CIRED                            18th International Conference on Electricity Distribution                     Turin, 6-9 June 2005

                            IN JINAN POWER DISTRIBUTION SYSTEM

                 Yutian LIU      Jiachuan SHI                                            Xia QIAN
         School of Electrical Engineering, Shandong University                     Jinan Electric Power Company
                               Jinan 250061, China                                         Jinan 250012, China
                        liuyt@sdu.edu.cn                                            qianx@hdpi.com.cn

INTRODUCTION                                                         Based on engineering practice and current technology, some
                                                                     measures are outlined for solving these problems. They are
The main reasons of voltage and reactive power problems in           paying attention to the reactive power planning, building the
Jinan power distribution system are unreasonable network             reactive power management network, enforcing the power
structure, large load change ranges, lack of reactive power          transmission network, keeping the voltage regulating
compensation and voltage regulation means. Based on                  equipments in good condition, improving the distribution
engineering practice and current technology, some measures           network, adding reactive power compensation and voltage
are presented for solving these problems. An optimal voltage         regulation devices, paying attention to voltage measurement,
regulation method is proposed for medium and low voltage             improving local reactive power compensation for large users,
(from 10kV to 0.4kV) distribution network. The regulating            enforcing harmonic management and keeping high available
frequencies of no-load tap changers and capacitor banks are          rate of capacitors. With these measures being taken, the
properly limited. Operating conditions are selected to cover         voltage qualified rate is improved and the active power loss
load variations during a long period of time. The                    is decreased. However, it is need to research in depth the
combinatorial optimization problem about tap-changers and            reactive power compensation and voltage management in
shunt capacitors is solved by tabu search technology to get          medium and low voltage distribution networks to ensure the
the global minima or acceptable results in reasonable time.          voltage quality of 0.4kV end users.
The simulation results of a practical network in Jinan, China,
show that the method is effective in improving the voltage
                                                                     VOLTAGE REGULATION METHOD

                                                                     Voltage regulation and/or reactive power optimization aims
VOLTAGE AND REAVTIVE POWER CONDITION IN                              to improve voltage profiles and reduce active power losses by
JINAN POWER SYSTEM                                                   regulating reactive power flow distribution. Currently, little
                                                                     or no generator is installed in medium and low voltage
Jinan city is the capital of Shandong province in east China.        distribution systems in Jinan, China. Therefore, the regulation
With increasing rapidly in recent years, the biggest electric        means in distribution systems are mainly transformer tap-
power load is about 2200MW in 2004. The electric power is            changers and shunt capacitors, which are discrete variables.
mainly supply from three power plants and one 500kV
substation. There are 11 220kV substations, 24 110kV                 Voltage regulation and/or reactive power optimization in
substations. More than 400 10kV lines supply about 1500              distribution system in this paper is to form a regulation
10kV/0.4kV power distribution transformers in the urban              scheme for locations and sizes of capacitors and positions of
district.                                                            transformer tap-changers to improve voltage profiles and
                                                                     minimize active power loss and. Therefore, it is a constrained
With social and economic development in Jinan city, the load         combinatorial optimization problem.
difference between peak and valley time becomes so large
that the voltage regulation is not enough to meet all the            In most of the papers on voltage regulation and/or reactive
operating conditions. There exist usually lower voltage in           power optimization in 10kV or higher voltage distribution
peak time and higher voltage in valley time in the power             networks [1-5], on-load tap-changers and shunt capacitors can
transmission and distribution networks. The main reasons are         be adjusted several times a day according to load variations.
unreasonable network structure, large load change ranges,            However, most of the transformers are of no-load tap
lack of reactive power compensation and voltage regulations.         changers and fixed shunt capacitors cannot be adjusted


Session No 3
CIRED                             18th International Conference on Electricity Distribution                                              Turin, 6-9 June 2005

automatically in the medium and low voltage (10kV-0.4V)                 solutions, Pl, is assessed by the following membership
distribution network in Jinan, China. Therefore, a medium               function
voltage (10kV) distribution feeder based voltage regulation
method aimed to improve secondary side voltage (0.4kV)                                 0                                  Pl > ( 2 Pl_ori − Pl _ min )
profiles is presented in this paper. The regulating frequency                                                                                                (2)
                                                                                        Pl − Pl _ ori
of tap-changers and shunt capacitors is limited to several                    Floss   =                             Pl_ min ≤ Pl ≤ ( 2 Pl_ori − Pl _ min )
                                                                                        2( Pl _ min − Pl _ ori )
times a year by considering the variation of loads during a                            1                                         Pl < Pl_ min
long period of time.                                                                   

Voltage Profiles Assessment                                             Where Floss is the membership function for active power loss,
                                                                        Pl-ori is the active power loss before optimization
A trapezoid membership function is introduced to evaluate               compensation and regulation, Pl_min is the minimum active
the network voltage profiles. The two main aims of                      power loss in some imagined operating condition.
var/voltage optimization, decreasing active power loss and
improving voltage regulation, may conflict. Generally,                  Voltage Imbalance Assessment
reactive power optimization trends to minimize active power
loss by increasing voltage to upper voltage limit, which is not         Voltage imbalance is common in distribution networks
acceptable in practice considering variations of loads and              because of the unbalanced impedances, single-phase loads,
source node voltage. If the voltage constraints are treated as          phase-to-phase loads and unbalanced three-phase loads.
“hard” constraints, that is, no voltage violation is allowed; it        According to the symmetrical component theory, an
may be hard to find a feasible solution, especially considering         unbalanced system can be decoupled into three balanced
two or more operating conditions. Thus, the voltage profiles            systems, that is, positive sequence, negative sequence and
constraints are considered as an objective by introducing a             zero sequence system.
membership function to evaluate voltage eligibility of each             As a part of power quality, the unbalance factor for each load
node and voltage profiles. The membership function for each             should not be bigger than 1.3% in China. The unbalance
node voltage is described as                                            factor for three-phase line-to-line voltage readings is defined
                                                                        by the IEC as
                  Vi − Lupper
                  upper

                   L1 − Lupper
                                 (L upper
                                    0         < Vi < L1 )

                          0
                                                                                                             1 − 3 − 6β
                  V − Llower                                     (1)                              ε=                                                         (3)
                  i
   Fvolt (Vi ) =  lower

                   L1 − Llower
                                  (L  lower
                                      0       < Vi < Lupper
                                                      1       )                                              1 + 3 − 6β
                          0
                                  (L  lower
                                      1       ≤ Vi ≤ L1 )

                                                                                                            Vab + Vbc + Vca
                                                                                                              4     4     4
                                 (others)                                                        β=                                                          (4)
                                                                                                          (V    2
                                                                                                               ab   + Vbc + Vca
                                                                                                                        2     2

Where, Fvolt(Vi) is the voltage eligibility membership function
for node i, L0upper and L0lower are unacceptable voltage limits,        Where ε is the unbalance factor, Vab, Vbc and Vca are the line-
L1upper and L1lower are acceptable voltage margins, respectively.       to-line voltages, respectively.

The value for voltage profiles assessment is the mean of the            The membership function for the ε is defined as
membership values of all nodes.
                                                                                               1                           ( εi ≤ ε1 )
Active Power Loss Assessment                                                                   
                                                                                                x − ε0
                                                                              Funbal ( ε i ) =                             ( ε1 < εi < ε0 )                  (5)
Different from the membership function for voltage profiles                                     ε1 − ε 0
evaluation, there is not a standard or limit for active power                                  0                           ( εi ≥ ε0 )
loss [4]. The active power losses in a radiating distribution
                                                                        Where εi is the unbalance factor of node i, ε0 is an
network will decrease if all reactive power loads are fully
                                                                        unacceptable unbalance factor limit and ε1 is an acceptable
compensated and tap-changers adjusted to their upper voltage
                                                                        unbalance factor margin.
limits. In such an imagined operating condition, power flow
is calculated and the active power loss, Pl_min, is considered as
                                                                        Capacitor Placement
the minimum power loss. Each active power loss of trial


Session No 3
CIRED                                18th International Conference on Electricity Distribution                     Turin, 6-9 June 2005

                                                                         technology. The process of the tabu search adopted is
Properly placed shunt capacitors can improve power factor                described as following.
and voltage quality, and reduce reactive power demand so                 i)    Read in initial data, including impedance of feeders,
that active power loss. The nodes with violated voltage-                       loads, regulation variables and inequality constraint
unbalance limits or much varying reactive power demands are                    conditions. Code the regulation variables.
considered as to compensate with automatic-switched
capacitors (ASC). While fixed shunt capacitors (FC) are                  ii)   Generate initial solution. Set the regulation variables
placed according to the following sensitivity coefficient of                   randomly without breaking the constraint conditions,
active power loss to reactive power injection under the                        including power flow restriction. Calculate the objective
heaviest load condition                                                        function f(X), and set best solution vector Xopt as X
                                                                               which is consisting of Tk and Qcj.
                                ∂Ploss                                   iii) Generate a group of trail solutions, X1, X2, … , Xk, by
                     SCi =                                    (6)
                               ∂Qinject _ i                                   “move” from X, and calculate the responding values of
                                                                              objective function, f(X1), f(X2), …, f(Xk).
Where SCi is the sensitivity coefficient, Ploss is the active            iv) Search neighborhoods. Get the best one, X* , from the
power loss of the branch, Qinject_i is the reactive power                    trail solutions. Update X with X*, if X* is not in the
injection at node i.                                                         Tabu list, or X* fits the aspiration criteria. Try the next
                                                                             solution, if the former one cannot update X.
The nodes with highest sensitivity coefficient are pre-selected
as the nodes to be compensated.                                          v)    Update Tabu list. Push the record of reversed move into
                                                                               the Tabu list, which is a FIFO (First-In-First-Out) stack.
Objective Function                                                       vi) Update Xopt with X*, if f(X*) is better than f(Xopt).
                                                                         vii) Terminate condition. Stop optimization and output
The objective function including voltage profiles, active
                                                                              results if f(Xopt) has not been improved for several
power loss and voltage unbalance under operation condition
                                                                              iterations or the given maximum iteration number is
m is represented as
                                                                              met. Otherwise go to step iii) to continue.

        Fobj _ m = (α Fvolt + β Floss + γ Funbal )m          (7)
                                                                         Software Package

Where Fobj_m is the objective function for operating condition           A Client/Server software package based on the proposed
m, α, β and γ are weight factors for the three objectives.               approach is implemented in C++ with man–machine
Thus, the voltage regulation problem is described as the                 interactive procedures, which has following distinct features.
following                                                                i) Tabu search optimization technology is utilized in the
                                                                            software to solve the optimization problem. Compared
                         M                                                  with the traditional mathematical optimization algorithms,
         Max Fobj = ∑ ( Fobj _ m )                                          the heuristic searching and optimizing technique can
                        m =1
                                                                            avoid trapping in local optimums and get the global
         s.t.     Tk min ≤ Tk ≤ Tk max                        (8)           optimum with high probability or acceptable solutions in
                  0 ≤ Qcj ≤ Qcj max                                         reasonable time.
                                                                         ii) The package employs two kinds of databases. Customer
Where M is the number of operating conditions, Tk is the ratio               database is used to store the data of various customer
of transformer k, Qcj is the reactive power of capacitors at                 networks, including node data, line data and transformer
node j. The restriction of power flow is not listed here.                    data. Common database is used to store the information of
                                                                             common devices, including line common data and
Optimization Algorithm                                                       transformer common data. The data of customer database
                                                                             can be inputted by users or selected from common
To get the global minima or acceptable results in reasonable                 database directly. Users can get data via ODBC from the
time, above-mentioned combinatorial optimization problem                     SCADA system, which enhances the online optimization
can be solved by a genetic algorithm or tabu search                          ability of the software package.


Session No 3
CIRED                               18th International Conference on Electricity Distribution                    Turin, 6-9 June 2005


SIMULATION RESULTS                                                      The main reasons of voltage and reactive power problems in
                                                                        Jinan power distribution system are outlined. Proper measures
The effectiveness of the proposed method is verified by the             are presented based on engineering practice and current
application to a practical distribution feeder shown as Fig.1 in        technology.
Jinan, China. Data are sampled by the SCADA system every
                                                                        An optimal voltage regulation method is proposed for
15 minutes.
                                                                        medium and low voltage distribution networks. The reactive
                                                                        power optimization problem is formed as a multi-object
                                                                        optimization problem, which aims to improve voltage
                                                                        profiles, decrease active power losses and balance voltage
                                                                        imbalance. Membership functions are introduced to balance
                                                                        different objectives. The regulating frequencies of no-load
                                                                        tap-changers and capacitor banks are properly limited.
                                                                        Operating conditions are selected to cover load variations
                                                                        during a long period of time. The combinatorial optimization
                                                                        problem is solved by tabu search technology to get the global
                                                                        minima or acceptable results in reasonable time. The
                                                                        simulation results of a practical network show that the method
                                                                        is effective in improving the voltage profiles.

                Fig.1 Distribution network structure
                                                                        [1] Y. Liu, et al. 2002, “Optimal voltage/var control in
Three-phase power flow [5] is calculated by a forward-                      distribution systems”, Int J Elec Power, vol.24, 271-276.
backward sweeping algorithm. All transformers in this                   [2] Z. Hu, et al. 2003, “Volt/Var control in distribution
                                                                            systems using a time-interval based approach”, IEE P-
network have no-load tap-changers (NLTC), which are of 5
                                                                            Gener Transm D, vol.150, 548-554.
positions (1±2.5%×2). All the original positions are 3, that is,
                                                                        [3] S. F. Mekhamer, et al. 2003, “Application of fuzzy logic
all the ratios are 1 p.u. The phase-to-neutral voltage upper and
                                                                            for reactive-power compensation of radial distribution
lower limits for 220V distribution networks are 7% (235.4V)                 feeders” IEEE T Power Sys, vol.18, 206-213.
and -10% (198V) respectively.                                           [4] C. Su, et al. 1996, “A new fuzzy-reasoning approach to
The node (6000#) with most unbalanced reactive power                        optimum capacitor allocation for primary distribution
demands is selected to compensate with nonsymmetrical auto-                 systems”, Proceedings Industrial Technology, IEEE
switched capacitors. So the reactive power loads at this node               pp.237-241.
                                                                        [5] W. Lin, et al. 1999, “Three-phase unbalanced distribution
are almost fully compensated.
                                                                            power flow solutions with minimum data preparation”,
Some of optimization results are given as following. Tap                    IEEE T Power Sys, vol.14, 1178-1183.
changers of all the transformers are regulated from position 3
to position 1, that is, to decrease the ratios from 1 to 0.95 in
p.u. Three transformers with biggest sensitivity values are
selected to be compensated. They are node 6010#, 6007# and
6008#. The compensation capacities are 18, 12 and 18 kVar,
respectively. Transformer 6010# is of the smallest VA
capacity (250kVA), whose impedance is bigger than that of
the others (315kVA or 400kVA). After optimization, the
voltage profiles are improved quite effectively with small
active power loss decrease. In other words, all the voltages at
different conditions are in the permitted range.


Session No 3

To top