Energy Consumption Forecast Based on Coupling PSO-GPR

Consumption for energy has a great influence on sustainable development of economy and society. The scientific and rational energy development strategy can effectively guarantee social and economic stability and orderly development and energy development strategy is inseparable from the right to accurately predict energy demand and analysis. In order to improve the accuracy of prediction of energy consumption, hybrid forecasting model for energy consumption based on Gaussian process (GPR) and Particle Swarm (PSO) is proposed. Firstly, PSO algorithm is employed to optimize two parameters in covariance function, and then the initial value of parameters are obtained; next time series are trained in GPR model for energy consumption. Under the Bayesian framework, parameters in covariance function can be optimizing again. Finally, we can forecast energy consumption in well-trained GPR model, and the results can be compared with the auto-regressive integrated moving average model and exponential smoothing models. The results show that the proposed model has good stability and high prediction accuracy. It is suitable to be applied in forecasting consumption for energy.