Improved Satin Bowerbird optimization algorithm to optimize school buildings with Elman neural networks and life cycle cost: a case study in China

Abstract

ABSTRACT Improving the energy efficiency, indoor air quality, and energy storage of public buildings, especially schools, is crucial due to the associated energy consumption and costs. In this study, a retrofitting approach is presented, which employs the Improved Satin Bowerbird optimization algorithm, Elman neural network, and life cycle cost analysis to optimize the insulation thickness of the exterior walls and roof of two school building models in China. The study focuses on two objective functions, namely annual heating load and indoor thermal discomfort due to overheating, using five design variables. To estimate the objective function values, 95 Elman neural networks are trained and verified, with models 1 and 2 achieving r2 values of 0.9851 and 0.9827, respectively. The optimization technique is employed to specify the Efficient fronts, with the choice criteria being to minimize the life cycle cost. Compared to the starting condition, model 1 reduces yearly heating load and life cycle cost by 27.5% and 29.5%, respectively. Similarly, model 2 lowers yearly heating load and life cycle cost by 22.1% and 25.3%, respectively. The results demonstrate the success of the approach in enhancing energy efficiency and reducing costs in school buildings, both of which are crucial areas for public investment.