Title :
Multi-variable time series forecasting for thermal load of air-conditioning system on SVR
Author :
Xuan, Zhou ; Qing-dian, Liu ; Guo-qiang, Liu ; Jun-wei, Yan ; Jian-cheng, Yang ; Lie-quan, Liang ; Wei, Hu
Author_Institution :
School of Mechanical & Automotive Engineering, South China University of Technology, Guangzhou 510640, China
Abstract :
Single-variable time series forecasting method is the major forecasting method for thermal load time series forecasting method of air-conditioning system, without considering other factors, such as outdoor weather parameters and relevant operating parameters. Therefore, a multi-variable forecasting method for thermal load based on Support Vector Regression (SVR) was proposed in this paper. The impacts from different factors on thermal load predicting accuracy were also analyzed by Statistical Product and Service Solutions (SPSS). Then, the SVR prediction model for air-conditioning load was set up whose hyper-parameters were optimized by particle swarm optimization algorithm and Expected Error Percentage (EEP), Mean Bias Error (MBE) were used to evaluate the prediction accuracy for four models of different input variables. In addition, the operating data were divided into subsets to model due to the load characteristics. The simulation results showed that the prediction accuracy of this multi-variable method improved the accuracy of 7.4%, compared to single-variable method.
Keywords :
Accuracy; Atmospheric modeling; Load modeling; Predictive models; Support vector machines; Thermal loading; Time series analysis; Particle Swarm Optimization (PSO) algorithm; Support Vector Regression; multi-variable time series forecasting method; thermal load of air conditioning system;
Conference_Titel :
Control Conference (CCC), 2015 34th Chinese
Conference_Location :
Hangzhou, China
DOI :
10.1109/ChiCC.2015.7260952