基于大数据结构方程模型及分位数回归的用户需求响应分析

User’s demand response analysis based on big data structure equation model and quantile regression

研究用户用电行为是电力公司把握市场动态、提供差异化服务的基础。传统的用电行为分析大多基于用户电表信息制定电价及需求响应机制,对用户的基本信息及反馈意见等缺乏深度挖掘,该文提出一种基于多维用户信息的用户需求响应电量分析方法,针对用户数据规模大、维度高的特点,采用均值处理压缩数据规模;通过Kendall系数及结构方程模型分别对用户需求响应电量与用户的基本信息进行相关性分析,分别得到用户需求响应关联因素的直接影响路径及结构影响路径;基于此,利用分位数回归模型对新增用户的需求响应电量进行概率预测。算例分析表明:用户的需求响应电量与基本信息之间存在结构影响路径;通过对用户需求响应电量的概率预测,能较好地把握用户需求响应电量的分布,为电力公司制定需求响应机制及精细化服务提供指导。

Studying the behavior of uses is the basis to grasp market dynamics and provide customized services for power companies. Traditional analysis for residential consumption mainly focuses on formulating tariff and demand response mechanism based on smart meters information, lack of deep mining of basic information and feedback of users. This paper proposes an analytical method for the user demand response based on multi-dimensional residential user information. Considering the large scale and high dimension of user data, the mean value method is used to compress the data scale. The Kendall coefficients and Structure Equation Model(SEM) are used to analyze the correlation between the user demand response power and user’s basic information, and the direct affecting path and structural path of the residential user demand response related factors are obtained respectively. Then, the quantile regression model is used to predict the potential demand response of new users. Case study shows that there exists structural influence path between the user demand response power and basic information. By predicting the probability of for the user demand response, the user demand response power distribution can be well captured to provide guidance for power companies to develop demand response mechanism and refined services.