基于ARMA-ARCH模型和BP模型的短期风速组合预测方法研究

Short-Term Wind Speed Combined Prediction Based on ARMA-ARCH Model and BP Model

随着风力发电的大规模并网,由风速的波动引起的网侧不稳定现象越来越显著。为了提高风电场风速预测的精度,首先建立了ARMA模型,利用拉格朗日乘数法检验ARMA模型残差的条件异方差效应,从而建立ARMA-ARCH模型;其次建立BP神经网络预测模型;最后分别以固定权和时变权方差-协方差(MV)法将ARMA-ARCH模型和BP模型进行优选组合预测。为验证模型的适应性,分别以西班牙某风电场2016年8月和2017年1月的风速数据进行建模仿真。仿真结果表明:组合预测模型的预测结果更优,且时变权组合预测精度更高;对于单一模型来说ARMA-ARCH模型的预测精度要高于BP模型,而ARMA模型的预测精度最低。

With the large-scale integration of wind power generation,instability of the grid side caused by wind speed is becoming more and more significant.Accurate prediction of wind speed is very important to reduce the phenomenon of abandoned wind.In order to improve the accuracy of wind speed prediction,this paper,first of all,establishes an ARMA model,where by using the Lagrange multiplier method,the ARMA model residual effect variance is tested so as to establish the ARMA-ARCH model.Second,the paper establishes a BP neural network prediction model.The optimum combination prediction is performed on the ARMA-ARCH model and the BP model by the fixed weight method and time-varying weight variance covariance(MV)method respectively.To verify the adaptability of the model,modeling is performed according to the data of a certain Spain wind farm respectively in August 2016 and January 2017.The simulation results show that the combined prediction model has better results,and variable weight combination has even higher prediction precision.In addition,the prediction accuracy of the ARMA-ARCH model is higher than that of the BP model for a single model,while the ARMA model has the lowest prediction accuracy.