Interval forecasting for wind speed using a combination model based on multiobjective artificial hummingbird algorithm

Short-term wind speed prediction is critical for enhancing the efficiency of wind power systems and assuring the stability and continuity of power generation and the host electricity markets. Various methods are available to improve the performance of wind speed prediction. However, these methods use traditional point forecasting and neglect the limitations of individual models, which cannot handle uncertainty in system operation. We propose a combined interval forecasting method that combines multiobjective artificial hummingbird algorithm, interval forecasting, and individual forecasting methods. As our proposal integrates various forecasting models including autoregressive integrated moving average, bidirectional long short-term memory, long short-term memory, and gated recurrent unit, it overcomes the limitations of single models and enhances the prediction accuracy. Experimental results show that the forecasting performance of the proposed combined interval forecasting model is considerably higher than that of similar models.