User group feature extraction method, system, equipment and medium

• Main Authors: LI JIACHEN, QU XUN, WANG XIAOFENG, CAO ZENGXIN, WANG TIANYI, YAN CHEN, YAN SIQI, HE YANBIN, YU QIAN, SU BIAO, JIANG SHAN, CHI JIANFENG, WANG ZIHE, CHANG YUAN, TIAN ZIJIAN, LU LU, GUAN TAO, QU HENG, YANG ERLE, GAO MENGYA
• Format: Patent
• Language: Chinese; English
• Published: 10.10.2023
• Subjects: CALCULATING; CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTINGELECTRIC POWER; COMPUTING; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; ELECTRIC DIGITAL DATA PROCESSING; ELECTRICITY; GENERATION; PHYSICS; SYSTEMS FOR STORING ELECTRIC ENERGY; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention belongs to the field of electrical load prediction, and particularly relates to a user group feature extraction method, system and device and a medium. According to the method, users with similar power consumption behavior characteristics are represented as a graph structure, the size of the correlation between the users is measured by using a Pearson's coefficient, time and space characteristics in the correlation are captured by adopting a mode of combining causal convolution and graph convolution, and the method has a relatively good capturing capability for the time and space relation in a user group. The load prediction precision of a single user and a user group is improved, and the improvement effect of the proposed load prediction method is enhanced along with the improvement of the prediction time step length. 本发明属于用电负荷预测领域，具体涉及一种用户群体特征提取方法、系统、设备和介质。将具备相似用电行为特征的用户表征为图结构，利用皮尔逊系数衡量用户之间相关关系的大小，并采用因果卷积与图卷积结合的方式来捕捉其中的时间与空间特征，对用户群体中的时空关系具备较好的捕捉能力，单个用户与用户群体的负荷预测精度均得到了提升，同时随着预测时间步长的提升，所提出的负荷预测方