Photovoltaic Power Prediction Based on EMD-BLS Model

Abstract

While the energy consumption structure is changing in this world, the consumption rate of renewable energy is gradually increasing. This paper takes photovoltaic power as the research object, analyses the power curve characteristics in different climatic change situation and the correlation between different weather factors and photovoltaic(PV) power output, and then puts forward a photovoltaic power prediction model based on empirical mode decomposition-broad learning system (EMD-BLS). First, the PV power historical sequence needs to be reconstructed after pretreatment, then the EMD method is used to decompose the reconstructed output sequence. The broad learning system model is established for each subsequence. Finally, the prediction results of each sub-series are superimposed to get the prediction of photovoltaic generation. The model is tested with a genuine data of photovoltaic power generation in a region of our country. Compared with the traditional broad learning system (BLS), extreme learning machine (ELM), LSTM, and other prediction models, The prediction error of the proposed model is lower, which can effectively improve the prediction accuracy of photovoltaic power generation.