A Fast Market Competition Approach to Power System Unit Commitment Using Expert Systems - Intelligent Computing for Sustainable Energy and Environment - page 442

A Fast Market Competition Approach to Power System
Unit Commitment Using Expert Systems

Lijun Xu 1 , 2 ,KangLi 2 , and Minrui Fei 1

1 Shanghai Key Laboratory of Power Station Automation Technology,
School of Mechatronical Engineering and Automation, Shanghai University, Shanghai, China
2 School of Electronics, Electrical Engineering and Computer Science
Queen's University Belfast, University Road, Belfast, BT9 5AH, UK
ruby [e-mail], [e-mail], [e-mail]

Abstract. This paper proposes a novel two-layer hierarchical market competi-
tion algorithm (THMCA) combined with expert system for the unit commit-
ment(UC) problem in power systems. Two hierarchical population individuals
are defined in the algorithm, namely the holding companies and the subsidiary
companies which altogether form conglomerates. Market competitions among
these conglomerates lead to the convergence to a monopoly at the end, resulting
in an optimal solution of the above problem. In the meanwhile, expert system is
used to produce several expert rules for heavy constraints handling not only in
the pre-scheduling process and in the THMCA process as well, ensuring that the
positions of all companies are feasible and near-optimal solutions to the UC prob-
lem. The algorithm is shown to have a fast execution speed for UC application
and the comparison simulation results on a power system with up to 100 gener-
ating units have demonstrated the effectiveness on cost reduction of the proposed
method.

1 Introduction

The unit commitment (UC) [1-5] is a complex combinatorial optimization problem to
determine the power generation schedule of units in order to satisfy load demand at
minimum cost in power systems. A variety of constraints on the generating units must
be satisfied, including the time-dependent ones. Therefore, to solve the nonlinear, non-
convex, large scale, mixed integer UC problem is a challenging task. It should also
be noted that to solve this problem can be computationally expensive for large power
systems.
To tackle this issue, several intelligent optimization methods can be adopted, for
example genetic algorithm (GA) [6], particle swarm optimization (PSO) [7], clone evo-
lutionary algorithm [8], bacterial foraging (BF) [9] and shuffled frog leaping algorithm
(SFLA) [10] etc. GA is one of the early proposed evolutionary algorithms which has
found many successful applications. However, it is computationally expensive, and the
convergence cannot be guaranteed. The comparison results [7] show that the PSO is
more efficient than GA. But unfortunately, in solving the UC problem by PSO, many
particles in a randomly created swarm are not feasible solutions because of their vio-
lation of constraints, and especially to meet the minimum up/down time limits. Bacte-
rial foraging (BF) algorithm is a feature selection method based on a heuristic search