数据分辨率影响风电功率预测精度分析

Analysis of the Influence of Data Sampling on the Short-Term Wind Power Prediction Accuracy

目前对风电功率短时预测的研究主要集中在预测方法上,而缺乏对数据本身特性的探讨。从实测数据出发,呈现3种典型分辨率5 min、10 min、15 min,并结合Elman神经网络算法对超短期(4 h)和短期(24 h)的风力发电机输出功率进行预测分析。结果表明:分辨率为10 min的原始数据对风电输出功率的超短期预测具有更好的结果,15 min分辨率的数据对风电功率的短期预测结果更佳。采用合理分辨率的数据后,能够有效地提高风电功率的预测精度。

At present,the study of short-term wind power prediction focuses mainly on the prediction method,but little attention is paid to the characteristics of the data itself on the wind power prediction accuracy. Based on the measured data this paper presents three typical resolutions of 5 min,10 min and 15 min,and combined with the Elman neural network algorithm conducts the forecast analysis for a single 4 h and24 h actual wind power output. The results shows that for the ultra-short-term prediction the temporal resolution of 10 min has better results while for the short-term prediction the 15 min resolution has better results. And with adoption of the reaso-nable resolution data,the prediction accuracy of wind power can be improved effectively.