Adaptive Ensemble Ultra-Short Term Forecasting of Wind Power Considering Curtailment Data Identification

Accurate wind power forecasting plays an important role in secure and economic operation of power system with renewables. This paper proposes an adaptive ensemble ultra-short term wind power forecasting approach considering wind curtailment data identification. Firstly, the data processing is carried out to identify the wind curtailment data by the horizontal and vertical quartile method. Secondly, the basic wind power forecasting models based on ELM, XGBoost and LSTM are established and trained respectively. Then, the weights of basic wind power forecasting models are adaptively assigned based on the errors on the verification set. Multiple basic forecasting models are adaptively integrated to realize wind power forecasting. Finally, the statistical analysis of the probability distribution of forecast error is carried out under multivariate conditions. The effectiveness of the proposed method is verified by case analysis.