Research on the Photovoltaic Power Prediction Based on Optimization by Extreme Learning Machine

Abstract

We propose a prediction method based on optimization of photovoltaic power on Extreme Learning Machine. To predict the level of PV power min, the method uses Extreme Learning Machine (ELM) which were constructed spring, summer, autumn and winter four prediction model. We selected a group of important factors affecting the construction of photovoltaic power as the predictive model input feature. Then, we learned the relationship between selected features and cross-validation to predict function mapping on error. And, we use an integrated search strategy Improved Chaos particle swarm to optimize ELM Model parameters. Last, Ningxia predictions of a PV power plant measured data to verify the validity of the proposed method.