You are on page 1of 6

PMU Optimal Placement using Sensitivity Analysis for

Power Systems Fault Location


P. Mohammadi, Student member, IEEE and S. Mehraeen, Member, IEEE

Abstract—This article presents a novel algorithm to find optimal location optimization. References [3], [7]-[8] provide optimal
sets of Phasor Measurement Units (PMUs) in power systems PMU location using optimization algorithms and graph theory.
using measurement sensitivity analysis aiming for fault detection Many other research works utilize the popular “one-bus-
without multi-estimation. The algorithm generalizes the spaced deployment strategy” introduced by [9]. Multiple
impedance method in fault detection through optimizing PMU
heuristics and mathematical algorithms have been proposed in
utilization in order to detect a fault with desired precision in
interconnected power systems. By deriving bus voltage and [1]-[11] that consider multi-channel PMUs. Multi-channel
currents sensitivity indices to the fault location and impedance, PMU refers to the assumption that a PMU installed on a bus
possible deviations of the estimated fault location and/or has enough channels to sense bus voltage and lines current.
impedance due to measurement noise, accuracy, precision limits, The issue of PMUs’ optimal number and locations will be
or simply the inability of a measurement point to sense a fault is more complicated when the network normal observability and
evaluated. Therefore, the algorithm can solve Optimal PMU fault observability are considered [4]-[8]. A normal observable
Placement (OPP) for desired fault detection precision based on system is when voltage phasors for all the system buses are
these indices for various points of measurement observing faults available, while fault observable is a system during fault
in the system. Finally, avoiding multi-estimation guarantees the
where voltage phasors of all buses and current phasors at any
unique mapping between measurements of the selected PMU
sets and faults throughout the system. The proposed algorithm is end of all lines are determinable [6]. When many approaches
performed on the IEEE 7-bus and 14-bus benchmark systems are proposed to solve the Optimal PMU Placement (OPP)
and the fault location capability is evaluated through neural problem for power system normal observability, there are a
networks. very limited number of studies targeting OPP for fault
observability. It is worth mentioning that normal observability
Index Terms—Phasor Measurement Unit (PMU), Optimal PMU does not guarantee fault observability [6]. Optimal PMU
Placement (OPP), Fault observability, Fault location, Voltage placement for fault observability is introduced by [6]-[8]. The
sensitivity index, Current sensitivity index, Multi-estimation. available approaches used to solve OPP apply different
algorithms for observability constraints while the important
I. INTRODUCTION
issue of measurement sensitivity and its impact on fault
Synchronized Phasor Measurement Units (PMUs) roles in location, that is of paramount importance, is not very well
power systems operation, control, and protection are studied [1].
prominent and constantly developing [1], [2]. The traditional This paper considers PMU direct measurements with
supervisory control and data acquisition (SCADA) systems adequate channels for voltage and currents measurements. The
collect data from the remote terminal units (RTUs) that are introduced sensitivity indices can be used to assess the quality
mostly available in substations. With the global positioning of a bus location and its measurements in power systems. Here
system (GPS) and by employing PMUs, accurate and time- the value of these indices is used to solve OPP and develop a
synchronized measurement signals are now available. This fault location algorithm; these indices can also be used as
expands the pertinent applications beyond control and criteria for other system analyses such as network fault
monitoring aims [1], [2]. These applications include accurate vulnerability, voltage stability, contingency studies, etc., that
fault location [3], state estimation [4], normal and fault are mostly fault-related. Moreover, measurement precision or
observabilities [5]-[6], and post-contingency analysis [11] as inaccuracy originating from the current transformers (CTs,)
well as static analysis, identifying network dynamics, transient potential transformers (PTs,) and PMUs can be considered in
stability prediction and control, voltage and frequency stability the sensitivity indices.
[13], etc. These PMU applications can play an important role The remainder of this paper is organized in the following
in paving the way toward safe operational smart grids that order: Section II presents the power system fault analysis
employ renewable and distributed energy resources. along with the proposed sensitivity indices. In Section III, the
Due to PMU and its required infrastructure costs, recently proposed algorithm is explained using IEEE 7-bus test system.
a significant amount of research has been dedicated to PMU

Authors are with the school of Electrical Engineering and Computer EPEC 2015 London, ON, Canada
Science, Louisiana State University, Electrical Engineering Building, South 978-1-4799-7664-5/15/$31.00 ©2015 IEEE
Campus Dr., Baton Rouge, LA 70803. Contact authors:
pmoham1@tigers.lsu.edu, smehraeen@lsu.edu.

244 IEEE
Section IV contains results for OPP and fault location using Voltage and current rates of changes in all system buses,
neural network. Finally, conclusive remarks are in Section V. i.e., potential observant buses, can be calculated using original
II. POWER SYSTEM MEASUREMENT SENSITIVITY a)
ANALYSIS
The approach presented in this paper is built upon the
classical fault analysis and is considered for three-phase b)
symmetrical systems. However, the approach can be easily
generalized to single-phase and unsymmetrical networks [12].
The fault in power systems changes the structure of the system c)
where its location and impedances are unknown.
Subsequently, previously known system states, impedance
matrix (ܼ଴ ), and admittance matrix (ܻ଴ ) should be altered to
accommodate the fault. A fault is referred to value ‫ ܨ‬ൌ d)
൫݈௙ ǡ ‫ܦ‬ǡ ܴ௙ ൯ where ͳ ൑ ݈௙ ൑ ‫ ܮ‬is the line number with ‫ ܮ‬being
the total number of lines in the power system, Ͳ ൑ ‫ ܦ‬൑ ͳ is
the normalized distance of the fault with respect to one of the e)
௟௘௡௚௧௛ሺ௟௣ሻ
line end buses where ‫ ܦ‬ൌ , and Ͳ ൑ ܴ௙ ൑ ܴ௠௔௫ is
௟௘௡௚௧௛ሺ௟௞ሻ
the fault line-to-ground resistance in the single-phase
equivalent circuit with ܴ௠௔௫ being the maximum fault
impedance of interest. The line exposed to the fault is located
Figure1. Steps for ܼ௕௨௦ modification. Z0 through Z4 are the steps of
between network buses ݈ and ݇ that are unknown due to the change in Zbus
random nature of the fault.
Figure 1.a illustrates the unfaulty network with known Zbus as well as ‫ ܦ‬and ܴ௙ which will be explained next.
impedance matrix ܼ଴ , voltages, and currents. Also, the figure
depicts the system with fault with impedance matrixܼସ. In A. Voltage Sensitivity Indices
order to have the needed variables, four steps are required to Voltage change in observant bus ݄ due to fault ‫ ܨ‬ൌ
modify ܼ଴ and obtain ܼସ as explained below and are presented ൫݈௙ ǡ ‫ܦ‬ǡ ܴ௙ ൯ is presented in (1). Using the chain rule on ߂ܸ௛ ,
in Fig. 1 [12]. Each of these steps results in an impedance
voltage sensitivity indices are defined as derivatives of ‫ ܦ‬and
matrix subscripted by the step number:
ܴ௙ with respect to ߂ܸ௛ as
Z1: Remove the transmission line between buses ݈ and ݇ by ିଵ
డ௱௏೓ ିଵ డ஽ ோ ௏ డ௱௏೓ డோ೑
adding the line’s negative impedance (െܼ௟௞ ) between ஽௏
ܵ௛ǡி ൌቀ ቁ ൌ ,ܵ௛ǡி೑ ൌ ൬ ൰ ൌ . (3)
డ஽ డ௱௏೓ డோ೑ డ௱௏೓
buses;
Z2: Add ሺͳ െ ‫ܦ‬ሻ ൈ ܼ௟௞ between bus ݇ and new bus (‫;)݌‬ Equations in (3) can be achieved by calculating derivatives
Z3: Add ‫ ܦ‬ൈ ܼ௟௞ between bus ݈ and existing bus ‫;݌‬ of ߂ܸ௛ with respect to ‫ ܦ‬and ܴ௙ and using the inverse
Z4: Add ܴ௙ between bus ‫ ݌‬and ground reference node; operator. Differentiation of ܸ௣௥௘௙ with respect to ‫ ܦ‬and ܴ௙ can
be performed by considering (2). In the following, the
Using standard fault analysis, the observant bus ݄ voltage
expanded ܼଷ ሺ݄ǡ ‫݌‬ሻ and ܼଷ ሺ‫݌‬ǡ ‫݌‬ሻ are the result of the step-by-
changes (when fault ‫ ܨ‬occurs at bus ‫ )݌‬can be described as
step parametric impedance matrix manipulations.
௓య ሺ௛ǡ௣ሻ
߂ܸ௛ ൌ ൈ ܸ௣௥௘௙ (1) ൫௓మ ሺ௛ǡ௣ሻି௓మ ሺ௛ǡ௟ሻ൯ൈሺ௓మ ሺ௣ǡ௣ሻି௓మ ሺ௟ǡ௣ሻሻ
௓య ሺ௣ǡ௣ሻାோ೑ ܼଷ ሺ݄ǡ ‫݌‬ሻ ൌ ܼଶ ሺ݄ǡ ‫݌‬ሻ െ
௓మ ሺ௣ǡ௣ሻା௓మ ሺ௟ǡ௟ሻିଶൈ௓మ ሺ௣ǡ௟ሻା஽ൈ௓೗ೖ
where ܼଷ ሺ݄ǡ ‫݌‬ሻ is the ሺ݄ǡ ‫݌‬ሻ entree of the impedance matrix of ൫௓మ ሺ௣ǡ௣ሻି௓మ ሺ௣ǡ௟ሻ൯ൈሺ௓మ ሺ௣ǡ௣ሻି௓మ ሺ௟ǡ௣ሻሻ
Z3, ܼଷ ሺ‫݌‬ǡ ‫݌‬ሻ is the system Thevenin impedance seen from bus ܼଷ ሺ݄ǡ ‫݌‬ሻ ൌ ܼଶ ሺ‫݌‬ǡ ‫݌‬ሻ െ
௓మ ሺ௣ǡ௣ሻା௓మ ሺ௟ǡ௟ሻିଶൈ௓మ ሺ௣ǡ௟ሻା஽ൈ௓೗ೖ
‫݌‬, and ܸ௣௥௘௙ is the pre-fault voltage at the point of fault in the
From transition in matrix impedances ܼଵ to ܼଷ , one can
network. With the assumption of linear voltage drop along the
conclude that for any fault ܼଶ ሺ‫݌‬ǡ ‫݌‬ሻ is the only ‫ܦ‬-dependent
transmission lines between buses and by ignoring line
variable in ܼଷ ሺ݄ǡ ‫݌‬ሻ and ܼଷ ሺ‫݌‬ǡ ‫݌‬ሻ as
capacitances to avoid complexity, ܸ௣௥௘௙ can be calculated as
ܼଶ ሺ‫݌‬ǡ ‫݌‬ሻ ൌ ܼଵ ሺ݇ǡ ݇ሻ ൅ ሺͳ െ ‫ܦ‬ሻ ൈ ܼ௟௞ .
ܸ௣௥௘௙ ൌ ܸ௟ ൅ ሺͳ െ ‫ܦ‬ሻ ൈ ሺܸ௟ െ ܸ௞ ሻ . (2)
Thus, considering ܼଶ ሺ‫݌‬ǡ ‫݌‬ሻ derivatives of ܼଷ ሺ݄ǡ ‫݌‬ሻ and
ܼଷ ሺ‫݌‬ǡ ‫݌‬ሻ with respect to ‫ ܦ‬are

245
డ௓య ሺ௛ǡ௣ሻ ൫௓మ ሺ௛ǡ௣ሻି௓మ ሺ௛ǡ௟ሻ൯ൈ௓೗ೖ 

డ஽ ௓భ ሺ௞ǡ௞ሻା௓మ ሺ௟ǡ௟ሻିଶൈ௓మ ሺ௣ǡ௟ሻା௓೗ೖ

డ௓య ሺ௣ǡ௣ሻ ൫௓భ ሺ௞ǡ௞ሻାሺଵିଶ஽ሻ௓೗ೖ ି௓మ ሺ௟ǡ௟ሻ൯ൈ௓೗ೖ 


ൌ .
డ஽ ௓భ ሺ௞ǡ௞ሻା௓మ ሺ௟ǡ௟ሻିଶൈ௓మ ሺ௣ǡ௟ሻା௓೗ೖ

It should be mentioned that these derivatives with respect


to ܴ௙ are zero, but ܴ௙ should be considered in imposing chain
rule on (1).
Figure 2. Bus ݄ line ݄‫ ݑ‬in faulty system
B. Current Sensitivity Indices
In a similar manner to voltage sensitivity indices, current It should be mentioned that for cases where fault is on the line
஽ூ ோ೑ ூ
sensitivity indices are defined for any fault ‫ ܨ‬in the system whose current is measured, ܵ௛௣ǡி and ܵ௛௣ǡி are calculated with
as: ‫ ݌‬ൌ ݊ ൅ ͳ due to an additional bus at the fault location.
ିଵ
஽ூ డ௱ூ೓ೠ ିଵ డ஽ ோ ூ
೑ డ௱ூ೓ೠ డோ೑ Equations (3) and (4) present observant bus ݄ voltage and
ܵ௛௨ǡி ൌቀ ቁ ൌ ,ܵ௛௨ǡி ൌ൬ ൰ ൌ (4)
డ஽ డ௱ூ೓ೠ డோ೑ డ௱ூ೓ೠ current sensitivity indices with respect to fault location ‫ ܦ‬and
where ݄ is the observant bus and ‫ ݑ‬is the adjacent bus impedance ܴ௙ for any fault ‫ ܨ‬ൌ ൫݈௙ ǡ ‫ܦ‬ǡ ܴ௙ ൯. Let ‫ܨ‬ሺ݈௙ ሻ ൌ
connected to ݄ by transmission line ݄‫ݑ‬. The maximum ൫݈௙ ǡ Ǥ ǡ Ǥ ൯ represent all faults on system line ݈௙ with varying
ோ ௏
number of current sensitivity indices for each bus ݄ is equal to ஽௏
Ͳ ൑ ‫ ܦ‬൑ ͳ and Ͳ ൑ ܴ௙ ൑ ܴ௠௔௫ . Therefore, ܵ௛ǡிሺ௟ ೑
, ܵ௛ǡிሺ௟ ,
೑ሻ ሻ ೑
the number of lines connected to that bus. Figure 2 illustrates ೑ ோ ூ
஽ூ
an example of a line current in the state of fault. ܵ௛௨ǡிሺ௟೑ሻ
, and ܵ௛௨ǡிሺ௟ ሻ
are observant bus ݄ sensitivity indices

Since ߂ܸ௛ is available for any ݄ within the network, for all possible faults on line ݈௙ . Hence, all observant bus (݄)
according to the standard power system fault analysis, line measurement sensitivities can be evaluated for all possible
current changes can be expressed as faulty lines (݈௙ ). Subsequently, any observant bus ݄
߂ܸ௛ െ ߂ܸ௨ measurement can be qualified to detect faults on a group of
߂‫ܫ‬௛௨ ൌ ൌ ܻ௛௨ ൈ ሺ߂ܸ௛ െ ߂ܸ௨ ሻ system lines, and the final possible PMU set should be
ܼ௛௨
optimized in a way to cover all system lines regarding
ൌ ܻଶ ሺ݄ǡ ‫ݑ‬ሻ ൈ ሺ߂ܸ௨ െ ߂ܸ௛ ሻ (5) measurement sensitivity for fault detection. On the other hand,
a unique function mapping between the PMU set’s
where ܼ௛௨ is the line impedance and ܻଶ ሺ݄ǡ ‫ݑ‬ሻ is the
measurements and system faults is possible as long as there is
admittance matrix ሺ݄ǡ ‫ݑ‬ሻ entree which corresponds to ܼସ
no multi-estimation. Multi-estimation is a condition where
according to Fig. 1. The faulted network admittance matrix
different faults in the power system cause similar measured
can be obtained by matrix manipulations similar to the
values in a set of observant buses with available precisions.
procedure explained for impedance matrix transition. This
Exhaustive search is used in this paper to guarantee that the
process results in a function for ܻଶ elements, many of which selected PMU set’s measurements, that satisfy the sensitivity
are not a function of ‫ ܦ‬or ܴ௙ . Five elements that are ‫ܦ‬- criteria, have distinguishable values for all possible faults
dependent and one element that is ܴ௙ -dependent are obtained, throughout the power system.
డ௒మ డ௒మ
for which and are calculated as
డ஽ డோ೑ III. PROPOSED ALGORITHM FOR OPP
డ௒మ ሺ௟ǡ௟ሻ డ௒మ ሺ௟ǡ௣ሻ ି௒೗ೖ The developed sensitivity analysis is carried out in this
ൌെ ൌ
డ஽ డ஽ ஽మ section for IEEE 7-bus test system. The proposed algorithm is
డ௒మ ሺ௞ǡ௞ሻ డ௒మ ሺ௞ǡ௣ሻ ೗ೖ ௒ explained using the test system with its single-line diagram
ൌെ ൌ ሺଵି஽ሻ మ
డ஽ డ஽ depicted in Fig. 3. The system consists of 3 generators and 10
డ௒మ ሺ௣ǡ௣ሻ ଵ ଵ transmission lines connecting buses to one another. By using
ൌ ቀሺଵି஽ሻమ െ ቁ ൈ ܻ௟௞
డ஽ ஽మ load flow, the network’s normal operation states are obtained
డ௒మ ሺ௣ǡ௣ሻ ିଵ and after-fault values are calculated using the proposed
ൌ .
డோ೑ ோ೑ మ method.
Fault is analyzed in all the network lines with given ‫ ܦ‬and
Using chain rule on (5), current sensitivity indices in (4) ܴ௙ resolutions by considering ܶܲ ஽ as “target precision for ‫”ܦ‬
are
and ܶܲோ೑ as “target precision for ܴ௙ .” Fault location range is
ିଵ
஽ூ
ܵ௛௨ǡி ൌቀ
డ௒మ ሺ௛ǡ௨ሻ
ሺ߂ܸ௛ െ ߂ܸ௨ ሻ ൅ ቀ
డ௱௏ೠ

డ௱௏೓
ቁ ܻ௛௨ ቁ Ͳ ൑ ‫ ܦ‬൑ ͳ on a power line and, thus, for a given ܶܲ ஽ ൑ ͳ,
డ஽ డ஽ డ஽ ଵ
fault can be detected to be on one of ವ equally-spaced points
ିଵ ்௉
ோ ூ
೑ డ௒మ ሺ௛ǡ௨ሻ డ௱௏ೠ డ௱௏೓ on any power line. Also, if fault resistance range of interest is
ܵ௛௨ǡி ൌ൬ ሺ߂ܸ௛ െ ߂ܸ௨ ሻ ൅ ൬ െ ൰ ܻ௛௨ ൰ .
డோ೑ డோ೑ డோ೑
Ͳ ൑ ܴ௙ ൑ ܴ௠௔௫ , for the given ܶܲோ೑ the fault resistance can be

246
ோ೘ೌೣ
detected as any of ೃ equally-spaced resisstances between 0 respect to total ‫ ܦ‬ൈ ܴ௙ plane is prresented in Fig. 5 for all
்௉ ೑
observant buses and all faulty lines, and a minimum of 90% is
and ܴ௠௔௫ p.u. Subsequently, the desired upper limits for
considered in this paper for satisfactory sensitivity indices.
sensitivity indices introduced in (3) and (4) can be calculated
This can be used to convert Fig. 5 to a binary matrix form for
as
ೃ “sensitivity of ‫ ܦ‬with respect to Volttage (ܵ‫ ”)ܸܦ‬as
்௉ವ ோ ௏ ்௉ ೑ ்௉ವ
஽௏
ܵ௛ǡி ൑ ൌ ߝ஽௏ , ܵ௛ǡி೑ ൑ ൌ ߝோ೑௏ ஽ூ
, ܵ௛௨ǡி ൑ ൌ
்௏ா ೇ ்௏ா ೇ ்௏ா ಺
ೃ೑
ோ೑ ூ ்௉
ߝ஽ூ , and ܵ௛௨ǡி ൑ ൌ  ߝோ೑ூ (6)
்௏ா ಺

Figure3. IEEE 7-bus system

where ܸܶ‫ ܧ‬௏ and ܸܶ‫ ܧ‬ூ are total vector errors in voltage and
current measurements, respectively. In this section ܶܲ஽ ൌ
ͲǤͲͳ and ܶܲோ೑ ൌ ͲǤͲͷ are the desired resolutions in the fault
detection algorithm. That is, the final fauult location using
resulted PMU sets from the proposed algorithm should not
have errors more than 1% and 5% from fault’s actual ‫ ܦ‬and
ܴ௙ . Therefore, one has ߝ஽௏ ൌ ͳͲ, ߝோ೑௏ ൌ ͷͲ, ߝ஽ூ ൌ ͳͲ, and
ߝோ೑ூ ൌ ͷͲ.
The latest version of IEEE C37.118-2011 standard that Figure 4. Bus 4 voltage sensitivities for F= (7,Ͳ ൑ ‫ ܦ‬൑ ͳ,Ͳ ൑ ܴ௙ ൑ ͳ)
deals with PMU devices and measurements compliances, Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ
testing conditions, and requirements determines that a PMU ‫ې Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳۍ‬
accuracy expressed as Total Vector Error (TVE) must be ‫ۑ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳێ‬
‫ێ‬ ‫ۑ‬
smaller that 1% in steady state and 3% in dynamic condition ܵ‫ି଻ܸܦ‬௕௨௦ ൌ ‫ۑͲ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳێ‬
under given testing conditions [15]. Recent advances in PMU ‫ۑ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳێ‬
algorithms show that estimated voltage or current phasor error ‫ۑ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ ͳ Ͳ Ͳێ‬
can be as low as 10-4, i.e., 0.01% [15]. Also, IEEE C57.13 ‫ے ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳۏ‬
standard for instrument transformers requirements has where logic 1 in any (݄,݈௙ ) entree shows that bus ݄ is qualified
஽௏
recognized 0.3% as a reasonable error limit for current and to observe faults on line ݈௙ regarding ܵ௛ǡிሺ௟ ೑ሻ
൑ ߝ஽௏ criteria
voltage transformers. Thus, accuracy ranges of 1%, 0.1%, and ோ ௏ ோ ூ
0.01% are considered in this study for current and voltage with over 90% coverage. Similarly for ܵ௛ǡி೑ , ܵ௛௨ǡி
஽ூ ೑
, and ܵ௛௨ǡி ,
measurements and final results are provided for those ranges. corresponding binary matrices can be calculated which are
Figures 4a.1 and 4a.2 depict voltage ߂ܸସ magnitude and ܴܵ௙ ܸ, ܵ‫ܫܦ‬, and ܴܵ௙ ‫ܫ‬. It should be mentioned that for
angle for all possible faults on line 7 connecting bus 1 to 4. sensitivities with respect to line currents, a bus with multiple
Voltage sensitivity indices with respect to ‫ ܦ‬and ܴ௙ are lines should meet the condition mentioned in (6) for at least
calculated and illustrated in Fig. 4b.1 and 4b.2, respectively. one of its connected lines measurements.
஽௏ ோ೑ ௏ From the above discussion, all sensitivity final binary
Subsequently, ܵସǡிሺ଻ሻ and ܵସǡிሺ଻ሻ are depicted in Fig. 4c.1 and
matrices can be calculated. Figure 6 illustrates the proposed
4c.2 with their region meeting condition (14) colored in blue. logic after deriving sensitivity binary matrices. An OR logic is
Undesired sensitivities depicted in red are due to faults that applied on ܵ‫ ܸܦ‬and ܵ‫ܫܦ‬. An AND logic is used between the
cause low impacts on voltage change with changes in ‫ ܦ‬and resultant ܵ‫ ܫܸܦ‬and ܴܵ௙ ܸ‫ ܫ‬as an observant bus ݄ should meet
ܴ௙ . The projection of the desired sensitivity on the ‫ ܦ‬ൈ ܴ௙ both criteria to detect both ‫ ܦ‬and ܴ௙ in a fault incident.
plane represents values for ‫ ܦ‬and ܴ௙ for which sensitivity Finally, the final sensitivity decision-making matrix ܵ‫ܴܦ‬௙ is
indices satisfy (6). The percentage of this projection with

247
ͳ Ͳ Ͳ ͳ ͳ ͳ ͳ ͳ ͳ Ͳ detection estimation error, Neural Network (NN) is employed
‫ͳۍ‬ ͳ ͳ Ͳ ͳ Ͳ ͳ ͳ Ͳ Ͳ‫ې‬ to create the unique function mapping between the
‫Ͳێ‬ ͳ ͳ ͳ ͳ ͳ ͳ ͳ ͳ ͳ‫ۑۑ‬
‫ێ‬ measurements set and related system faults. A NN is designed
ܵ‫ܴܦ‬௙ ൌ ‫Ͳێ‬ Ͳ Ͳ ͳ ͳ ͳ ͳ Ͳ ͳ Ͳ‫ ۑ‬. for each system in each case consisting of two parts. The first
଻ି௕௨௦
‫Ͳێ‬ Ͳ Ͳ ͳ ͳ ͳ ͳ ͳ ͳ Ͳ‫ۑ‬ locates the faulty line ݈௙ using the utilized PMU set
‫Ͳێ‬ ͳ ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ‫ۑ‬
‫Ͳۏ‬ measurements. Afterwards, based on the detected ݈௙ , another
Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ Ͳ ͳ‫ے‬
NN is employed to estimate faults ‫ ܦ‬and ܴ௙ . That is, in the
second stage ‫ ܮ‬number of NNs are trained to estimate faults’
locations and resistances for each line. It should be mentioned
that fault maximum resistance (ܴ௠௔௫ ) is selected to be small
SDV Next combination
SDI Sensitivity
D Sensitivity SDVI Combination with
SDRf
SRfVI full coverage
SRfV
SRfI
Rf Sensitivity

MEVI1 N
FM Combinations
Set-1
For new
Buses
MEV
MEI .. 1 1 .. Coverage? Y
Set-2
MEVIi
i
i

Figure 6. Algorithm logic diagram

(0.1) so that loads can be ignored in the provided results.


Otherwise, the load information may be needed to detect the
fault accurately which does not fit in this context.
Figure 5. Percentages of satisfactory ‫ܦ‬-voltage sensitivity indices for
஽௏
all faulty lines per each observing bus regarding ܵ௛ǡி ൑ ߝ஽௏ ൌ ͳͲ
Table I presents OPP results for IEEE 7-bus for cases
(ܶܲ ஽ ൌ ͲǤͲͳ,ܶܲோ೑ ൌ ͲǤͲͷ) and (ܶܲ஽ ൌ ͲǤͲͷ,ܶܲோ೑ ൌ ͲǤͲͷ)
Using final sensitivity matrix (ܵ‫ܴܦ‬௙ ), PMU sets are and multiple combinations of measurement precisions
generated with the condition that all system lines are covered (ܸܶ‫ ܧ‬௏ ǡ ܸܶ‫ ܧ‬ூ ).
by available PMUs in each set. Later, each PMU set is TABLE I
checked for multi-estimation for all buses and their related IEEE 7-bus OPP and NN for various target precisions and different
lines in which their sensitivity criteria is maintained. Multi- measurements accuracies
estimation process results in a symmetrical ‫ ܮ‬ൈ ‫ ܮ‬matrix
IEEE 7-bus OPP and NN - ࡾࢌ max 0.1 p.u.
( ୦ ) in which lines that have multi-estimation with each
NN correct estimation %
other are assigned 0. Multi-estimation is carried out for both ܸܶ‫ ܧ‬௏ ܸܶ‫ ܧ‬ூ #PMUs PMU Sets
݈௙ D Rf
voltage and currents for each bus, and just one voltage or
ࢀࡼࡰ ൌ ૙Ǥ ૙૚, ࢀࡼࡾࢌ ൌ ૙Ǥ ૙૞
current is adequate to have no multi-estimation, represented
by logic 1, in the related entree in final matrix  ୦ . Again, 10-2 10-2 3 1,2,3-2,3,5 99.8
100 100
100 100
in buses with multiple lines one line is enough to not have a 99.8 100
multi-estimation since it makes the fault distinguishable. 10-3 10-2 3 1,2,3-2,3,5 100
98.3 100
Finally, for each set of PMU to cover all system lines for fault 10 -3
10 -3
1 1-2-3-5 99.5
100 100
observability without multi-estimation, the condition 100 100
‫ڀ‬௜‫א‬௦௘௧  ୧ ൌ  ൌ ͳ should meet. ࢀࡼࡰ ൌ ૙Ǥ ૙૞, ࢀࡼࡾࢌ ൌ ૙Ǥ ૙૞
100 100
IV. RESULTS AND DISCUSSION 10-2 10-2 1 3-5 98.3
100 100
The proposed sensitivity analysis is performed on IEEE 7- 99.1 100
10-3 10-2 1 3-5 99.1
91.6 100
bus and 14-bus standard test cases consisting of 3 and 2
-3 -3 100 100
generators and 10 and 21 lines, respectively. The developed 10 10 1 1-2-3-4-5-6 98.3
100 100
algorithm is applied to find OPP solution for these systems for
various target precisions and measurements accuracies. The One can see that in cases with ܶܲ஽ ൌ ͲǤͲͳ and ܸܶ‫ ܧ‬ூ ൌ ͳͲିଶ
algorithm finds all possible sets fully covering the system with at least 3 PMUs are required in each case resulting in 2 PMU
the minimum number of PMUs. Also, in order to examine the sets. Sets are separated by “-” in the table and “,” separates
efficiency of the proposed methodology in the final fault different buses in a set. However, by improving ܸܶ‫ ܧ‬ூ or by

248
using ܶܲ஽ ൌ ͲǤͲͷ, the minimum required number of PMUs is algorithm is proposed to find minimum number of PMUs and
1 for all cases. The last three columns of the table present PMU sets covering all system faults regarding sensitivity
percentages of correct estimations of NN. Underlined PMU criteria. On the other hand, in order to achieve a unique
sets colored in red are used to train the NN. One can see that function mapping between a set of measurements and all
NN correctly estimates the faulty line in more than 98% of system faults, multi-estimation has been checked guaranteeing
cases of all possible faults, with given target precisions distinguishable values for all faults in the system with given
resolutions, in all system lines. The last two columns show target precisions. Results are provided for multiple cases in
that NN accurately estimates faults’ locations and resistances two test systems. Examining resulted PMU sets using
with error less than target precisions for all cases except of designed NNs showed high percentages of correct fault
two. The top values are due to the average of percentages for location estimation.
faults on all lines and the below values refer to the minimum
correct estimation percentage of resulted values. Table II REFERENCES
provides results for IEEE 14-bus showing similar trends. In [1] Albuquerque, R.J.; Paucar, V.L., "Evaluation of the PMUs
Table II, one can see the importance of measurements Measurement Channels Availability for Observability Analysis," Power
Systems, IEEE Trans. on , vol.28, no.3, pp.2536,2544, Aug. 2013-9
precision in the minimum number of required PMUs and OPP [2] Manousakis, N.M.; Korres, G.N.; Georgilakis, P.S., "Taxonomy of
PMU Placement Methodologies," Power Systems, IEEE Transactions
TABLE II
on , vol.27, no 2, pp.1070,1077, May 2012-12
IEEE 14-bus OPP and NN for various target precisions and different
[3] Pereira, R.A.F.; da Silva, L.G.W.; Mantovani, J.R.S., "PMUs optimized
measurements accuracies
allocation using a tabu search algorithm for fault location in electric
power distribution system," Transmission and Distribution Conference
IEEE 14-bus OPP and NN - ࡾࢌ max 0.1 p.u.
and Exposition: Latin America, 2004 IEEE/PES , vol., no., pp.143,148,
NN correct estimation % 8-11 Nov. 2004-3
ܸܶ‫ ܧ‬௏ ܸܶ‫ ܧ‬ூ #PMUs PMU Sets
݈௙ D Rf [4] Phadke, A.G.; Thorp, J.S.; Karimi, K.J., "State Estimation with Phasor
ࢀࡼࡰ ൌ ૙Ǥ ૙૚, ࢀࡼࡾࢌ ൌ ૙Ǥ ૙૞ Measurements," Power Engineering Review, IEEE , vol.PER-6, no.2,
pp.48,48, Feb. 1986-4
2,5,7,9,12,13,14
2,5,8,9,12,13,14 [5] Baldwin, T.L.; Mili, L.; Boisen, M.B., Jr.; Adapa, R., "Power system
2,6,7,9,12,13,14 98.25 100 observability with minimal phasor measurement placement," Power
10-2 10-2 7 2,6,8,9,12,13,14 99.8
95.0 100 Systems, IEEE Transactions on , vol.8, no.2, pp.707,715, May 1993-5
2,7,9,11,12,13,14
2,8,9,11,12,13,14 [6] Kavasseri, R.; Srinivasan, S.K., "Joint placement of phasor and
2,6,9 99.1 100 conventional power flow measurements for fault observability of power
-3 -2
10 10 3 2,9,12 99.2 systems," Generation, Transmission & Distribution, IET , vol.5, no.10,
2,9,13 91.6 100
pp.1019,1024, October 2011-6
98.25 100
10-3 10-3 1 5 99.8 [7] Pokharel, S.P.; Brahma, S., "Optimal PMU placement for fault location
95.0 100
in a power system," North American Power Symposium (NAPS), 2009 ,
ࡰ ࡾࢌ
ࢀࡼ ൌ ૙Ǥ ૙૞, ࢀࡼ ൌ ૙Ǥ ૙૞ vol., no., pp.1,5, 4-6 Oct. 2009.
-2 -2 2,6-2,12-2,13 100 99.5 [8] Geramian, S.S.; Abyane, H.A.; Mazlumi, K., "Determination of optimal
10 10 2 5,6-5,12-5,13 94.1 PMU placement for fault location using genetic algorithm," Harmonics
100 91.6
1,6-2,6-2,10- and Quality of Power, 2008. ICHQP 2008. 13th International
2,11-2,12-2,13 Conference on , vol., no., pp.1,5, Sept. 28 2008-Oct. 1 2008-2
-3 -2 100 100
10 10 2 2,14-5,6-5,10- 86.6 [9] Kai-Ping Lien; Chih-Wen Liu; Chi-Shan Yu; Joe-Air Jiang,
5,11-5,12-5,13- 100 100
5,14 "Transmission network fault location observability with minimal PMU
100 100 placement,"Power Delivery, IEEE Transactions on , vol.21, no.3,
10-3 10-3 1 1-2-3-4-5 95.4 pp.1128,1136, July 2006
100 100
[10] Pereira, R.A.F.; da Silva, L.G.W.; Kezunovic, M.; Mantovani, J.R.S.,
solution. It is shown that by increasing ܸܶ‫ ܧ‬ூ to ͳͲିଷ , the "Improved Fault Location on Distribution Feeders Based on Matching
number of required PMUs drops to 1 from 7 when During-Fault Voltage Sags," Power Delivery, IEEE Transactions on ,
maintaining a good percentage of correct fault location vol.24, no.2, pp.852,862, April 2009
[11] Emami, R.; Abur, A., "Robust Measurement Design by Placing
estimation. Synchronized Phasor Measurements on Network Branches," Power
V. CONCLUSION Systems, IEEE Transactions on , vol.25, no.1, pp.38,43, Feb. 2010-11
[12] J. J. Grainger, W. D. Stevenson, Power System Analysis, New York:
Measurements precision are of paramount importance in McGraw-Hill, Inc., International Editions 1994, pp. 283-467
solving the OPP problem. A sensitivity analysis is developed [13] Nazaripouya, H.; Mehraeen, S., "Optimal PMU placement for fault
to calculate sensitivity indices describing buses observation observability in distributed power system by using simultaneous
voltage and current measurements," Power and Energy Society General
quality for faults in the system. This sensitivity analysis, based Meeting (PES), 2013 IEEE , vol., no., pp.1,6, 21-25 July 2013-16
upon the generalized impedance method fault location, relates [14] Korkali, M ; Abur, A., "Optimal Deployment of Wide-Area
a bus topographical location and measurements precision to its Synchronized Measurements for Fault-Location Observability," Power
observing values quality. Thus, based on these qualitative Systems, IEEE Transactions on , vol.28, no.1, pp.482,489, Feb. 2013
[15] Barchi, G.; Macii, D.; Petri, D., "Synchrophasor Estimators Accuracy:
indices, one can judge whether a bus is qualified to observe A Comparative Analysis," Instrumentation and Measurement, IEEE
faults on a line with desired target precisions for ‫ ܦ‬and ܴ௙ . An Transactions on , vol.62, no.5, pp.963,973, May 2013.

249

You might also like