Investigating the performance of genetic algorithm and particle swarm for optimizing daylighting and energy performance of offices in Alexandria, Egypt

Abstract

PurposeThis study aims to compare the performance of two nature-inspired metaheuristics inside Grasshopper in optimizing daylighting and energy performance against brute force in terms of the resemblance to ideal solution and calculation time.Design/methodology/approachThe simulation-based optimization process was controlled using two population-based metaheuristic algorithms, namely, the genetic algorithm (GA) and particle swarm optimization (PSO). The objectives of the optimization routine were optimizing daylighting and energy consumption of a standard reference office while varying the urban context configuration in Alexandria, Egypt.FindingsThe results from the GA and PSO were compared to those from brute force. The GA and PSO demonstrated much faster performance to converge to design solution after conducting only 25 and 43% of the required simulation runs, respectively. Also, the average proportion of the resulted weighted sum optimization (WSO) per case using the GA and PSO to that from brute force algorithm was 85 and 95%, respectively.Originality/valueThe work of this paper goes beyond the current practices for showing that the performance of the optimization algorithm can differ by changing the urban context configuration while solving the same problem under the same design variables and objectives.