Uncertainty Quantification in Photovoltaic Power Output Forecasting Based on Long Short-Term Memory Neural Networks and Kernel Density Estimation

To improve photovoltaic power output forecasting accuracy and quantify forecasting uncertainty, this paper proposes an intraday forecasting method for uncertain photovoltaic power output based on long short term memory (LSTM) neural networks and kernel density estimation (KDE). First, considering that photovoltaic power output is mainly affected by weather conditions, the K means clustering algorithm is used to classify three weather types. Then, an intraday photovoltaic power output forecasting model is constructed by combining interval prediction based on KDE with the LSTM model. Based on the point forecasting results of the LSTM neural network, interval forecasting results are generated using KDE under given confidence levels. Finally, the effectiveness of the model is verified using operating data from an actual photovoltaic power station. Experimental results show that the proposed method improves the forecasting performance of photovoltaic power output.