Ultra-short-term Power Forecasting of Wind Farm Cluster Based on Spatio-temporal Graph Neural Network Pattern Prediction

Abstract

Timely and accurate wind farm cluster power prediction is of great significance to the stability of the power system. Due to the strong randomness and uncertainty of wind, traditional prediction methods cannot meet the requirements of power prediction tasks. Moreover, many methods ignore the temporal and spatial correlation between wind farms, so it is difficult to achieve more accurate prediction. In this paper, we propose a power prediction method based on the prediction of the cluster output pattern. We use the spatio-temporal graph neural network to extract the spatio-temporal correlation between wind farms. We base the cluster power prediction problem on a graph instead of using methods such as griding to simplify the spatial correlation between power farms. Experiments show that our proposed method is superior to other methods of real wind farm cluster power dataset.