LARGE-SCALE UNIT COMMITMENT USING A HYBRID GENETIC ALGORITHM

In the new competitive electricity supply industry, there is a renewed interest in algorithms that can provide savings in operation costs. A n optimal scheduling of generators can provide substantial annual savi ngs in fuel costs, but this highly constrained non-linear mixed intege r optimisation problem can only be fully solved by complete enumeratio n, a process which is not computationally feasible for realistic power systems. In the recent past, evolutionary computation techniques have been applied to the solution of the unit commitment problem, but when implemented as stand alone systems, they suffer from computational ti me limitations, especially when the systems are scaled up. This paper proposes a hybrid genetic algorithm incorporating a priority list unit ordering scheme to solve the generator scheduling problem. Test resul ts on networks with up to 110 generators are presented and the results demonstrate the viability of the hybrid GA method for unit commitment in realistic power systems. Copyright (C) 1996 Elsevier Science Ltd