Research And Application Of Solar Radiation Forecasting Technique Based On Decomposition-Clustering-ensemble Learning Paradigm

Due to energy crisis and environmental problems, it is very urgent to find alternative energy sources nowadays. Therefore, a novel hybrid ensemble learning model integrating the ensemble empirical mode decomposition(EEMD), least squares support vector regression(LSSVR), gravitational search algorithm(GSA)and K-means clustering algorithm(Kmeans), based on the principle of/decomposition-clustering-ensemble0, is first proposed for short-, medium- and long-term solar radiation forecasting in this paper. This hybrid ensemble learning paradigm is specifically designed to solve the problem of solar radiation modeling with high volatility, complexity and irregularity. In this hybrid ensemble learning paradigm,the first application of a competitive decomposition algorithm, EEMD, is used to decompose the original solar radiation time series into the intrinsic mode function(IMFs) and a residual term. Secondly, the LSSVR model, which is optimized by GSA, is implemented to forecast the IMFs and residual components separately. Thirdly, Kmeans clustering algorithm is utilized to distinguish the difference between the prediction results of all components, based on its inherent properties, and it is divided into different clusters. Finally, these components predicted results of different clusters are aggregated into an ensemble result as final prediction, using other different GSA-LSSVR tools. For illustration and verification purposes, the proposed learning paradigm is used to predict solar radiation in Beijing. Empirical results demonstrate that the proposed EEMD-LSSVR-K-LSSVR learning paradigm statistically outperforms other benchmark models(including popular single models and similar hybrid ensemble models) in both prediction accuracy(in terms of level and directional measurement) and effectiveness(in terms of robustness), indicating that it is a promising tool to predict complicated time series with high volatility and irregularity.