Paper Conference

Proceedings of Building Simulation 2017: 15th Conference of IBPSA


Using an Adaptive Meta-Model Evolutionary Algorithm for Mixed-Integer Type Building Design Optimization for Building Simulation 2017 Conference

Weili Xu1, Khee Poh Lam1, Omer T. Karaguzel1
1School of Architecture, Carnegie Mellon University, Pittsburgh, U.S.

Abstract: The adaptive meta-model evolutionary algorithm has demonstrated its excellent convergence and faster speed than the conventional algorithms in building design evaluations (Xu et al., 2016). However, this algorithm is built on integer-type problems while most of the building designs have mixed integer-type parameters. In this study, numerical calculations are added in the operators of the algorithm. These updates allow the algorithm to work on mixed integertype problems. The updated algorithm is applied to a large office building with 350 thermal zones. The design of the daylight shelf and building-integrated photovoltaic (BiPV) systems are included in the design parameters. The results indicate that despite its high capital costs, BiPV system is favored by most of the optimal designs. On the contrary, the light shelf system is not favored in this integrated building design optimization for various reasons such as the low daylight harvesting potential from the high efficiency lighting technology and the objectives of the study.
Pages: 1274 - 1283