基于序列分解的母线负荷降噪预测方法  

A noise reduction forecasting method of bus load based on sequence decomposition

作　　者：杨坚 赵洁[2] 汤义勤 蒋旭 唐佳杰 张怀勋 YANG Jian;ZHAO Jie;TANG Yiqin;JIANG Xu;TANG Jiajie;ZHANG Huaixun(State Grid Taizhou Power Supply Company,Taizhou,Zhejiang 318000,China;School of Electrical Engineering and Automation,Wuhan University,Wuhan 430072,China;State Grid Nantong Power Supply Company,Nantong,Jiangsu 226000,China)

机构地区：[1]国网浙江省电力有限公司台州供电公司,浙江台州318000 [2]武汉大学电气与自动化学院,武汉430072 [3]国网江苏省电力有限公司南通供电分公司,江苏南通226000

出　　处：《浙江电力》2023年第12期81-87,共7页Zhejiang Electric Power

基　　金：国网浙江省电力有限公司科技项目(5211TZ1900S4)。

摘　　要：新型电力系统背景下,分布式电源和用户侧行为的多样性使母线负荷稳定性不足,对负荷短期预测提出了新的挑战。为此,提出一种基于序列分解的母线负荷降噪预测方法,将变分模态分解方法的构造与分解规则应用到母线负荷序列分解中,针对序列分解后的余项,利用局部加权回归方法进行平滑处理,实现母线负荷降噪预测。基于某地区母线负荷有功功率实测数据,构建循环神经网络对降噪后的母线负荷进行预测,结果表明该方法能够去除母线负荷序列噪声,序列趋于光滑且保留了原始母线负荷序列的特征,具有优良的预测曲线和精确的预测结果。In the context of new power systems,the diverse distributed power sources and user-side behavior have introduced instability in bus loads,thus presenting a fresh challenge for short-term load forecasting.In response to this challenge,a noise reduction forecasting method for bus load based on sequence decomposition is proposed.The construction and decomposition rules of the variational mode decomposition(VMD)are applied to the sequence de⁃composition of bus load.The residual terms following the sequence decomposition are smoothed using locally weighted regression(LWR)to forecast the noise reduction of bus load.Based on the measured active power data of bus load in an area,a recurrent neural network(RNN)is constructed to forecast the bus load after noise reduction.The results indicate that this method effectively eliminates the noise of the bus load sequence,and the sequence tends to be smooth while retaining the characteristics of the original bus load sequence.Moreover,it yields excellent forecasting curves and accurate forecasting results.

关 键 词：母线负荷预测 母线负荷降噪 变分模态分解 循环神经网络 局部加权回归 

分 类 号：TM715[电气工程—电力系统及自动化]