基于误差修正的极端学习机超短期风速预测

Ultra-short Term Wind Speed Prediction for Extreme Learning Machine Based on Error Correction

超短期风速预测对风电场功率预报系统的建立和运行至关重要。针对具有较大随机波动性的风速预测,研究了一种基于误差修正的极端学习机(ELM)超短期风速预测方法。利用ELM模型对风速进行初步预测,并利用由此得到的误差数据样本建立差分自回归滑动平均模型(ARIMA),进行误差预测,最后使用预测误差对风速的初步预测值进行补正,从而求得最终预测值。仿真实验结果表明,该方法在风速超短期预测中的可行性及有效性。

Ultra-short term wind speed prediction is important for the establishment and operation of a wind power forecasting system.Aimed at wind speed with evident fluctuation property and randomness,an ultra-short-term wind speed prediction method for extreme learning machine（ELM）based on error correction is proposed.ELM are used to predict wind speed preliminarily.With the error as original data,using auto-regressive integrated moving average to predict the error.Finally,prediction error is used to correction the wind speed to forecast the wind speed.Simulation results show feasibility and effectiveness of the method in ultra-short term prediction of wind speed.