Short-Term Prediction Model of Output Power of Photovoltaic Power Generation System Based on Improved LSTM Neural Network and Traditional Regression Algorithm

To address the impact of photovoltaics on the power system, a short-term prediction model of the output power of a photovoltaic power generation system based on an LSTM neural network and a traditional regression algorithm is proposed. It can be directly predicted using the historical output power value and past and predicted daily weather information. We analyze various output power factors and screen out suitable parameters as the input data. Given the defects of the traditional BP neural network, it will lead to the local minimum point and slow convergence speed, the use of increasing the momentum term and variable learning. It is improved by the method of combining rates. Finally, we verify the effectiveness of the proposed model by predictive analysis.