基于非线性独立成分估计的分布式光伏窃电数据增强方法

Data Augmentation Method for Distributed Photovoltaic Electricity Theft Based on Non-linear Independent Components Estimation

由于分布式光伏窃电的强隐蔽性和稽查资源的有限性,导致电力部门掌握的窃电样本数量不足,限制了窃电检测的精度。为此,提出了一种基于非线性独立成分估计(NICE)的分布式光伏窃电数据增强方法。首先,利用多个可逆函数将窃电样本映射成服从高斯分布的隐变量,并通过逆变换将其反向重构成新的窃电样本。然后,提出了3种典型的光伏窃电模型,并针对窃电样本的数据特征构建了卷积神经网络作为分类器。最后,通过仿真算例和实际算例验证了所提方法的有效性和适应性。仿真结果表明,NICE能够同时兼顾样本的形状和分布特征,生成的窃电样本能够显著提升不同分类器的性能。

Due to the strong concealment of distributed photovoltaic electricity theft and the limitation of inspection resources,the number of electricity theft samples mastered by the power utility is insufficient,which limits the accuracy of electricity theft detection.Therefore,a data augmentation method for distributed photovoltaic electricity theft based on the non-linear independent components estimation(NICE)is proposed.First,the multiple inversible functions are used to map the electricity theft samples into latent variables that obey the Gaussian distribution,and they are reconstructed into new electricity theft samples by inverse transformations.Then,the three typical photovoltaic electricity theft models are proposed,and the convolutional neural network is constructed as a classifier according to the data characteristics of electricity theft samples.Finally,the effectiveness and adaptability of the proposed method are verified by a practical example and a simulation example.Simulation results show that NICE is able to take into account both the shape and distribution characteristics of the samples,and the generated electricity theft samples can significantly improve the performance of different classifiers.