Development of a back-propagation neural network and adaptive grey wolf optimizer algorithm for thermal comfort and energy consumption prediction and optimization
Lu Li,Yunfei Fu,Jimmy C.H. Fung,Huamin Qu,Alexis K.H. Lau
DOI: https://doi.org/10.1016/j.enbuild.2021.111439
IF: 7.201
2021-12-01
Energy and Buildings
Abstract:Heating ventilation and air conditioning (HVAC) systems provide a comfortable indoor thermal environment, but in the process of attaining appropriate indoor thermal comfort levels, they usually entail high energy consumptions. It is therefore imperative to balance thermal comfort value with energy consumption. However, such research currently faces two problems: one, it is difficult to obtain accurate parameters pertaining to the indoor environment of buildings, particularly near heat source areas; two, it is the diametrical nature of having to simultaneously maintain thermal comfort and keep energy consumption low. Therefore, this study aims to propose a rapid thermal comfort level prediction and optimization algorithm, as well as a method to minimize the energy consumption using only a computational fluid dynamic (CFD) database that is compact in size. Firstly, CFD is used to implement the database that stores data on indoor airflow and temperature distributions of different building structures and indoor conditions. Next, using the database as a basis, a back-propagation neural network (BPNN) is developed to predict the thermal comfort level. The adaptive grey wolf optimizer (GWO) algorithm is then applied to optimize the thermal comfort value, and the latest control methods: the artificial neural network (ANN)-genetic algorithm (GA) and ANN-particle swarm optimizer (PSO) algorithm are compared against the BPNN-based adaptive GWO method. The results show that the BPNN-based adaptive GWO algorithm can rapidly predict the thermal comfort level and have strong optimization ability. Meanwhile, 1.01% of energy savings are achieved.
energy & fuels,construction & building technology,engineering, civil