AUTHOR=Long Huan , Geng Runhao , Zhang Chen TITLE=Wind Speed Interval Prediction Based on the Hybrid Ensemble Model With Biased Convex Cost Function JOURNAL=Frontiers in Energy Research VOLUME=10 YEAR=2022 URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.954274 DOI=10.3389/fenrg.2022.954274 ISSN=2296-598X ABSTRACT=

This study proposes a combination interval prediction based hybrid ensemble (CIPE) model for short-term wind speed prediction. The combination interval prediction (CIP) model employs the extreme learning machine (ELM) as the predictor with a biased convex cost function. To relieve the heavy burden of the hyper-parameter selection of the biased convex cost function, a hybrid ensemble technique is developed by combining the bagging and stacking ensemble methods. Multiple CIP models with random hyper-parameters are first trained based on the sub-datasets generated by the bootstrap resampling. The linear regression (LR) is utilized as the meta model to aggregate the CIP models. By introducing the binary variables, the LR meta model can be formulated as a mixed integer programming (MIP) problem. With the benefit of the biased convex cost function and ensemble technique, the high computational efficiency and stable performance of the proposed prediction model is guaranteed simultaneously. Multi-step ahead 10-min wind speed interval prediction is conducted based on actual wind farm data. Comprehensive experiments are carried out to verify the superiority of the proposed interval prediction model.