The Real-time Forecast of Photovoltaic Output Based on Empirical Mode Decomposition

The randomness and volatility of photovoltaic (PV) output brings difficulties to do real-time forecast. This paper proposed a new method called EMD-SVM-GA to forecast PV output real-time. Three steps were needed to carry out the method. Firstly, decompose historical PV output data sequence into several Intrinsic Mode Functions (IMF) and a Residual Component based on Empirical Mode Decomposition (EMD) theory. Secondly, construct Support Vector Machine (SVM) models for every component and search best parameters by Genetic Algorithm. Then superpose all forecast results to obtain the final data. Examples show that EMD-SVM-GA method has better performance and higher accuracy.