基于数据增强的配电网运行态势预警方法  

A Method for Distribution Network Operation Situation Warning Based on Data Augmentation

作　　者：李晓露 陈思羽 柳劲松 林顺富 LI Xiaolu;CHEN Siyu;LIU Jinsong;LIN Shunfu(College of Electrical Engineering,Shanghai University of Electric Power,Yangpu District,Shanghai 200090,China;State Grid Shanghai Electric Power Research Institute,Hongkou District,Shanghai 200437,China)

机构地区：[1]上海电力大学电气工程学院,上海市杨浦区200090 [2]国网上海市电力公司电力科学研究院,上海市虹口区200437

出　　处：《电网技术》2025年第3期1207-1216,I0090-I0092,共13页Power System Technology

基　　金：国家自然科学基金项目(51977127);国家电网公司科技项目(52094022001G)。

摘　　要：新型电力系统背景下,配电网运行态势的及时预警是保障其安全稳定运行的重要前提。针对配电网运行态势预警的量测数据不充分问题,该文提出一种基于数据增强的配电网运行态势预警方法。首先,基于支持向量数据描述识别初始预警边界,结合模糊隶属度函数,增强数据样本密度对边界识别的影响,准确识别初始预警边界;其次,建立基于改进深度卷积生成对抗网络的配电网运行状态电气量数据增强模型,其损失函数中引入边界形态偏差量、样本概率分布偏差量以及态势预测回归关系曲线偏差量,以修正配电网的运行态势预警边界;然后,提出态势预警方法的判别依据,得到运行点向边界的行动趋势,在线实现配电网运行态势预警。最后,基于改进IEEE 33节点和IEEE 123节点算例验证了所提方法能够有效提高配电网运行态势预警的速度和精度。In the context of the new power system,timely warnings of the operation situation of the distribution network are an important prerequisite for ensuring the safety and stability of distribution network operation.This paper proposes a distribution network operation situation warning method based on data augmentation in response to the problem of insufficient measurement data for distribution network operation situation warning.Firstly,based on the support vector data description,the initial warning boundary is identified,combined with the fuzzy membership function to enhance the influence of data sample density on boundary recognition and accurately determine the initial warning boundary.Secondly,the data augmentation model of distribution network operation status electrical quantity based on improved deep convolutional generative adversarial networks is established.The loss function introduces deviation of boundary shape,deviation of sample probability distribution,and deviation of situation prediction regression relationship curve to correct the operation status warning boundary of the distribution network.Then,the discrimination basis for the situation warning method is proposed,and the trend of action towards the boundary at the operating point is obtained to achieve distribution network operation situation warning online.Finally,the improved IEEE 33-bus system validation and IEEE 123-bus system validation show that this method can effectively improve the speed and accuracy of distribution network operation situation warnings.

关 键 词：配电网 态势预警 边界识别 数据增强 生成对抗网络 

分 类 号：TM721[电气工程—电力系统及自动化]