A Brief Comparative Study of Uncertainty

Modeling Techniques in Power System

Allah Wasaya Intisar Ali Sajjad
Department of Electrical Engineering Department of Electrical Engineering
University of Engineering and Technology Taxila University of Engineering and Technology Taxila
Taxila, Pakistan Taxila, Pakistan
[email protected] [email protected]

Rehan Liaqat Muhammad Muzaffar Iqbal

Department of Electrical Engineering Department of Electrical Engineering
University of Engineering and Technology Taxila FAST University Lahore Campus
Taxila, Pakistan Lahore, Pakistan
[email protected] [email protected]

Abstract—Modern power systems tend to have large As sources of these uncertainties are different, different
penetrations of renewable energy resources (RERs). Since power therefore different types of uncertainty handling techniques
generation from these resources is intermittent due to uncertain have been developed for power systems so for. These
climate conditions, therefore their rising integration poses approaches can be classified as the possibilistic approach,
several challenges for power system engineers to ensure stable
probabilistic approach, information gap decision theory
grid operation. To overcome these challenges, the impact of such
variable sources on the system needs to be judiciously analyzed. (IGDT), hybrid probabilistic-possibilistic approach, interval
2020 IEEE 23rd International Multitopic Conference (INMIC) | 978-1-7281-9893-4/20/$31.00 ©2020 IEEE | DOI: 10.1109/INMIC50486.2020.9318090

This paper presents different approaches to handle uncertain optimization, robust optimization and some recent techniques
parameters related to RERs, particularly solar and wind (fuzzy logic, scenarios creation, 2-stage scheduling schemes,
energies, with a critical assessment. Many models and worst-case scenario method, etc.) as shown in Fig. 1 [2]. These
approaches are critically assessed, particularly focusing on past techniques have been used for modeling the uncertainty of
and recent techniques applied to determine the available distributed generation resources and load.
renewable energy at a specific interval of time. Moreover, the
relative features of various models are compared while
highlighting their potential applications. The discussion
presented in this work will support researchers in choosing
appropriate models for simulation studies in power systems with
high penetrations of distributed RERs. It will also help outline
the needs and requirements of novel models for uncertainty
modeling in energy systems.

Keywords— uncertainty modeling, information gap decision

theory,Monte Carlo simulation, probabilistic modeling, renewable
energy resources, robust optimization

I. INTRODUCTION
In recent days, the demand for renewable energy is rising
with the rise in the global population and restrictions imposed
on greenhouse gas emissions. This scenario needs reshaping
the traditional ways of energy generation, transmission and Fig. 1. The uncertainty modeling approaches
distribution to meet the energy requirements. The legacy
power system is being reconstructed by the introduction of The objective of different types of approaches is to see the
renewable energy resources (RERs) such as solar and wind. As influence of intermittent input parameters on various output
these sources are uncertain, intermittent and weather parameters of the system. Different functions are used for
dependent, therefore their higher penetrations cause difficulty modeling the uncertainty of variable input parameters in
in stable and economic operation of the power system. Besides different methods. In the case of probabilistic methods,
the intermittency of RERs, there are some other uncertainty probability density function (PDF) is used while possibilistic
sources in the power system such as unplanned methods employ membership function. This paper gives brief
outages/interruptions, transmission capacity variations, details for various uncertainty handling techniques in power
changes in fuel price and energy price, varying market rules, systems. The organization of the paper is as follows.
time-varying load requirements and other so on [1]. During Probabilistic methods are presented in Section II. Modeling of
power system planning and operation, all these uncertainties possibilistic uncertainty is discussed in Section III and Section
are taken into account to ensure the reliable and efficient IV presents the hybrid probabilistic-possibilistic approach.
operation of the power grid. Interval optimization is briefly overviewed in Section V.
Concept of uncertainty handling through information gap
Mathematical uncertainty arises due to calculation and theory is explored in Section VI while Section VII presents
observation errors and it causes differences in true/measured robust optimization. Some recent algorithms are discussed in
and estimated value [1]. A lot of traditional and deterministic Section VIII and Section IX concludes the paper.
approaches exist to deal with different types of uncertainties.

978-1-7281-9893-4/20/$31.00 ©2020 IEEE

 Fig. 2. Probabilistic approaches of uncertainty modeling

II. PROBABILISTIC METHODS 3) Pseudo-sequential MCS

In this approach, a multivariant function ‫ ݕ‬ൌ ݂ሺ‫ݔ‬ሻ is Pseudo-sequential MCS is faster than sequential and non-
considered in which ‫ ݔ‬ൌ  ሼ‫ݔ‬ଵ ǡ ‫ݔ‬ଶ ǡ ‫ݔ‬ଷ ǡ ǥ ǡ ‫ݔ‬௡ ሽ is a vector sequential MCS schemes because it is a hybrid of these two
containing variable inputs having known PDF. Now the techniques. A pseudo sequential MCS based method is used in
question arises that how we can determine the PDF of ‫ݕ‬. The [8] to observe the customers’ nodal reliability and reserve
Weibull PDF usually used for generating wind speed patterns deployment with high penetration of PV. Generally, a large
is defined by (1). Similarly, normal PDF which is generally number of random numbers are generated for MC but this
employed for load variations is given by (2). MCS technique generates results quickly.

ߚ ఉିଵ ‫ ݔ‬ఈ B. Analytical Approaches

݂ሺ‫ݔ‬ሻ ൌ ‫ݔ‬ ‡š’ ൤െ ൬ ൰ ൨ (1)
ߙఉ ߚ The analytical approaches for uncertainty modeling are
ͳ ିሺ௫ିఓሻమ divided into two groups. One group of these approaches relies
݂ሺ‫ݔ‬ሻ ൌ ݁ ଶఙమ (2) on linearization while the 2nd group of analytical methods is
ߪξʹߨ based on PDF approximation as shown in Fig. 2. Gram-
Here ߤ and ߪ represent the mean and standard deviation Charlier series, Convolution method, Cornish-Fisher
respectively while ߙ and ߚ are scale and shape parameters of expansion, cumulant method, Taylor series, Edgeworth
Weibull PDF respectively. expansion and first-order second-moment methods are based
A. Monte Carlo Simulation on linearization while unscented transformation and point
estimation method are examples of PDF approximation-based
Several methods of probabilistic uncertainty analysis can methods. The technical merits and demerits of some
be found in the literature. These methods mainly use Monte techniques are given in Table I.
Carlo Simulation (MCS) to solve the problem of variable
III. POSSIBILISTIC METHOD
power generation due to distributed generators (DGs). A
combination of MCS and Taylor series expansion is employed The idea of uncertainty modeling through the possibilistic
in Ref. [3] to assess the reliability of the bulk power system. approach was given by ƒ†‡Š[9]. In this method, uncertain
MCS has further three types: sequential MCS, non-sequential parameters having fuzzy boundaries were demonstrated by
MCS and pseudo sequential MCS as shown in Fig. 2. linguistic categories. This technique is applied by finding out
the membership function of the output variable with known
1) Sequential MCS input membership function. ߙ-cut and defuzzification methods
For the determination of posterior distribution, Sequential are used for it. In the presence of uncertainties, the impact of
Monte Carlo (SMC) schemes are assessed as the most effective DGs’ operation on active losses was analyzed and load supply
approaches. The SMC methods are flexible and easy to ability of distribution network was determined in [10]. In [11],
implement. The influence of wind on the power distribution the authors performed the harmonic analysis by introducing a
system is precisely examined by SMC which is build up by a harmonic load flow in presence of power generation through
time-based wind storm sampling scheme [4]. In [5], the authors wind farms.
deployed the SMC technique for simulating the adaptability of
the power systems with wind farms. For the reliability IV. JOINT PROBABILISTIC-POSSIBILISTIC METHODS
estimation of a complex power system, MCS is used in [6]. The combined probabilistic-possibilistic method is useful
2) Non-sequential MCS to model certain input data probabilistically and remaining data
Non-sequential MCS, also known as the state sampling possibilistically. The coding of such type of problem is done
method, is an MCS based approach. In this technique, the by two loops. The outer loop is based on the probabilistic
probability of each component state is evaluated by sampling method while the inner loop is based on the possibilistic
its probability of component lying on that state for any state of method. In Possibilistic-Monte Carlo, the inner loop is of
the system in a group of all component states. It is mainly used possibilistic approach (ߙ-cut method) and the outer is Monte-
for the estimation of risk in the power system. A non-sequential Carlo. For Possibilistic-Scenario based scheme, scenarios are
MCS based optimization tool was developed in [7] for optimal generated in the outer loop and the Possibilistic loop runs
dispatch of thermal power. The objective of this study was to completely for each scenario generation.
reduce environmental pollution in the presence of variable
wind power.
 TABLE I. SUMMARY OF UNCERTAINTY MODELING ATTRIBUTES

Group Examples Main Idea Advantages Disadvantages

Accurate for complex
Probabilistic (based on
MCS Real state simulation problems having large Very time consuming
simulation)[12]
amounts of data.
High order instants
Probabilistic (analytical) Cumulants Based on linearization Fast cannot be attained
correctly.
It is faster than MCS and the
Only the mean value is
Scenario-based methods Scenarios creation selection of scenarios is
achieved at the output.
important for accuracy in it.
Probabilistic (based on Probability distribution Simple and computationally Only valid for PDF-
Point estimation method
different scenarios) function approximation efficient. based problems.
The uncertain variable
Probability distribution
Two-point estimation method Accurate and fast determines the execution
function approximation
time.

Fuzzy membership Member function of variable Simulation takes a long

Possibilistic Į-cut Method
function (FMF) used output can be attained. time.

Uncertainty modeling of Both probabilistic and

Joint probabilistic- Fuzzy-Monte Carlo, Fuzzy-
Joint probabilistic- possibilistic uncertainties can It is time-consuming.
possibilistic scenario based
possibilistic approaches be modeled.

More effective when Intervals correlation

Interval Analysis Method of interval analysis Using intervals
intervals exist. cannot be modeled.
Information gap theory Predicted values of Decision making becomes
IGDT It is more complex.
(IGT) [13] parameters are used easy in severe intermittency.
Effective when a specific Non-linear problems are
Roust optimization [14] Robust optimization method Using intervals
interval exists. difficult to solve.

VI. INFORMATION GAP DECISION THEORY

Various applications of this combined approach can be
found in the literature. In [15], an assessment tool is The IGDT is adopted for those problems that cannot be
developed by this hybrid method to analyze the active solved by PDF due to less historical data. In [13], the
losses due to uncertainty caused by DG integration. authors applied IGDT to overcome the variation in wind
power generation. Ref. [17] presents a model of
V. INTERVAL OPTIMIZATION uncertainties using the non-probabilistic information gap
In this uncertainty handling method, when the system theory for short term scheduling of generation companies.
has intermittent input parameters then upper and lower In [18], the authors introduced an IGDT model to meet the
bounds are each variable are taken that can have customer load demand by helping network operators in the
representation through an interval. Assuming a selection of specific supplying resources. Similarly, a non-
multivariable function of the form ݂ሺ‫ݔ‬ଵ ǡ ‫ݔ‬ଶ ǡ ‫ݔ‬ଷ ǡ ǥ Ǥ ‫ݔ‬௡ ሻ and probabilistic IGDT was applied in [19] to model
݈‫ ݅݋‬൑xi ൑ upi with upper and lower bounds upi and loi uncertainty in the day-ahead price of electricity in the
respectively. electricity market.

An interval arithmetic approach is used to see the VII. Robust Optimization

variation in load demand that is based on probabilistic Robust optimization is considered as an effective and new
distribution. approach for solving uncertainty problems in case of lack
ିሺ௫ିఓሻʹ
of information. In [20], the authors developed a strategy of
ͳ
ܲ‫ ܾ݋ݎ‬ൌ න ‫ͳܣ‬ ݁ ଶఙ మ
(3) scheduling energy generation in microgrid using robust
ߪξʹߨ optimization for system stability and cost minimization.
  Various applications of this scheme can be found in the
௕ మ ௖ ିሺ௫ିఓሻమ
ͳ ‫ ݔ‬െ ܽ ିሺ௫ିఓሻ literature. A home energy management scheme is
ܲ‫ ܾ݋ݎ‬ൌ ቈන ݁ ଶఙమ ൅ න ݁ ଶఙమ
ߪξʹߨ ௔ ܾ െ ܽ ௕
presented in [14] considering the uncertainty of PV
ௗ మ (4) storage. A flexible robust optimization method is proposed
‫ ݔ‬െ ݀ ିሺ௫ିఓሻ
൅න ݁ ଶఙ మ
቉ in [21] for economic power dispatch in multiple intervals
௖ ܿെ݀ of time. Uncertainty of battery charging and discharging is
where a, b, c and d are edge points of the slope in the modeled using a distinct robust optimization [22]. In [23],
interval of range a to d. Coordination of directional the authors did microgrid planning by considering the
overcurrent relay is among the major problems in the uncertainty of load, renewable generation and market
power system. In [16], the authors proposed an interval price.
analysis technique to cope with this protection problem.
 TABLE II. SOME RECENT UNCERTAINTY MODELS

Model
Sr. no. Method Features Ref.
Applications

Power produced • Day-ahead scheduling is done in the first stage

1 2-stage scheduling schemes [24]-[25]
from RESs. • Hours ahead scheduling is done 2nd stage to
adjust possible errors
Day-ahead price
Determining worst-case forecasting, wind • Upper and lower bounds of uncertain inputs are
2 [26]
scenario speed determined
• High system stability is achieved

Heat and power • Distance between reference and net distribution

3 By estimating Kernel density demand in the is determined through Kullback-Liebler [27]
renewable energy divergence
system • Historical data is required

Using Hyper-Heuristics Generation of

4 [28]
(HHs) renewable power • Used for heuristic selection

Load demand, load

5 Fuzzy logic • On each uncertain parameter, a specific fuzzy [29]-[30]
forecasting
membership function is applied
Load power, wind • For load – Normal distribution
6 Scenarios creation speed, solar • For wind -- Weibull distribution [31]-[32]
insolation • For solar insolation –Beta Distribution

Using discrete states • Generation of different PDFs for each source

7 Solar irradiance, [33]
combination • Splitting up each one-hour continuous
wind speed
distribution into N number of states
• The location and weighing factors are evaluated
Hong’s point estimation
8 Solar irradiance, for each random variable. [34]
method (PEM) method
wind speed • The variance, mean and skewness coefficient are
determined for central moments

VIII. RECENT APPROACHES REFERENCES

Power system stability and operation is affected by large [1] J. Zhu, Optimization of Power System Operation: Second
renewable energy penetration. Some new approaches to Edition. 2015.
overcome the uncertainty problem mainly arise due to RERs [2] A. Soroudi and T. Amraee, “Decision making under uncertainty
in energy systems: State of the art,” Renew. Sustain. Energy
have been discussed in Table II. Rev., vol. 28, pp. 376–384, 2013.
[3] Y. Zhao, F. Fan, J. Wang, and K. Xie, “Uncertainty analysis for
IX. CONCLUSION bulk power systems reliability evaluation using Taylor series
Recent smart developments, higher penetration of the and nonparametric probability density estimation,” Int. J.
intermittent RERs and other uncertain parameters have risen Electr. Power Energy Syst., vol. 64, pp. 804–814, 2015.
[4] G. Li et al., “Risk analysis for distribution systems in the
the uncertainty in the energy system. Researchers are northeast U.S. under wind storms,” IEEE Trans. Power Syst.,
looking for better modeling approaches to tackle the vol. 29, no. 2, pp. 889–898, 2014.
uncertainty while solving various problems related to the [5] X. Han, Y. Qu, P. Wang, S. Member, and J. Yang, “Four-
power system. A brief comparative analysis of different Dimensional Wind Speed Model for Adequacy Assessment of
uncertainty modeling approaches was presented in this Power Systems With Wind Farms,” pp. 1–8, 2012.
[6] “Electrical power supply and utilizafion (scientific, technical),”
paper while taking into account their various features and no. January, p. 2000, 2000.
applications. Effectiveness of different uncertainty [7] F. Vallée, C. Versèle, J. Lobry, and F. Moiny, “Non-sequential
modeling methods such as possibilistic, probabilistic, joint Monte Carlo simulation tool in order to minimize gaseous
possibilistic-probabilistic, interval analysis, IGDT, robust pollutants emissions in presence of fluctuating wind power,”
optimization and some recent techniques was presented to Renew. Energy, vol. 50, pp. 317–324, 2013.
[8] Y. Ding, P. Wang, L. Goel, R. Billinton, and R. Karki,
resolve emerging uncertainty problems. Additionally, the “Reliability assessment of restructured power systems using
merits and drawbacks of these methods were also presented. reliability network equivalent and pseudo-sequential simulation
This discussion of this comparative study can be helpful for techniques,” Electr. Power Syst. Res., vol. 77, no. 12, pp. 1665–
power system researchers to choose and develop better 1671, 2007.
uncertainty modeling methods for the power system [9] L. A. Zadeh, “A THEORY OF POSSIBILITY *,” pp. 9–34,
1999.
simulation studies and for the advancement of scientific
[10] A. Soroudi, M. Ehsan, R. Caire, N. Hadjsaid, and S. Member,
research in this domain. “Possibilistic Evaluation of Distributed Generations Impacts on
Distribution Networks,” vol. 26, no. 4, pp. 2293–2301, 2011.
ACKNOWLEDGMENTS [11] A. A. Romero, C. Samudio, H. Zini, and G. Rattá, “Modeling
The authors would like to thank Postgraduate Studies Uncertainties in the Harmonic Distortion Calculation in Power
Systems , due to Wind Farms,” pp. 1–6, 2012.
office of Department of Electrical Engineering, University [12] Y. Zhang, J. Wang, and Z. Li, “Uncertainty Modeling of
of Engineering and Technology Taxila, Pakistan for Distributed Energy Resources: Techniques and Challenges,”
providing the research facilities. Curr. Sustain. Energy Reports, vol. 6, no. 2, pp. 42–51, 2019.
[13] A. Soroudi, “Information Gap Decision Theory based OPF with system energy loss minimization,” IEEE Trans. Power Syst.,
HVDC Connected Wind Farms,” p. 8040, 2015. vol. 25, no. 1, pp. 360–370, 2010.
[14] P. Zhao, H. Wu, C. Gu, and I. Hernando-Gil, “Optimal home [34] S. Mohammadi, B. Mozafari, S. Solimani, and T. Niknam, “An
energy management under hybrid photovoltaic-storage Adaptive Modified Firefly Optimisation Algorithm based on
uncertainty: A distributionally robust chance-constrained Hong’s Point Estimate Method to optimal operation
approach,” IET Renew. Power Gener., vol. 13, no. 11, pp. 1911– management in a microgrid with consideration of
1919, 2019. uncertainties,” Energy, vol. 51, pp. 339–348, 2013.
[15] Z. F. Liu, G. H. Huang, and N. Li, “A dynamic optimization
approach for power generation planning under uncertainty,”
Energy Sources, Part A Recover. Util. Environ. Eff., vol. 30, no.
14–15, pp. 1413–1431, 2008.
[16] A. S. Noghabi, H. R. Mashhadi, and J. Sadeh, “Optimal
Coordination of Directional Overcurrent Relays Considering
Different Network Topologies,” vol. 25, no. 3, pp. 1348–1354,
2010.
[17] R. S. Gencos, B. Mohammadi-ivatloo, S. Member, H.
Zareipour, S. Member, and A. Indices, “Application of
Information-Gap Decision Theory,” vol. 28, no. 2, pp. 1093–
1102, 2013.
[18] A. Soroudi and M. Ehsan, “IGDT Based Robust Decision
Making Tool for DNOs in Load Procurement Under Severe
Uncertainty,” pp. 1–10, 2012.
[19] M. Kazemi, B. Mohammadi-ivatloo, and M. Ehsan, “Risk-
Constrained Strategic Bidding of GenCos Considering Demand
Response,” pp. 1–9, 2014.
[20] R. Wang, P. Wang, and G. Xiao, “A robust optimization
approach for energy generation scheduling in microgrids,”
Energy Convers. Manag., vol. 106, pp. 597–607, 2015.
[21] Á. Lorca and X. A. Sun, “Adaptive Robust Optimization With
Dynamic Uncertainty Sets for Multi-Period Economic Dispatch
Under Significant Wind,” IEEE Trans. Power Syst., vol. 30, no.
4, pp. 1702–1713, 2015.
[22] M. R. Sarker, H. Pandžiü, and M. A. Ortega-Vazquez, “Optimal
operation and services scheduling for an electric vehicle battery
swapping station,” IEEE Trans. Power Syst., vol. 30, no. 2, pp.
901–910, 2015.
[23] A. Khodaei, S. Bahramirad, and M. Shahidehpour, “Microgrid
Planning Under Uncertainty,” IEEE Trans. Power Syst., vol. 30,
no. 5, pp. 2417–2425, 2015.
[24] X. Wu, X. Wang, S. Member, C. Qu, S. Member, and A. Sets,
“A Hierarchical Framework for Generation Scheduling of
Microgrids,” vol. 29, no. 6, pp. 2448–2457, 2014.
[25] Y. Zhang, J. Le, F. Zheng, Y. Zhang, and K. Liu, “Two-stage
distributionally robust coordinated scheduling for gas-
electricity integrated energy system considering wind power
uncertainty and reserve capacity configuration,” Renew.
Energy, vol. 135, pp. 122–135, 2019.
[26] S. Fliscounakis, P. Panciatici, F. Capitanescu, and L. Wehenkel,
“Contingency ranking with respect to overloads in very large
power systems taking into account uncertainty, preventive, and
corrective actions,” IEEE Trans. Power Syst., vol. 28, no. 4, pp.
4909–4917, 2013.
[27] J. Yang, J. Liu, Z. Fang, and W. Liu, “Electricity scheduling
strategy for home energy management system with renewable
energy and battery storage: A case study,” IET Renew. Power
Gener., vol. 12, no. 6, pp. 639–648, 2018.
[28] E. K. Burke et al., “Hyper-heuristics: A survey of the state of
the art,” J. Oper. Res. Soc., vol. 64, no. 12, pp. 1695–1724,
2013.
[29] J. Soares, M. A. Fotouhi Ghazvini, Z. Vale, and P. B. de Moura
Oliveira, “A multi-objective model for the day-ahead energy
resource scheduling of a smart grid with high penetration of
sensitive loads,” Appl. Energy, vol. 162, pp. 1074–1088, 2016.
[30] A. K. Shukla and P. K. Muhuri, “Big-data clustering with
interval type-2 fuzzy uncertainty modeling in gene expression
datasets,” Eng. Appl. Artif. Intell., vol. 77, no. July 2018, pp.
268–282, 2019.
[31] S. Talari, M. Yazdaninejad, and M. R. Haghifam, “Stochastic-
based scheduling of the microgrid operation including wind
turbines, photovoltaic cells, energy storages and responsive
loads,” IET Gener. Transm. Distrib., vol. 9, no. 12, pp. 1498–
1509, 2015.
[32] M. Pilz and L. Al-Fagih, “Recent advances in local energy
trading in the smart grid based on game-theoretic approaches,”
IEEE Trans. Smart Grid, vol. 10, no. 2, pp. 1363–1371, 2019.
[33] Y. M. Atwa, E. F. El-Saadany, M. M. A. Salama, and R.
Seethapathy, “Optimal renewable resources mix for distribution