International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]

ISSN: 2349-6495(P) | 2456-1908(O)

Study of State Estimation Using Weighted Least

Squares Method
Majdoub Meriem, Sabri Omar, Cheddadi Bouchra, Belfqih Abdelaziz, Boukherouaa
Jamal,El Mariami Faissal, Cherkaoui Nazha
Laboratory of Electric Systems and Energy, ENSEM, University Hassan II of Casablanca, Morocco

Abstract—Power state estimation constitutes the core of critical operational functions of a power grid such as real-
the on-line security analysis function. The challenge time security monitoring, load forecasting, economic
number one of a state estimator is to provide the optimal dispatch, and load frequency control.
estimates of system state with minimum of measurement Most of network applications use the real-time data
data. This paper describes weighted least squares state provided by the state estimator. Therefore, an optimal
estimation method and investigates how the efficiency of performance of state estimation output is the ultimate
WLS state estimation changes according to 4 parameters: concern for the system operator. This need is particularly
number of measurements, measurement type, more in focus today due to deregulated and congested
measurement weight and level of noise. Different systems and smart grid initiatives. The output of the state
simulation cases are tested on 3-bus system and IEEE 14- estimator nearly represents a true state of the system.
bus system. The results show that accurate estimates of However, discrepancies may occur due to incomplete
system state can be obtained with minimum of measurements, meaning many variables are not measured
measurement data on condition to choose a good or data is not available, inaccurate network parameters,
combination of accurate measurements with a minimum and errors in measurements [1].
of voltage measurements and power injection Most state estimation programs in practical use are
measurements and these data should be properly formulated as overdetermined systems of non-linear
distributed throughout the system. For best results, the equations and solved as weighted least-squares(WLS)
two factors (weight and noise) must be combined to problems [2].
obtain the best estimation. Indeed, the most accurate This paper describes Weighted Least Squares method for
measurements (lower level of noise) should have greater state estimation of power system, investigates its
weight compared to bad measurements (higher level of characteristics and observes the effect of 4 parameters
noise), specially voltage measurements due to their big (Number of measurements, measurement type,
impact. measurement weights and level of noise) on the quality of
Keywords—Level of noise, Measurement type, state estimation. Both simple power system case (3 bus)
Measurement weight, Number of measurements, and a larger power system IEEE 14 bus test cases are
Voltage measurement,weighted least squares state utilized.
estimation method.
II. WLS METHOD
I. INTRODUCTION The starting equation for the WLS state estimation
Electric power system deals with the generation, algorithm is:
transmission, and distribution of electric energy. The = ℎ( ) + (1)
efficient and optimum economic operation and planning, where: z is the (mx1) measurement vector; x is an (nx1)
along with security of electric power systems, have state vector to be estimated; h is a vector of nonlinear
always occupied an important position in the power functions that relate the states to the measurements; and e
industry. In order to achieve these objectives, it is is an (mx1) measurement error vector. Clearly, m must be
essential for power engineers to accurately monitor the grater then n in order to have measured the n states and
power system operating states. An essential tool for have additional information to provide redundancy, m>n.
monitoring the power system is state estimation. In The measurement errors ei are assumed to satisfy the
energy control centers, power system state estimation is following statistical properties:
carried out in order to provide best estimates of what is First, the errors have zero mean: E(ei) = 0, i = 1, ..., m
happening in the system based on real-time measurement Second, the errors are assumed to be independent, such
and a predetermined system model. It is required in the that the covariance matrix is diagonal.

www.ijaers.com Page | 55
International Journal of Advanced Engineering
Engi Research and Science (IJAERS) [Vo
[Vol-3, Issue-8, Aug- 2016]
349-6495(P) | 2456-1908(O)
ISSN: 2349
Cov(e) = E (e, eT ) = R = diag{σ12 , σ22 , ….,σm2} (2) III. SIMULALATION RESULTS
The solution to the state estimation problem
pr can be This section presents a study of WLS state
formulated as a minimization of follow llowing objective estimation characteristics ththrough the observation of
function: the effect of the 4 followiwing parameters (Number of
measurements, measureme
ment type, measurement
∑ ℎ ℎ (3)
weight and level of noise)se) on the accuracy of state
To find the minimization of this objectiv
ctive function the estimation. The simulatio tions are tested on two
derivative should be set to zero. The dederivative of the systems: a simple 3-buss system and IEEE 14-bus
objective function is denoted by g(x): system presented below:
• Case Study Utilizing
ing a Three-Bus System
ℎ 0 (4) A simple case study off 33-bus system is shown in
where: H(x)= called the measure
urement Jacobian Figure2. Bus 1 is the refere
erence bus, bus 2 is the load
bus, and bus 3 is the generator
ge bus. The network
matrix. Ignoring the higher order termss of the Taylor
data are shown in the same fig
figure.
series expansion of the derivative of the objective
functions yields an iterative solution as sho
hown below:
!
# $ ℎ (5)
Where the gain matrix, G, is defined as:
%
$ (6)
For the first iteration of the optimization th
the measurement
function and measurement Jacobian shoul ould be evaluated
Fig.2: Case Stud
tudy of 3-bus system
at flat voltage profile, or flat start. A flat
at start refers to a
state vector where all of the voltage mag agnitudes are 1.0 • Case Study Utilizing
ing IEEE 14-bus system
per unit and all of the voltage angles are 0 degrees. In The system is shown in figu
igure 3. The network data files
conjunction with the measurements, thee next n iteration of can be downloaded from Power Systems Test Case
the state vector can be calculated again an and again until a Archive [6].
desired tolerance is reached [3,4].
The flowchart [5] of WLS method is shown wn in figure 1:

Fig. 3: IEEE 114 Bus Test case

For both test cases the me measurement data are chosen
from Newton Raphson loa load flow results [7,8] and
consist of three kinds of measurements: voltage
magnitudes, real and react active power injections, real
and reactive power flows.
f Weight of all
measurements is assumed ed 1. The true values of
voltage magnitude and angle are from Newton
Raphson load flow result ults. To compare the state
estimate accuracy of th the following simulations,
mean absolute percenta
ntage error (MAPE) is
introduced as follows [9]:
Fig. 1: The flow chart of WLS Method
M ./ 0/
'()* ∑+- , , 1 100% (34)
+ ./

www.ijaers.com Page | 56
International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]
ISSN: 2349-6495(P) | 2456-1908(O)
Where, (4 is the actual value and 54 is the variables: voltage magnitude(MPAEV) and voltage
calculated value. A smaller value of MAPE angle(MAPEθ).
indicates a more accurate state estimation result. 3.1.1 Simulation results for 3 bus system:
3.1. Effect of number of measurements:
The robustness of state estimation can be Table.1:Cases studies with different number of
guaranteed only if the number of the available measurements
measurements is high enough and properly m η V Pinj Qinj Pflow Qflow
distributed throughout the system. A measure of the Case1 21 4,2 3 3 3 6 6
number of measurements may be denoted by the
Case2 3 0 0 6 6
redundancy factor η, which is defined as [9]: 15 3
Case3 0 3 3 6 3
7 8+9 :+:; Case4 5 1 1 0 0 2 2
6 (35)
7 8+9 :+:; + <= Case5 <5 All possible combinations are tried
We will analyze the influence of the degree of The results are shown below in table2 and 3:
redundancy through 5 different cases for both 3-
bussystem and IEEE 14-bus system. The
comparison is set according to the two state

Table.2:WLS state estimation of voltage magnitude / different redundancy degree/3 bus system
True Value Estimated value of voltage magnitude by WLS (P.U) / 3 bus System
Bus ID
(P,U) Case1 Case2 Case3 Case4 Case 5
1 1 1 1,000002 1
2 0,8898 0,889791 0,889795 Gain 0,889796 Gain
3 1,05 1,05 1,050004 matrix 1,050006 matrix
MAPEV (%) 0,0004 0,0004 is Singular 0,0003 is Singular
Number of Iterations 5 5 5

Table.3:WLS State estimation of voltage angle/ Different redundancy degree/3bus system

True Value Estimated value of voltage angle by WLS (degree)/ 3 bus System
Bus ID
(degree) Case1 Case2 Case3 Case4 Case 5
1 0,0000 0,000000 0,000000 0,000000
2 -13,3116 -13,311451 -13,311332 -13,311422
Gain matrix Gain matrix
3 -4,2380 -4,237860 -4,237850 -4,237997
is Singular is Singular
MAPEθ (%) 0,0033 0,0035 0,0001
Number of Iterations 5 5 5

For the latest case (Case5), all possible

combinations of measurements types were tried but 3.1.2. Simulation results for IEEE 14- bus system:
all have not converged, this verify the observability Table.4: Cases studies with different number of
condition (m>=n). It is a necessary but not measurements
sufficient condition. In fact, we have a counter- m η V Pinj Qinj Pflow Qflow
example: Case 3, even if the number of Case1 120 4,4 14 14 14 38 40
measurements (m=15) is greater than number of
Case2 95 3,5 1 8 8 38 40
states (n=5), the algorithm has not converged. On
the other hand, for case 2 with the same number of Case3 41 1,5 1 8 8 12 12
measurements the algorithm has converged, Case4 27 1,0 1 6 6 7 7
indicating that WLS state estimation is affected by Case5 <27 All possible combinations are tried
the combination of measurements types chosen.
The results are presented in figures 4 and 5:

www.ijaers.com Page | 57
International Journal of Advanced Engineering
Engi Research and Science (IJAERS) [Vo
[Vol-3, Issue-8, Aug- 2016]
349-6495(P) | 2456-1908(O)
ISSN: 2349
As seen, increased redundanancy improves the accuracy of
the estimation, but not unifor
formly. In fact, although case 4
has a lower redundancy degree d a better solution is
obtained compared with casesses 2,3 which have a higher η.
We conclude, that the WLS LS state estimation is affected
not only by the number of measurements
me but also by other
correlated factors like meas
easurement type, measurement
location, measurement error…r….
In the following point, we will study the effect of
measurement type.
Fig.4: Comparison of WLS state estimati
ation of voltage 3.2. Effect of measurementt ty
type:
magnitude with different redund
ndancy The WLS algorithm is test ested on 6 different cases of
combinations of measureme ments types for both 3 bus
1 2 3 4 5 6 7 8 9 10 11 12 13 14
system and IEEE 14 bus systestem.
-3 3.2.1. Simulation results forfo 3 bus system:
TrueValue(P,U) Table.5:Cases studies with
ith different combinations of
Case1(120meas) measurem
ements types
-8 Case2(95meas) m η V P
Pinj Qinj Pflow Qflow
Case3(41meas)
Case4(27meas)
Case1 21 4,2 3 3 3 6 6
-13 Case2 18 3,6 0 3 3 6 6
Case3 15 3 3 0 0 6 6
Case4 12 2,4 0 0 0 6 6
-18
Fig.5: Comparison of WLS state estimati
ation of voltage Case5 9 1,8 3 3 3 0 0
angle with different redundancyy degree
d Case6 6 1,2 0 3 3 0 0

The nearest case to the true value is casee 1,

1 that’s obvious The results are shown in table
bles 6 and 7 below:
because we have a large number of measurements
(m=120).
Table.6: WLS State estimation off vo voltage magnitude /different combinations of measuremements types/3bus system
True
Bus ID Case1 Case2 Case3 Case4 Cas
ase5 Case6
Value(P,U)
1 1 1 0,999993 1,000002 1,000004 1 0,99998
2 0,8898 0,889791
791 0,889783 0,889795 0,889798 0,88
,889788 0,889766
3 1,05 1,05 1,049994 1,050004 1,050006 1,04
,049998 1,049979
MAPEV (%) 0,0004 0,0011 0,0004 0,0004 0,00
,0005 0,0026
Number of Iterations 5 6 5 6 5 6

Table.7: WLS State estimation of voltage angle /different combinations of measurement

ents types/3bus system
True Value
Bus ID Case
se1 Case2 Case3 Case4 Case5 Case6
(degree)
1 0 0 0 0 0 0 0
2 -13,3116 -13,311
11451 -13,311649 -13,311332 -13,311276 -13
13,311519 -13,312072
3 -4,238 -4,237
3786 -4,237919 -4,23785 -4,237833 -4,
4,237857 -4,238018
MAPEθ (%) 0,003
033 0,0019 0,0035 0,0039 0,0034 0,0004
Number of Iterations 5 6 5 6 5 6

Tables 6 and 7 shows that for cases (2,4

2,4 and 6) where deduce that the presence ce of voltage measurements
voltage measurements are missing, more
mo number of isnecessary for an efficient
ent execution of the program
iterations (6 iterations) is required and
a estimation without problems.
accuracy is lower compared with the othe
ther cases. So, we

www.ijaers.com Page | 58
International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]
ISSN: 2349-6495(P) | 2456-1908(O)
3.2.2. Simulation results for IEEE 14- bus system: injection measurements with some power flow
Table.8: Cases studies with different combinations of measurements.
measurements types In practice we usually use a redundancy factor η
m η V Pinj Qinj Pflow Qflow >=1,5. So, to be closer to the reality, we will
Case1 120 4,4 14 14 14 38 40 consider in the next studies, two models as follow:
- For 3 bus system: 7 measurements are taken
Case2 106 3,9 0 14 14 38 40
throughout the network (1V, 2 Pinj, 2 Pflow, 2
Case3 92 3,4 14 0 0 38 40
Qflow).- For 14 bus system: 41 measurements are
Case4 78 2,9 0 0 0 38 40 taken throughout the network (1V, 8Pinj,
Case5 42 1,6 14 14 14 0 0 8Qinj,12Pflow, 12 Qflow).
Case6 28 1,0 0 14 14 0 0 3.3. Effect of measurements weights:
As defined previously the measurement error
The results are presented in figures 6 and 7. covariance matrix R is a diagonal matrix of
TrueValue(P,U) Case1(120meas) Case2… Case5 measurement variances constituted by weights. So,
Wi(weight) =1/ σ²i, where σ²i assumed error
variance of measurement “i”.
1.05
In this point, we will study the effect of
measurements weights on the state estimation by
WLS.
0.95 In the previous simulations, we supposed that all
1 2 3 4 5 6 7 8 9 10 11 12 13 14 measurements had the same weight which was set
to 1. Now, two simulations will be presented: one is
Fig 6: Comparison of WLS state estimation voltage angle setting the same weight for the different
according to different combinations of measurements measurements, we only change his value. Another,
types different weights are tried according to the type of
measurements.
1 2 3 4 5 6 7 8 9 1 0 1True
1 1Value(degree)
2 13 14
3.3.1. Same weight for all measurements:
Case1 (120meas) 3 simulations are tried with different weights (σ=1,
-7.8 σ=0.1 and σ=0.001), for the two systems (3 bus and
Case2
IEEE 14 bus).The 3 bus system results are shown in
(106meas)
Case5 table 9 and table10.
(42meas) Table.9:Estimated value of voltage magnitude with the
same weight for all measurements
-17.8 Bus True
Fig 7: Comparison of WLS state estimation voltage angle ID Value(P,U) σ=1 σ=0,1 σ=0,001
according to different combinations of measurements 1 1 1,000001 1,000001 1,000001
types
2 0,8898 0,889797 0,889797 0,889797
For cases 3 and 4, Gain matrix is singular,so the 3 1,05 1,050007 1,050007 1,050007
program doesn't converge. In case 6, only power MAPEV (%) 0,0004 0,0004 0,0004
injection measurements are utilized, the program Iterations 5 5 5
diverges: Number of iterations >1000 and MAPE> Table.10: Estimated value of voltage angle with the same
100%. It means that the presence of power injection weight for all measurements
measurements without voltage measurements may Bus True Value
lead to convergence problems. ID (degree) σ=1 σ=0,1 σ=0,001
A good accurate solution is obtained in case 5.
1 0 0 0 1,000001
Therefore, the combination of voltage
2 -13,3116 -13,31149 -13,31149 0,889797
measurements with power injection measurements,
is better than the combination of voltage 3 -4,238 -4,237957 -4,237957 1,050007
measurements with power flow measurements. As MAPEθ (%) 0,001 0,001 0,0004
result, an optimal combination should necessary Iterations 5 5 5
contains a minimum number of voltage and power

www.ijaers.com Page | 59
International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]
ISSN: 2349-6495(P) | 2456-1908(O)
According to those results, we deduce that 3.3.2. Different weights according to the type of
Whatever the value of the weight as it is the same measurement:
for all measurements, the result doesn’t change. The 6 cases are tested for 3 bus and 14 bus systems; results are
same results are obtained for IEEE 14 bus system. shown below.
3.3.2.1. Simulation results for 3 bus system

Table.11: Estimated value of voltage magnitude with different weights according to measurements type
Bus True Measurements Variance [V, Power injection, Power flow]
ID Value(P,U) [0.1, e-6, e-6] [e-6, 0.1, e-6] [e-6, e-6, 0.1] [0.1, 0.1, e-6] [0.1, e-6, 0.1] [e-6,0.1,0.1]
1 1 1,000117 1 1 1,000004 1,000001 1
2 0,8898 0,889919 0,889796 0,889796 0,889801 0,889797 0,889796
3 1,05 1,050116 1,050006 1,050006 1,05001 1,050007 1,050006
MAPEV (%) 0,0121 0,0003 0,0004 0,0005 0,0004 0,0003
Iterations 6 5 5 5 5 5

Table.12:Estimated value of voltage angle with different weights according to measurements type
Bus True Value Measurements Variance [V, Power injection, Power flow]
ID (degree) [0.1, e-6, e-6] [e-6, 0.1, e-6] [e-6, e-6, 0.1] [0.1, 0.1, e-6] [0.1, e-6, 0.1] [e-6,0.1,0.1]
1 0 0 0 0 0 0 0
2 -13,3116 -13,307962 -13,311422 -13,311513 -13,311301 -13,311492 -13,311507
3 -4,238 -4,237056 -4,237997 -4,237947 -4,237964 -4,237942 -4,237961
MAPEθ (%) 0,0223 0,0001 0,0013 0,0008 0,0014 0,0009
Iterations 6 5 5 5 5 5

In the first case, voltage magnitude measurement has the

lowest weight compared with power injection and power 1 2 3 4 5 6 7 8 9 10 11 12 13 14
TrueValue(P,U) [0,1,e-6,e-6]
flow measurements. As seen in the tables 11 and 12, this
[e-6,0,1,e-6] [e-6,e-6,0,1]
weight combination requires a higher number of iterations [0,1,0,1,e-6] [0,1,e-6,0,1]
for convergence and yields the biggest deviation from the -8.6 [e-6,0,1,0,1]
true value for both states: voltage magnitude and voltage
angle. On the other hand, when voltage measurement has
a greater weight, the results are better.
3.3.2.2. Simulation results for IEEE 14 bus system:
TrueValue(P,U) [0,1,e-6,e-6] -18.6
[e-6,0,1,e-6] [e-6,e-6,0,1] Fig.9: State estimation of voltage angle with different
[0,1,0,1,e-6] [0,1,e-6,0,1]
[e-6,0,1,0,1] weights according to measurements type
1.03 We conclude that voltage magnitude error produce a large
deviation on the state estimation. So, it would be
interesting to choose voltage measurements with small
errors and to minimize noise disturbance which could
0.93 affect measurements' quality for this type.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 3.4. Effect of level of noise:
Fig.8: State estimation of voltage magnitude with In this point we will study the effect of noise on the
different weights according to measurements type accuracy of WLS state estimation. Two simulation cases
are tested: one assuming the same level of noise for all
In those figures 8 and 9, it is noticed that the simulations measurements type and the other by changing the level of
result of the first case are far from the true values noise according to the type of measurement.
(MAPEV=7% and MAPE θ= 16%). Also, the 3.4.1. Same level of noise for all measurements:
convergence of that case requires a higher number of 3.4.1.1. Simulation results for 3 bus system:
iterations: 11 counter only 5 iterations for the other cases.
This joins the results of 3 bus system simulations.

www.ijaers.com Page | 60
International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]
ISSN: 2349-6495(P) | 2456-1908(O)
Table.13: Estimated value of voltage magnitude with the
same level of noise for all measurements TrueValue(P,U) Whithoutnoise
True 1%noise 3%noise
Bus
Value Without 1% 3% 6% 1.09 6%noise
ID
(P,U) noise noise noise noise
1 1 1,000001 1,009977 1,029931 1,059867 1.04
2 0,8898 0,889797 0,899511 0,91896 0,948179
3 1,05 1,050007 1,059945 1,079826 1,109657
0.99
MAPEV(%) 0,0004 1,0121 3,037 6,0764 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Iterations 5 5 5 5
Fig.10: Estimated value of voltage magnitude with the
Table.14: Estimated value of voltage angle with the
same level of noise for all measurements
samelevel of noise for all measurements
True -1.36
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Bus TrueValue(P,U)
Value Without 1% 3% 6%
ID Whithoutnoise
(degree) noise noise noise noise -6.36 1%noise
1 0 0 0 0 0 3%noise
2 -13,3116 -13,311 -13,1620 -12,8726 -12,4613 6%noise
-11.36
3 -4,238 -4,2379 -4,1997 -4,1252 -4,0182
MAPEθ (%) 0,001 0,9036 2,6614 5,1857
-16.36
Iterations 5 5 5 5
Fig.11: Estimated value of voltage angle with the same
The accuracy of the state estimation changes level of noise for all measurements
proportionally to the level of noise applied to the 3.4.2. Different level of noise according to the
measurements. measurements type:
3.4.1.2. Simulation results for IEEE 14 bus system: 6 different cases are studied, to assess the effect of noise
As seen in figures 10 and 11, the more the level of noise applied to the various measurement types with different
applied to measurements is important (6%), the more the levels on the estimation quality.
deviation between estimation results and the true values is 3.4.2.1. Simulation results for 3 bus system:
bigger (MAPE V=4%).
Table.15: WLS state estimation of voltage magnitude with different level of noise for measurements according to their type
True Value %Noise [V, Power injections, Power flow]
Bus
(P,U)
ID
[6%, 1%, 1%] [1%, 6%, 1%] [1%, 1%, 6%] [6%, 6%, 1%] [6%, 1%, 6%] [1%,6%,6%]
1 1 1,059797 1,012466 1,007988 1,061857 1,058127 1,01011
2 0,8898 0,952286 0,900441 0,894964 0,952605 0,948169 0,895545
3 1,05 1,107075 1,062506 1,060556 1,109166 1,107887 1,062749
MAPEV (%) 6,146 1,2112 0,7948 6,2929 5,9619 0,957
Iterations 4 5 5 5 5 5
Table.16:WLS State estimation of voltage magnitude with different level of noise for measurements according to their type
Bus True value %Noise [V, Power injections, Power flow]]
ID (degree) [6%, 1%, 1%] [1%, 6%, 1%] [1%, 1%, 6%] [6%, 6%, 1%] [6%, 1%, 6%] [1%,6%,6%]
1 0 0 0 0 0 0 0
2 -13,3116 -11,800027 -13,743553 -13,307602 -12,332684 -11,916925 -13,904794
3 -4,238 -3,843578 -4,24089 -4,344744 -3,88559 -3,974423 -4,388224
MAPEθ (%) 9,3068 0,0682 2,5187 8,3155 6,2194 3,5447
Iterations 4 5 5 5 5 5

As noticed in tables 12 and 13, for cases (1,4 and 5) deviation from the true value is important for both states
voltage measurements are affected with high level of (voltage magnitude and voltage angle). On the other hand,
noise (6%) compared with other types, therefore the

www.ijaers.com Page | 61
International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]
ISSN: 2349-6495(P) | 2456-1908(O)
for cases (2,3 and 6), where voltage measurements are the effect is not uniform. In fact, satisfying solution
less affected by noise, the accuracy of estimation is better. may be obtained without redundancy η=1, on the
3.4.2.2. Simulation results for IEEE 14 bus system: other hand the system may be unobservable even
TrueValue(P,U) with high degree of redundancy which means that
[6%,1%,1%]
[1%,6%,1%] state estimation is affected by other correlated
1.1
[1%,1%,6%] factors as measurement type, measurement location,
[6%,6%,1%]
measurement error…
The results show also the importance of voltage
1.05 measurements compared with the other types:
therefore, their presence is indispensable for an
efficient execution of WLS state estimation
1 program without problems and they should be
1 2 3 4 5 6 7 8 9 10 11 12 13 14 accurate as possible because voltage measurement
error produces a large deviation in final results.
Fig 12: WLS state estimation of voltage magnitude with
The study of the effect of measurement weights and
different level of noise according to measurements type
noise, depicts that those factors must be combined
to obtain the best estimation. Indeed, the most
1 2 3 4 5 6 7 8 9 10 11 12 13 14
accurate measurements (lower level of noise)
TrueValue(P,U) should have greater weight compared to bad
[6%,1%,1%] measurements (higher level of noise).
-7.07 [1%,6%,1%]
[6%,6%,1%]
[6%,1%,6%] REFERENCES
[1] Mukhtar Ahmad , "Power System State Estimation",
Artech House , 2013.
[2] A. Monticelli, “State Estimation in Electric Power
-17.07 Systems: a Generalized Approach”, New York:
Fig 13: WLS state estimation of voltage angle with Kluwer, 1999.
different level of noise according to measurements type [3] Ali Abur and Antonio Gomez Exposito,” Power
System Estimation:Theory and Implementation”,
For voltage magnitude, the difference between cases Marcel Dekker, Inc., 2004. 1989.
is more pronounced than voltage angle's estimation. [4] Nursyarizal Mohd Nor, Prof. Dr. Ramiah Jegatheesan
The results are similar to those obtained in the case and Ir.Perumal Nallagowden, “Newton-Raphson
of assigning different weights to measurements by State Estimation Solution Employing Systematically
type. In fact, when voltage measurement is affected Constructed Jacobian Matrix”, World Academy of
by a higher level of noise 6% (Cases: 1,4 and 5), the Science, Engineering and Technology, pp.42, 2008.
error become more important for both states [5] Pankaj Tripathi, Jinendra Rahul and Nim Anurag
MAPEV and MAPEθ. On the other hand, with a Radhamohan “A Weighted Least Square Technique :
lower noise 1%( cases 2,3 and 6) the error is small. For Assessment of State Estimation of Power
We conclude that the two factors (weight and noise) System” Skit Research Journal, Volume 5, Issue 2 :
must be combined to obtain the best estimation. 2015 .
Indeed, the most accurate measurements (lower [6] Abdelaziz Belfqih, Faissal El Mariami, Jamal
level of noise) should have greater weight compared Boukherouaa, « Stratégie de manœuvre pour la
to bad measurements (higher level of noise). conduite des réseaux électriques », congrès
International d’analyse numérique et d’optimisation,
IV. CONCLUSION Essaouira ,Novembre 2013.
This paper describes Weighted Least Squares state [7] Abdelaziz Belfqih, Faissal El Mariami, Hicham
estimation method, investigates its characteristics Medromi, Jamal Boukherouaa, N. Belaguide and A.
and observes the effect of 4 parameters (Number of Sayouti, « Improving the performance of a real THT
measurements, measurement type, measurement network in the Casablanca region », International
weight and level of noise) on the quality of state journal of applied information systems, New York-
estimation. USA, December 2013.
The simulations show that increased redundancy
improves the accuracy of the estimation, but the

www.ijaers.com Page | 62
International Journal of Advanced Engineering Research and Science (IJAERS) [Vol-3, Issue-8, Aug- 2016]
ISSN: 2349-6495(P) | 2456-1908(O)
[8] Jiaxiong Chen and Yuan Liao “State Estimation and
Power Flow Analysis of Power Systems”, journal of
computers ,vol,7, NO.3,March2012.
[9] D. P. Kothari, I. J. Nagrath, “ Modern Power System
Analysis”, Tata McGraw-Hill Education India
Private Limited, New Delhi, 2008.

www.ijaers.com Page | 63