School of Electrical and Power Engineering, Hohai University, Nanjing 211100, China

Aiming at the problems of insufficient use of sky image information and large errors in ramp power forecasting, which lead to the limited predictive performance improvement, an ultra-short-term photovoltaic power forecasting method based on multi-level sky image features and broad learning is proposed. Firstly, the multi-level features of the ground-based sky image are extracted as the image features of the power forecasting model. At the same time, the cloud coverage and cloud change rate are introduced as image features of the ramp recognition model. Secondly, combined with the historical power data, the photovoltaic power forecasting model and the ramp recognition model based on the broad learning are developed. Finally, if the ramp recognition result is a non-ramp event, the forecasting results are obtained according to the power forecasting model, but if the ramp recognition result is a ramp event, the power forecasting model is incrementally updated using the historical data related to the ramp event, and the forecasting results are obtained based on the updated power forecasting model. The experimental results show that the proposed method can effectively improve the forecasting accuracy of ultra-short-term photovoltaic power.

This work is supported by National Natural Science Foundation of China (No. 52077062).