Ultra-short-term Interval Prediction Model for Photovoltaic Power Based on Bayesian Optimization

Photovoltaic power generation has great potential to replace the traditional coal-fired power generation which has high pollution and high energy consumption. However, photovoltaic power generation has great uncertainty, so how to accurately predict the photovoltaic power generation is an urgent issue to be solved. Currently, Artificial Intelligence provides a promising way to improve prediction accuracy and reliability. In addition, compared with deterministic prediction, probabilistic prediction can analyze the probability distribution of photovoltaic power generation at a certain time period to obtain more favorable results. In this paper, based on a deep learning algorithm, a probabilistic algorithm of time series prediction based on Bayesian optimization is proposed. The input data are optimized by the Bayesian networks, and deterministic results are predicted by using the LSTM network. A case study of prediction results shows that the probabilistic prediction interval of photovoltaic power under 15%-95% incredible conditions can be effectively obtained. The actual operation data of the photovoltaic power station are used to verify and the prediction results show that the precision of the optimized data is increased by about 20%, which shows good engineering application