Hybrid Genetic Algorithm and Support Vector Regression in Cooling Load Prediction

Abstract

This study develops a novel methodology hybridizing genetic algorithms (GAs) and support vector regression (SVR) and implements this model in a problem forecasting hourly cooling load. The aim of this study is to examine the feasibility of SVR in building cooling load forecasting by comparing it with back-propagation neural networks (BPNN) and the autoregressive integrated moving average (ARIMA) model. To build an effective SVR model with predictive accuracy and generalization ability, real value GAs are adopted to automatically determine the optimal hyper-parameters for SVR. The experimental results demonstrate that the hybrid model provides better prediction capability than the BPNN and ARIMA models, and therefore is considered as a promising alternative method for forecasting building hourly cooling load.