基于ARIMA模型误差修正的小波神经网络风速短期预测

Short-term forecasting of wind speed based on wavelet neural network and ARIMA model of error correct

风电场风速短期预测对于风力发电具有重要意义。本文首先根据国家标准《风电场风能资源评估方法》推荐的方法对风速数据进行预处理,修正不合理的数据以及插补丢失的数据:接着运用小波神经网络对预处理之后的数据进行预测,并对预测之后的残差形成的随机序列建立ARIMA误差预测模型,最后用预测的误差来修正风速预测结果。将以上方法用于某小型风电场实测数据,并将运算结果与小波神经网络的预测进行比较,MAPE降低了46.97%,结果表明基于ARIMA模型误差修正的小波神经网络明显改善了风速预测的精度,可有效应用于短期风速预测。

Short term wind speed forecasting is very important to wind power generator. Firstly, the wind speed data collected by wind farms is preprocessed according to the national standard methodology of wind energy resource assessment for wind farm recommended method so as to revise the unreasonable data and interpolation of missing data. Then the data is used to forecast the wind speed by using wavelet neural network, establishing the error prediction model of ARIMA by using the errors stochastic series. Lastly, using the forecasted errors to update the forecasted results.The above method for a small wind farm measured data and comparing the results with the forecasting results by using wavelet neural network, MAPE reduced by 46.97 %, the results show that the wavelet neural network based on the ARIMA model error correction can obviously improve the prediction precision of the wind speed forecasting, and can be used to short-term wind forecast effectively.