A New Technique For Applying Genetic Algorithms To Power System Security Optimisation Problems

Abstract

This aper presents a new technique for harnessing the power of Anetic Algorithms, (GA), for the optimisation of power system transient security and economy. The paper summarises the 'classical' application of GAS to the problem, and discusses a number of factors which contribute to the poor performance of the resulting optimiser, both in terms of solution quality and computational load. The main concerns raised are then addressed through a novel, mathematically justified, encoding strategy based on the implicit inclusion of important constraint equations, together with eqlicit adherence to the Building Block Theorem which has been previously offered as an important pre-requisite for the convergence of GAs[l]. One of the most important properties of the new technique is that it achieves its high performance without the need for any assumptions or ap roximations, thus guaranteeing the quality of the final sofkon. Indeed it is claimed that the reduced computational effort and improved search strategy results in the location of better solutions than those obtainable using any existing technique. However, the design of the algorithm also provides a number of routes by which auxiliary information can be incorporated at the core of the optimisation process to achieve faster convergence at the possible expense of solution quality for liations where execution time is of prime importance. %st the algorithm has been designed for fast execution on a single processor machine, significant care has been exerted to ensure that parallel implementation can be easily achieved using either distributed processing on a cluster of single-processor workstations, or an appropriate multiprocessor machine. The paper concludes by presenting comparative results for a classical GA and one based on the new technique, both applied to a small power system for which com lete knowledge of the 'best' solutions is available. &ese results clearly demonstrate the advantages of the presented method.