OPTIMAL SELECTION OF PMU LOCATIONS FOR ENHANCED STATE

ESTIMATION IN SMART GRID

M. Rihan1, M. Anas Aness1, M. Danish Anees1, M. Ahmad1 and M. S. Beg2
1
Electrical Engineering Department, Aligarh Muslim University, Aligarh, India
2
Electronics Engineering Department, Aligarh Muslim University, Aligarh, India
[email protected],[email protected],[email protected],mirzasalimb
[email protected]

Abstract: Installation of modern sensing devices is an An upcoming technology is the synchrophasor

integral requirement in the transition of a conventional measurement of voltage and current in the power grid
power grid to smart grid. Phasor measurements syn- by phasor measurement units (PMUs). These measure-
chronized through GPS presents one such alternative to ments are synchronized via the Global positioning sys-
enhance the monitoring and control feature in the power tem (GPS). In fact some utilities in the world have al-
grid. Due to the high financial implications, these units ready installed PMUs in their networks on a limited
should be deployed optimally in the grid. The methods scale. PMUs provide dynamic visibility into the opera-
reported in the literature for optimal placement usually tion of power grid at a much faster rate as compared to
provide multiple solutions. This paper presents a criteri- the arrangement in conventional grid [3].
on for selection of the best solution out of multiple High cost of PMU as well as technical and eco-
combinations. The criterion is based on improving the nomic requirements of the associated communication
state estimation results by the installation of phasor circuitry limits their widespread installation. However it
measurement units. The reason behind selection of this is still possible to derive the benefits of PMU installa-
criterion is that all the downstream applications of con- tion if some of the buses are monitored directly by in-
tingency analysis; steady-state, voltage, and dynamic as stallation of PMU and some are monitored indirectly, by
well as the optimal power flow are dependent on the the PMUs installed on connected buses. Determining a
accuracy of the state estimation results. Because of its minimum number of PMUs to make the system observ-
key role in energy management system, state estimation able is often referred to as Optimal PMU Placement
is set to be a vital function in the development of smart (OPP) problem.
grid. The OPP is a well researched problem and an ex-
tensive review of the algorithms reported in the litera-
Keywords: Smart Grid, State Estimation, Phasor Meas- ture is available in [4]. The methods reported mostly
urement, PMU. minimize the number of PMUs as a cost effective meas-
ure. However a PMU placement scheme should ap-
INTRODUCTION proach the problem in a more general and dynamic per-
spective in which both economical and technical aspects
Power grids throughout the world have evolved as huge are incorporated. Furthermore most of the OPP algo-
interconnected networks spread over large geographical rithms do not provide a unique solution. Hence some
areas. This arrangement has worked satisfactorily for a qualitative parameters should be developed to compare
long time but in recent years it has been experienced these solutions and provide a unique placement strategy.
that the conventional model of power grid has reached This paper presents a criterion for placement of
its limit. It is becoming increasingly difficult to main- PMUs based on best state estimation results. State esti-
tain a secure operation of the grid. Therefore a paradigm mation (SE) is a key function of energy management
shift is required in the way power grid has been moni- system and all the downstream applications of contin-
tored and controlled. It necessitates utilization of mod- gency analysis; steady-state, voltage, and dynamic as
ern sensing, communication and computation tools at all well as the optimal power flow are dependent on the
the stages in the grid. Smart grid is the name given to accuracy of the SE results. Therefore any improvement
such a modern and futuristic power grid and it is ex- in the SE solution will be helpful in making the power
pected to operate with enhanced efficiency, stability, grid more aware of its operating state, a fundamental
and flexibility [1]. requirement of the smart grid.
However a smart grid will not be a new structure Section 2 of the paper explains the effect of
altogether rather it will gradually emerge from the exist- phasor measurements on state estimation. Section 3 pre-
ing infrastructure with the introduction of new technol- sents the formulation of algorithm for hybrid state esti-
ogies at different levels of the grid. Therefore the utili- mation in the presence of both conventional and phasor
ties should start deploying and managing the existing measurements. Section 4 presents the criterion for
and emerging technologies in a better way on the way to placement of phasor measurement units. The case study
a smart grid [2]. results and discussion are presented in section 5. The

560
last section of the paper presents the conclusions drawn between measurements and state variables is linear and
from the present work. a non-iterative least square solution may be used to de-
termine the system state [6]. The PMU provides a syn-
STATE ESTIMATION AND PHASOR chronized measurement of the voltage phasor at a bus
MEASUREMENT and the current phsors through the branches associated
incident on it. The measurements are synchronized via
The electric power system consists of a large number of the GPS signal with a frequency of about 30 samples
interconnected elements forming a geographically huge per second. Due to technical and economical constraints
and complex dynamic system. Because of its nature the it may not be feasible to install PMUs at every bus of
system experience disturbances of varying scale as the the system. However a system can still be made observ-
operating conditions change. A set of imminent disturb- able by placing PMUs only on few selected buses.
ances is referred to as contingencies. Power system se- Therefore the existing SE can be improved by using
curity is defined as the ability of the system to survive data from a few PMUs installed at critical locations.
plausible contingencies without interruption to the pow-
er supply. In order to maintain a normal secure opera- HYBRID STATE ESTMATION
tion it becomes mandatory to keep a watch on the power
system. The operating conditions of a power system at Hybrid State Estimator is used to estimate the states of a
any given instant can be determined if in addition to the system monitored by conventional and Phasor Meas-
network topology, the complex voltage phasor at every urement Units. As discussed in the previous section, one
system bus are known. Since the set of complex voltage of the remarkable properties of Synchrophasor meas-
phasors completely specifies the system, it is referred to urement is to convert non linear state estimator to a lin-
as the system state. State estimation function utilizes ear state estimator. This means that the relation between
telemetered measurements to generate an optimal esti- the measured quantities and the system state is linear
mate of the system state. and states can be calculated (rather than being estimat-
For reliable state estimation, it is necessary that ed) using measured data).
the number of measurements should be greater than the A hybrid SE may be implemented by utilizing either of
number of states. This condition is called the observa- the following two approaches:
bility criterion. Apart from providing an optimal state 1. Calculating the states using synchrophasor (PMU
estimate the estimator also detect and filter out gross based) measurements only and running conventional
errors in the measurement set (bad data detection) and state estimation by utilizing these calculated values.
detect the topology errors in the network configuration. This is a two step process.
Traditionally, the input measurements to the state 2. Mixing the synchrophasor measurements with con-
estimator were provided by the remote terminal units ventional measurements and estimating the system
(RTUs) at the substations. These measurements did not states while giving more weightage to Synchro-
contain the phase angles due to the difficulty associated phasor measurements.
with the synchronization of measurements. Consequent- The second approach has been utilized in the present
ly the phase angle was estimated with the slack bus as work.
reference. However with the advent of PMUs this diffi- The state estimation model is as follows:
culty can be removed as the PMU measures voltage and z = h(x) + v (1)
current phasors synchronized through GPS. where z is the mൈ1 vector of measurements provided by
It is expected that the availability of large SCADA (bus voltage magnitudes, branch active and
amounts of phasor measurements will completely reactive power flows, and bus active and reactive injec-
change the nature of SEs. A major advantage of obtain- tions) and by PMUs (bus voltage phasors and branch
ing complex measurements of voltages and currents is current phasors), x is the nx1 state vector (n = 2N൏m),
that the estimator equations will be linear and the esti- h(x) is a mൈ1 non linear vector function relating meas-
mation will not require iterations like the present esti- urements to states, and v is the mൈ1 measurement error
mator. This SE can be updated quite often maybe in vector. The errors are assumed to be independent and
seconds (rather than minutes today) and because all the uncorrelated, having a Gaussian distribution with zero
measurements are time stamped, the time skew of the mean E(v) = 0 and covariance matrix R = cov(v) = di-
measurements will always be known resulting in much ag(σ12…..σm2) where σi is the standard deviation of the ith
more accurate SE results. It will also be possible to ex- measurement. In this work a reference free non-linear
change these SE results among neighbouring power WLS state estimator has been used [7]. The SE algo-
operators or up to the next level control centre [5]. rithm considers traditional as well as synchronized
In the conventional SE, the relationship between phasor measurements and utilizes rectangular coordi-
measurements and state variables is non linear and it has nates for the first iteration and polar coordinates for the
to be linearised and then the solution to WLS problem is next iterations. The polar coordinate formulation of cur-
obtained by using iterative technique. However if a rent phasors is a better choice than the rectangular coor-
PMU is placed at every bus of the system the relation dinate formulation, because the magnitude and phase of

561
a phasor quantity as measured and computed in a PMU šୠ ୲୰୳ୣ is actual value of the b-th state variable achieved
are largely independent variables. from load flow.
In order to find out the optimal SE solution, the follow- N is number of system buses.
ing objective function of square errors is minimized: 2N – 1 is number of state variables.
J(x)=[‫ ܢ‬െ ‫])ܠ(ܐ‬T‫ି܀‬૚ [‫ ܢ‬െ ‫])ܠ(ܐ‬ (2) In terms of variance MSE is given as
ଶ୒ିଵ ଶ
The normal equation that is solved in order to find out ൫šොୠ ୣୱ୲ ൯
the state estimation solution is
෍ (7)
ʹ െ ͳ
۵൫‫ ܓ ܠ‬൯ο‫ ܓܠ‬ൌ ۶‫ ܂‬൫‫ ܓ ܠ‬൯‫ି܀‬૚ ൣ‫ ܢ‬െ ‫ܐ‬ሺ‫ ܓ ܠ‬ሻ൧ (3) ୠୀଵ
Where, k is the iteration index, xk is the solution vector A 7-bus system of fig. 1 is used to obtain PMU loca-
ப୦ሺ୶ሻ tions for minimum state estimation error.
at iteration k, ۶ሺ‫ܠ‬ሻ ൌ ப୶ is the measurement Jacobi-
an matrix ሺ ൈ ሻǡ and ۵൫‫ ܓܠ‬൯ ൌ ۶‫ ܂‬൫‫ ܓ ܠ‬൯‫ି܀‬૚ ۶൫‫ ܓ ܠ‬൯ is CASE STUDY
the gain matrix ሺ ൈ ሻǤ
A seven bus test network as shown in fig. 1 is used to
PMU PLACEMENT CRITERION find out PMU locations for best state estimation solution
[7]. It is assumed that three PMUs are available for in-
A suitable criterion is essential for state estimation. stallation on the 7 bus network. It is assumed that PMUs
Considering which as the objective function for optimi- available have sufficient channel capacity to monitor all
zation algorithm, optimal placement of certain number the branches incident on a particular bus. Various com-
of PMUs is carried out. To evaluate the accuracy of a binations are possible for installation of these three
process, an indicator called Mean Square Errors can be PMUs. For all these combinations SE algorithm was run
used which is expressed as. and MSE was determined. The buses combinations and
σొ ౛౩౪ ି୶ ౪౨౫౛ ൯మ there corresponding MSE values are listed in table 1.
౟సభ൫୶౟ ౟
 ൌ (4) It may be observed from table 1 that the minimum mean
୒ square error is 0.099104 p. u. which is available when
The outputs of state estimation program, i.e. the esti-
mated state variables are obtained from the values of all the three PMUs are placed at branches 1, 6, and 7.
of the measuring devices through the Weighted Least Therefore the available PMUs should be placed at these
Squares method. On the other hand, for simulation stud- three branches to obtain best state estimation results.
ies according to the system planning scheme for state
estimation process, errors of measuring devices are ac-
quired randomly through normal distribution with a Bus Bus Bus
mean error equal to zero and then summed up with the
actual value of the desired variable, which is achieved
from load flow. Assuming Zmeas as the measured value
obtained from a measuring device and ZTrue as the actual
value of the measured quantity, the measured values, Bus Bus
with defining e as random measuring error, can be ex-
pressed as
 ୘୰୳ୣ ൌ  ୫ୣୟୱ ൅ ‡ (5)
In system planning, therefore, one-time running of the Bus Bus
state estimation cannot be a proper indicator for evaluat-
ing the accuracy of state estimation for the placement
problem of PMUs. Hence, the MSE equation, as defined Power flow measurement
above, would not be an appropriate method for evaluat-
Power injection measurement
ing the accuracy of state estimation in the problem of
PMU placement. Thus, the average of squared state es-
timation errors should be utilized for minimizing the
error. In other words, the error of state estimation is Fig. 1 Seven bus test system.
random and its mean value is calculated using the itera-
tion process [8]. CONCLUSION
ୣୱ୲ ଶ
ଵ଴଴଴ ଶ୒ିଵ ൫šୠǡ୧ െ šୠ ୲୰୳ୣ ൯ State estimation will form one of the most important
σୠୀଵ
 ൌ ෍ ʹ െ ͳ (6) energy management functions to be enhanced in order
ͳͲͲͲ
୧ୀଵ to make the conventional grid smart. Inclusion of phasor
where measurements in the conventional state estimation will
šୠǡ୧ ୣୱ୲ is estimated value of the b-th state variable in the enhance its performance significantly. Due to high cost
i-th state estimation process. of phasor measurement units their optimal utilization is
mandatory. However most of the methods reported for

562
optimal PMU placement minimize the number of PMUs REFERENCES
and do not offer a criterion to choose PMU placement
locations if there are multiple solutions providing mini- [1] Amin M. (2007): ‘Preventing Blackouts’, Scientific
mum number PMUs. In the present work one such crite- American, pp. 60-67.
rion has been presented to decide the best locations for
the available PMU such that the state estimation error is [2] Collier S E (2009): ‘Ten steps to a smarter grid’,
minimized. IEEE Industry Applications. 16(2), pp. 62-8.

Branch combinations Mean square error (p.u.) [3] Terzija V.r, Valverde G., Cai D., Regulski P.,
5,6,7 0.1167 Madani V., Fitch J., Skok S., Begovic M. M.,
4,6,7 0.1089 Phadke A. (2011): ‘Wide-AreaMonitoring,
4,5,7 0.1206 Protection, and Control of Future Electric Power
4,5,6 0.1093 Networks’, Proceedings of the IEEE, 99(1), pp. 80-
3,6,7 0.1072 93.
3,5,7 0.1205
3,5,6 0.1093 [4] Manousakis M. N., Korres G. N., Georgilakis P. S.
3,4,7 0.1206 (2012): ‘Taxonomy of PMU placement
3,4,6 0.1090 methodologies’, IEEE Transactions Power System,
27(2), pp. 1070-1077..
3,4,5 0.1212
2,6,7 0.1085
[5] Bose A. (2010): ‘Smart transmission grids and their
2,5,7 0.1208
supporting infrastructure’, IEEE Transactions Smart
2,5,6 0.1077 Grid, 1(1), pp. 11-19.
2,4,7 0.1208
2,4,6 0.1081 [6] Ree J. D. L., Centeno V., Thorp J. S., Phadke A. G.
2,4,5 0.1214 (2010): ‘Synchronised phasor measurement
2,3,7 0.1197 applications in power systems’, IEEE Transactions
2,3,6 0.1086 Smart Grid, 1(1), pp. 20-27.
2,3,5 0.1198
2,3,4 0.1202 [7] Manousakis K. G. N and M. Nikolaos (2012): ‘A
1,6,7 0.0991 state estimator including conventional and
1,5,7 0.1280 synchronized phasor measurements’, Elsevier
1,5,6 0.1213 Computers and Electrical Engineering, 38(2), pp.
1,4,7 0.1271 294–305.
1,4,6 0.1118
1,4,5 0.1062 [8] Karami Z. A., Rajabi M. H., Abadi and Haji M. E.
1,3,7 0.1220 (2012), ‘Optimal Placement of a Defined Number of
1,3,6 0.1135 Phasor’, in Proc. IEEE 2nd Iranian Conference on
1,3,5 0.1067 Smart Grids (ICSG), Tehran, 2012.
1,3,4 0.1224
1,2,7 0.1195
1,2,6 0.1128
1,2,5 0.1053
1,2,4 0.1143
1,2,3 0.1165
2,6,7 0.1085
2,5,7 0.1208
2,5,6 0.1077
2,4,7 0.1208
2,4,6 0.1081
2,4,5 0.1214
2,3,7 0.1197

Table 1: Branches Combination.

563