Abstract This paper identifies the data requirements for out that real-time measurements such as voltage,
real-time monitoring and control of distribution currents, and power flows taken from various points in
systems. It points out that in addition to having the system are needed for distribution system
supervisory control and data acquisition on switches monitoring and control. In section three, real-time data
and control equipment, methods are needed to obtain processing methods are reviewed and the advantages of
an accurate estimation of data needed for feeder State Estimation (SE) method is outlined. In section
automation functions. A meter placement method is four, a meter placement method is proposed for
proposed for this purpose. It is shown that the placement of the meters needed to obtain the real-time
measurements from such a metering scheme can be measurements for SE. Test results and conclusions are
used with a state estimator to provide the real-time presented in sections five and six, respectively.
data needed for real-time monitoring of distribution
system. 11. REAL - TIME DATA REQUIREMENTS FOR DA
Key Words: Distribution Systems, State Estimation, Real-time data required for real-time monitoring
Real-time monitoring. and control of a distribution system is mainly
determined by the functions to be automated in the
I. INTRODUCTION system. Economic considerations usually put limits on
the number of the functions that can be automated.
Utility interest in Distribution Automation (DA) Table 1 shows the commonly selected functions [2,3].
for monitoring and control of various devices at the
substation and at the feeder level is continuously Table 1: Commonly Selected DA functions
growing since the early demonstration projects
indicated its feasibility and potential benefits. As 5 u bs t a t i on Au t v m a t i on Feeder Automation
Fault location and isolation Fault location and isolation
utilities enter the implementation phase, the selection,
prioritization, and integration of automation functions, Service restoration Feeder Kestorntion
such as feeder switching for restoration and reconfigu- Trfmr. load balancing Feeder Reconfiguration
ration, volt/var control, and load management, Power factor control Volt/Var Control
become more important [1,2]. However, methods for
obtaining the real-time data needed for on-line
implementation of these automation functions has not
yet received proper attention.
Early demonstration projects focused on
technological feasibility of the individual DA functions
and the data required for these functions are either Data requirements for the functions listed in Table
directly measured in real-time or obtained from 1 can be grouped as follows:
historical load data [3,4]. New methods proposed for
implementation of these DA functions assume the (i) Status of Switches: Status of circuit breakers at the
availability of the required real-time data [5-81. Recently substation needs to be monitored and controlled for
new methods, such as short-term load forecasting [9], fault detection and isolation a t substation level.
power flow [lo], and state estimation [ll],have been Similarly, status of switches on feeders, such as line
proposed to obtain the necessary real-time data, mainly reclosers, and sectionalizers need to be monitored and
the load data. controlled for fault detection and isolation. Switch
The next section of this paper looks at some of the status information is also used in determining the
most commonly implemented DA functions and topology of the system in real-time. This topology
identifies their real-time data requirements. It points information is used by other control functions.
229
easily be generalized for detecting changes in switch desired performance. Here, rather than adopting a
status in a given feeder. comprehensive formulation of this problem, such as
the one developed for transmission systems [15], we
zone mO developed a heuristic approach based on the following
observations:
230
Rule 3: Put meters on normally open tie switches that where
are used for feeder switching. These meters can also be W is the Jacobian of the measurement functions
of current type. Voltage measurements at both ends of F is the Jacobian of the functions of interesting
these tie switches are also desirable for monitoring and quantities, f(x),
control of volt/var control devices from the substation W is a diagonal covariance matrix containing the
and/or dispatch center [8]. Note that although these variances 02i of the measurements zi, i = 1,...,m.
meters won't contribute to the accuracy of SE while the 2
tie-switches are open, they will do so when these Variances Oyi occupy the diagonal entries of the
switches are closed for feeder reconfiguration or covariance matrix ItY. The variances cr2i of the
restoration. measurements needed in above calculations can be
obtained by using measurement accuracies azi, since
The proposed method is a good compromise
between the accuracy and the computational simplicity;
it does not guarantee the optimality of the solution, but
it is computationally simple. The method selects both Note that the smaller the variance of a quantity,
the type and location of meters and tries to improve the the more accurate the estimation of corresponding
accuracy of the data needed for both the feeder interesting quantity will be. Hence, we define a system
switching and the volt/vax control. We will refer to the accuracy index as:
metering scheme this method produces as the basic
metering scheme. Note that this method is specially
tailored for radial feeders, a moire comprehensive
approach is needed for meshed networks.
and use it to rank the measurements in the basic
Meter Ranking measurement set Z o . To determine the ranking,
measurements in Z o are eliminated one by one from
Since in feeder operation, data needed for feeder this set as follows. First measurements are taken out of
switching is considered more important than the load the available measurement list one at a time and the
distribution data, it may be hard to justify economically resulting change in a ( z ) is calculated. The measurement
placing all the meters the above method will suggest. which causes the least change in a ( z ) is then actually
To help user eliminate some of the meters from this eliminated from the measurement set and the
basic metering scheme, we adapted Koglin's methad elimination process is repeated until all the
[13] to rank these meters. The ranking will indicate the measurements are eliminated and ranked.
order with which the meters are to be eliminated. User can use this ranking to reduce the meter
Ranking is based on the contribution of a selection and to reach a compromise between the
measurement to the accuracy of the quantities that we performance and the cost; the basic metering scheme
want to estimate which are called interesting quantities, will yield the best performance but will cost the most.
y. For example, for feeder switching, functions, current As one eliminates meters, cost will decrease, but the
flows at switching locations are the interesting performance will be sacrificed as the variances 4 i ' s
quantities. For a given metering scheme z, we can use will get bigger. After selecting a reduced metering
SE to estimate these interesting quantities since they scheme, a SE based simulation can be used to assess the
can be expressed in terms of the system state variables, x performance of the reduced measurement set. This
(usually chosen as the node voltage magnitudes and process is illustrated in the next section. A three-phase
angles) as SE should be used in simulations especially if the load
unbalance between the phases in the system is severe
y* = &(x) i = l-..k. (1) (greater than 30%), otherwise balanced feeder
Accuracy with which the SE will estimate these approximations can be used for three phase feeders, to
quantities can be expressed in terms of the variances of simplify especially the variance calculations.
these quantities,
V. TEST RESULTS
O$ i=l...k.
To calculate these variances, we assume that the The proposed rule based meter placement method
measurement errors are normally distributed, i.e., is implemented on a DEC Workstation environment
and tested using four different size feeders. The results
will be pven here for an IIEEE test feeder.
zi = h(x) + ei i=l.-.m (2)
The test feeder is a 34 node, 23 kV, 3-phase radial
where h,(x) is the measurement functions that defines E E E test feeder [12]. A one-line diagram of the feeder is
measurement zi in terms of the state x, and e; given in Fig.2 with the nodes renumbered to make the
represents the measurement er Tor. Then, the illustration of the results easier. The feeder is pre-
covariance matrix of interesting quantities can be dominantly three-phase with some single-phase
calculated from [14] laterals and has both spot and distributed loads. For test
purposes, distributed line section loads are lumped
R, =F(H~R-'H)-~F~ (3) equally at terminal nodes of the line sections. This
23 1
meter
0 0 switch
v) fuse
nominal load data is used in simulations. The voltage Reduced Metering Scheme
at the substation is held at the assumed measured
value of 1.0 p.u. Forecasted loads and actual The goal here is to reduce the basic metering
measurements are assumed to have 30% and 3% scheme obtained above for the case in which feeder
accuracy, respectively. A three-phase SE with full switching and switch status monitoring are the only
Jacobian [ll]is used for calculating the estimate of the objectives. Therefore, the interesting quantities are only
interesting quantities and their variances. Test results the current flows at switch locations, yo - y6.
can be summarized in the following two parts. For this case, the meters of the basic metering
scheme Zo are first ranked by using the proposed
Basic Meter Placement ranking procedure. The results of this ranking is
summarized in Fig.4 which shows the order with
The test feeder is assumed to have three line which the measurements are eliminated to determine
switches, three fuses and two tie switches as shown in the ranking and the corresponding changes in the
Fig. 2. Basic meter placement considers feeder system accuracy index. The elimination order is from
switching, switch monitoring, and load estimation. right to left, starting with m?
Thus, currents through the six switches, yo - y5 in the
figure, and the loads are chosen as the interesting
quantities. By applying the rules of the proposed meter
placement method, meters are placed on the test feeder
as follows. First, six meters are placed at the switch/fuse
locations - meters mo - m5 in Fig.2 - using the first rule.
Using the second rule, two more meters (mg & m7 in
the figure) are placed on the feeder to distribute the
total load evenly into meter zones. Finally, to satisfy
the third rule, two more meters are placed at the open-
tie switch locations TSWl and TSW2 shown in Fig. 2. o ! ; : : : : : ; I
tmo tm3 t m 4 +mi +m2 +m6 +m5 t m 7
All of these meters are of current type except mo which
measures the power supplied to the feeder at the Figure 4:meter ranking with respect to accuracy index
substation end. Meters are assumed to be measuring all
the phases individually. The results in Fig.4 indicate that the first five
For this basic metering scheme, variances of measurements have relatively low impact on the
interesting quantities of current flows at the switch
system accuracy index, and hence they can be
locations are given in Tb1.3 (except that of yo which is eliminated. The variances of interesting quantities
always measured). Tb1.3 shows that for all yis o;i < o:i
corresponding to this new reduced metering scheme is
where & is the variance of measurement yi. This given in Tb1.4. Note that this reduced metering scheme
result indicates that, with this metering scheme, uses only three measurements (mol m3, m4) at switch
currents through the six switches will be estimated
locations, and there are no measurements a t fuse
more accurately by using SE than by just directly
measuring them. Furthermore, variance calculations locations of y1, y2 and y5.
for loads indicate that the ten biggest loads will be We tested the performance of this reduced
estimated with accuracies ranging between 11% and metering scheme as follows. First, since yo, y3, and y4
25%. are measured but not y1, y2, and y5, the main concern is
about the estimation of these non metered interesting
Table 3: Variances with basic metering Table 4: Variances with reduced metering
232