palm Springs, CA, USA, 21–25 February 2010; pp. 458–465.
[CrossRef]
130. Yong, J.Y.; Ramachandaramurthy, V.K.; Tan, K.M.; Mithulananthan, N. Bi-directional electric vehicle fast
charging station with novel reactive power compensation for voltage regulation. Int. J. Electr. Power
Energy Syst. 2015, 64, 300–310. [CrossRef]
131. Azzouz, M.A.; Shaaban, M.F.; El-Saadany, E.F. Real-Time Optimal Voltage Regulation for Distribution
Networks Incorporating High Penetration of PEVs. IEEE Trans. Power Syst. 2015, 30, 1–12. [CrossRef]
132. Nour, M.; Ali, A.; Farkas, C. Evaluation of Electric Vehicles Charging Impacts on A Real Low Voltage Grid.
Int. J. Power Eng. Energy 2018, 1–6.
133. Zhang, K.; Xu, L.; Ouyang, M.; Wang, H.; Lu, L.; Li, J.; Li, Z. Optimal decentralized valley-filling charging
strategy for electric vehicles. Energy Convers. Manag. 2014, 78, 537–550. [CrossRef]
134. Su, J.; Lie, T.; Zamora, R. Modelling of large-scale electric vehicles charging demand: A New Zealand case
study. Electr. Power Syst. Res. 2019, 167, 171–182. [CrossRef]
135. Staudt, P.; Schmidt, M.; Gärttner, J.; Weinhardt, C. A decentralized approach towards resolving transmission
grid congestion in Germany using vehicle-to-grid technology. Appl. Energy 2018, 230, 1435–1446. [CrossRef]
136. Wang, Z. Grid Power Peak Shaving and Valley Filling Using Vehicle-to-Grid Systems. IEEE Trans. Power Deliv.
2013, 28, 1822–1829. [CrossRef]
214
Energies 2020, 13, 4675
137. Colmenar-Santos, A.; Linares-Mena, A.-R.; Borge-Diez, D.; Quinto-Alemany, C.-D. Impact assessment of
electric vehicles on islands grids: A case study for Tenerife (Spain). Energy 2017, 120, 385–396. [CrossRef]
138. Dagdougui, H.; Ouammi, A.; Dessaint, L.A. Peak Load Reduction in a Smart Building Integrating Microgrid
and V2B-Based Demand Response Scheme. IEEE Syst. J. 2019, 13, 3274–3282. [CrossRef]
139. Martinenas, S.; Knezović, K.; Marinelli, M. Management of Power Quality Issues in Low Voltage Networks
using Electric Vehicles: Experimental Validation. IEEE Trans. Power Deliv. 2016, 32, 1. [CrossRef]
140. Akhavan-Rezai, E.; Shaaban, M.F.; El-Saadany, E.F.; Karray, F. Managing Demand for Plug-in Electric Vehicles
in Unbalanced LV Systems with Photovoltaics. IEEE Trans. Ind. Inform. 2017, 13, 1057–1067. [CrossRef]
141. Farahani, H.F. Improving voltage unbalance of low-voltage distribution networks using plug-in electric
vehicles. J. Clean. Prod. 2017, 148, 336–346. [CrossRef]
142. Jiménez, A.; García, N. Voltage unbalance analysis of distribution systems using a three-phase power flow
ans a Genetic Algorithm for PEV fleets scheduling. In Proceedings of the 2012 IEEE Power Energy Society
General Meeting 2012, San Diego, CA, USA, 22–26 July 2012; pp. 1–8. [CrossRef]
143. Jabalameli, N.; Su, X.; Ghosh, A. Online Centralized Charging Coordination of PEVs With Decentralized
Var Discharging for Mitigation of Voltage Unbalance. IEEE Power Energy Technol. Syst. J. 2019, 6, 152–161.
[CrossRef]
144. Gallardo-Lozano, J.; Milanés-Montero, M.I.; Guerrero-Martínez, M.A.; Romero-Cadaval, E. Electric vehicle
battery charger for smart grids. Electr. Power Syst. Res. 2012, 90, 18–29. [CrossRef]
145. Rauchfuss, L.; Foulquier, J.; Werner, R. Charging station as an active filter for harmonics compensation
of smart grid. In Proceedings of the 2014 16th International Conference on Harmonics and Quality of
Power (ICHQP), Bucharest, Romania, 5–28 May 2014; Institute of Electrical and Electronics Engineers (IEEE);
pp. 181–184.
146. Ali, A.; Raisz, D.; Mahmoud, K. Voltage fluctuation smoothing in distribution systems with RES considering
degradation and charging plan of EV batteries. Electr. Power Syst. Res. 2019, 176, 105933. [CrossRef]
147. García-Villalobos, J.; Zamora, I.; Knezović, K.; Marinelli, M. Multi-objective optimization control of plug-in
electric vehicles in low voltage distribution networks. Appl. Energy 2016, 180, 155–168. [CrossRef]
148. Brinkel, N.B.G.; Gerritsma, M.K.; AlSkaif, T.A.; Lampropoulos, I.; van Voorden, A.M.; Fidder, H.A.;
van Sark, W.G.J.H.M. Impact of rapid PV fluctuations on power quality in the low-voltage grid and mitigation
strategies using electric vehicles. Int. J. Electr. Power Energy Syst. 2020, 118, 105741. [CrossRef]
149. Magdy, G.; Nour, M.; Shabib, G.; Elbaset, A.A.; Mitani, Y. Supplementary Frequency Control in a
High-penetration Real Power System by Renewables Using SMES Application. J. Electr. Syst. 2019,
15, 526–538.
150. Liu, L.; Kong, F.; Liu, X.; Peng, Y.; Wang, Q. A review on electric vehicles interacting with renewable energy
in smart grid. Renew. Sustain. Energy Rev. 2015, 51, 648–661. [CrossRef]
151. Teng, F.; Aunedi, M.; Strbac, G. Benefits of flexibility from smart electrified transportation and heating in the
future UK electricity system. Appl. Energy 2016, 167, 420–431. [CrossRef]
152. Dallinger, D.; Wietschel, M. Grid integration of intermittent renewable energy sources using price-responsive
plug-in electric vehicles. Renew. Sustain. Energy Rev. 2012, 16, 3370–3382. [CrossRef]
153. Dallinger, D.; Gerda, S.; Wietschel, M. Integration of intermittent renewable power supply using
grid-connected vehicles—A 2030 case study for California and Germany. Appl. Energy 2013, 104, 666–682.
[CrossRef]
154. Lund, H.; Kempton, W. Integration of renewable energy into the transport and electricity sectors through
V2G. Energy Policy 2008, 36, 3578–3587. [CrossRef]
155. Li, Y.; Li, K. Incorporating Demand Response of Electric Vehicles in Scheduling of Isolated Microgrids With
Renewables Using a Bi-Level Programming Approach. IEEE Access 2019, 7, 116256–116266. [CrossRef]
156. Carrión, M.; Zárate-Miñano, R. Operation of renewable-dominated power systems with a significant
penetration of plug-in electric vehicles. Energy 2015, 90, 827–835. [CrossRef]
157. Saber, A.Y.; Venayagamoorthy, G.K.; Member, S. Plug-in Vehicles and Renewable Energy Sources for Cost
and Emission Reductions Plug-in Vehicles and Renewable Energy Sources for Cost and Emission Reductions.
IEEE Trans. Ind. Electron. 2016, 58, 1229–1238. [CrossRef]
158. Kordkheili, R.A.; Pourmousavi, S.A.; Savaghebi, M.; Guerrero, J.M.; Nehrir, M.H. Assessing the Potential of
Plug-in Electric Vehicles in Active Distribution Networks. Energies 2016, 9, 34. [CrossRef]
215
Energies 2020, 13, 4675
159. Ahmadian, A.; Sedghi, M.; Mohammadi-Ivatloo, B.; Elkamel, A.; Golkar, M.A.; Fowler, M. Cost-Benefit
Analysis of V2G Implementation in Distribution Networks Considering PEVs Battery Degradation. IEEE Trans.
Sustain. Energy 2017, 9, 961–970. [CrossRef]
160. Arias, N.B.; Hashemi, S.; Andersen, P.B.; Traeholt, C.; Romero, R. Distribution System Services Provided by
Electric Vehicles: Recent Status, Challenges, and Future Prospects. IEEE Trans. Intell. Transp. Syst. 2019, 20,
4277–4296. [CrossRef]
161. Zecchino, A.; Knezović, K.; Marinelli, M. Identification of conflicts between transmission and distribution system
operators when acquiring ancillary services from electric vehicles. In Proceedings of the 2017 IEEE PES Innovative
Smart Grid Technologies Conference Europe (ISGT-Europe), Torino, Italy, 26–29 September 2017.
162. Gerard, H.; Puente, E.I.R.; Six, D. Coordination between transmission and distribution system operators in
the electricity sector: A conceptual framework. Util. Policy 2018, 50, 40–48. [CrossRef]
163. Zeb, M.Z.; Imran, K.; Khattak, A.; Janjua, A.K.; Pal, A.; Nadeem, M.; Zhang, J.; Khan, S. Optimal Placement
of Electric Vehicle Charging Stations in the Active Distribution Network. IEEE Access 2020, 8, 68124–68134.
[CrossRef]
164. Zhang, H.; Moura, S.J.; Hu, Z.; Song, Y. PEV Fast-Charging Station Siting and Sizing on Coupled Transportation
and Power Networks. IEEE Trans. Smart Grid 2018, 9, 2595–2605. [CrossRef]
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
216
MDPI
St. Alban-Anlage 66
4052 Basel
Switzerland
Tel. +41 61 683 77 34
Fax +41 61 302 89 18
www.mdpi.com