我国工业电力用户价值画像模型构建与应用研究

在新型冠状病毒感染肺炎疫情对我国电力市场造成巨大冲击的宏观背景下，为进一步提升我国供电企业营销服务资源配置效能，最大化撬动供电企业的综合效益，笔者开展了我国工业电力用户价值画像模型研究。本文对电力用户价值进行了分析和定义，从安全稳定价值(S)、经济效益价值(E)、契约信用价值(C)与有序用电价值(O)四个维度，构建了我国工业电力用户价值评级SECO指标模型，并集成智能算法中的RST(粗糙集理论)与数据挖掘技术中的PAM(围绕中心点切割聚类算法)，构造了一种半监督自动化用户价值识别、预测与特征展示模型，模型包括基于RST的指标体系设计、基于Gower相异度系数与PAM的用户价值评级，以及基于用户画像的价值特征展示三大模块。其中，为增强聚类分析结果的科学性与可靠性，采用霍普金斯统计量进行聚类趋势判断，利用间隔统计量输出理论最佳聚类数目，运用轮廓系数评估模型效果与识别误判样本。以我国南方电网公司下属某供电企业电力用户数据进行模型测试与应用研究，得到具有较高解释性与区分度的用户细分方案，表明本模型是一套可行有效的用户价值评级与特征可视化工具。

Research on the Construction and Application of the Customer Value Portrait Model of Industrial Power Enterprise in China

Under the macro background that the novel Coronavirus epidemic had a huge impact on China’s electric power market, in order to further improve the efficiency of marketing service resource allocation of China’s power supply enterprises and maximize the comprehensive benefits of power supply enterprises, a study is carried out on the user value portrait model of China’s industrial power. The power user value is analyzed and defined, and the SECO indicator model of China’s industrial power user value evaluation is constructed from four dimensions: safety ＆ stability value (S), economic benefit value (E), contract credit value (C) and ordered power value (O), also at the same time, RST (Rough Set Theory) in intelligent algorithm and PAM (Partitioning Around Medoids) in data mining technology are combined to construct a semi-supervised automatic user value identification and feature display model. The model includes three modules: RST based index model design, user value evaluation based on Gower distance and PAM, and value feature visualization via user portraits. Among them, in order to enhance the scientificity and reliability of the cluster results, Hopkins statistics is used to access the clustering tendency, gap statistics is used to output the optimal number of cluster approach, and silhouette coefficient is used to evaluate the model effect and identify the misjudged samples. The model is tested and applied to the user data of a power supply enterprise affiliated to China Southern Power Grid, and a highly explanatory and persuasive user segmentation scheme is obtained, which indicates that the model is a set of feasible and effective user value rating and feature visualization tool.