基于MTF维度提升和残差网络的窃电识别方法

Power theft identification method based on MTF dimension promotion and residual network

针对智能电网中的窃电检测问题,通过分析用户日用电量,发现正常用户用电具有季节波动和假期关联特性,窃电用户呈现杂乱无序状态。为此,提出了一种基于马尔可夫转移场的一维到二维图像转换方法,从多个时间尺度挖掘用电特征,再利用引入残差模块的卷积神经网络进行窃电用户识别。在国家电网提供的数据集上进行实验,所提方法具有94.31%的准确率,验证了其有效性和可行性。

To address the problem of electricity theft detection in smart grids,this paper proposes a method based on Markov transfer field for one-dimensional to two-dimensional image conversion.By analyzing daily electricity usage of clients,it is found that normal clients have seasonal fluctuation and holiday correlation characteristics,while electricity theft clients show disordered states.The proposed method mines electricity usage features from multiple time scales,and then uses convolutional neural networks with residual modules for electricity theft client identification.Experiments are conducted on a dataset provided by the State Grid Corporation of China,and the proposed method achieves an accuracy of 94.31%,which verifies its effectiveness and feasibility.