Smart Campus Energy Management System: Advantages, Architectures, and the

Impact of using Cloud Computing

Hanaa TALEI* Mohamed ESSAAIDI

School of Science and Engineering ENSIAS School
AlAkhawayn University in Ifrane Rabat, Morocco
Ifrane, Morocco [email protected]
[email protected]

Driss BENHADDOU
Engineering Technology Department
University of Houston
Houston, TX, USA
[email protected]

Abstract—A Microgrid is a subset of smart grid, a small-scale Microgrids enable the implementation of energy
electrical system powered with renewable energy resources that generators at the low voltage distribution; they are energy
can operate either in a connected or a disconnected mode to/from
systems able to work both in connected or islanded modes
the main grid. Given that universities are very important
to/from the main power grid. Another novel aspect of smart
electricity consumers, using an academic Microgrid will solve
many problems regarding energy usage[1]. The purpose of this grid is that it strives toward supplying energy using renewable
paper, is to describe Microgrid components with an emphasis on distributed energy resources (DER). These DERs can be
energy management (EMS). Given its vital role, the paper implemented at the customer site allowing them to actively
presents different architectures of an EMS and discusses how participate in the production of energy. A mere
cloud computing can be incorporated to the Microgrid implementation of DER in a home for instance does not
architecture to improve the EMS efficiency. The paper necessarily make the home a Microgrid. To become a
concludes with presenting results of EMS data collection using Microgrid, a customer should implement enough DER to be
Kaa IoT platform.
self-sufficient and be able to operate independently from the
Keywords—Microgrid, Energy Management System, Agent, power grid in certain conditions.
Cloud Computing, IoT.
An example of a Microgrid is a smart campus: a
university campus that implements different types of DERs
[23]. However the power supplied by DERs can’t be ensured
I. INTRODUCTION continuously; the new Microgrid needs a scheduling
Gusage
iven the importance of the electrical grid and the electricity
increase rate, various research projects are carried
management software, called Energy Management System
(EMS), which will manage the DERs for the smart campus to
out to improve the current electrical system. Smart grid, as a use energy efficiently [2]. EMS functions include economic
leading project, is a cyber-physical system aiming at dispatch algorithms, monitoring, metering, control, protection,
improving the current power grid by integrating many and stability and power quality analysis. These functions are
information and communication technologies (ICT) features implemented at different levels of Microgrids depending on
to monitor and control how electricity is generated, their timing requirements.
transmitted, and distributed. Among the novel aspects of smart
Cloud computing has been considered as platform to
grid is the use of Microgrids.
implement the EMS functions of smart grid [10], however, the
Permission to make digital or hard copies of part or all of this work for personal or impact of cloud computing on implementing different EMS
classroom use is granted without fee provided that copies are not made or distributed for
profit or commercial advantage and that copies bear this notice and the full citation on function is not well studied. The purpose of this paper is to
the first page. Copyrights for components of this work owned by others than ACM must present an empirical study of a Cloud-based EMS on a
be honored. To copy otherwise, distribute, republish, or post, requires prior specific
permission and/or a fee. Request permissions from [email protected]. Microgrid requirements. The paper put an emphasis on
ICSDE '17, July 21–23, 2017, Rabat, Morocco
© 2017 Association for Computing Machinery.
ACM ISBN 978-1-4503-5281-9/17/07…$15.00
https://doi.org/10.1145/3128128.3128129
implementing its architecture using Kaa platform and study • Energy Management System (EMS): The processor of
the impact of delay on different EMS functions in the cloud. the Microgrid orchestrating its operations by coordinating
DESs, DGs and the loads to come out with a reliable,
The paper is structured as follows: Section II describes economic, sustainable, environmentally friendly energy
the relationship between a Microgrid and EMS. Section III self-productive system.
presents various EMS architectures. The impact of adding • Advanced IT Infrastructure: this consists of a set of
cloud computing to the system is presented in Section IV. sensors, meters, software that will collect and process
Section V will propose an EMS Kaa-based architecture and real-time data for the user to make some decisions such
the performance evaluation, finally conclusions and future has turning off an air conditioner when a sensor detects
work are presented in Section VI. no motion in the house.

The relationship between a Microgrid and an EMS is a

II. ENERGY MANAGEMENT SYSTEMS AND MICROGRID: mutual one: this fundamental computer based software will
WHAT RELATIONSHIP guarantee the following Microgrid characteristics: scheduling
generators using a forecasting process, adjusting voltage,
meeting the demand load, guarantying a smooth switch
CERTS, Consortium for Electric Reliability Technology
between islanded and connected modes from/to main grid,
Solutions, defined a Microgrid as a self-productive electricity
minimizing emissions, and maximizing the Microgrid
system able to connect and disconnect itself from the main
utilization efficiently[6].
grid seamlessly without causing any disturbance to the loads
within the system[2]. Although not universal, [3], [4] and [5] III. ENERGY MANAGEMENT SYSTEMS ARCHITECTURES: A
presents the main components of a Microgrid as shown in REVIEW
figure1:
The use of Microgrids will lead to a revolution in the
current electrical system: a system suffering from various
problems such as blackouts that can often lead to huge
damages. For instance, in 2011, Japan had a 9.0 magnitude
devastating earthquake that damaged the district power
supply. Despite the miserable city situation, Sendai Microgrid
provided hospitals and nursing home residents with power for
three consecutive days[7].

For a Microgrid to sustainably maintain its functions, it

needed a system to supervise its operations: an Energy
Management System that will coordinates the operations of
Figure 1: Microgrid Main Components the Microgrid components listed above: generators, storage,
controls, and coupling point. To understand how an EMS
• Distributed Generators (DSs): Consists of power work, figure 2 is a general illustrative architecture presenting
generators such as wind turbines, photovoltaic panels the EMS as a Microgrid processing unit that uses different
installed near the end-users. However in order to algorithms to analyze forecast and historical data, optimize the
maximize benefit, combined heat and power plants will electricity supply/demand, aggregate the different forecast
be the heart of the Microgrid. data of different sources and last provide the user with
• Distributed Energy Storage (DES): A storage system, interfaces to monitor, view and control the
such as batteries, is needed when energy produced by the consumption/production.
main grid is cheap or when there an excessive energy
production from the DSs. The stored energy can be used
during peak time and also during blackouts.
• Controllable Loads: this consists of electrical devices,
such as electrical vehicles, that can adjust their electricity
usage during peak time.
• Point of Common Coupling (PPC): consists of a
meeting point for power generation, transmission and
customer interface with the main utility.
 Microgrid as it will collect real time data used to monitor the
Microgrid operation status and react to any disturbance that
happens in the system[4]. MCC will also gather information
sent from different Microgrid Local Controllers, which are
associated with distributed energy resources/storage and
control loads, and make important decisions.

Depending on what approach used, local controllers can

exhibit different functions control types: in a centralized
approach an LC will follow the commands sent by an MCC,
while it will take decisions locally in a decentralized
approach. For instance, following an increase of market
Figure 2: Microgrid EMS- A General Architecture electricity price, an information will be sent from the market
operator to the MCC, the latter will use an optimization
To achieve the Microgrid main objectives (supply power, algorithm and send signals to distributed energy resources to
participate in energy market, manage critical load, etc.), an start serving while the decision is taken by DER local
EMS can supervise and control the system either in a controllers in a decentralized approach.
centralized or a decentralized mode[5]; this can be achieved
using a hierarchical structure as represented in figure 3: In a centralized approach, a Microgrid needs to be a
powerful system capable of processing real time data, issued
by different DER, DES and loads, in predefined time slots [5].
Figure 4 presents the information flow between MCC and LCs
where the information exchanged between MCC and LC is
carried through a 2 communication infrastructure. Depending
on the market prices and DER production level, DER LC will
send the power production level and the load LC will issue
request for the demand. Upon the reception of these
information, MCC which is using an optimization algorithm,
will signals DER to start the production, indicates the load
distribution and send the market prices for the next time slot.
Figure 3: A Microgrid Control Hierarchical Architecture

This three level hierarchical structure is made up of a:

• Distributed Network Operator (DNO) and Market

Operator (MO): In an area made of a network of
Microgrids, DNO will be a management aggregator for real-
time information and commands sent by different
Microgrids and the main grid. As for the MO, it will be a
system exchanging data between a Microgrid and an
electricity market in a specific area[4].
• Microgrid Central Controller (MCC): an interface Figure 4: Microgrid Centralized Control- Information Flow
between DNO/MO and the local controllers within a
Centralized approach has many advantages, such as ease
Microgrid and also a coordinators of Local Controllers.
of implementation. However, as the number of devices
• Local Controllers (LCs): depending on what approach
increases, amount of exchanged information increases and the
used, centralized or distributed, LCs will be responsible of
communication channel along with computation process can
monitoring and controlling the distributed energy resources
become slow (bottleneck) [4] .
and the controllable loads such as water heaters.
Unlike the centralized paradigm, a decentralized or
Ideally an Energy Management System will be an entity
distributed approach aims at maximizing the autonomy for the
of the Microgrid Central Controller. This latter has a two way
Microgrid DER and LC units. This goal is achieved by using
communication medium with the DNO and MO to exchange
intelligent controllers that can make their own decision
important information such as electricity supply and
without getting commands from the Microgrid central
participate in energy market. MCC will play a vital role in the
controller. LCs will also be able to communicate between IV. ENERGY MANAGEMENT CHALLENGES AND CLOUD
each other, constructing an intelligent LC network, to COMPUTING: WHAT ADDED VALUE
maximize the production of the corresponding units and
optimize the performance of the Microgrid. To achieve this
goal, the Microgrid needs a system made up several entities Through the implementation of a distributed architecture,
that need to collaborate between each other. Agent-based smart grid will provide end users with a reliable, efficient and
approach is meant to split a main problem into sub-ones and cost effective power management. To achieve this goal and
deal with every sub-problem separately; this is what makes allow the user to use smart grid applications from any place
the agent-based model a key candidate to distributed control and any time, there is a critical need of using advanced
modes and an environment where agents can operate techniques to avoid any technical problems as the system is
regardless of their location: grid level, Microgrid level[8] [9]. source of heterogeneous data [10]. Cloud computing, as an
emerged new technique, is a good model candidate to consider
One novel aspect about using a multi-agent EMS control as it meant to provide on-demand access to computing
system, in comparison to other distributed paradigms, is that resources capable of handling vast data in a cost-effective
every agent has the intelligence to predict a next action and way[11].
influence the other agent neighbors. The communication
between agents can be done using Agent Communication Cloud computing has some main characteristics and
Language (ACL) which can be used not only to exchange services that are very useful for smart grid
value, such as power production level, but also actions and applications[12][13]. The main services can be categorized as
commands to execute in the next time slot [5]. Exchanging follow:
information between local neighbors will imply transferring
• Infrastructure as a service (IaaS): an infrastructure that
less information than the centralized approach. The
provides resources on behalf of its users that can be
computations that are carried by the MCC will also be
adjusted on demand; this includes virtual machines,
executed at the level of agents which will alleviate the load on
storage, load balancers, servers, network….
the entire network. It is worth mentioning that an MCC can
still play an important role in a decentralized approach mainly • Platform as a service (PaaS): A cloud computing
service that exempt the user from the installation of
taking over an LC in case of a control failure[4].
hardware and software components needed to develop
Despite its advantages such as modularity and flexibility and run a new applications usually through a web service
of a plug and play environment, decentralized approach interface. This service includes development tools, web
suffers from some security flows that are sometimes hard to servers, and databases.
troubleshoot given the structure of the architecture. Besides • Software as a service (SaaS): a model that hosts all the
this, in power systems, the communication channel is devoted applications and provide them to users upon request. The
to exchange information between an EMS and Local agents. main advantage of this service is eliminating the need to
This infrastructure need to be carefully redesigned to allow install and run applications in one own servers which will
neighboring agents exchange information which is an save hardware installation, maintenance, license,
upgrading process that is expensive[4]. Figure 5 presents an installation and support costs.
inside Microgrid decentralized EMS control layered
architecture: Given the importance of energy management in smart
grids, various new features were added to the architecture to
solve some management issues; these tools include load
balancing, dynamic pricing, home energy management
system, and building energy management system. However,
to ensure a reliable, cost effective real time energy
management, there are other problems in the current system
that can be addressed by using cloud computing features:

1. A need for a Scalable Platform: A Microgrid is a system

made up of a dynamic number of sensors sending huge
amount of data that need to be analyzed rapidly to meet a
real time demand/supply. Given its dynamic elasticity,
resource pooling (will avoid restrictions during peak
Figure 5: A Microgrid Decentralized EMS Architecture
time), and on demand self-service (will allow a customer
to keep track of energy usage and pricing) characteristics,
 cloud computing infrastructure is an ideal one for from different sources such as sensors and weather
dynamic and computationally huge data applications that forecasting offices and apply machine learning algorithms
need scalable resources[10][14]. that can be used in an academic Microgrid to adapt to the
2. A need for Fast Demand Response: [12] propose a new context of users (students, staff and faculty) to optimize
architecture called Cloud Based Demand Response the operation of the system.
(CDR) meant for a quick response time by changing the
!master/slave architecture representing utility-customer The current smart grid architecture is suffering from
into utility-smart different problems related to the use of: a limited bandwidth, a
meters\HomeEnergyManagementSystem . CDR will lead controlled demand response by the utility, a master-slave
to a secure, scalable and reliable demand response by architecture… Various research experiments proved that cloud
using data-centric communication, publisher/subscriber computing applications are very useful for smart grid energy
and topic-based group communication[10]. management as presented in [12-19].
3. A need for an Intelligent Demand Management:
through the use of a software as a service cloud based
V. Kaa Cloud based EMS Architecture
platform, supporting heterogonous devices from different
The implementation of an ideal Campus Microgrid EMS
layers in the smart grid, and a dynamic demand response
platform should carefully consider the following research
(described in [15]), will come up with a system capable of
challenges: the cost of implementing the EMS platform, the
periodically forecasting demand hence make intelligent
ability to customize the services provided, the scalability of
demand management avoiding peak time burden[10].
the platform for other campuses, and last the system
4. A need for a Dynamic Bandwidth Allocation (DBA):
performance platform in terms of delay[20]. The last feature is
among the problems in the traditional electrical system is
very important to consider mainly for some smart grid
the peak time where the demand is high and customers
sensitive delay applications as stated in table 1 (According to
are served depending on their priorities and available
the draft of European Telecommunications Standards Institute
resources; when the supply is less than the demand,
(ETSI) technical report on M2M communication for the smart
blackouts occur. In [16], X.Li and J.Chung Lo proposed a
grid [21]):
cloud computing scheduling algorithm that proved the
interaction between distributors and consumers. The
Smart Grid Application Max Response
algorithms is named pricing and peak aware scheduling Time
(PPAS) and it is based on dynamic pricing and setting a
peak threshold that can be dynamically assigned so that to Protection 1 to 10 ms
avoid exceeding energy usage at any day time. Results of
the proposed algorithm experiments proved efficient Control 100 ms
energy management of cloud computers along with the
energy distributor[10]. Monitoring 1s
5. A need for a Power Monitoring System (PMS): [17]
Metering/Billing 15 mins
propose a monitoring system for smart grid users, using
advanced cloud services solution based on service- On load Tap changer Seconds
oriented architecture and enterprise service bus to achieve
an early warning system and an efficient, on-demand, Stability and power flow analysis Seconds
scalable power monitoring service[10].
6. A need to meet Customers Energy Requirements: Economic Dispatch 15 mins to Days
virtualization in cloud computing applications can be ahead
used to integrate external energy sources and storage
devices with the internal ones constructing an energy Demand Side Management Day/hours
exchange mechanism between different Microgrids and
hence meeting the consumers demand as proposed in the Table 1: Smart Grid Applications Response Time
framework in [10][18].
To quantify the response time for some delay sensitive
7. A need to use a robust and scalable data analytics
smart grid applications, we run some experiments based on a
platform capable of managing an always increasing
free open source cloud based IoT platform called Kaa. The
volume data by collecting, categorizing (structured, semi-
structured and structured), storing, modeling, and
predicting solutions for Microgrid clients [19]. As
depicted in figure 2, an EMS analyzes big data coming

Figure 6: Experiment Architecture using Kaa Single node

 choice of this platform was based on many features stated in The measured delay meets table 1 requirements except for
[22]; it is briefly a complete end to end internet of things the protection and control applications which we propose to
cloud based middleware platform that can be installed as a run on the edge in future work. Moreover, after 24 hours
single node, cluster nodes or as a sandbox. Initially we running time, Kaa server performance decreased as it stop
considered Kaa as a single Linux node using the architecture responding from time to time and access to Cassandra
presented in figure 6 but we opted for Kaa Sandbox as it is a database was not possible when data exceed the max storage
pre-configured system that has all Kaa components in a virtual capacity allocated for the virtual box. For all these reasons, we
environment. will use Kaa as a cluster of Linux nodes in the coming work.

The main goal of the experiment is to use Kaa sandbox to

ingest time series into Cassandra and measure the TCP delay VI. Conclusion
using wireshark using the design presented in figure 7:

To better understand the information flow between a The world of smart grid is a huge research project that
client and Kaa server, we captured the data flow using attracted the attention of scientists from different research
wireshark presented in figure 8: areas. The study of wireless sensors, Internet of things, cloud
As for the delay we run our experiment for more than 24
computing and big data analytics are all key factors to
hours by increasing the number of clients every 15 min
(maximum number of clients used is 40); our main goal was consider to understand smart grid from a computer science
to watch the delay as we increase traffic and check if we will perspective. The main purpose of this paper is to present the
meet the response time in table 1. We computed the delay, importance of an energy management system in a Microgrid
variance and standard deviation using wireshark TCP delta and how cloud computing is contributing on the improvement
time which is the time since previous frame in a TCP steam; of the EMS performance. We used a cloud based IoT platform
called Kaa to quantify the delay for some smart grid delay
sensitive applications. Our future work will be about the
installation of a real IoT test bed using Kaa platform as a
cluster of Linux nodes.

REFERENCES

[1] KyungGyu Park, Yoonkee Kim, SeonMi Kim,

KwangHo Kim, WookHyun Lee, and HwaChoon
Park, “Building Energy Management System based on
Smart Grid,” 2011 IEEE 33rd Int. Telecommun.
Energy Conf., pp. 1–4, 2011.

[2] Q. Qin, Z. Chen, and X. Wang, “Overview of micro-

Figure 7: Time Series Data Ingestion into Cassandra using Kaa Sandbox
grid energy management system research status,”
2012 Power Eng. Autom. Conf., pp. 1–4, Sep. 2012.

[3] X. Fang, S. Misra, G. Xue, and D. Yang, “Smart Grid

— The New and Improved Power Grid: A Survey,”
IEEE Commun. Surv. Tutorials, vol. 14, no. 4, pp.
944–980, 2012.
Figure 8: Data Flow between a Client and Kaa server [4] I. Introduction, “Energy Management Systems in
Microgrid Operations.pdf,” Electr. J., vol. 25, no. 8,
results are presented in table 2:
2012.
15 min 30 min 45 min 60 min 24h [5] F. Katiraei, R. Iravani, N. Hatziargyriou, and A.
Delay(s) 0.300329 0.301395 0.302119 0.304418 0.304442
Dimeas, “Microgrids management,” IEEE Power and
Variance 0.209095 0.209719 0.21682 0.219676 0.214944 Energy Magazine, vol. 6, no. 3, pp. 54–65, 2008.

Standard 0.457270 0.457951 0.465639 0.468696 0.463620 [6] I. D. Serna-Suarez, G. Ordonez-Plata, and G. Carrillo-
Deviation Caicedo, “Microgrid’s Energy Management Systems:
Table 2: Client/Kaa Server Average TCP delay, Variance and A survey,” 2015 12th Int. Conf. Eur. Energy Mark.,
Standard Deviation
pp. 1–6, 2015.

[7] K. Hirose, T. Shimakage, N. T. T. Facilities, J. T.

 Reilly, and R. Associates, “The Sendai Microgrid Systems,” in Conference on Electrical Power
Operational Experience in the Aftermath of the Distribution Networks, 2011.
Tohoku Earthquake : A Case Study,” pp. 1–6.
[19] J. P. D. Comput, M. D. Assunção, R. N. Calheiros, S.
[8] G. Rohbogner, S. Fey, U. J. J. Hahnel, P. Benoit, and Bianchi, M. A. S. Netto, and R. Buyya, “Big Data
B. Wille-Haussmann, “What the term agent stands for computing and clouds : Trends and future directions,”
in the Smart Grid definition of agents and multi-agent J. Parallel Distrib. Comput., vol. 79–80, pp. 3–15,
systems from an engineer’s perspective,” pp. 1301– 2015.
1305, 2012.
[20] M. Abdullah, A. Faruque, K. Vatanparvar, and G. S.
[9] M. Meiqin, D. Wei, and L. Chang, “Multi-agent based Member, “Energy Management-as-a-Service Over
simulation for Microgrid energy management,” 8th Fog Computing Platform,” vol. 3, no. 2, pp. 161–169,
Int. Conf. Power Electron. - ECCE Asia, pp. 1219– 2016.
1223, 2011.
[21] T. Report, “Applicability of M2M architecture to
[10] S. Bera, S. Misra, and J. J. P. C. Rodrigues, “Cloud Smart Grid Networks ; Impact of Smart Grids on
Computing Applications for Smart Grid: A Survey,” M2M platform,” vol. 1, pp. 1–58, 2012.
IEEE Trans. Parallel Distrib. Syst., vol. 26, no. 5, pp.
1477–1494, 2015. [22] P. P. Ray, “ScienceDirect A survey of IoT cloud
platforms,” Futur. Comput. Informatics J., pp. 1–12,
[11] “Final Version of NIST Cloud Computing Definition 2017.
Published | NIST.” [Online]. Available:
https://www.nist.gov/news- [23] H. Talei, B. Zizi, M. R. Abid, M. Essaaidi, D.
events/news/2011/10/final-version-nist-cloud- Benhaddou and N. Khalil, "Smart campus microgrid:
computing-definition-published. Advantages and the main architectural components,"
2015 3rd International Renewable and Sustainable
[12] H. Kim, Y.-J. Kim, K. Yang, and M. Thottan, “Cloud- Energy Conference (IRSEC), Marrakech, 2015.
based demand response for smart grid: Architecture
and distributed algorithms,” in 2011 IEEE
International Conference on Smart Grid
Communications (SmartGridComm), 2011, pp. 398–
403.

[13] A. S. B. C. Tak, B. Urgaonkar, “Cloudy with a

Chance of Cost Savings,” IEEE Trans. PARALLEL
Distrib. Syst., vol. 24, no. 6, pp. 1223–1233, 2013.

[14] C. Y. W. Chen, K. H. J. Liu, and W. H. C. Hsu,

“Implementation of Smart Power Management and
Service System on Cloud Computing,” in
International Conference on Ubiquitous Intelligence
and Computing and 9th International Conference on
Autonomic and Trusted Computing, 2012, pp. 924–
929.

[15] V. P. Y. Simmhan, S.Aman, A.Kumbhare, S.Stevens,

Q.Zhou, “Cloud-Based Software Platform for Big
Data Analytics in Smart Grids,” Computing in Science
& Engineering, 2013.

[16] X. Li and J. Lo, “Pricing and Peak Aware Scheduling

Algorithm for Cloud Computing,” in IEEE PES
Innovative Smart Grid Technologies (ISGT), 2011.

[17] L. Ji, W. Lifang, and Y. Li, “Cloud Service based

intelligent power monitoring and early-warning
system,” in IEEE Innovative Smart Grid Technologies
(ISGT )-Asia, 2012, pp. 1–4.

[18] G. G. P. Nezamabadi, “Electrical Energy Management

of Virtual Power Plants in Distribution Networks with
Renewable Energy Resources and Energy Storage