A Genetic Algorithm for Mixed-Integer Optimisation in Power and Water System Design and Control

Abstract

This chapter presents a dual mutation Genetic Algorithm (GA) which employs individual mutation schemes for different types of decision variables in optimal design and control problems. The GA is integrated with nuclear power station whole plant design and performance optimisation and optimal water system quality and quantity control problems. The objective of the nuclear power station design problem is to improve the overall performance of the thermal cycle of a nuclear power plant by optimising both station design and operation using integrated evolutionary search and conventional optimisation techniques. The objective of the water system control problem is to find the best trajectories of control variables through the controlled time period with minimum cost. The problems pursued are in the class of mixed-integer, nonlinear constrained optimisation problems. After an initial parametric study of various adaptive search and classical optimisation techniques to determine their relative potential within a search space characterised by heavy nonlinear constraints, a hybrid approach has been developed. This firstly utilises a genetic algorithm (GA) as a preprocessor to identify a feasible region within the search space before employing a dual-mutation GA strategy to search the space of mixed-integer variables. A linear programming optimisation routine then periodically searches from the best GA points with the design/control configuration fixed to return an optimal solution in terms of system performance.