Paper Conference

Proceedings of eSim 2008: 5th Conference of IBPSA-Canada

     

MULTIOBJECTIVE OPTIMIZATION OF BUILDING DESIGN USING GENETIC ALGORITHM AND ARTIFICIAL NEURAL NETWORK

Laurent Magnier, Liang Zhou, Fariborz Haghighat

Abstract: Building optimization, especially using multiple objectives, is a time-consuming process. The GAINN approach presented in this study first uses simulationbased artificial neural network to characterize building behaviour, and then combines it with a genetic algorithm for optimization. This process has proven to enable fast and reliable optimization. GAINN was improved in this study by integration of multiobjective evolutionary algorithms (MOEAs). Two new MOEAs named NSGAINN and PLAGUE were designed specifically for the presented methodology. They are both based on NSGA-II but take advantage of extremely quick calculations. They were tested over bench test functions, and compared with NSGA-II based on dominated space measure. Results will be presented and discussed here. Finally, a past case study using GAINN methodology will be re-optimized with developed MOEA. Improvement in results compared to classical weighted sum method will be shown and discussed.
Pages: 86 - 93
Paper:
esim2008_086_093