Short term wind power prediction for regional wind farms based on spatial-temporal characteristic distribution

Abstract

Accurate regional wind power prediction is of great significance to the wind farm clusters integration and the economic dispatch of the regional power grid. The complex spatiotemporally coupled characteristics between multiple wind farms bring challenges to wind power prediction (WPP) of regional wind farm clusters. In this context, this paper proposes a regional WPP method using spatiotemporally multiple clustering algorithm and hybrid neural network to learn the potential spatial-temporal dependencies of regional wind farms. In which, a long-term daily power curve similarity method is proposed to identify spatially correlative wind power plants in long-term. Furthermore, the spatio-temporal wind farm sub-clusters are dynamically recognized by the similar fluctuation trend of short-term power sequences. On this basis, a spatial-temporal integrated prediction model consisting of the improved convolutional neural network (I–CNN) and the bidirectional long short-term memory (BILSTM) network is established for spatio-temporal sub-cluster based on point clouds distribution. Finally, the effectiveness of the proposed regional wind power forecasting framework is validated by using the Wind Integration National Dataset Toolkit, and the results show that the method improves accuracy effectively.