A Comprehensive Analysis of Supervised Learning Techniques for Electricity Theft Detection

Abstract

There are many methods or algorithms applicable for detecting electricity theft. However, comparative studies on supervised learning methods for electricity theft detection are still insufficient. In this paper, comparisons based on predictive accuracy, recall, precision, AUC, and F1-score of several supervised learning methods such as decision tree (DT), artificial neural network (ANN), deep artificial neural network (DANN), and AdaBoost are presented and their performances are analyzed. A public dataset from the State Grid Corporation of China (SGCC) was used for this study. The dataset consisted of power consumption in kWh unit. Based on the analysis results, the DANN outperforms compared to other supervised learning classifiers such as ANN, AdaBoost, and DT in recall, F1-Score, and AUC. A future research direction is the experiments can be performed on other supervised learning algorithms with different types of datasets and suitable preprocessing methods can be applied to produce better performance.

Electricity theft detection for energy optimization using deep learning models

“…Accuracy, precision, area under the receiver operating characteristic curve (AUC) score, F 1 score, and recall are the performance metrics utilized in this paper to assess the performance of the proposed model 19 . These measures are the ones that are most frequently used in the ETD literature 46–49 . To calculate the aforementioned metrics, we need to calculate the confusion matrix.…”

“…19 These measures are the ones that are most frequently used in the ETD literature. [46][47][48][49] To calculate the aforementioned metrics, we need to calculate the confusion matrix. Confusion matrix consists of four elements: false positives (FP), true positives (TP), false negatives (FN), and true negatives (TN).…”

A novel deep learning technique to detect electricity theft in smart grids using AlexNet

“…The outcomes reveal it enhanced handling of attack circumstances. The approach has also been recommended for future use in utility corporations [9]. Another study conducted in [8] offered a comparison to determine an appropriate for electricity theft using deep neural network technology, which supplied Recall, F1 score, and AUC.…”

“…One limitation of this proposed technique is state-based methods of performing ETD, which require more hardware and are more expensive. A game theory-based strategy for smart houses to reduce peak energy costs is presented in [9,10]. The suggested technique also establishes coordination between smart appliances and minimizes energy losses.…”

Ứng Dụng Mạng Nơ-Ron Tích Chập Để Phát Hiện Hành Vi Ăn Trộm Điện Dựa Trên Dữ Liệu Từ Công Tơ Thông Minh

“…Có thể thấy độ chính xác AUC đạt được là 92,2%, MAP@100 là 99,2%, MAP@200 là 98,2%. Chúng tôi cũng so sánh với những mô hình khác được huấn luyện và kiểm thử trên cùng bộ dữ liệu của Tổng công ty lưới điện Trung Quốc tại Bảng 3, kết quả cho thấy các mô hình học máy như AdaBoost được Bohani và các cộng sự [11] huấn luyện cho kết quả tương đối thấp chỉ đạt AUC 53,46%, F1-Score 13,34%. Cũng tại nghiên cứu đó kết quả của mạng nơ-ron sâu cho thấy đây là một hướng khả quan khi AUC được cải thiện lên 69%, F1-Score 45,85%.…”