Research on Photovoltaic Power Generation Prediction Based on Ensemble Learning DBO-LSTM-EBRB

Abstract

Photovoltaic power generation represents a pivotal form of renewable energy. Accurate power prediction is therefore of paramount importance to the advancement of the energy revolution. This paper examines the Yumara Solar System PV plant in the central region of Australia as a case study. The research begins with the use of the k nearest neighbor algorithm (KNN) and extreme gradient boosting algorithm (XGBoost) to pre-process the PV power generation data from the PV plant. This is followed by the creation of line graphs, which illustrate the changes in variables in the PV power generation data from the PV plant in 2022. Subsequently, this paper employs the dung beetle optimization (DBO) algorithm to optimize the long short-term memory (LSTM) network and construct the DBO-LSTM model. The DBO-LSTM model is further integrated with the extended belief rule base (EBRB) model under the integrated learning framework. The weight assignments are reasonable, in accordance with the model features, for the purpose of constructing the DBO-LSTM-EBRB model for predicting the capacity of photovoltaic power generation. This paper presents experimental results that demonstrate superior predictive performance of the DBO-LSTM-EBRB model in comparison to the traditional LSTM and EBRB models. The R2 value achieved for the DBO-LSTM-EBRB model ranges from 0.950 to 0.978.