Reservoir system optimisation using a penalty approach and a multi-population genetic algorithm

Abstract

A multi-population genetic algorithm (GA) was used to optimise a system of two reservoirs that supplies monthly varying demands and environmental flow requirements. Optimisation aimed at minimising the penalty resulting from non-supply of water and the occurrence of low reservoir storage states that would limit non-consumptive utilisation of water in the reservoirs. Four cases were analysed viz. Case I: Reservoir capacities and demands were fixed and the operating rules were optimised; Case II: Demands were fixed and the reservoir capacities and operating rules were optimised; Case III: Reservoir capacities were fixed and the demands and operating rules were optimised; and Case IV: Reservoir capacities, demands and operating rules were optimised. The genetic algorithm obtained reasonable solutions for all cases. A detailed analysis of Case IV obtained several high-performance solutions of varied sizes and supply capabilities. This analysis revealed specific limitations of supply reliability and the expected storage states of one of the reservoirs. The analysis also obtained the ranges within which the optimal monthly operating rules for the system are expected.


Water SA Vol.29(3) 2003: 273-280