基于自适应移动平滑与时间卷积网络误差修正的风电功率预测  

Wind Power Forecasting Based on Error Correction Using Adaptive Moving Smoothing and Time Convolution Network

作　　者：孙蓉 李强 罗海峰 窦迅[2] 邓叶航 SUN Rong;LI Qiang;LUO Haifeng;DOU Xun;DENG Yehang(State Grid Jiangsu Electric Power Co.,Ltd.Research Institute,Nanjing 211103,Jiangsu Province,China;Nanjing Tech University,Nanjing 211816,Jiangsu Province,China)

机构地区：[1]国网江苏省电力有限公司电力科学研究院,江苏省南京市211103 [2]南京工业大学,江苏省南京市211816

出　　处：《全球能源互联网》2022年第1期11-22,共12页Journal of Global Energy Interconnection

基　　金：国网江苏省电力有限公司科技项目(基于实时资源和运行数据的高精度新能源功率在线预测,J2020128)。

摘　　要：为了解决风电功率预测误差对电力系统调度运行影响的问题,提出一种基于自适应移动平滑(adaptive movement smoothing,AMS)和时间卷积网络(temporal convolutional network,TCN)误差修正的风电功率预测方法。该方法首先利用变分模态分解和TCN提取风电功率的时空特性,得到初步预测结果;然后利用AMS模型对预测误差序列进行自适应平滑处理,降低误差的波动性;最后利用TCN模型提取预测误差的时间特性,对初步预测结果进行修正,提高预测的精度和稳定性。基于辽宁双子台和内蒙古克什克腾旗两个风电场的实测数据进行了实验对比分析,相较于其他方法,采用所提风电功率预测方法在15 min、30 min和1 h时间尺度下得到的预测结果,平均绝对误差降低50.0%以上,平均相对误差降低10.0%以上,验证了所提方法的有效性。To mitigate the impact of wind power prediction errors on power system dispatching operations,we proposed a wind power prediction method based on error correction using adaptive moving smoothing(AMS),and time convolution network(TCN).First,the spatio-temporal characteristics of wind power were extracted using variational modal decomposition,and TCN,to obtain preliminary prediction results.Subsequently,adaptive smoothing was applied to the prediction error series using the AMS model,to reduce the volatility of the error.Lastly,the temporal characteristics of the prediction error were extracted using the TCN model,to correct the preliminary prediction results,and improve the accuracy and stability of the prediction.The experimental comparison analysis was conducted based on the measured data from two wind farms in Shuangzitai,Liaoning,and Hexigten Banner,Inner Mongolia.Compared with other methods,the prediction results obtained at 15 min,30 min and 1 h time scales,using the proposed method,improved the mean absolute error by more than 50.0%,and the mean relative error by more than 10.0%.Therefore,the superiority of the method proposed in this paper was verified.

关 键 词：误差修正 风电功率预测 时间卷积网络 自适应移动平滑 

分 类 号：TM614[电气工程—电力系统及自动化]