摘要
针对目前配电网用户负荷数据高维度时序数据特征提取难、交叉数据聚类处理难、负荷数据精准标签化难等问题,文章提出面向用户负荷数据的基于降噪自编码器和改进粗糙模糊K均值的特征提取与标签定义模型(feature extraction and label definition model based on DAE and improve RFKM,FLMbD-iR)。FLMbD-iR通过降噪自编码器对原始用户负荷数据进行深度特征提取后,利用基于类簇规模不均衡度量的粗糙模糊K均值进行聚类,处理聚类中簇间交叉数据存在误差的缺陷,最后构建描述指标对典型日负荷曲线进行标签定义。实验采用美国电力负荷数据进行仿真模拟,实验结果显示本方法在用户负荷数据聚类处理上效果显著。
Aiming at the problems of difficult feature extraction of high-dimensional time series data of user load data,difficult cross-data clustering processing,and difficult accurate labeling of load data in distribution network,this paper proposes a feature extraction and label definition model based on DAE and improve RFKM(FLMbD-iR)for user load data based on denoising autoencoder and improved rough fuzzy K-means.After the deep feature extraction of the original user load data by the denoising autoencoder,FLMbD-iR uses the rough fuzzy K-means based on the imbalanced measure of cluster sizes,and deals with the error of the cross data between clusters in the clustering.Finally,the description index is constructed to label the typical daily load curve.The experiment uses the American power load data for simulation.The experimental results show that this method has a significant effect on the clustering processing of user load data.
作者
刘礼
杨佳轩
强仁
龚钢军
陆俊
武昕
LIU Li;YANG Jiaxuan;QIANG Ren;GONG Gangjun;LU Jun;WU Xin(Beijing EngineeringResearch Center of Energy Electric Power Information Security(North China Electric Power University),Changping District,Beijing 102206,China)
出处
《电力信息与通信技术》
2024年第7期35-44,共10页
Electric Power Information and Communication Technology
基金
国家重点研发计划项目(2022YFB3105101)。
作者简介
刘礼(1998),男,博士研究生,从事深度学习、配电网智能配用电等研究工作;杨佳轩(1997),男,博士研究生,从事综合能源系统、智能配用电等研究工作;强仁(1997),男,博士研究生,从事信息物理耦合、能源路由控制等研究工作;通信作者:龚钢军(1974),男,教授,从事能源电力信息技术等研究,gong@ncepu.edu.cn。
