A Novel Electricity Theft Detection Strategy Based on Dual-Time Feature Fusion and Deep Learning Methods

Qinyu Huang, Zhenli Tang (), Xiaofeng Weng, Min He, Fang Liu, Mingfa Yang and Tao Jin ()
Additional contact information
Qinyu Huang: Department of Electrical Engineering, Fuzhou University, Fuzhou 350116, China
Zhenli Tang: Fujian YILI Information Technology Co., Ltd., Fuzhou 350001, China
Xiaofeng Weng: Fujian YILI Information Technology Co., Ltd., Fuzhou 350001, China
Min He: Fujian YILI Information Technology Co., Ltd., Fuzhou 350001, China
Fang Liu: Fujian YILI Information Technology Co., Ltd., Fuzhou 350001, China
Mingfa Yang: Department of Electrical Engineering, Fuzhou University, Fuzhou 350116, China
Tao Jin: Department of Electrical Engineering, Fuzhou University, Fuzhou 350116, China

Energies, 2024, vol. 17, issue 2, 1-18

Abstract: To enhance the accuracy of theft detection for electricity consumers, this paper introduces a novel strategy based on the fusion of the dual-time feature and deep learning methods. Initially, considering electricity-consumption features at dual temporal scales, the paper employs temporal convolutional networks (TCN) with a long short-term memory (LSTM) multi-level feature extraction module (LSTM-TCN) and deep convolutional neural network (DCNN) to parallelly extract features at these scales. Subsequently, the extracted features are coupled and input into a fully connected (FC) layer for classification, enabling the precise detection of theft users. To validate the method’s effectiveness, real electricity-consumption data from the State Grid Corporation of China (SGCC) is used for testing. The experimental results demonstrate that the proposed method achieves a remarkable detection accuracy of up to 94.7% during testing, showcasing excellent performance across various evaluation metrics. Specifically, it attained values of 0.932, 0.964, 0.948, and 0.986 for precision, recall, F1 score, and AUC, respectively. Additionally, the paper conducts a comparative analysis with mainstream theft identification approaches. In the comparison of training processes, the proposed method exhibits significant advantages in terms of identification accuracy and fitting degree. Moreover, with adjustments to the training set proportions, the proposed method shows minimal impact, indicating robustness.

Keywords: deep learning; electricity theft detection; feature fusion; parallel model (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1996-1073/17/2/275/pdf (application/pdf)
https://www.mdpi.com/1996-1073/17/2/275/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jeners:v:17:y:2024:i:2:p:275-:d:1313481

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

More articles in Energies from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().