基于时空关联矩阵的配电台区反窃电预警方法

针对配电台区窃电用户难发现、窃电量预估不准确的问题，提出了一种基于时空关联矩阵的配电台区反窃电预警方法。首先，构建配电台区数据清洗方法，采用线性插值算法对低压户表采集的缺失数据进行补正，以消除配电台区量测数据缺失对模型的影响。其次，构建配电台区窃电分析算法，通过台区线损波动率、线损与电流差异曲线的变点时间进行关联分析，从而判断台区是否存在窃电行为。再次，构建窃电用户的时空关联分析模型，通过变点、离群点和关联检测分析窃电嫌疑用户的时空分布特征，并计及用户窃电时间和用电容量等特性，提供预估窃电量。最后，通过实例验证了所提方法的有效性和实用性。

Early Warning Method of Electricity Anti-theft in Distribution Station Area Based on Spatiotemporal Correlation Matrix

Aiming at the problems that it is difficult to find electricity theft users and accurately estimate the quantity of electricity theft, an early warning method of electricity anti-theft in the distribution station area based on the spatiotemporal correlation matrix is proposed. Firstly, the data cleaning method in the distribution station area is constructed, and the linear interpolation algorithm is used to fill in the missing data of the low-voltage household meter, which eliminates the influence of missing measurement data in the distribution station area on the model. Secondly, the analysis algorithm of electricity theft in the distribution station area is built, which can diagnose whether there is electricity theft behavior in the station area through the correlative analysis of line loss volatility, line loss and the time of change-points on the current difference curve. Thirdly, the analysis model of spatiotemporal correlation is proposed. It analyzes the temporal and spatial distribution characteristics of users suspected of electricity theft through change-points, outlier points and correlation detection, and provides estimated quantity of electricity theft with characteristics of electricity theft time and power consumption capacity. Finally, the effectiveness and practicability of the proposed method are verified by an example.