Optimal Placement of PMU S Using Greedy Algorithm and State Estimation

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

t

1s IEEE International Conference on Power Electronics. Intelligent Control and Energy Systems (ICPEICES-2016)

Optimal Placement of PMUs using Greedy


Algorithm and State Estimation

Varsha Jaiswall, Siddhartha Sankar Thakur2 and Biswaranjan Mishra3


I School of Electrical and Electronics, LovelyProfessional University, Phagwara, India
23
, Electrical Engineering Department, National Institute of Technology, Durgapur, India
2
E-mail: Ivarsha.19642@lpu.co.in. sst@ee.nitdgp.ac.in.3biswaranjan.ee@gmail.com

Abstract-This paper proposes a new technique of state Accordingly, emphasis has been given in literatures
estimation (SE) for electric power systems. In the proposed for optimal placement of PMU in the system, retaining
scheme, the Phasor Measurement Units (PMU) are first complete observability to minimize its nwnber [7-10].
placed optimally using Greedy Aigorithm for cost reduction, Baldwin et al [11] has also presented a novel method to
while complete observability of system is also obtained. The
obtain optimal placement of PMU for a given power
SE uses a linear measurement model to obtain the estimated
system so that the number ofPMUs are reduced while the
states directly, without any iteration, thereby improves the
system is fully observable.
quality of the estimated data base. To reveal the efficacy of
the proposed scheme it has been tested on standard IEEE 5-
The measurements from PMUs are transmitted over
bus, 14-bus, 30-bus, 57-bus and 118-Bus test systems and the different communication channels through suitable
test results are presented. modem. The measurement vector thus obtained is given as
Keywords-Phasor Measurement Unit; Synchronized an input to the state estimator whose function is to find out
Measurement; Optimal PMU Placement; Greedy Algorithm the best estimate of the power system state variables.

I. INTRODUCTION 11. STATEMENT OF SE PROBLEM


The novel concept of SE in power systems was The WLS is the most popular and widely used
introduced by Schweppe et al. [1-3]. It is the process of technique for solution of SE problem.
assigning a set of values to unknown system state
variables based upon some criterion, that provides the best A. SE using WLS
estimate of the system states by either maximizing or For a power system with rn-dimensional measurement
minimizing that criterion. Traditionally, Weighted Least vector z and n-dimensional state vector x may be
Square (WLS) technique is used to solve the problem of modeled as,
SE which makes use of nonlinear iterative method. But the X
above technique is quite complex and also computational Z=fC )+e (1)
time is considerably higher than that is necessary for SE where, f(x) is a m-dimensional vector of power tlow
using synchronized measurements from PMU, which uses
equations and e is the rn-dimensional noise vector.
linear filtering method.PMU is a instrwnent, that provides
In WLS technique, objective is to minimize the sum
synchronized measurements of complex bus voItage and
of squares of weighted deviations of actual measurements
complex line currents [4-5].
The biggest advantage of PMU over analog meters is from the estimated measurements, represented by J( x)

(
that through the use of dock pulse of G lobal Positioning and is known as measurement residual. Thus,
Satellite (GPS), all the PMUs in a power system can be [z;eas - X)]2
synchronized in time thus elevating the power system J(x)= I
i=l O'i
(2)
monitoring, control and protection standards to a much
higher level. It also improves the bad data detection and where, Nm is the nwnber of independent
identification, thereby uplifting the accuracy of the measurements and (J;2 is the variance for the ith
estimated states. Timing pulses from the GPS system are measurement.
used for obtaining synchronization by simuItaneous Equation (2) can be written in compact form as
sampling of voItage and current phasors [6].
The biggest disadvantage of analog meters is their J(x)=[z- h(.x)fR-1[z- h(x)] (3)
dependency upon time. When a disturbance occurs at any where R is known as the measurement error co­
substation, they may or may not trigger at that very variance matrix.
instant. Hence, the disturbance may remain immeasurable. Minimization of equation (3) yields an iterative
This makes the use ofPMUs a must for avoidance of such solution as
problems. But the problem associated with the use of
PMU is its high cost. Llx = [HT R-1H r HT R-1t,z (4)

978-1-4673-8587-9/16/$31.00 ©2016 IEEE [1]


t
1s IEEE International Conference on Power Electronics. Intelligent Control and Energy Systems (ICPEICES-2016)

Where, His known as the Jacobian matrix and the f1z


is the vector of deviations of actual measurements from
E = {I,0 lf that particular
Otherwise
bus is observable
(7)
,
the estimated measurements. Simultaneously, row and column corresponding to the
PMU bus are deleted from the Orpmu matrix. When all the
B. Proposed Method 0/SE
elements of visibility matrix become 1 then the complete
In this paper a novel scheme of SE is proposed. Here, system is said to be observable.
the overall problem is subdivided into parts, viz. optimal For an example, IEEE 5-bus test system with 7 lines
placement of PMUs to reduce expenditure for its is considered. For this system the Orpmu is given as
deployment maintaining complete observability of the 1 1 1 0 0
system and optimal fiItering. 1
1 o
C. Optimal Placement 0/PMUs
Orpmu = (8)
o 1
As discussed above, PMU provides us with the o 1 0 1 1
voItage and current phasor measurements. So, if PMU is Therefore, the problem of optimal placement ofPMU
placed at any bus it not only provides us with the voItage for the above example can be formulated as
of that bus but also the branch current phasors of all the 1 1 1 0 0 x1 1
s
lines connected to that bus. So on placingPMU at one bus
not only that bus but all the connected buses also become 2
i=
>i 1 x2
visible. Thus, there is no necessity of placing PMU at Min l 1 ° x3 2 (9)
these buses resuIting in cost reduction. ° 1 x4
As per G reedy algorithm [12] PMUs are placed at the
X
° ° s
buses with maximum connectivity, so that all the
connected buses also become automatically observable. Here we can see that bus number 2 has highest
This results in a vast decrease in the number of PMUs connectivity, as it is connected with all the other buses
required for complete observability. present in the system. So if we place a PMU at this bus
In order to execute this algorithm a matrix Orpmu is then the whole system will become observable.
determined, whose elements are defmed as
D. Optimal Filtering using Synchronized
_ {Iif i and j are connected or i=j
Phasor Measurements
Orpmu - . (5)
° otherwlse
The final outputs of thePMUs are transmitted through
The Orpmu is called the connectivity matrix, having
various communication channels and are used as inputs to
dimension ofnbus x nbus where nbus is the number of bus in
state estimator, whose function is to find out the best
the system. It is obtained by converting elements of
estimate of state variables using the linear measurement
admittance matrix into binary form. For optimal PMU
model
placement the objective is to minimize LXi, where Xi is the
Z= C.V+TJ (10)
PMU buses (if PMU is placed at that bus then its element
Where, Z is complex measurements vector, Cis an Nm
is 1 else 0) such that,
X Ns coefficient matrix whose elements are either 1 or ° or
Orpmu'X 2 [l]nbusxl function of line parameters and connectivity of buses. TJ is
a complex random variable vector with properties of white
Where, X= [x J , X2 ... xnbuS] noise i.e.
If above inequality is satisfied then it means that the E(TJJ = ° Vi (11)
whole system is observable (if any element of the matrix and correlation vector is defined as,
on RHS is ° then it means that particular bus is
if i*- j
{O,
unobservable) . E[TJ,·TJj]= (12)
2 2
In order to find the bus with maximum connectivity (JRi +(J/,
sum of each column of the connectivity matrix Orpmu is
determined. Column number with highest value of this where, (j�i and a;, represents variance of real and
sum corresponds to the bus with maximum connectivity imaginary part of error T]i.
and aPMU is placed at that bus. This means As equation (12) is linear in nature, the estimated
Sum= IOrpmu (6) complex bus voItage vector is obtained directIy without
any iteration as
will be a (1 x nbus) matrix. When PMU is placed at
any bus the visibility matrix is updated which is again a (1 V= [CT R;lCr'CT R;'Z (13)
x nbus) matrix whose elements E [0, 1] depending upon Which is similar to that obtained for WLS technique
observability of that bus, i.e. but since here C is a linear coefficient matrix therefore

[2]
t
1s IEEE International Conference on Power Electronics. Intelligent Control and Energy Systems (ICPEICES-2016)

solution can be obtained in single iteration only, thus under different normal operating conditions. However, for
reducing the complexity of the problem. the sake of brevity, some of the important resuIts are
Another advantage ofPMU over conventional method portrayed only for the stated simulated conditions.
of state estimation is its low redundancy level. Where, the The resuIts of optimized PMU placement for the five
ratio of number of measurements input to the system to IEEE test systems using G reedy algorithm for full
the number of states to be estimated is defined as observability are shown in Table 1. This table reveals the
Redundancy. effectiveness of the G reedy algorithm in reducing the
number of PMUs retaining full observability of the
III. COMPUT ATIONAL PROCEDURE systems for SE. Table 2 shows the properties of the five
For implementation of the proposed scheme the main test systems along with number of measurements and
computational steps are as folIows: redundancy used for simulation studies. From this table it
Step J: Read the available line data. is clear that SE using PMU requires much less
Step 2: Determine the bus connectivity matrix using measurements compared to WLS technique, where
the admittance matrix. redundancy used in general is 2.0.
Step 3: Find out the highest degree bus and place a The comparison of estimated active and reactive
PMU at it. powers at different important buses of the five IEEE test
Step 4: If two or more bus es have same connectivity systems with true values are shown in figures 1 to 10. In
then randomly choose any one of them and all these figures it is clear that estimated values are very
placePMU at that bus. close to the true values of active and reactive powers.
TABLE I: OPTIMAL PMU PLACEMENT
Step 5: Update the visibility matrix.
Step 6: Check if all the buses are observable or not. If Bus System NO.ofPMU PMU Placement Buses
not, then go to 3rd step, else the resuIt ofPMU IEEE 5- bus I 2
placed buses is given as an input to the state IEEEI4-bus 5 4,6,2,9 and 7
estimator. IEEE 3O-bus II 6, 1 0, 12,27,2, 15,24,3, 19,9 and 25
IEEE 57-bus 21 9, 13,38, 1 ,4,4 1 ,6, 15,24,29,32,36,
Step 7: Using the available data carry out SE.
56, 19,2 1 ,27,30,39,46,50 and 53
Step 8: Print the results.
IEEEI 18-bus 40 50, 1 0 0, 13,80, 18,38,60, 1 ,6,33,85,
92, 1 05, 16,57,63,66,70,77,96, 1 1O,
IV. SIMULATION RESULTS
20,24,28,3I,4 1 ,47,2, 1 0,22,30,35,
The efficacy of the proposed scheme has been 45,52,54,69,7 1 ,75,89 and 86
TABLE 2: REDUNDANCY USED FOR SIMULATION STUDlES
evaluated by testing it on IEEE 5-bus, 14-bus, 30-bus, 57-
bus and 118-bus test systems under simulated operated Test No.of No.ofComplex State Variables Redundancy
conditions and the test resuIts have been compared with System Lines Measurements (Voltage

the true values obtained from successive load tlow Magnitudes and
Phase Angles)
solutions.
5 7 4+ 1=5 9 I.I I I I
A. Description ofSimulation 14 20 20+5=25 27 1.85I9
30 41 43+ 1 1=54 59 1.8305
The simulated operating conditions have been 57 80 70+2 1=9 1 1 13 1.6I 06
obtained by linearly varying the load at each bus from 179 179+40=2 19 235 1.8638
70% to 120% in 30 time steps. The system jitter is Note: Number of eomplex Measurements=Number of eomplex Ime
eurrent measurements of eonneeted liDes plus number of eomplex
considered by a normally distributed random tluctuation
voltage measurement ofbuses at whieh PMU are plaeed.
o f zero mean and standard deviation of 2% of the
trend component. The system power factor is
O�
considered to be constant so that the reactive power
follows the active counterpart. The total load change is o.l

considered to be shared among the generators as per �


� Ol
e
participation factors. The true values of the Q

measurements are obtained by running successive i 01

0
load tlows. The simulated measurements of c o m p l e x
b u s v o I t a g e s a n d c o m p l e x line f10ws a r e obtained �
.01
by adding a normally distributed error of zero mean and
standard deviation of 2% of the true values. .02

00�
B. Discussions of Results � 10 , ."
T""«kl
The performance of the proposed technique of SE has
Fig. I: Comparison of Aetive Power at 2nd Bus ofIEEE 5 Bus
been tested extensively on all of five IEEE test systems

[3]
t
1s IEEE International Conference on Power Electronics. Intelligent Control and Energy Systems (ICPEICES-2016)

- OONe)
0.4' -QfPMU)

.,
TtIIIII1(I
Fig. 6: Comparison ofReactive Power at 25'h Bus ofIEEE 30 Bus
Fig. 2: Comparison ofReactive Power at 2"d Bus ofIEEE 5 Bus
�==�
03------------------ --�� --

0.25 02
C-:;=�J
0.2
01S

l 0.1
� O.o!

i
.! -'0
D
.05

� .0;
.0 ;$
.Q.2
-'0.25
o '0 1�
o 10 1� 20 T
TIi'M(l(I
Fig. 7: Comparison of Active Power at 42"d Bus ofIEEE 57 Bus
Fig. 3: Comparison of Active Power at 12'h Bus ofIEEE 14 Bus
03r-----------------1=����
0.3
I--�-�;:======;l
Q(trU13:)
j - Q(PMU) 02 1 :::��J� l
-:­
0.2 --
I D1

"'- ...
:::>

0.1 0
;:-c

i ·01
l 0
. .,
.e ·0.1
j
IZ
. .o2
a........
/\ -..

:8l -03
'Cl:
·0.2

.0
o 10 20 30 ��---5�- -�1�O�--�1��--���-�·
Time(k) TlmeOO

Fig. 4: Comparison ofReactive Power at 12'h Bus ofIEEE 14 Bus Fig. 8: Comparison ofReactive Power at 42"d Bus ofIEEE 57 Bus

Fig. 5: Comparison of Active Power at 25'h Bus ofIEEE 30 Bus Fig. 9: Comparison of Active Power at 1 0 1 'hBus ofIEEE 1 18 Bus

[4]
t
1s IEEE International Conference on Power Electronics. Intelligent Control and Energy Systems (ICPEICES-2016)

[2] F. C. Schweppe and D. Rom ,"Power system static state estimation,


part 11: Approximate Model, " IEEE Trans. Power App. Syst., vol.
PAS-89,pp. 125- 130, 1970
[3] F. C. Schweppe, "Power system static state estimation, part lll:
Implementation, " IEEE Trans. Power App. Syst., vol. PAS-89, pp.
130- 135, 1970
[4] AG. Phadke, J.S. Thorp and KJ. Karimi, "State estimation with
phasor measurements", IEEE Transactions on Power Systems,
Vol.!, No.l, 1986J. Clerk Maxwell, A Treatise on Electricity and
Magnetism, 3rd ed.,vol. 2. Oxford: Clarendon, 1892,pp.68-73.
[5] A G. Phadke, "Synchronized phasor measurements in Power
Systems", in Proc. IEEE Computer Applications in Power, vol. 6,
no. 2,pp. 1 0- 1 5,Apr. 1993.
[6] Phadke AG "Synchronized Phasor Measurements - A Historical
Overview", 0-7803-7525-4/02 © 2002 IEEE,pp.476-479
[7] Reynaldo F. Nuqui, Member, IEEE, and Arun G. Phadke, Life
Fig. 1 0: Comparison ofReactive Power at 1 0 1 thBus oflEEE 1 18 Bus Fellow, IEEE, " Phasor Measurement Unit Placement Techniques
for Complete and Incomplete Observability" IEEE Transactions On
Power Delivery, Vol. 20,No. 4,October 2005.
V. CONCLUSION
[8] Preksha P. Dalawai and A R. Abhyankar, Member, IEEE
This paper has presented a new method of optimal 'Placement of PMUs for Complete and Incomplete Observability
using Search technique' 978-1 -4799-2275-8/ 13/$3 1.00 ©20 13
placement of PMU to reduce number of costly PMUs, IEEE
maintaining complete observability of the systems. At [9] V. Madani, M. Parashar, J. Giri, S. Durbha, F. Rahmatian, D. Day,
fiItering step a linear measurement model is used to M. Adamiak, and G. Sheble, "PMU placement considerations a
enhance the efficacy of the estimation. The use of linear roadmap for optimal PMU placement, " in Power Systems
Conference and Exposition(PSCE),20 1 1 IEEE/PES,March,pp. 1-7
measurement model provides the estimated states directly, [ 1 0] D. Dua, S. Dambhare, R. K. Gajbhiye, and S.ASoman, "Optimal
without any iteration. The performance of the proposed multistage scheduling of PMU placement: An ILP approach , "
scheme has been tested on IEEE 5-bus, 14-bus, 30-bus, IEEE Trans. Power DeI., vol. 23, no. 4,pp. 18 12- 1 820,Oct. 2008
57-bus and 118-bus test systems and compared with true [lI] T. L. Baldwin, L. Mili, J. M. B. Boisen, and R. Adapa, "Power
system observability with minimal phasor measurement placement,
values under various sirnulated conditions. The test resuIts "IEEE Trans. Power Syst., vol. 8, no. 2,pp. 707-7 15,May 1993
presented reveal the effectiveness of the proposed scheme. [ 12] Thomas H. C. , CharIes E. L., Ronald L. R. and Clifford S.,
"Introduction to A1gorithms", Third Edition, The MIT Press, July,
REFERENCES 2009.

[I] F. C. Schweppe and J. Wildes, "Power system static state


estimation, part I: Exact model, " IEEE Trans. Power App. Syst.,
vol. PAS-89,pp. 120-125, 1970.

[5]

You might also like