基于集成机器学习的电力系统窃电行为辨别方法

Identification Method for Power System Stealing Behavior Based on Integrated Machine Learning

电力系统中窃电行为对线损计算和电网经济运行造成不利影响,严重影响电力市场的经济秩序。基于此,本文提出了一种基于集成机器学习的电力系统窃电辨别方法,以快速准确的辨别窃电行为。首先,该方法建立了基于改进神经网络的电力系统窃电行为特征提取模型,利用实测数据与神经网络全变量重构值进行比较,以获取用电数据的抽象行为特征;然后,利用提取出的窃电行为特征,采用集成机器学习方法进一步推导出电力系统窃电行为辨别方法;最后,基于实际数据集,验证了所提算法具有较高的检测灵敏度和分类精度。

The behavior of stealing electricity in the power system has a negative effect on the calculation of line loss and the economic operation of the power grid,and seriously affects the economic order of the power market.Based on this,this paper proposes an integrated machine learningbased identification method for electric power system theft,in order to identify electric power theft behavior quickly and accurately.Firstly,an improved neural network-based feature extraction model of power system stealing behavior is established.The measured data is compared with the reconstructed value of neural network to obtain the abstract behavior characteristics of power consumption data.Then,based on the extracted characteristics of electricity stealing behavior,the integrated machine learning method is used to further deduce the power system electricity stealing behavior identification method.Finally,based on the actual data set,the proposed algorithm is verified to have high detection sensitivity and classification accuracy.