Abstract:

Abstract Electricity theft is a major concern for electric power distribution companies. The data set from the State Grid Corporation of China (SGCC) is preprocessing; first, order dataset by date, second, remove the empty record from the dataset, third, missing values by linear interpolation, and finally, imbalanced data handling technique. Then find feature extraction including monthly average, slope, moment and standard deviation, Variance, Peak to Peak, Energy Entropy, Skewness, Crest Factor, Total harmonic distortion, Log Energy, and Kurtosis for all months with gather. After finding features of the dataset, Deep Convolution Neural Network (DCNN) applied DCNN with A map is classified using a convolution layer to extract features, followed by a softmax layer. DCNN is used for data classification in energy theft or non-theft. Finally, the calculated accuracy achieved 100%, which is quite promising in comparison to other reported categorization schemes. The number and date of the theft were then calculated for each of the records in which the theft occurred.

Keywords: