You are on page 1of 5

Advanced Materials Research Submitted: 2014-05-15

ISSN: 1662-8985, Vols. 986-987, pp 1360-1364 Accepted: 2014-05-19


doi:10.4028/www.scientific.net/AMR.986-987.1360 Online: 2014-07-18
© 2014 Trans Tech Publications, Switzerland

A Review of Regional Reactive Power Optimization Techniques


Wei Wei1, Lijie Ding2,Yanjiao Liu3
1
State Grid Electric Power Research Institute, Sichuan Electric Power Company,
Chengdu,610000, China
2
State Grid Electric Power Research Institute, Sichuan Electric Power Company,
Chengdu, 610000, China
3
Control and Computer Engineering School, North China Electric Power University,
Beijing,102206,China
email: Liuyanjiao_job@163.com

Keywords: regional power grid; reactive power optimization; planning algorithm; intelligent
algorithms

Abstract. With increasingly complex of the power grid structure and increasingly user requirements
of power quality, Power grid voltage reactive power optimization is still the difficult points in
power system operation control. This paper introduces the general optimization methods of
multi-objective reactive power optimization, intelligent algorithm, the development of the hybrid
method, and their respective advantages, disadvantages and improvement; It also analyzes and
summarizes the simplification of search space for reactive power optimization and the key issues
and research development trend of the real-time reactive voltage control system.

Introduction
Reactive power optimization is a basic problem of grid scheduling planning and operation; It has
vital significance to ensure the power quality and power grid operation of the economy. Since the
application of SCADA system and power grid scheduling, on-line reactive power optimization has
become the basic advanced application of dispatching automation system. In power grid operation,
the output adjustment of generator, transformer tap gear adjustment and through cutting of capacitor
reactive power compensation equipment is the basic means for the control of reactive. Due to large
number of these variables, the confounding control mode of continuous and discrete binding, it is a
typical NP problem. To solve this problem, various classical methods and intelligent planning
algorithm is applied to which different method of characteristics, scope of application and the
solving basic conditions each are not identical, it is difficult to briefly comment on the merits.
This article points out the problem of reactive power optimization studies and the direction of
their research by analyzing the basic solving principle of the common classical method and the
basic solving idea of reactive power optimization problem, comparing the existing problems of
different algorithms and their scope in reactive power optimization problem solving; combining
Reactive Power Optimization actual needs with EMS systems development.

A review of reactive power optimization algorithm


1. Research Status of classical methods and improvements
Classic method basic principle of reactive power optimization is first determining an initial
point, and then, constantly amending current solution according to the specified trajectory
optimization until the satisfactory results are obtained, the calculation ends The main algorithms is
planning algorithms, the algorithms use the objective function and constraint conditions of first or
second order derivative as the main information of finding the optimal solution.

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications, www.ttp.net. (#90182954, University of Wollongong, Wollongong, Australia-28/05/17,07:41:49)
Advanced Materials Research Vols. 986-987 1361

1.1 The classification of classical algorithm


(1) Nonlinear programming classes
Gradient method is used earlier for solving optimal power flow and voltage and reactive power
optimization problem. This method uses control variables as the optimization of the negative
gradient direction, Newton's method to calculate the trend to follow changes in the state variables,
the penalty function to handler Inequality boundaries. Although there is strict mathematical
reasoning, the actual application convergence is slow. We can’t even find the optimal solution.It
will appear the phenomenon of “sawtooth” near the most advantages. Later second derivative
method has developed, such as Newton method and a series of non-linear programming method.
Generally, the nonlinear programming method’s calculation speed is slow and in a poor
coordination between control variables.
(2) Linear programming classes
Usually, the objective function of control variables the partial derivative is used to establish the
linear programming model. Advantages: reliable convergence. Difficulties: adjustment of
linearization step; How to reduce the error caused by linearization of the model.
Reference [1] proposes a linear programming linear step dynamic adjustment strategy in
reactive power optimization problem. The strategy solves the step size selection of linear reactive
power optimization to some extent.
(3) Interior point method
The interior point method involves nonlinear programming and linear programming. Difficulties
in the point method: there is no general rule to select the initial dual variables and Fix the obstacles
parameters which is based on the original dual interior point algorithm; it requires strict control of
the intermediate step so that the iteration variable is within the feasible region when Newton
iterative solution method is selected; the processing of discrete variable, the processing of inequality
constraints and the solution of the higher-order correction equation and so on.
1.2 The key issues of classical algorithm
(1) The processing of inequality constraints
One of the difficulties in reactive power optimization classical algorithm is the processing of
inequality constraints. Commonly-used method is to construct a slack variable and a penalty
function and so on. Reference [2] choose a collection filter penalty function instead of the
traditional approach to process a large number of inequality constraints in the objective function, it
effectively avoid the impact of the algorithm convergence and convergence rate caused by the
penalty factor. Thus solving the problem of oscillation in an iterative process which led to the
convergence of the algorithm is difficult. But it also increases the complexity of the algorithm.
(2) The processing of discrete variable
It often directly cut result for a simple rounding in the optimization of planning algorithms, and
may misses the optimal solution. Another method is using a discrete integer branch and bound
method to find the optimal solution. However, the computational efficiency is not high, when the
variable dimension increases, a significant increasing will occurs in computation time. Later a
continuous algorithm is introduced in the heuristic rules or penalty function and complementary
constraint method. In summary, although there is strictly theoretical supporting of the algorithm, it
is also easy to produce dimension disasters. And the complex real-time control of mathematical
model can’t meet the requirements.

Intelligent algorithm and its research status


In recent years, intelligent algorithm has been widely used in the global reactive power of
regional power grid voltage real-time optimization. They do not need to form a specified optimal
trajectory by using mathematical models of all kinds of information. They are applicability, and
easy to deal with discrete variables. Evolutionary algorithms such as simulated annealing algorithm,
genetic algorithms, particle swarm optimization, differential evolution algorithm; Tabu search
algorithm and the combination of these algorithms and so on.
1362 Energy Research and Power Engineering 2014

2.1 Evolutionary Algorithms


(1) Genetic Algorithms
The principle of this algorithm is to produce a set of initial solutions under various constraints
limited by the objective function into the fitness function to evaluate their quality, eliminate low
fitness, have a chance to adapt to its high degree of genetic characteristics to the next generation,
retain talented individuals, and appropriate variation.
(2) Particle Swarm Optimization(PSO)
It is a little later for the use of particle swarm optimization to the reactive power optimization.
This algorithm is relatively simple, fewer parameters and easier to achieve, but it is easy to fall into
local optimum. Dynamic inertia weight that is used by the improved PSO algorithm of reference [3]
is based on the particle fitness variance and the current search for the best position change of the
particle. Reference [4] provides a parallel adaptive particle swarm optimization algorithm. Firstly,
the initial populations are randomly divided into N sub-group and then optimize separately in each
subgroup to the proposed method, so that it can enable the parallel computing algorithms.
(3) Evolutionary Algorithms
Differential evolution algorithm is based on the theory of swarm intelligence optimization
algorithm. The algorithm produces a swarm intelligence guidance to optimize search by cooperation
and competition within groups of individuals. The use of based-on a simple differential mutation
and one competitive survival strategies, reduce the complexity of the genetic manipulation. The
algorithm is essentially a kind of greedy genetic algorithm based on real number encoding that
confirms the optimal thought.

Hybrid algorithm
Researchers have proposed a hybrid optimization algorithm for various algorithms existing
problem. A hybrid strategy that is used earlier is an solution that discrete variables and continuous
variables use intelligent algorithm and programming algorithm respectively and alternately.
Another strategy is to use a wide range of intelligent optimization algorithm, and then use planning
algorithm for local optimization.
Reference [5] combine the adaptive immune algorithm and forecast - correction interior point
method, use the AIA for a wide range of global optimization to find the best candidate for the
advantages, use it as an initial feasible point within the point method, and then search in the
neighborhood of certainty initial feasible point through the prediction - correction interior point
method for local, to improve the accuracy and speed of understanding.

The difficulty and key problem of reactive power optimization research


1. multi-objective processing method
Reactive power optimization are considering all kinds of operation indicators, such as safety
margin index, voltage quality indicators, running economy index, etc. reactive power optimization
objective function which is composed of these indicators tend to have different dimensional, and
there is no direct analytical relationship between each other, how to balance is very difficult.
(1) the multi-objective model transformation
Multiple objective function translates into a single target by using a model transformation
method. the original problem is simplified into single a objective optimization problem.
Reference [6] draws the contour corresponding to the various methods by summarizing the
features of all kinds of reactive power optimization of multi-objective model transformation, on the
basis of using the target space. From the geometric point of view, it discusses the model
transformation method to determine optimal direction, and refers to approximate the pareto optimal
solution set, studies the various kinds of model transformation method for the optimum solution of
geometric meaning.
Advanced Materials Research Vols. 986-987 1363

(2) Multi-objective evolutionary algorithm


The design of reference [7] contains two stage of multi-objective optimization and auxiliary
decision-making method to solve the model. In the optimization stage, this paper proposes an
adaptive multi-objective differential evolution algorithm, it chooses the improved not bad not bad in
genetic algorithm, and crowded distance sorting; In decision-making stage, the first is to use the
fuzzy c-means clustering algorithm for Pareto optimal set, and then evaluate the relative efficiency
of decision schemes that is used in every kind of super efficiency data envelopment analysis method
t, then help operators to extract the effective compromise solution.

Improve the efficiency of optimization algorithm


2.1 Decomposition coordination calculation
(1) partitioning method
The grid partition is based on the sensitivity, the electric distance. Reference [8] use the
partition of power system according to the voltage - reactive sensitivity, makes the regional node
voltage reactive power coupling the weakest. First find out the voltage - reactive sensitivity between
nodes under the normal operation, search threshold partition algorithm for node partition system,
use node splitting method to decompose system into several smaller network.
(2) the equivalent method after the partition
It includes NORTON equivalence, ward equivalent method and the technology of REI
equivalence. Reference [9] reduced the REI equivalence partitions of external network, and
compared to the centralized optimization method and the reactive power optimization based on
Ward equivalent area parallel algorithms.
(3) coordinate calculation after the partition between different districts
To solve the augmented Lagrangian function in inseparable problem in reactive power
optimization decomposition-coordination model, reference [10] proposed a near-the-center of all
steps decomposition algorithm based on the center adjacent algorithm. The algorithm put forward a
kind of applicable to special equality constraint optimization problem, which can realize calculate
all the points. Numerical example results show that the proposed algorithm can realize optimizing
reactive decomposition coordination calculation, and the computational efficiency is much higher
than the decomposition-coordination algorithm based on auxiliary problem principle.
2.2 Simplify the search space
In the large area power grid with many nodes, the reactive power optimization model is
complex, with increasing control variables, if considers the entire network of reactive power
compensation point and transformer as variables of optimization algorithm, the search space
increases rapidly, often cannot finish in limited time to fully search the solution space, which is
difficult to meet the requirements of real-time online.
In reference [11], when particle swarm optimization algorithm is used to solve the transient
stability of multi-objective optimal power flow, if encountered in the initialization invalid solution,
the repeat variation for efficient solution is running until the particles, also to a certain extent,
improve the efficiency of algorithm.

Conclusion
This article analyzes the common classic methods, the basic principle of solving its in reactive
power optimization problem solving thinking, compares to different algorithms and their
application scope in reactive power optimization problem solving problems; Pointed out that the
current power grid dispatching operation of reactive power optimization calculation efficiency is
not high. Besides, common algorithm search step length and the search direction with the
combination of the reactive power optimization is still inadequate, these problems restrict the effect
of optimization calculation.
Reactive power optimization problem is a typical NP problem, with the further development of
computer hardware and intelligent algorithm, it is based on the distributed computing technology
1364 Energy Research and Power Engineering 2014

and intelligent algorithm combined with reactive power optimization problem further, reactive
power optimization of online computation efficiency will be further improved.

References
[1] qishan-fufeng-meixian ZongWei Qin Lijun. Power system reactive power optimization in linear
programming problem of linear step dynamic adjustment strategy [J]. Power grid technology, 2006,
30 (18) : 40-44.
[2] Yang Shuo Zhou Jing Yang, li qiang, etc. Based on filter collection points in reactive power
optimization research [J]. Power system protection and control, 2011, 19.
[3] Badar A Q H, S, B Junghare a. s. Reactive power control using dynamic Particle Swarm
Optimization for real power loss minimization [J]. International Journal of Electrical power &
Energy Systems, 2012, 41 (1) : 133-136.
[4] Liu Shicheng Zhang Jianhua, Liu Zongqi. Parallel adaptive particle swarm algorithm application
in power system reactive power optimization [J]. Power grid technology. 2012, 4 (1) : 108-112.
[5] bypassed his Li Honglu, 8 XinYu. Based on adaptive immune algorithm and forecast -
correction interior-point method of reactive power optimization [J]. Journal of tianjin university,
2008, 41 (2) : 168-174.
[6] Li Zhihuan Li Yingong, reactive power optimization of multi-objective model conversion
method of contour analysis [J]. Journal of electrotechnics, 2012, 27 (6) : 153-160
[7] Qiu Wei Zhang Jianhua, Liu Nian. Considering environmental factors and voltage stability of
multi-objective optimal power flow [J]. Journal of electrotechnics, 2012, 27 (2) : 190-197.
[8] Chen Canxu Liu Mingbo. Reactive power optimization decomposition coordination calculation
node partition method [J]. Power system protection and control. 2010 (017) : 1-6.
[9] Liu Zhiwen Liu Mingbo. Reactive power optimization based on REI equivalence region parallel
computing [J]. Power grid technology, 2012, 36 (3).
[10] li zhi, Yang Honggeng. One for decomposition coordination all points near the center of
reactive power optimization algorithm [J]. Proceedings of the csee, 2013, 33 (001) : 77-83.
[11] Ye Chengjin Huang Minxiang. Considering the transient stability of multi-objective optimal
power flow [J]. Proceedings of the csee, 2013, 33 (10) : 137-145..

You might also like