Deep attention ConvLSTM-based adaptive fusion of clear-sky physical prior knowledge and multivariable historical information for probabilistic prediction of photovoltaic power

Abstract

Photovoltaic (PV) power is an important way to utilize solar energy. Accurate PV power forecast is crucial to the large-scale application of PV power and the stability of electricity grid. This paper proposes a novel method for short-term photovoltaic power forecast using deep attention convolutional long short-term memory (ConvLSTM) network and kernel density estimation (KDE). ConvLSTM has strong ability to automatically extract the information from multiple variables related to PV power and the temporal information from historical PV data. Considering that clear-sky solar radiation is an important physical prior knowledge in PV power generation, attention mechanism is innovatively introduced into ConvLSTM to dynamically and adaptively adjust the weights of the physical prior knowledge and the historical PV data in forecast. KDE is used for estimating the joint probabilistic density function and giving the probabilistic confidence interval. Experiments in the three-year actual photovoltaic power data of two provinces in Belgium verified the effectiveness of the proposed method. Comparison experiments were made with naive persistence and other eight conventional methods including autoregressive integrated moving average with exogenous variable (ARIMAX) model, convolutional neural network (CNN), long short-term memory (LSTM) network, multilayer perceptron (MLP), support vector regression (SVR), extreme learning machine (ELM), classification and regression tree (CART) and gradient boosting decision tree (GBDT). Experiments show that the proposed method can realize the optimal fusion of the historical data and clear-sky prior knowledge, and significantly improve the PV power forecast accuracy in all seasons of a year.