Non-crossing quantile probabilistic forecasting of cluster wind power considering spatio-temporal correlation

Probabilistic forecasting plays an important role in the safety, stability and operation of power system. The traditional quantile regression method of non-parametric probability forecasting has the problem of crossing-quantile. Besides, current neural network methods for wind farm cluster power forecasting often overlook the spatio-temporal correlation among related wind farms. To solve these problems, a cluster power forecasting model (CFM) considering spatio-temporal correlation is proposed in this paper. A novel spatial pattern attention (SPA) combining the advantages of convolutional neural network and attention mechanism is used to extract the spatial information. An improved multi-horizon quantile recurrent neural network (IMQ-RNN) and an improved non-crossing quantile regression (INCQR) strategy are used as the output module of CFM to produce high quality forecasting results. Numerical simulations are conducted by using public real-world data from the Global Energy Forecasting Competition 2014. The results show that the proposed model has excellent performance in both deterministic forecasting and probabilistic forecasting.