Clean Energy Consumption Forecast Based on GA-LSSVR Hybrid Learning Paradigm

In order to improve the forecasting accuracy for clean energy consumption with inherently high complexity, a hybrid learning paradigm integrating genetic algorithm (GA) and least squares support vector regression (LSSVR), i.e., GA-LSSVR model, is formulated in this study. In this learning paradigm, LSSVR, as a powerful artificial intelligence tool, is employed to forecast clean energy consumption, furthermore, GA is employed to determine the parameters in LSSVR. Taking the Chinese hydropower energy as sample, empirical results indicate that the GA-LSSVR model significantly outperforms other benchmark models, including Artificial neural network (ANN), Autoregressive integrated moving average (ARIMA) and a set of hybrid models based on LSSVR and other searching methods (e.g., particle swarm optimization (PSO) and simulated annealing (SA)), in both level prediction accuracy and directional forecasting. The GA-LSSVR learning paradigm can be extended as an effective prediction technique for other complex data.