A Novel Deep Learning Fusion Model for Probabilistic Prediction of Photovoltaic Power

Abstract

When a high percentage of photovoltaic power is connected to the power system, the volatility and non-smoothness of the photovoltaic power output can seriously affect the safe and stable operation of the power system. Accurate PV power prediction is of great importance to the safe and economic operation of the power system and to the dispatch of the power system. PV power probability interval prediction can effectively quantify the uncertainty of PV power prediction, which can provide more comprehensive information for power system decision makers compared to traditional point prediction models and help power system risk control and decision making. To address the shortcomings of low prediction accuracy of a single model, a CNN-BiLSTM fusion prediction model is constructed based on convolutional neural network (CNN) and bi-directional long and short-term memory neural network (BiLSTM) in this paper. Finally, the quantile regression model and kernel density estimation method are used to obtain reliable probability interval prediction results. Simulations were carried out using PV power data under different weather types and compared with other probabilistic interval prediction models based on a variety of evaluation metrics. The results show that the QR-CNN-BiLSTM PV power output probability prediction model performs better.