Electricity Theft Detection in Power Grids with Deep Learning and Random Forests

Abstract

As one of the major factors of the nontechnical losses (NTLs) in distribution networks, the electricity theft causes significant harm to power grids, which influences power supply quality and reduces operating profits. In order to help utility companies solve the problems of inefficient electricity inspection and irregular power consumption, a novel hybrid convolutional neural network-random forest (CNN-RF) model for automatic electricity theft detection is presented in this paper. In this model, a convolutional neural network (CNN) firstly is designed to learn the features between different hours of the day and different days from massive and varying smart meter data by the operations of convolution and downsampling. In addition, a dropout layer is added to retard the risk of overfitting, and the backpropagation algorithm is applied to update network parameters in the training phase. And then, the random forest (RF) is trained based on the obtained features to detect whether the consumer steals electricity. To build the RF in the hybrid model, the grid search algorithm is adopted to determine optimal parameters. Finally, experiments are conducted based on real energy consumption data, and the results show that the proposed detection model outperforms other methods in terms of accuracy and efficiency.

A framework to select a classification algorithm in electricity fraud detection

“…Nagi et al 19 used a SVM classifier to detect fraud for a power system in Malaysia whilst in India, Depuru et al 20 used smart meter data in their SVM classification study. Coma-Puig et al 21 used k-NN and SVM classifiers, amongst others, to evaluate electricity data from Spain, whilst Li et al 3 used several classifiers, including a hybrid random forest classifier, to evaluate data from Ireland. In each study, performance metrics were used to assess which classifier could be considered the best.…”

“…As a result, researchers have chosen to include several performance measures when reporting the assessment of a classifier. Coma-Puig et al 21 used the performance measures recall and f-measure whilst Ghori et al 7 and Li et al 3 opted to include precision as one of their measures. There is little consensus on which metric is best; however, there is consensus that more than one metric should be reported.…”

“…Revenue that is lost due to the difference between electricity supplied and electricity purchased is partitioned into two classes. The first class resulting from transmission and other infrastructural limitations is labelled as technical losses, whilst the second class, the majority of which are a result of meter tampering or bypassing, is labelled as non-technical losses [1][2][3] (NTL). Estimates of losses worldwide are in the billions 4,5 of US dollars and suppliers of electricity have expressed concern over these losses and the sustainability of the supply 6,7 .…”

“…Notably these include support vector machines (SVM), naïve Bayesian (NB) methods and k-nearest neighbour (k-NN) classifiers -the algorithms used in this study. The field has not stagnated; recent computational methods include convolutional neural networks 3 and ensemble-based classifiers 7 whilst time series methods have been explored 2 . In many of these studies, the common methodological approach is to assess the classifier by considering results from a test data set captured in a confusion matrix summary and reported as a performance measure.…”

“…In addition, the false negatives and false positives should be zero. The confusion matrix, in Table 1, was used in studies by Messinis et al 12 and Li et al 3 to assess classification models, whilst Guo et al 13 addressed the problem for data sets which are highly imbalanced.…”