Optimal Ensemble Forecasting Method for One-Day Ahead Hourly Wind Power Forecasting

This paper proposes an optimal ensemble method for short-term wind power forecasting. Ensemble forecasting method that incorporates several single models to improve prediction error has been widely applied in renewable energy forecasting. In this paper, a k-means method is used to assort wind power and wind speed data into five different types. Five different machine learning models are created and then used to produce initial prediction. The swarm intelligence methods, including particle swarm optimization (PSO), salp swarm algorithm (SSA) and whale optimization algorithm (WOA), are used to optimize the weight allocation for each single model. The final prediction is then generated using the weighted sum of each single prediction model. A wind power generation system that is located in Changhua, Taiwan is used to validate the proposed method. Testing results show that the proposed method provides more stable and accurate prediction than each single model. The proposed method also allows more accurate predictions compared to Least Absolute Shrinkage and Selection Operator (LASSO) and ridge regression methods.