Window-Varying Particle Filter for Parameter Identification of Space Thermal Model

Abstract

Due to the large proportion of heating, ventilating, and air-conditioning (HVAC) devices in electricity demand, energy conservation can be realized by applying appropriate control strategies on HVAC loads. For this purpose, a proper space thermal model is needed, in which parameter identification is an essential problem. The identification results of parameters are usually sensitive to measurement noise and initial value using conventional particle filter method. To overcome these issues, a modified particle filter algorithm, called window-varying particle filter (WVPF), is proposed to identify time-varying parameters of space thermal model by adaptively changing the prediction time window in this paper. Furthermore, a measurement system is developed to collect data for parameter identification and to verify the accuracy of space thermal model. Experiment and simulation results confirm the desirable performance of the proposed method for parameter identification.