Energy management and design of centralized air-conditioning systems through the non-revisiting strategy for heuristic optimization methods

Abstract

It is getting more and more popular to apply heuristic optimization methods, like genetic algorithm (GA) and particle swarm optimization (PSO), to handle various engineering optimization problems. In this paper, optimization problems of typical centralized air-conditioning systems were solved by the non-revisiting (Nr) strategy, which was proposed to be incorporated into the common heuristic methods for improving the optimization effectiveness and reliability. This approach can store the evaluated fitness values in an archive with minimal computer memory, detect the revisits and prevent them from re-evaluating. It is particularly useful for the problems formulated by dynamic simulation or detailed modeling with very demanding computational time for function evaluation. The non-revisiting strategy can facilitate the search of the global optimum by its parameter-less adaptive mutation capability. In the optimization problems of central air-conditioning systems, it was found that the NrGA and NrPSO could search better solutions at a limited number of function evaluations than the conventional GA and PSO did. The ultimate goal is to determine the required parameters for optimal design and energy management. The proposed strategy can be applied to similar types of air-conditioning or engineering optimization problems, and possibly incorporated into other kinds of heuristic optimization methods.