Electricity Theft Detection Based-On Deep Learning

Electricity theft detection issue has drawn lots of attention during last decades since it caused lots of economic losses in our country and worldwide.Timely identification of the electricity theft in the power system is crucial for the safety and availability of the system.Although sustainable efforts have been made,the detection task remains challenging and falls short of accuracy and efficiency,especially with the increase of the data size.Recently,deep learning-based methods have achieved better performance in comparison with traditional methods,which employ handcrafted features and shallow-architecture classifiers.In this paper,we proposed three kinds of different deep learning structures which called 1D-DesenNet,Multi-scale DenseNet and BPIE-BiSRNN networks based on convolutional neural networks(CNNs)and recurrent neural networks(RNNs),in order to capture the long-term and short-term periodic features within the sequential data.We compare the proposed approaches with the classical algorithms,and the experimental results demonstrate that the our approaches can significantly improve th e accuracy of the detection.Therefore,the cost of manually detecting can be greatly reduced,and compared with classical machine learning methods,the deep learning methods we proposed improved the performance a lot,the logloss reduced from 0.343 to 0.256,and the AUC value increased.from 0.756 to 0.871.