Intra-Hour Photovoltaic Generation Forecasting Based on Multi-Source Data and Deep Learning Methods

Abstract

Global issues pertaining to climate change have necessitated the rapid deployment of new energy sources, such as photovoltaic (PV) generation. In smart grids, accurate forecasting is essential to ensure the reliability and economy of the power system. However, PV generation is severely affected by meteorological factors, which hinders accurate forecasting. Various types of data, such as local measurement data, numerical weather prediction, and satellite images, can reflect meteorological dynamics over different time scales. This paper proposes a novel data-driven forecasting framework based on deep learning, which integrates an advanced U-net and an encoder-decoder architecture to cooperatively process multi-source (time series recording and satellite image) data. The adaption of the neural networks to the data sources and the collaborative learning of both spatial and temporal features boost the model accuracy. The experimental results for 50 real-world PV power stations indicate that the proposed framework features a higher accuracy than that of other baseline models.