Optimal simplified thermal models of building envelope based on frequency domain regression using genetic algorithm

Simple and effective building energy models are essentially needed for many applications, such as building performance diagnosis and optimal control, etc. The energy model involves two important parts of building components, i.e., building envelopes and building internal mass. This paper presents a methodology for parameter optimization of 3R2C thermal network model of building envelopes (composed of three resistances and two capacitances) based on frequency domain regression using genetic algorithm (GA). First, the theoretical frequency characteristics of heat transfer through building envelope are calculated using detailed physical description within the frequency range of concern. Second, the frequency characteristics of the simplified 3R2C model are calculated with random values of individual resistances and capacitances which constrain to total thermal resistance and capacitance. Then, the errors between the theoretical frequency characteristics and the frequency characteristics of the simplified model are calculated. Finally, GA estimator is developed to optimize the parameters of the simplified model, allowing the frequency responses of the simplified model match the actual heat transfer through building envelope the best. Various case studies are conducted also to validate the parameter optimization method of the simplified 3R2C model. The accuracy of simplified models for constructions of different weights is studied.