You are on page 1of 4

32

Comparison of Several PMU Placement Algorithms for State Estimation

Xu Dongjie, He Renmu, Wang Pen& Xu Tao

North China Electric Pow1:r University (Beijing), China

measurement of the system state. The greatest. benefit


ABSTRACT coming from its unique capability to provide real time
Synchronized phasor measurement units (PMUs) synchronized measurements.
have evo’lved into a practical tool for measurement of Figure 1 shows a conceptual picture of a phasor
power system voltage and current phasors. It can measurement unit system.
enhance many present applications such as state

estimators, stability controls, and remedial action


schemes, and they can also serve as disturbance
monitors. Since the quantities measured by PMUs are
voltage and current phasors, the linear relation between
them holds when modeling the branches in the network.
This feature permits linear state estimation process, thus
avoiding repetitive manipulations with large matrices in
iterative procedure as it is in the traditional case. This
significantly reduces the computational time and errors
level. This paper describes several methods for Phasor Fig . I Conceptual Diagram of a Synchronized Phasor Measuring Systcrn
Measurement Unit (PMU) placement with the aim of
linear static state estimation of power system networks. PMU placement in each substation allows for direct
These methods are depth first, graph theoretic
measurement of the state of the network. However, a
procedures and simulated annealing method as well as
ubiquitous placement of PMUs is rarely conceivable
tabu search method and Genetic algorithms.
due to cost or non-existence of communication facilities
The effectiveness and flexibility of these proposed
in some substations. Nonetheless, the ability of PMUs
algorithms are illustrated with numerical simulation
to measure line current phasors allows the calculation
using some IEEE test system.
of the voltage at the other end of the line using Ohm’s

0. Introduction Law. Some power system experts showed that optimal


placement of PMUs requires only 115 to 114 of the
The phasor measurement unit (PMU) is a power number of network buses to ensure observability.
system device capable of measuring the synchronized It is possible to reduce the numbers even further if
voltage and current phasor in a power system. PMUs are placed for incomplete observability. In this
Synchronicity among phasor measurement units (PMUs) approach, PMUs are placed sparingly in such a way as
is achieved by same-time sampling of voltage and to allow unobserved buses to exist in the system. The
current waveforms using a common synchronizing technique is to place PMUs so that in the resulting
signal from the global positioning satellite (GPS). The system the topological distance of unobserved buses
ability to calculate synchronized phasors makes the from those whose voltages are known is not too great.
PMU one of the most important measuring devices in
The CNX of this overall scheme is the interpolation of
the future of power system monitoring and control.
any unobserved bus voltage from the voltages of its
PMUs improve the monitoring and control of power
neighbors.
systems through accurate, synchronized and direct

0 2004 The Institution of Electrical Engineers.


Printed and published by the IEE, Michael Faraday House, Six Hills Way, Stevenage, SGI 2AY
1. Linear S t a t i c S t a t e Estimation 1. Ohm’s Law, U=IR: Any bus that is incident to an
observed line connected to an observed bus is observed
State estimation provides the platform for advanced
(the known current in the line, the known voltage at the

-
security monitoring applications in control centers. For
observed bus, and the known resistance of the line
example, contingency analysis and optimal power flow
determines the voltage at the bus).
rely on the state of the power system derived from the
state estimator. For this reason, state estimation is
perhaps the most important PMU application.
2 . Ohm’s Law, I=U/R Any line joining two
Since the quantities measured by PMUs are voltage
observed buses is ohkerved (the known voltage at b?th
and current phasors, the linear relation between them
observed buses and the known resistance of the line
holds when modeling the branches in the network. This
determines the current on the line).
feature permits linear state estimation process, thus
avoiding repetitive manipulations with large matrices in
iterative procedure as it is in the traditional case. This
significantly reduces the computational time and errors
level. 1 3. Kirchoffs Law: If all the lines incident to an

Now we are studying a novel approach for power observed bus are observed, except one, then all of the
system small signal stability analysis and control which lines incident to that bus are observed (the net current
combines the state-of-the-art small signal analysis tools, flowing through a bus is zero).
data mining, and the linear state estimation techniques.
Applying linear state estimation, we can get the load
flow solution almost in real time, on this condition, we
can predict the eigenvalues and small signal stability.

2. PMU Placement Rules Topological observability approach is widely


considered as a first step towards subsequent optimising
The power system monitoring problem was first techniques (bisection search, simulated annealing
studied as a variation of the well known dominating set method, an so on), these last ones proving quite
problem. Let G = (V, E) be a graph. A set S V is a c expensive in calculation time if adopted for large
dominating set of G if every vertex in V \ S has at least electric power systems. Results very close to the
one neighbor in S. The cardinality of a minimum optimal ones have been obtained by pure topological
dominating set of G is the domination number y (G). search, based on simple rules:
a

.;fib
d

Fig. 2 : Dominalion Example


Fig. 3 Graphical e ~ p l a n a l i ~ofn PMU placement rules:
a) rules 1-2; b) rule 3; e ) rule 4.

Let S = {a, c}. Notice that the remaining vertices of the Rule 1:
graph are adjacent to at least one vertex in S . Therefore, Assign one voltage measurement to a bus where a PMU
S is a dominating set of the graph G has been placed, including one current measurement to
each branch connected to the bus itself (Figure 3 a).
Rules: Rule 2:
. j

Assign one voltage pseudo-measurement to each node individual. Finally, each hit in a new individual is
reached by another equipped with a PMU (Figure 3 a). flipped, again with a predetermined probability.
Rule 3: As an optimization technique, Genetic Algorithms
Assign one current pseudo-measurement to each branch simultaneously examine and manipulate a set of
connecting two buses where voltages are known possible solutions. The power of GA’s comes from the
(Figure 3 b). This allows interconnecting observed fact that the technique is robust, and can deal
zones. successhlly with a wide range of problem areas,
Rule 4: including those which are difficult for other methods to
Assign one current pseudo-measurement to each branch solve. GA’s are not guaranteed to find the global
where current can be indirectly calculated by current optimum solution to a problem, hut they are generally
Kirchhoff law (Figure 3 c). This applies when current good at finding “acceptably g o o d solutions to
balance at one node is known, i.e. if the node has no problems. Where specialized techniques exist for
power injections (pure transit node). In fact, in this case solving particular problems, they are likely to out
if we know N-l currents incident to the node, the last perform GA’s in both speed and accuracy of the final
current can be directly inferred. result. Another drawback of Genetic Algorithms is that
they are not well suited to perform finely tuned search,
3. Algorithms but on the other hand they are good at exploring the
solution space since they search from a set of designs
Depth first method uses only Rules from 1 to 3 (it and not from a single design.
does not consider pure transit nodes). The first PMU is Simulated annealing method is a very famous way to
placed at the bus with the largest number of connected find solutions in many kinds of optimization. However,
branches. If there is more than one bus with this in spite of its ability to find the global solution in most
characteristic, one is randomly chosen. Following cases, there is a week point: it may need too much time
PMUs are placed with the same criterion, .until the in finding solutions.
complete network observability is obtained. In recent years, it is well known that tabu search(TS)
Graph Theoretic method is similar to the depth first is superior to other meta-heuristic approaches such as
algorithm, except for taking into account pure transit SA and GA. TS allows to find out solutions with high
nodes (Rule 4). accuracy and less computational effort. The idea of TS
.Below we will discuss three meta-heuristic methods: is based on the hill-climbing method that is one of local
SA, GA and TS. search. The hill-climbing method usually gets stuck in
The encoding used is a binary representation of each local minima. So to overcome this drawback, TS
unit’s starting period concatenated together. This gives corresponds to the hill-climbing method with some
a suitable representation for use with a genetic restricted search directions.
algorithm. After generating a starting population The major distinction of TS compared to other
consisting of individuals generated at random, for a heuristic search methods is that it exploits data
predetermined number of iterations the genetic structures of the search history as a condition of the
algorithm performs roulette wheel selection to choose a next moves. Rather than randomly exploring the search
pair of individuals. Single point crossover is then space like the others, TS uses a short-term memory,
applied, followed by the application of a mutation named recency-based memory, to store recent search
operator. Selection favours the better individuals by areas and uses a long-term memory, named
ensuring that each individual’s probability of being frequency-based memory, to store frequency of
selected is proportional to the fitness of that individual. searching in each area. Data in frequency-based
The probability of crossover taking place is memory are scaled by a positive constant Q which is
predetermined and if crossover does occur, single-point then used as penalty values. The purpose of such
crossover is applied from a random position in the penalty is to diversiEj future search into regions which
are rarely visited. We find that all of these methods can find the
All three meta-heuristic techniques were able to find optimum solution.
the optimum solution to a small problem. When
presented with larger problems, tabu search and 5. Conclusion
simulated annealing produced good quality solutions
whereas a genetic algorithm produced lower quality Comparison of Several PMU Placement Algorithms
solutions despite many tests during the small problem for State Estimation are described in this paper, The
to hy and ascertain the hest combination of parameters. effectiveness and flexibility of these proposed
In addition, the GA runs took longer due to the algorithms are illustrated with numerical simulation
overheads involved in processing a population of using some IEEE test system. now we attempt to use
chromosomes, rather than a single individual. In the function of linear state estimation in small signal
conclusion, for this problem the SA and TS algorithms stability analysis and control.
produce similar solutions and both outperform the GA.

6. Acknowledgment
4. Numerical simulation
The authors gratefully acknowledge Dr. Federico
In this section, two IEEE test systems were used to Milano, for his excellent simulation software PSAT.
verify the effect of these PMU optimal placement
methods.
7. References

1. T. L. Baldwin, L. Mili, M. B. Boisen, R. Adapa,


“Power System Observability with Minimal Phasor
Measurement Placement”, IEEE Transactions on
Power Svstems, Vo1.8, No. 2, May 1993, pp.
701-715.
2. D.I. Brueni, “Minimal PMU Placement for Graph
Observability”, Master of Science Thesis, Virginia
Polytechnic Institute and State University,
Decemherl993.
3. R. 0. Burnett, M. M. Butts, S. S . Sterlina, “Power
system Applications for Phasor Measurement
Fig. 4 : IEEE 14 Bus system
Units”, IEEE Computer Applications in Power,
Janury 1994, pp. 8-13.
4. R.O. Burnett, M.M Butts, T.W. Cease, V. Centeno,
G. Michel, R.J. Murphy, A.G. Phadke,
“Synchronized Phasor Measurement of a Power
System Event”, IEEE Transactions on Power
Svstems. Vol. 9, No. 3, August 1994, pp.
1643-1649.
5. . A.G. Phadke, “Synchronized Phasor Measurements
in Power Systems,“ IEEE Computer A d c a t i o n s
in P
April 1993.owel,

Fig. 5 ; IEEE 39 Bus system

You might also like