passive harmonic filter planning in a power system
Post on 07-Jul-2018
227 views
TRANSCRIPT
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
1/11
208 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 1, JANUARY 2009
Passive Harmonic Filter Planning in a Power SystemWith Considering Probabilistic Constraints
Gary W. Chang , Senior Member, IEEE , Hung-Lu Wang , Student Member, IEEE , Gen-Sheng Chuang, andShou-Yung Chu , Student Member, IEEE
Abstract—This paper presents a new method for planning
single-tuned passive harmonic filters to control harmonic voltage
distortion throughout a power system. In the problem, the prob-
abilistic characteristics of the harmonic source currents andnetwork harmonic impedances in the filter planning are taken into
account. The objective is to minimize the total filter installation
cost, while the harmonic voltage limits and filter component
constraints are satisfied with predetermined confidence levels. Toobtain the optimal size of each filter component of the planning
problem, the proposed procedure is first to find the candidate filter
buses based on the sensitivity analysis. Next, the formulated prob-
ability-constrained problem is transformed into a deterministicnonlinear programming problem and is solved by a genetic-algo-
rithm-based optimizer. The proposed solution procedure is tested
with an actual distribution network and is verified by the conven-
tional deterministic approach and by the Monte Carlo simulation.
Numerical experiences show that the proposed method yields
favorable results compared with the other two approaches.
Index Terms—Chance-constrained programming model, geneticalgorithm (GA), passive harmonic filter, sensitivity analysis.
I. INTRODUCTION
IN the power system, nonlinear loads may be categorizedas the harmonic current or voltage source types. Therefore,
the effectiveness of the shunt passive harmonic filter for har-
monic and reactive power compensation of a nonlinear load
depends highly on the harmonic source type [1]. In addition,
the interactions between the grid voltages and the nonlinear
loads may slightly change the harmonic current injections [2].
However, as suggested by IEEE Std. 519 [3], electric utilities
are responsible for controlling individual and total harmonic
voltage distortions (i.e., IHDv and THDv) at their network
buses, while the customers are responsible for controlling their
harmonic current injections into the utility network. Therefore,
it is practical to control harmonic voltage distortion throughoutthe utility power network by the placements of passive filters at
selected candidate buses instead of installing filters at specified
nonlinear load buses. Assume that the harmonics-producing
loads are under balanced operations. The distributed nonlinear
loads can be modeled as aggregated harmonic current injections
Manuscript receivedJanuary 25, 2008; revised June01, 2008. Current versionpublished December 24, 2008. Paper no. TPWRD-00028-2008.
The authors are with the Department of Electrical Engineering, Na-tional Chung Cheng University, Min-Hsiung, Chia-Yi 621 Taiwan (e-mail:[email protected]).
Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRD.2008.2005371
at certain buses and the direct current injection method can be
applied to perform the linear harmonic analysis.
When considering passive harmonic filter planning in a
power system, most approaches adopt the deterministic model
for network harmonic impedance and harmonics-producing
loads [4]–[8]. However, the deterministic approach for har-
monic studies may fail to model the actual behaviors of
nonlinear loads and network harmonic impedances. In many
cases, harmonic currents produced by nonlinear loads may have
probabilistic characteristics due to the variation of load levels
which change over time [9]–[11]. For the passive harmonic
filter planning, if harmonic currents and the system impedance
are regarded as deterministic, the capacities of installed passive
harmonic filters may not be properly sized, which also leads
to excessive cost. Therefore, the probabilistic characteristics
of harmonic currents and impedances of the system should be
considered when performing filter planning.
In recent years, major standards for controlling power system
harmonics have set up limits for harmonic current injections
based on probabilistic characteristics. For instance, IEC Std.
61000-3-6 considers the extensive influences of harmonic cur-
rents generated by electrical equipment and requires that the
electrical equipment work normally within 95% probabilityvalue of maximum daily and weekly allowed harmonic currents
[12]. That is, the harmonic currents cannot exceed the allowed
limit of 95% of the time during measurement. IEEE Std. 519 also
suggests considering the probabilistic constraints for harmonic
currents generated by nonlinear loads and the standard interprets
that the harmonic level can exceed the recommended limits in a
short time without causing damage to the equipment [3].
In [13], the authors have developed a probabilistic approach
for planning passive harmonic filters connected to an indus-
trial power system. To tackle the more complex system-wide
filter planning problem, which requires efficiently controlling
the harmonic voltage and voltage distortion at each network bus,the authors extend a previously developed method for planning
the single-tuned passive filters in a power system, where the
probabilistic characteristics of the system parameters are taken
into account. The planning problem is formulated as a proba-
bilistic-constrained optimization problem and is then solved to
find the optimal sizes of filter components. The objective func-
tion to be minimized is the total filter component cost. The con-
straints are the individual and total harmonic voltage distortion
limits imposed on the nonlinear load, tuned frequency varia-
tion limits of the filter, and the root mean square (rms) voltage
and current limits imposed on the filter capacitor and inductor,
respectively.
0885-8977/$25.00 © 2008 IEEE
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
2/11
CHANG et al.: PASSIVE HARMONIC FILTER PLANNING IN A POWER SYSTEM 209
Fig. 1. Schematic diagram for the th order of harmonic current source at bus injected into the power system.
Fig. 2. Thevenin equivalent circuit for a passive filter connected to bus
To solve the probabilistic optimization problem, the authors
propose a practical two-phase solution procedure. The sensi-
tivity analysis-based placement procedure is first used to quickly
locate the best candidate filter buses. Next, the linear approx-
imation method is employed to obtain the approximate mean
value and variance of each probabilistic constraint function, and
the chance-constrained programming-based model is proposed
to transform the probabilistic problem into a deterministic one.
The deterministic programming problem is then solved by a
GA-based optimizer that is built upon the Matlab environment.
Finally, the proposed method is tested for an actual distribution
system to show its usefulness.
II. HARMONIC ANALYSIS
Figs. 1 and 2 show a typical -bus network with an th order
of a harmonic current source at bus and the equivalent network
after the passive filter installed at bus , respectively. Before the
passive filter is installed, the th order of the harmonic voltage
at each bus is given in (1)
(1)
where is the th order of the harmonic
transfer impedance between buses and .
If there are more than one harmonic current sources of the
th order existing in the network, the corresponding harmonic
voltage at any bus becomes
(2)
where are the bus numbers of the harmonic
sources.
As shown in Fig. 2, after the single-tuned passive filter is con-nected to bus for controlling the th order harmonic current
in the system, the harmonic current drawn by the passive filter
is given in (3)
(3)
where
(4)
and where is the filter tuned harmonic
order, is the base apparent power; and is the reactive
power capacity of the filter at the fundamental frequency. Also,
is the Thevenin equivalent voltage at bus
before siting the passive filter and is
the driving point harmonic impendence at bus before placing
the passive filter.
The new harmonic voltage at any bus for the th harmonic
after the filter installed at bus is then obtained by
(5)
where is the harmonic transfer
impedance between buses and before siting the passive
filter and is the harmonic voltage at bus
. If more filters are installed at different buses, the harmonic
voltage at bus is given
(6)
where are bus numbers of passive filters for
the th harmonic.When considering the probabilistic characteristics of the pas-
sive harmonic filter planning problem, the real and imaginary
components of the harmonic current and the system harmonic
impedance are treated as random variables. Then, (5) and (6)
can be rewritten as
(7)
where represents probabilistic
parameters and is the solution variable (i.e., filter
harmonic impedance) to be determined for the passive filter.
III. PROBABILISTIC APPROACH OF THE PASSIVEHARMONIC FILTER PLANNING
A. Linear Approximation for a Function of Random Variables
Let be a function of random vari-
ables. After expanding the in a Taylor’s series about the mean
value, the following is obtained:
(8)
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
3/11
210 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 1, JANUARY 2009
where is the mean value of the corresponding random vari-
able . From the aspect of engineering applications, it is prac-
tical to approximate the mean value and the variance of by
the transmission of variance formulae, as shown in (9) and (10),
to simplify the calculation of and [14], [15]
(9)
(10)
where is the covariance between and . If
and are uncorrelated for all and , the second term at the
right-hand side of (10) can be ignored.
B. Siting Index for Determining Passive Filter Locations
The first step of the solution procedure is to find the best can-
didate bus one at a time among existing capacitor busses in the
network for siting passive filters while minimizing the system
voltage distortion of a specific harmonic order under considera-
tions [16]. Based on sensitivity analysis, the best filter bus can
be identified by the siting index
(11)
where and where is the
highest order of harmonic under considerations and is the
th harmonic voltage change at bus after the filter is installed
at bus . Also
(12)
and
(13)
where and
are the mean value and the variance of the corre-
sponding probabilistic parameter. The detailed derivation of
the siting index is given in Appendix A. Equation (11) implies
the effectiveness for controlling voltage distortion across the
network corresponding to the filter placement at bus . The top
priority bus is the filter bus that yields the least value among allsiting indices.
Fig. 3. Applying the limit to the probabilistic harmonic voltage.
Fig. 4. Harmonic voltage estimation in the confidence interval range.
C. Chance Constraints for Harmonic Voltage at Each Bus
Figure 3 illustrates the immunity of harmonic pollutions at
each network bus in the system, where is the recommendedharmonic voltage limit and represents the actual har-
monic voltage. In is the vector of known parame-
ters with probabilistic characteristics for the harmonic passive
filter and is the vector of unknown variables.
Let denote the probability of event . By observing
Fig. 3, the probability constraint of the harmonic voltage ex-
pressed by the chance-constrained programming model of (14)
is to guarantee that the bus voltage is satisfied within the re-
quired probability level given by
(14)
where is the limit level of the probability.
For convenience, Fig. 3 can be expressed as a confidence in-
terval shown in Fig. 4, which includes the predetermined con-
fidence level. The permissive value of the harmonic voltage at
the point of common coupling is then estimated. Therefore, the
probabilistic constraint of the harmonic voltage expressed by
(14) can be replaced by the deterministic constraint given in
(15). In this situation, the probabilistic optimization problem is
treated as a deterministic nonlinear programming problem
(15)
In (15), and represent the mean value andthe variance of the probabilistic harmonic voltage, respectively.
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
4/11
CHANG et al.: PASSIVE HARMONIC FILTER PLANNING IN A POWER SYSTEM 211
is the tabulated value of the inverse cumulative distri-
bution function of the standard normal distribution evaluated at
. By substituting the mean value and the variance of (9) and
(10) into (15), the deterministic constraint corresponding to the
probability confidence interval of the harmonic voltage shown
in Fig. 4 is determined. The proof of (15) can be found in [13,
Appendix].
IV. PROBLEM FORMULATION
In the probabilistic passive harmonic filter planning problem,
the objective function is to minimize the filter installation cost.
The constraints of the problem include IEEE-519 individual
harmonic voltage and total harmonic voltage limits at each
bus, tuned frequency variation limits of the passive filter due
to manufacturing errors and environmental changes, as well as
the IEEE-1531 capacitor rms voltage and inductor rms current
limit of the filter [17]. The details of the problem formulation
will be listed.
A. Constraints
1) Voltage Harmonics and Voltage Distortion Constraints:
The IEEE-519 individual harmonic voltage distortion (IHDv)
and total harmonic voltage distortion (THDv) constraints im-
posed on any network bus after the filter placement are
(16)
for , and
(17)
where is the fundamental voltage at bus and are
usually 3% and 5%, respectively.
For convenience, substituting (7) into (16) to obtain the
squared function, the individual harmonic voltage limit can be
expressed by (18)
(18)
Similarly, the constraint of the total harmonic voltage distortion
at any network bus of (18) can be written as
(19)
The specified confidence levels of individual harmonic voltage
and total harmonic voltage distortion can be defined by the
chance-constraints of (20) and (21), respectively
(20)
(21)
where and represent the specified confidence levels of
the individual harmonic current and the total harmonic current
injections, respectively.
According to the linear approximation of (9), the mean valueof given in (18) at the th harmonic order can be written as
(22)
where
is the vector of mean values of associated probabilistic
parameters of . With (10), the approximate variance of (18) at
the th harmonic order then becomes
(23)
where
(24)
(25)
The corresponding coefficients used in (23)–(25) are given in
Appendix B.
2) RMS Voltage Limit Across the Capacitor of the Filter:
Under the specified confidence level, the chance-constraint of
the rms voltage across the capacitor can be expressed by
(26)
where
(27)
(28)
where is the capacitor rated rms voltage of the th filter
and is the recommended percentile limit of the rated voltagevalue; represents the specified confidence level of the rms
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
5/11
212 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 1, JANUARY 2009
voltage across a filter capacitor at the filter bus and is the
tuned harmonic order.
The mean value and variance of the rms voltage across the
capacitor of the th filter then can be approximated by
(29)
and
(30)
where
(31)
(32)
3) RMS Current Limit Flowing Through the Filter: Similarly,
the chance-constraint of the rms current flowing through the in-
ductor can be expressed by (33)
(33)
where
(34)
(35)
where is the rated inductor rms current of the th filter and
is the recommended percentile limit of the rated current value
and represents the specified confidence level of the rms
current flowing through the filter inductor at the filter bus.
The mean value and variance of the rms current flowing
through the th filter then can be approximated by
(36)
and
(37)
where
(38)
(39)
4) Filter-Tuned Frequency Variation Constraint: The manu-facturing and environment caused deviations for filter inductors
and capacitors are in the range of % to 3% and % to 12%
from their nominal values, respectively; the LC variation range
of the filter is usually between 0.92 and 1.06 at the system fun-
damental frequency [8]. Therefore, the deviation limits for any
tuned harmonic order will be controlled in the range; that is,
, where is the
exact tuned harmonic order and is the closest integer tuned
harmonic order for the th filter.
The chance-constraint corresponding to the th filter-tuned
frequency variation can then be expressed by
(40)
where is the specified confidence level of the filter tuned har-
monic order. Assume that the variation range of the tuned har-
monic order is normally distributed and the system frequency
is constant. The approximate probability distribution can be ex-
pressed by .
B. Probabilistic Optimization Problem
According to the described objective function and con-
straints, the problem formulation of the passive filter planning in
terms of each filter-tuned harmonic order becomes “Minimize”
(41)
subject to
(42)
(43)
(44)
(45)
(46)
where
and are unit costs associated with the fundamental re-
active power of the capacitor (i.e., ) and the inductor (i.e.,
) of the th filter, respectively; is the fundamental re-
active power of the th filter; , and repre-
sent the confidence levels, where the first four terms are 95% and
the last is 99.9%, respectively;
and .
V. SOLUTION PROCEDURE
The formulated filter planning problem is categorized as aprobabilistic-constrained optimization problem. After adopting
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
6/11
CHANG et al.: PASSIVE HARMONIC FILTER PLANNING IN A POWER SYSTEM 213
the chance-constrained programming model, the probabilistic
problem is transformed into a deterministic nonlinear program-
ming problem where the optimal solution by conventional non-
linear programming approaches is still very difficult to obtain
since the problem is highly nonlinear and nonconvex. Since
the genetic algorithm (GA)-based method can be used in con-
junction with local-improvement heuristics for nonlinear func-tion optimization and is very suitable for solving nontypical
nonlinear programming problems with relatively high solution
speed, the Genetic Algorithm Optimization Toolbox (GAOT)
from Matlab is adopted as the problem solver [18]. The solu-
tion procedure for GAOT typically includes creating an initial
population, evaluating all individuals, selecting a new popula-
tion, creating new solutions by the mutation and crossover, and
termination criteria check [18], [19].
The following list summarizes the major steps of the pro-
posed solution procedure for passive harmonic filter planning
in a power system.
Step 1) Input the probability parameters of harmonic source
data and harmonic transfer impedance of the net-work.
Step 2) Calculate the mean value and the variance of the har-
monic voltage, and determine the 95% confidence
interval value of individual and total harmonic volt-
ages at each bus.
Step 3) Check whether the 95% confidence value of each in-
dividual harmonic voltage violates the limit. If yes,
go to step 5. Otherwise, proceed to the next step.
Step 4) Check whether the 95% confidence interval of the
total harmonic voltage at each bus violates the limit.
If yes, proceed to the next step. Otherwise, stop.
Step 5) Find the filter siting index by (11) for each harmonicorder under consideration and for all network buses
one at a time and prioritize the indices. The bus with
the least value of (11) is the best candidate bus for
siting the filter.
Step 6) Convert the capacitor bus with the top priority of the
siting index to a filter bus for each harmonic order
under consideration.
Step 7) Formulate the probabilistic-constrained pro-
gramming problem of (41)–(46) and apply the
chance-constrained programming model to trans-
form the problem into a deterministic-constrained
problem.
Step 8) Solve the deterministic optimization problem corre-
sponding to (41)–(46) by the GAOT solver.
Step 9) Check whether all constraints are satisfied. If yes,
stop and output. Otherwise, proceed to the next step.
Step 10) Use the equivalent resistance approach of [7] to de-
termine the filter siting index for each harmonic
order under consideration and for all network buses
one at a time and prioritize the indices. The bus with
the least value of (11) is the best candidate filter bus
for that tuned harmonic order. Proceed to the next
step.
Step 11) For the candidate filter bus with an existing capacitor
bank, repeat step 6. Otherwise, for all commerciallyavailable capacitor sizes, sequentially start with the
Fig. 5. Eighteen-bus test system.
TABLE IPROBABILISTIC CHARACTERISTICS OF HARMONIC CURRENT SOURCES (%)
smallest unit and convert it into a passive filter at the
candidate bus.
Step 12) Repeat Steps 7) and 8) for each converted filter.
Step 13) Stop and output after the filter component sizes with
the least objective function value in Step 12) are
found.
VI. CASE STUDY
The proposed solution algorithm for passive filter planning istested by using an actual 18-bus distribution network, as shown
in Fig. 5 [20]. The base power of the sample system is 10 MVA
and all buses except 17 and 18 are 12.5-kV buses. In the system,
the harmonic current sources, the system impedances, and the
load at each bus have probabilistic characteristics.
Assume that the real and imaginary components for each
order of harmonic current of the nonlinear load are normally
distributed; the corresponding expected values and standard de-
viations are shown in Table I, where the correlation coefficients
related to real and imaginary components for each harmonic are
practically assumed to be 0.5 and the standard deviations of real
and reactive load power, P (%) and Q (%), are assumed to be
30% [10], [21]. Before the placement of passive filters, the IHDvand THDv at each bus are determined by PCFLO [20].
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
7/11
214 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 1, JANUARY 2009
Fig. 6. 95% confidence intervals for THDv at each network bus before placingthe filters.
Fig. 7. Largest individual harmonic voltages at the upper bound of the 95%confidence interval before placing the filters.
A. Results
The 95% confidence interval of the total harmonic voltage
and the upper/lower bounds of the 95% confidence interval of
individual harmonic voltage at 12.5-kV buses before siting fil-
ters are shown in Figs. 6 and 7, respectively. It is observed that
the upper limits of the largest values of the 95% confidence in-
terval of individual harmonic voltages for the 5th, 7th, 11th,
13th, 17th, and 19th harmonics are 7.5%, 4.9%, 3.11%, 2.0%,
2.7%, and 0.3%, respectively. The highest 95% confidence in-
terval of THDv is 9.92%, occurring at bus 14. Therefore, IHDv
and THDv at several network buses violate the IEEE-519 limits.
By applying the proposed solution procedure for filter plan-
ning, the component values and the tuned harmonic filters are
shown in Table II. Figs. 8 and 9 show the 95% confidence in-
terval of the IHDv and THDv at each 12.5-kV bus after siting
filters. It can be seen that IHDv and THDv at each network bus
are well controlled under IEEE-519 limits. After installing thepassive filters in the system, the average of the THDv expected
values at all 12.5-kV buses is reduced to 2.36%, and the highest
is 4.47%, occurring at bus 8. In addition, as listed in Table III,
the rms capacitor voltage and inductor current for each filter also
satisfy IEEE-1531 limits, where the capacitor rms voltage and
the inductor rms current limits are set to be 110% and 135%,
respectively.
B. Comparisons Between the Proposed And the Deterministic
Approaches
In the conventional deterministic approach, the proposed two-
phase solution procedure is used to solve the same problem [7],where a sensitivity analysis-based placement procedure is first
Fig. 8. Ninety-five percent confidence interval of THDV at each network busafter placing the filters.
Fig. 9. Largest individual harmonic voltages at the upper bound of the 95%confidence interval after placing the filters.
TABLE IIHARMONIC FILTER CANDIDATE BUSES AND COMPONENT SIZES
TABLE IIIFILTER CAPACITOR RMS VOLTAGE (%) AT THE UPPER
LIMIT OF THE 95% CONFIDENCE INTERVAL
used to quickly locate the best candidate filter buses among ex-isting capacitor buses for each harmonic order under consider-
ation. Then, the optimizer GAOT is employed for determining
the near-optimal or global optimal filter sizes while minimizing
filter installation cost and satisfying all associated constraints.
The system parameters are treated as the worst case that har-
monic currents and the system data are their mean plus three
times the standard deviation. Figs. 10 and 11 show the IHDv and
the THDv at each 12.5-kV bus before siting passive harmonic
filters. It is found that before placing the filters, the system is
severely polluted with 5th, 7th, 11th, and 13th harmonics. The
average THDv at all 12.5-kV buses is 6.46 %, and the highest is
10.75%, occurring at bus 14.
The results obtained by the conventional passive filer plan-ning method are shown in Table IV. Figs. 12 and 13 show the
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
8/11
CHANG et al.: PASSIVE HARMONIC FILTER PLANNING IN A POWER SYSTEM 215
Fig. 10. IHDv at each 12.5-kV bus before placing passive filters.
Fig. 11. THDv at each 12.5-kV bus before placing passive filters.
Fig. 12. IHDv at each 12.5-kV bus after placing passive filters.
Fig. 13. THDv at each 12.5-kV bus after placing passive filters.
IHDv and THDv, before and after installing passive filters, re-
spectively. It shows that the IHDv and THDv are well controlled
after placement of the filters. By comparing results shown in
Table V, the capacity of the passive filter obtained by the proba-
bilistic model is 2807.33 kVAR and by the deterministic model,
it is 4604.5 kVAR. The cost of the proposed approach is lowerthan that of the deterministic one.
TABLE IVFILTER COMPONENT SIZES OBTAINED BY USING THE DETERMINISTIC METHOD
TABLE VCOMPARISON OF THE PROPOSED AND DETERMINISTIC METHODS
Fig. 14. Largest values of the individual harmonic voltage distortion and totalharmonic voltage distortion (THD) at the upper bound of the 95% confidenceinterval at bus 9.
Fig. 15. Largest values of the individual harmonic voltage distortion and totalharmonic voltage distortion (THD) at the upper bound of the 95% confidenceinterval at bus 12.
C. Comparisons Between the Proposed Approach And Monte
Carlo Simulation
In the same manner, a Monte Carlo simulation program
developed by using Matlab is adopted to simulate the corre-
sponding probabilistic characteristics [22]. Simulations are
performed over 1000 times for the 18-bus test system. Figs. 14
and 15 show the expected values of the individual and total
harmonic voltage distortions (the 5th through the 19th harmonic
order) and the total harmonic voltage distortion (indicated by
THD) at buses 9 and 12, respectively, where results obtained
by the proposed linear approximation method are also shown
for comparisons. As indicated in Figs. 14 and 15, it is observedthat results obtained by the proposed method are very close to
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
9/11
216 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 1, JANUARY 2009
those obtained by Monte Carlo simulation. The absolute THD
differences between the two approaches for buses 9 and 12 are
0.21% and 0.35%, respectively.
VII. CONCLUSION
This paper presents a new method of harmonic passivefilter planning for controlling harmonic voltage distortion
by considering the probabilistic characteristics of harmonic
source currents and network harmonic impedances in the
power system. The theory and problem formulation for the
filter-planning problem are described in detail. In the solution
procedure, the linear approximation method, the sensitivity
analysis, the chance-constrained programming, and the genetic
algorithm-based approach are then employed to solve the
probabilistic optimization problem. The proposed solution
procedure has been tested by an actual distribution network. It
is shown that the proposed approach yields favorable results
in comparing those obtained by the conventional deterministic
approach and the Monte Carlo simulation.
APPENDIX
A. Siting Index for Determining Passive Filter Locations
In the unconstrained problem for passive filter planning, the
objective function is the sum of all harmonic voltages across the
power network after the filter placement and can be expressed
as
(A1)
where the th harmonic voltage at any bus before and after
filter placement are given in (A2) and (A3) as follows:
(A2)
(A3)
where is the index for harmonic current source location. Ac-
cording to (2) and (3), the following relationship holds:
(A4)
By substituting (A4) into (A3) and then into (A1), we have
(A5)
Representing (A5) with Taylor’s series around and trun-
cating the series at the linear terms, we have
(A6)
where
(A7)
and
(A8)
where and .
Therefore, the sensitivity factor for placement of the th filterbecomes
(A9)
where
(A10)
(A11)
and where and .
B. Corresponding Coefficients of (23)–(25)
(B1)
(B2)
where
(B3)
(B4)
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
10/11
CHANG et al.: PASSIVE HARMONIC FILTER PLANNING IN A POWER SYSTEM 217
(B5)
and
(B6)
(B7)
(B8)
(B9)
(B10)
(B11)
(B12)
(B13)
(B14)
(B15)
(B16)
(B17)
(B18)
(B19)
REFERENCES
[1] J. A. Pomlio and S. M. Deckmann, “Characterization and compensa-
tion of harmonics and reactive power of residential and commercialloads,” IEEE Trans. Power Del., vol. 22, no. 2, pp. 1049–1055, Apr.2007.
[2] Task Force on Harmonics Modeling and Simulation, “Modelingand simulation of the propagation of harmonics in electric powernetworks-part I: Concepts, models, and simulation techniques,” IEEE Trans. Power Del., vol. 11, no. 1, pp. 452–465, Jan. 1996.
[3] IEEE Recommended Practices and Requirements for Harmonic Con-trol in Electric Power Systems, IEEE Std. 519-1992.
[4] J. Arrillaga, N. R. Watson, and S. Chen , Power System Quality Assess-ment . New York: Wiley, 2000.
[5] D. Blume, J. Schlabbach, and T. Stephanblome , Voltage Quality in Electrical Power Systems, ser. Inst. Elect.Eng.Power andEnergy36.London, U.K.: IEE, 2001.
[6] T. H. Ortmeyer and T. Hiyama, “Distribution system harmonic filterplanning,” IEEE Trans. Power Del., vol. 11, no. 4, pp. 2005–2012, Oct.1996.
[7] G. W. Chang, S. Y. Chu, and H. L. Wang, “A new method of passiveharmonic filter planning for controlling voltage distortion in a powersystem,” IEEE Trans. Power Del., vol. 21, no. 1, pp. 305–312, Jan.2006.
[8] K. P. Lin, M. H. Lin, and T. P. Lin, “An advanced computer code forsingle-tuned harmonic filter design,” IEEE Trans. Ind. Appl., vol. 34,no. 4, pp. 640–648, Jul./Aug. 1998.
[9] IEEE Probabilistic Aspects Task Force, “Time-varying harmonics: PartI-characterizing measured data,” IEEE Trans. Power Del., vol. 7, no. 3,pp. 938–944, Jul. 1998.
[10] IEEE Probabilistic Aspects Task Force, “Time-varying harmonics: PartII-harmonic summation and propagation,” IEEE Trans. Power Del.,vol. 17, no. 1, pp. 279–285, Jan. 2002.
[11] T. Ortmeyer, W. Xu, and Y. Baghzouz, “Setting limits on time varyingharmonics,” in Proc. IEEE Power Eng. Soc. General Meeting, Jul.2003, vol. 2, pp. 13–17.
[12] Electromagnetic compatibility (EMC) Assessment of Emission Limits for Distorting Loads in MV and HV Power Systems-Basic EMC Publi-
cation, IEC Std. 61000-3-6, 1996-10.[13] G. W. Chang, H. L. Wang, and S. Y. Chu, “A probabilistic approach
for optimal passive harmonic filter planning,” IEEE Trans. Power Del.,vol. 22, no. 3, pp. 1790–1798, Jul. 2007.
[14] S. B. Vardeman , Statistics for Eng ineering Problem Solving. Boston,MA: PWS, 1994.
[15] H. S. Alfredo and H. T. Wilson , Probability Concepts in EngineeringPlanning and Design-Basic Principles. New York: Wiley, 1975.
[16] G. W. Chang, H. L. Wang, and S. Y. Chu, “Strategic placement andsizing of passive filters in a power system for controlling voltage dis-tortion,” IEEE Trans. Power Del., vol. 19, no. 3, pp. 1204–1211, Jul.2004.
[17] IEEE Guide for Application and Specification of Harmonic Filters,IEEE Std. 1531, 2003.
[18] C. R. Houck, J. A. Joines, and M. G. Kay, “A genetic algorithm forfunction optimization: A Matlab Implementation,” Raleigh, NC, NorthCarolina State Univ., , 1995.
-
8/18/2019 Passive Harmonic Filter Planning in a Power System
11/11
218 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 24, NO. 1, JANUARY 2009
[19] Genetic Algorithm Optimization Toolbox User’s Guide for use withMatlab 1999, Version 2, The Math Works Inc.,.
[20] “PCFLO User’s Manual,” 2007, Univ. Texas at Austin.[21] G. Carpinelli, T. Esposito, P. Varilone, and P. Verde, “First-order prob-
abilistic harmonic power flow,” Proc. Inst. Elect. Eng., Gen. Transm. Distrib., vol. 148, no. 6, pp. 541–548, Nov. 2001.
[22] R. Y. Rubinstein , Simulation and the Monte Carlo Method . Reading,MA: Addison-Wesley, 1981.
Gary W. Chang (M’94–SM’01) received the Ph.D. degree from the Universityof Texas at Austin in 1994.
He was with Siemens Power T&D, Brooklyn Park, MN, from 1995 to 1998.He is a Professorin theDepartment of Electrical Engineering at National ChungCheng University, Taiwan. His research interests include power systems opti-mization, harmonics, and power quality.
Dr. Chang is a member of Tau Beta Pi and a registered professional engineerin the state of Minnesota. He chairs the IEEE Task Force on Harmonics Mod-eling and Simulation.
Hung-Lu Wang (S’04) received the M.S.E.E. degree from National ChungCheng University, Chia-Yi, Taiwan, in 2002, where he is currently pursuing thePh.D. degree.
His research interests include power system harmonics and power quality.
Gen-Sheng Chuang received the M.S.E.E. degree from National Chung Cheng
University, Chia-Yi, Taiwan, in 2003, where he is currently pursuing the Ph.D.degree.
His areasof research interestsincludepower system optimizations,short-termresource scheduling, and power quality.
Shou-Yung Chu (S’04) received the M.S.E.E. degree from National ChungCheng University, Chia-Yi, Taiwan, in 2002, where he is currently pursuing thePh.D. degree.
His areas of research interests include power systems optimization and powersystem harmonics.