1

Impact of Load Uncertainty Modeling on the

Reliability of Power Systems Using Sequential
Monte Carlo Simulation

Abstract In this paper the impact load uncertainty on system

reliability is investigated using Monte Carlo Simulation. Hour by The objective of this paper is to study and evaluate the impact
hour simulation is made for specific load with different level of of the load uncertainty on the reliability of power system. The
uncertainty using normal random distribution at each hour. The objective will be achieved by comparing the reliability indices
unavailability of the power for a specific load in the simulation is of a test power system with different seniors where the load is
determine by using Gauss Seidel Power Flow method and check uncertain. The generation and the demand of the system will be
the voltage profile at each bus and checking the capacity of all changed every hour and the reliability will be evaluated by
available tie sets supply the load. Unavailability, failure frequency
using Monte Carlo Simulation. In general, the impact of the
and repair rate is calculated after the simulation is completed.
power demand increasing on the reliability of the system will
be simulated by one of two ways, the first case by assuming one
Index Terms Reliability, Monte Carlo, Load Uncertainty or more components will be overloaded and hence one load or
more will shaded until the power is return to be stable. The
second affect could be happen because the stress increases on a
I. INTRODUCTION specific component and that will affect the failure rate. In
practical, this affect could be prevented by increasing the
maintenance of that component which will increase the
I n the last decade many researches focused toward using
microgrid network due to its advantages over conventional
radial system, like decreasing the power losses by integrating
operation cost, but at specific point this case will affect the
reliability of the system and will be evaluated in this paper.
Distribution Generators (DG) since the load will be very close
to the source and it can be used to restore the power faster and II. PROBLEM DEFINITION
prevent cascade failures since the distribution systems can
operates even when it's islanded and also those DG can be The problem will be evaluating simple distribution power
renewable generators that will reduce the environmental impact system reliability using Monte Carlo simulation and determine
electricity generation [3]. But with this approach the the load reliability indices in normal operation case. Then, do
distribution system will be more complex and that makes the same study but with considering some sort of uncertainty in
evaluation the system reliability much harder due to increase of all the loads values and compare it with the previous result and
the complexity of the system and consume both time and effort see the effect of the load uncertainty in reliability indices.
to calculate reliability indices which is major concern for
planning and operation power systems [1]. The problem with this simulation is how to determine which
load to shade when there is overloaded line. To overcome this
One of the major parameters that affect the reliability problem a simple graph theory technique will be used by
assessment in future is the accuracy of the load consumption identifying all disjoint paths connected the load and the source
data which is always predicted with level of uncertainty. The and check if one of the paths is not overloaded and capable to
existing of this level of uncertainty could affect the system supply the load otherwise the load will not be served.
reliability indices before applying corrective actions like
building new generation plants or expanding the transmission The other issue with this simulation is how to measure the repair
systems in the right time. This paper will study the effect of the time if load point is going down. To overcome this problem root
load uncertainty on the reliability of power system. The paper cause analysis will be done to track the reason of the this
focuses more on the reliability of a fixed system without shortage and make the outage period equal to that component
changing the system topology. The study will be done by using repair time. If an outage occurs because of an overload case the
Monte Carlo Simulation (MCS) over one year period of time outage duration will be counted as one period of simulation
where the load is uncertain and flow stochastic process. since the load will be down until the simulation period is done.
Many papers discuss the uncertainty of the load foresting and
the impact of that in reliability like [4] where the author model
the uncertainty at a specific bus using normal distribution. To III. METHODOLOGY AND FORMULATION
evaluate specific system reliability in term of power capacity
and system stability can be achieved by using power flow as The main objective of the simulation is to impede sort of
explained in [6]. uncertainty to the whole loads in the system and study the
 2

effect. That why a data of one year hour by hour is needed either flow is performed using Gauss Seidel numerical method. Then
from history or forecasted data. The change will be investigated the voltage profile is checked to see either the load is supplied
from the load and transmission lines sides, that why a power or not. If the voltage is whiten the limit, the whole tie sets are
flow solution is needed. The source will be assumed to be checked to make sure at least one of them is not overloaded
sufficient and reliable all the time since the study focuses only otherwise the load will be considered not supplied. The final
on the distribution systems. The load will be counted as step is to count how many time the states are changed in the
unavailable if one of two conditions is happen; even if the whole process to find the frequency of each load.
voltage profile is less than allowable limit or the whole tie sets
lines between the source and the load are exceeded their IV. RESULTS AND ANALYSIS
capacity. The complete algorithm is shown in fig. 1.
The study will be conducted in a simple system with known
Start parameters obtained from Saadat's book [6]. The system has
three buses and two loads as shown in fig. 2 and table I. One
load will assume to be residential and the other commercial and
Get Data the load data has been modeled as described on [5] as equation
(1) and fig. 3.
t=1
L(t) = Pw Pd Ph (t) (1)
L1(t)*RN
The reliability indices of the component will be taken from
RBTS [4] as shown in table I.
yes
Isolate that RN>Pr?
Component
100% Reliable
no Power Supply
Power Flow

yes
Vi>Vmin? U=U+1
Line 2 Line 1
no
Line 3
yes
Itie>Imax?

no
no L2 L1
f=f+1 s(t)=s(t-1)?

yes
Fig. 2. The single line diagram of the studied system
t=t+1
The capacity of the lines will be assumed to be apple to supply
no the load at N-1 contingency. The supply bus is assumed to be
t=8760? 100% reliable and can supply the load all the time. Load 1 and
Load 2 are connected to bus 2 and bus 3 respectively. The load
yes flow is done in 10 iterations since the system has only 2 load
End buses and bus 1 assumed to be slack. While the MCS is
simulated up to 1000 times where each time represents one year
and at the end the average of all those years is taken. Using
Fig. 1. One year simulation flowchart Monte Carlo Simulation the load indices are calculated as seen
in table II and fig. 3-5.
In the first step the data is gathered including system
connection, impedances, thermal limits, failure rates, rapier
times, load profile/forecast for one year and mean standard
deviation of the normal destitution. After the simulation started
the load is multiplied with random number (RN) obtained from
normal distribution. Then another RN obtained from uniform
random distribution is compared with probability to fail (Pr) to
isolate the component from the system at that hour. Then a load
 3

4.00
One Day Load Shape
3.00
1.1
R_L1 2.00
Load (pu)

0.6 Res. R_L2

1.00
0.1 Com.
0.00
-0.4 0.20 0.10 0.05 Ref.

Fig. 5. Repair time of the two loads at different standard deviation

Fig. 3. Load shape of one day
Note when the standard deviation is 0.5 we didn't include it in
TABLE I. THE SYSTEM COMPONENTS DATA fig. 3-5 since the different is significant as shown in table II and
the system become not sufficient to supply the loads. The
Component Failure rate Repair Impedance unavailability and repair time are measured in hour, while the
(f/y) time (hour) (Ohm) failure rate is measured in fail per year.
Line 1 0.3 5 0.02+j0.04 The interesting result is that the unavailability of the loads are
Line 2 0.45 5 0.01+j0.03 in the same order for three cases and slightly higher when the
Line 3 0.5 5 0.0125+j0.025 standard deviation is 0.2. The failure rate is almost doubled
when the uncertainty is considered with standard deviation
TABLE II. THE RELIABILITY INDICES OF THE CASE STUDY equal to 0.05 and keeps increase with increasing the standard
deviation. The other interesting result is that the repair time
0.50 0.20 0.10 0.05 Ref. sd slightly decreases when the uncertainty level increases.
L2 L1 L2 L1 L2 L1 L2 L1 L2 L1
55.65 27.41 6.30 4.91 4.72 3.73 4.83 3.74 4.45 2.56 U V. CONCLUSIONS
50.71 24.06 3.85 2.75 2.46 1.62 2.26 1.97 1.38 1.21 F
The impact of load uncertainty on the power system reliability
1.10 1.14 1.64 1.78 1.92 2.30 2.13 1.90 3.23 2.11 R is investigated using sequential Monte Carlo Simulation. The
line capacity and the voltage profiles are considered in this
study to be the failure cause. The results shows the change on
the reliability indices when the load follow normal random
7.00
distribution with different variances. The simulation shows that
6.00 the failure rate is affected directly with increasing the
5.00 randomness of the load. The effect of load variance on the
unavailability is less than the failure rate. On the other hand, the
4.00 repair time is decrease slightly of almost the same on all cases.
U_L1
3.00 Further studies could be done based on this paper using the
U_L2 same technique by modeling more practical and complicated
2.00
systems and using different type of data. Furthermore, this
1.00 study could be useful for planning of power system and
0.00 evaluate the reliability of specific plan under specific load
0.20 0.10 0.05 Ref. forecast.

VI. REFERENCES
Fig. 3. Unavailability of the two loads with different standard deviation
[1] M. Benidris, J. Mitra, "Composite power system reliability assessment using
5.00 maximum capacity flow and directed Binary Particle Swarm Optimization,"
North American Power Symposium (NAPS), 2013 , vol., no., pp.1,6, 22-24
4.00 Sept. 2013

3.00
f_L1 [2] R. Billinton, D. Huang, "Effects of Load Forecast Uncertainty on Bulk,"
2.00 IEEE Transactions on Power Systems., Vol. 23, No. 2 May 2008
f_L2
1.00 [3] A. K. Basu , S. Chowdhury and S. P. Chowdhury , D. Ray, P. Crossley,,
0.00 "Reliability study of a micro-grid power system," Universities Power
Engineering Conference, 2008. UPEC 2008. 43rd International, vol., no.,
0.20 0.10 0.05 Ref. pp.1,4, 1-4 Sept. 2008

Fig. 4. Failure rate of the two loads with different standard deviation
 4

[4] N. Allan, R. Billintom, I. Sjariej, L. Goal, S. K. So, "Reliability Test System

for Education Purposes Basic Distribution System Dates and Results, "IEEE
Transactions on Power Systems, Vol. 6, n. 2, May 1991.

[5] P. Wang, R. Billinton, Time sequential distribution system reliability worth

analysis considering time varying load and cost models, IEEE Transactions on
Power Delivery, v.14, no. 3, Jul 1999, pp. 1046-1051.

[6] H. Saadat, "Power System Analysis", Mc Graw Hill, Inc., 2004.