Improving thermal performance of automatically generated floor plans using a geometric variable sequential optimization procedure

Abstract

This paper presents an approach for the optimization of floor plan designs. These are generated using a hybrid evolutionary approach, which produces alternative designs according to the user’s preferences and requirements. Once generated, an optimization algorithm is used to improve the thermal performance of each solution. The algorithm evaluates each possible transformation for several design variables in each floor plan, such as floor plan orientation and reflection, window orientation and size, overhang size, fin size, and wall translation. A geometric variable sequential optimization procedure is used to satisfy the user’s design strategy.A case study of a single-family house is carried out, where two design sets, with 12 alternative solutions each, are generated, assessed, and optimized according to thermal performance criteria. The results demonstrate that the thermal performance of the floor plans may improve by up to 41% for single-level solutions and 54% for two-level designs. When comparing solutions within each design set, the floor plan design which ranks first is 33% and 29% better than the worst design, in the first and second design set respectively.