ANN-based thermal control models for residential buildings

Abstract

This study aimed to develop an Artificial Neural Network (ANN)-based advanced thermal control method for creating more comfortable thermal environments in residential buildings. The proposed control method consisted of a thermal control logic framework with four thermal control logics therein, including two predictive and adaptive logics using ANN models, and a system hardware framework. The models were designed to achieve thermal comfort for living areas, taking into account not only air temperature, but also humidity or PMV as a control variable; and to reduce overshoots and undershoots of a control variable using ANN-based predictive and adaptive control. Incorporating IBPT (International Building Physics Toolbox) and MATLAB, a typical two-story single-family home in the U.S. was modelled for testing the performance of developed thermal control methods. Analysis revealed that ANN-based predictive and adaptive control strategies created more comfortable thermal conditions than did typical thermostat systems in terms of increased comfort period of air temperature, humidity, and PMV, and reduced over and undershoots. Thus, the proposed control methods using ANN can be concluded to have the potential for enhancing thermal comfort in residential buildings.