Using enhanced crow search algorithm optimization-extreme learning machine model to forecast short-term wind power

TL;DR: The simulation experiments revealed that the error interval and evaluation indexes of the ENCSA-ELM model outperformed the state-of-the-art wind power forecast methods, traditional machine learning models and ELM optimized by other algorithms.

Abstract: The strong volatility and randomness of wind power impact the grid and reduce the voltage quality of the grid when wind power is connected to the grid in large scale. The power sector takes the wind abandonment measures to ensure the grid voltage stability. The enhanced crow search algorithm optimization-extreme learning machine (ENCSA-ELM) model is proposed to accurately forecast short-term wind power to improve the utilization efficiency of clean energy. (1) The enhanced crow search algorithm (ENCSA) is proposed and applied to short-term wind power forecast. The convergence performance test revealed that the local development and global exploration capabilities of the ENCSA were enhanced, and the test result of the proposed ENCSA algorithm outperformed other well-known nature inspired algorithms and state-of-the-art CSA variants; (2) The output and input of the forecasting models were determined by analysis of the wind power impact factors and the wind power samples in autumn, winter and spring were forecasted by the ENCSA-ELM model; and (3) The forecast results were analyzed by multiple evaluation indexes. The simulation experiments revealed that the error interval and evaluation indexes of the ENCSA-ELM model outperformed the state-of-the-art wind power forecast methods, traditional machine learning models and ELM optimized by other algorithms. The RMSE value and MAPE value of the proposed model were controlled below 20% and 4%. Accurate wind power prediction maintains the voltage stability of power grid and increases the utilization efficiency of clean energy.