A Convolution–Non-Convolution Parallel Deep Network for Electricity Theft Detection
Abstract
:1. Introduction
- The features of the electricity consumption time series of normal users and electricity thieves were analyzed, and it was found that compared with those of electricity thieves, the electricity consumption time series of normal users have more obvious periodic characteristics;
- A novel CNCP-based method for electricity theft detection was proposed in which two heterogeneous deep neural networks were used to capture the features of the electricity consumption time series in different time scales effectively, therefore yielding better detection results.
2. Materials and Methods
2.1. Dataset
2.2. Data Preprocessing
2.3. Analysis of Electricity Consumption Features of Normal Residents and Electricity Thieves
2.4. A Novel CNCP-Network-Based Detection Method
2.5. Solution of Imbalance Data
3. Results
3.1. Optimal Structure Determination
3.2. Comparison with Other Methods
3.2.1. Parameters Setting of SVM, LR, and XGBoost
3.2.2. Parameters Setting of WD
3.2.3. Parameters Setting of HD
3.3. Complexity Analysis
4. Discussion
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Ali, S.; Yongzhi, M.; Ali, W. Prevention and Detection of Electricity Theft of Distribution Network. Sustainability 2023, 15, 4868. [Google Scholar] [CrossRef]
- Kabir, B.; Qasim, U.; Javaid, N.; Aldegheishem, A.; Alrajeh, N.; Mohammed, E.A. Detecting Nontechnical Losses in Smart Meters Using a MLP-GRU Deep Model and Augmenting Data via Theft Attacks. Sustainability 2022, 14, 15001. [Google Scholar] [CrossRef]
- Oprea, S.V.; Bâra, A.; Puican, F.C.; Radu, I.C. Anomaly Detection with Machine Learning Algorithms and Big Data in Electricity Consumption. Sustainability 2021, 13, 10963. [Google Scholar] [CrossRef]
- Khattak, A.; Bukhsh, R.; Aslam, S.; Yafoz, A.; Alghushairy, O.; Alsini, R. A Hybrid Deep Learning-Based Model for Detection of Electricity Losses Using Big Data in Power Systems. Sustainability 2022, 14, 13627. [Google Scholar] [CrossRef]
- Khan, Z.A.; Adil, M.; Javaid, N.; Saqib, M.N.; Shafiq, M.; Choi, J.G. Electricity Theft Detection Using Supervised Learning Techniques on Smart Meter Data. Sustainability 2020, 12, 8023. [Google Scholar] [CrossRef]
- Xia, X.; Lin, J.; Xiao, Y.; Cui, J.; Peng, Y.; Ma, Y. A Control Chart based Detector for Small-amount Electricity Theft (SET) Attack in Smart Grids. IEEE Internet Things J. 2021, 9, 6745–6762. [Google Scholar] [CrossRef]
- Leite, J.B.; Mantovani, J.R.S. Detecting and Locating Non-Technical Losses in Modern Distribution Networks. IEEE Trans. Smart Grid 2018, 9, 1023–1032. [Google Scholar] [CrossRef] [Green Version]
- Cui, X.; Liu, S.; Lin, Z.; Ma, J.; Wen, F.; Ding, Y.; Yang, L.; Guo, W.; Feng, X. Two-Step Electricity Theft Detection Strategy Considering Economic Return Based on Convolutional Autoencoder and Improved Regression Algorithm. IEEE Trans. Power Syst. 2021, 37, 2346–2359. [Google Scholar] [CrossRef]
- Lin, Y.; Wang, J. Probabilistic Deep Autoencoder for Power System Measurement Outlier Detection and Reconstruction. IEEE Trans. Smart Grid 2020, 11, 1796–1798. [Google Scholar] [CrossRef]
- Zanetti, M.; Jamhour, E.; Pellenz, M.; Penna, M.; Zambenedetti, V.; Chueiri, I. A Tunable Fraud Detection System for Advanced Metering Infrastructure Using Short-Lived Patterns. IEEE Trans. Smart Grid 2019, 10, 830–840. [Google Scholar] [CrossRef]
- Peng, Y.; Yang, Y.; Xu, Y.; Xue, Y.; Song, R.; Kang, J.; Zhao, H. Electricity Theft Detection in AMI Based on Clustering and Local Outlier Factor. IEEE Access 2021, 9, 107250–107259. [Google Scholar] [CrossRef]
- Biswas, P.P.; Cai, H.; Zhou, B.; Chen, B.; Mashima, D.; Zheng, V.W. Electricity Theft Pinpointing Through Correlation Analysis of Master and Individual Meter Readings. IEEE Trans. Smart Grid 2020, 11, 3031–3042. [Google Scholar] [CrossRef]
- Jokar, P.; Arianpoo, N.; Leung, V.C.M. Electricity Theft Detection in AMI Using Customers’ Consumption Patterns. IEEE Trans. Smart Grid 2016, 7, 216–226. [Google Scholar] [CrossRef]
- Song, C.; Sun, Y.; Han, G.; Rodrigues, J.J. Intrusion detection based on hybrid classifiers for smart grid. Comput. Electr. Eng. 2021, 93, 107212. [Google Scholar] [CrossRef]
- Hu, T.; Guo, Q.; Shen, X.; Sun, H.; Wu, R.; Xi, H. Utilizing Unlabeled Data to Detect Electricity Fraud in AMI: A Semisupervised Deep Learning Approach. IEEE Trans. Neural Networks Learn. Syst. 2019, 30, 3287–3299. [Google Scholar] [CrossRef]
- Messinis, G.M.; Rigas, A.E.; Hatziargyriou, N.D. A Hybrid Method for Non-Technical Loss Detection in Smart Distribution Grids. IEEE Trans. Smart Grid 2019, 10, 6080–6091. [Google Scholar] [CrossRef]
- Buzau, M.M.; Tejedor-Aguilera, J.; Cruz-Romero, P.; Gómez-Expósito, A. Detection of Non-Technical Losses Using Smart Meter Data and Supervised Learning. IEEE Trans. Smart Grid 2019, 10, 2661–2670. [Google Scholar] [CrossRef]
- Avila, N.F.; Figueroa, G.; Chu, C.C. NTL Detection in Electric Distribution Systems Using the Maximal Overlap Discrete Wavelet-Packet Transform and Random Undersampling Boosting. IEEE Trans. Power Syst. 2018, 33, 7171–7180. [Google Scholar] [CrossRef]
- Jindal, A.; Dua, A.; Kaur, K.; Singh, M.; Kumar, N.; Mishra, S. Decision Tree and SVM-Based Data Analytics for Theft Detection in Smart Grid. IEEE Trans. Ind. Inform. 2016, 12, 1005–1016. [Google Scholar] [CrossRef]
- Song, C.; Xu, W.; Han, G.; Zeng, P.; Wang, Z.; Yu, S. A Cloud Edge Collaborative Intelligence Method of Insulator String Defect Detection for Power IIoT. IEEE Internet Things J. 2021, 8, 7510–7520. [Google Scholar] [CrossRef]
- Hu, T.; Guo, Q.; Sun, H.; Huang, T.E.; Lan, J. Nontechnical Losses Detection Through Coordinated BiWGAN and SVDD. IEEE Trans. Neural Netw. Learn. Syst. 2021, 32, 1866–1880. [Google Scholar] [CrossRef] [PubMed]
- Aldegheishem, A.; Anwar, M.; Javaid, N.; Alrajeh, N.; Shafiq, M.; Ahmed, H. Towards Sustainable Energy Efficiency With Intelligent Electricity Theft Detection in Smart Grids Emphasising Enhanced Neural Networks. IEEE Access 2021, 9, 25036–25061. [Google Scholar] [CrossRef]
- Punmiya, R.; Choe, S. Energy Theft Detection Using Gradient Boosting Theft Detector With Feature Engineering-Based Preprocessing. IEEE Trans. Smart Grid 2019, 10, 2326–2329. [Google Scholar] [CrossRef]
- Yan, Z.; Wen, H. Electricity Theft Detection Base on Extreme Gradient Boosting in AMI. IEEE Trans. Instrum. Meas. 2021, 70, 1–9. [Google Scholar] [CrossRef]
- Bian, J.; Wang, L.; Scherer, R.; Woźniak, M.; Zhang, P.; Wei, W. Abnormal Detection of Electricity Consumption of User Based on Particle Swarm Optimization and Long Short Term Memory With the Attention Mechanism. IEEE Access 2021, 9, 47252–47265. [Google Scholar] [CrossRef]
- Takiddin, A.; Ismail, M.; Nabil, M.; Mahmoud, M.M.E.A.; Serpedin, E. Detecting Electricity Theft Cyber-Attacks in AMI Networks Using Deep Vector Embeddings. IEEE Syst. J. 2021, 15, 4189–4198. [Google Scholar] [CrossRef]
- Massaferro, P.; Martino, J.M.D.; Fernández, A. Fraud Detection in Electric Power Distribution: An Approach That Maximizes the Economic Return. IEEE Trans. Power Syst. 2020, 35, 703–710. [Google Scholar] [CrossRef]
- Gao, Y.; Foggo, B.; Yu, N. A Physically Inspired Data-Driven Model for Electricity Theft Detection With Smart Meter Data. IEEE Trans. Ind. Inform. 2019, 15, 5076–5088. [Google Scholar] [CrossRef]
- Ramos, C.C.O.; Rodrigues, D.; de Souza, A.N.; Papa, J.P. On the Study of Commercial Losses in Brazil: A Binary Black Hole Algorithm for Theft Characterization. IEEE Trans. Smart Grid 2018, 9, 676–683. [Google Scholar] [CrossRef] [Green Version]
- Lin, G.; Feng, X.; Guo, W.; Cui, X.; Liu, S.; Jin, W.; Lin, Z.; Ding, Y. Electricity Theft Detection Based on Stacked Autoencoder and the Undersampling and Resampling Based Random Forest Algorithm. IEEE Access 2021, 9, 124044–124058. [Google Scholar] [CrossRef]
- Takiddin, A.; Ismail, M.; Zafar, U.; Serpedin, E. Robust Electricity Theft Detection Against Data Poisoning Attacks in Smart Grids. IEEE Trans. Smart Grid 2021, 12, 2675–2684. [Google Scholar] [CrossRef]
- Aslam, Z.; Ahmed, F.; Almogren, A.; Shafiq, M.; Zuair, M.; Javaid, N. An Attention Guided Semi-Supervised Learning Mechanism to Detect Electricity Frauds in the Distribution Systems. IEEE Access 2020, 8, 221767–221782. [Google Scholar] [CrossRef]
- Buzau, M.M.; Tejedor-Aguilera, J.; Cruz-Romero, P.; Gómez-Expósito, A. Hybrid Deep Neural Networks for Detection of Non-Technical Losses in Electricity Smart Meters. IEEE Trans. Power Syst. 2020, 35, 1254–1263. [Google Scholar] [CrossRef]
- Zheng, K.; Chen, Q.; Wang, Y.; Kang, C.; Xia, Q. A Novel Combined Data-Driven Approach for Electricity Theft Detection. IEEE Trans. Ind. Inform. 2019, 15, 1809–1819. [Google Scholar] [CrossRef]
- Zheng, Z.; Yang, Y.; Niu, X.; Dai, H.; Zhou, Y. Wide and Deep Convolutional Neural Networks for Electricity-Theft Detection to Secure Smart Grids. IEEE Trans. Ind. Inform. 2018, 14, 1606–1615. [Google Scholar] [CrossRef]
- Song, C.; Jing, W.; Zeng, P.; Yu, H.; Rosenberg, C. Energy consumption analysis of residential swimming pools for peak load shaving. Appl. Energy 2018, 220, 176–191. [Google Scholar] [CrossRef]
- Song, C.; Jing, W.; Zeng, P.; Rosenberg, C. An analysis on the energy consumption of circulating pumps of residential swimming pools for peak load management. Appl. Energy 2017, 195, 1–12. [Google Scholar] [CrossRef]
- SGCC. Analysis on Abnormal Behavior of Electricity Customers. Available online: https://www.datafountain.cn/competitions/241 (accessed on 4 April 2022).
- Simonyan, K.; Zisserman, A. Very Deep Convolutional Networks for Large-Scale Image Recognition. In Proceedings of the International Conference on Learning Representations, San Diego, CA, USA, 7–9 May 2015. [Google Scholar]
Refs. | Algorithms | Core Ideas | Data Source | Normal Data | Theft Data |
---|---|---|---|---|---|
[6] | Control chart | The mean and variance of short-period normal data and short-period theft data are obviously different. | Simulated | Simulated | Simulated |
[7] | Control chart | The mean and variance of short-period normal data and short-period theft data are obviously different. | Simulated | Simulated | Simulated |
[8] | Convolutional autoencoder | The error of reconstructed theft data would be large using the autoencoder trained by short-period normal data. | IEEE 33-bus test system | Simulated | Simulated |
[9] | Probabilistic deep autoencoder | The error of reconstructed theft data would be large using the autoencoder trained by short-period normal data. | - | Simulated | Simulated |
[10] | FCM, K-means and SOM | Short-period normal data and short-period theft data will appear in different clusters. | Irish Smart Energy Trail dataset | Real | Simulated |
[11] | Clustering and local outlier factor | Short-period normal data and short-period theft data will appear in different clusters. | - | Real | Simulated |
[12] | Correlation analysis | The correlation between short-period normal data and short-period theft data would be significantly low. | Irish Smart Energy Trail dataset | Real | Simulated |
[13] | Consumption pattern | The specific theft mode will lead to a special abnormal state of data. | Irish Smart Energy Trail dataset | Real | Simulated |
Refs. | Algorithms | Data Source | Normal Data | Theft Data |
---|---|---|---|---|
[15] | Seven feature extraction networks + one discrimination network | Irish Smart Energy Trail dataset | Real | Simulated |
[16] | Breakout detection + SVM | Irish Smart Energy Trail dataset | Real | Simulated |
[17] | Six feature extraction methods + XGBoost | A private data set from Endesa in Spain | Real | Real |
[18] | Discrete wavelet transform + random undersampling boosting | - | Real | Real |
[19] | Decision tree + SVM | A private data set from USA | Real | Simulated |
Refs. | Algorithms | Data Source | Normal Data | Theft Data |
---|---|---|---|---|
[21] | Support vector description | Irish Smart Energy Trail dataset | Real | Simulated |
[22] | AlexNet + LightGBM | The State Grid Corporation of China | Real | Real |
GoogLeNet + AdaBoost | ||||
[23] | XGBoost, CatBoost, LightGBM | Irish Smart Energy Trail dataset | Real | Simulated |
[24] | XGBoost | Irish Smart Energy Trail dataset | Real | Simulated |
[25] | LSTM | A private data set from the University of Massachusetts | Real | Simulated |
[26] | Deep vector embeddings | Irish Smart Energy Trail dataset | Real | Simulated |
The State Grid Corporation of China | Real | Real | ||
[27] | Bayesian risk framework | A private data set from Montevideo | Real | Simulated |
A private data set from Montevideo | Real | Real | ||
[28] | Linear regression | A private data set from Southern California Edison (SCE) service territory | Real | Simulated |
[29] | Black hole algorithm | Two private datasets from a Brazilian electric utility | Real | Real |
[30] | Stacked autoencoder + random forest | Irish Smart Energy Trail dataset | Real | Simulated |
A private data set from China | Real | Real | ||
[31] | Deep autoencoder + feed forward neural networks | Irish Smart Energy Trail dataset | Real | Simulated |
[32] | Relational autoencoder + TripleGAN | Irish Smart Energy Trail dataset | Real | Simulated |
A private data set from Montevideo | Real | Real | ||
[33] | LSTM + MLP | A private data set from Endesa in Spain | Real | Real |
[34] | The maximum information coefficient + fast search and find of density peaks | Irish Smart Energy Trail dataset | Real | Simulated |
[35] | BP + CNN | The State Grid Corporation of China | Real | Real |
Our Method | Convolution + non-convolution deep network | The State Grid Corporation of China | Real | Real |
Branch | Code | Structure |
---|---|---|
Convolution branch [39] | C1 | VGG11 |
C2 | VGG13 | |
C3 | VGG16 | |
Non-convolution branch | N1 | fc64-relu-fc2-softmax |
N2 | fc128-relu-fc64-relu-fc2-softmax | |
N3 | fc256-relu-fc128-relu-fc64-relu-fc2-softmax | |
N4 | fc512-relu-fc256-relu-fc128-relu-fc64-relu-fc2-softmax | |
N5 | fc1024-relu-fc512-relu-fc256-relu-fc128-relu-fc64-relu-fc2-softmax |
Structure | TP | FP | FN | TN | Precision | Recall | F1Score | |
---|---|---|---|---|---|---|---|---|
Convolution branch | C1 | 0.7997 | 0.1776 | 0.0727 | 0.9045 | 0.8183 | 0.9167 | 0.8647 |
C2 | 0.853 | 0.1242 | 0.0276 | 0.9497 | 0.8729 | 0.9687 | 0.9183 | |
C3 | 0.7504 | 0.2268 | 0.122 | 0.8553 | 0.7679 | 0.8602 | 0.8114 | |
Non- convolution branch | N1 | 0.851 | 0.1263 | 0.0894 | 0.8878 | 0.8708 | 0.9049 | 0.8875 |
N2 | 0.8413 | 0.1359 | 0.0437 | 0.9336 | 0.8609 | 0.9506 | 0.9036 | |
N3 | 0.8594 | 0.1179 | 0.0425 | 0.9347 | 0.8794 | 0.9529 | 0.9146 | |
N4 | 0.8328 | 0.1444 | 0.0512 | 0.926 | 0.8522 | 0.9421 | 0.8949 | |
N5 | 0.6753 | 0.302 | 0.2857 | 0.6916 | 0.6910 | 0.7027 | 0.6968 | |
CNCP network | C1+N1 | 0.8656 | 0.1117 | 0.0771 | 0.9001 | 0.8857 | 0.9182 | 0.9017 |
C1+N2 | 0.9318 | 0.0454 | 0.0305 | 0.9668 | 0.9535 | 0.9683 | 0.9609 | |
C1+N3 | 0.9202 | 0.0571 | 0.0191 | 0.9682 | 0.9416 | 0.9797 | 0.9602 | |
C2+N1 | 0.8634 | 0.1138 | 0.0752 | 0.902 | 0.8835 | 0.9199 | 0.9013 | |
C2+N2 | 0.9302 | 0.0475 | 0.0155 | 0.9617 | 0.9514 | 0.9836 | 0.9672 | |
C2(3)+N3 | 0.9231 | 0.0478 | 0.0122 | 0.975 | 0.9508 | 0.9870 | 0.9685 | |
C2(5)+N3 | 0.9222 | 0.0591 | 0.0131 | 0.9738 | 0.9398 | 0.9860 | 0.9623 | |
C2(7)+N3 | 0.9184 | 0.0852 | 0.0149 | 0.9615 | 0.9151 | 0.9840 | 0.9483 | |
C3+N1 | 0.8706 | 0.1066 | 0.0635 | 0.9137 | 0.8909 | 0.9320 | 0.9110 | |
C3+N2 | 0.9158 | 0.0615 | 0.0143 | 0.963 | 0.9371 | 0.9846 | 0.9603 | |
C3+N3 | 0.9086 | 0.0686 | 0.0122 | 0.965 | 0.9298 | 0.9868 | 0.9574 |
Training = 60% | Training = 70% | Training = 80% | |||||||
---|---|---|---|---|---|---|---|---|---|
Precision | Recall | F1-Score | Precision | Recall | F1-Score | Precision | Recall | F1-Score | |
SVM | 0.5949 | 0.8545 | 0.7015 | 0.5967 | 0.8559 | 0.7032 | 0.5984 | 0.8570 | 0.7047 |
LR | 0.7391 | 0.6312 | 0.6809 | 0.7443 | 0.6367 | 0.6863 | 0.7457 | 0.6362 | 0.6866 |
RF | 0.8461 | 0.4040 | 0.5469 | 0.8667 | 0.3735 | 0.5220 | 0.8747 | 0.3522 | 0.5022 |
XGB | 0.8712 | 0.4892 | 0.6266 | 0.8787 | 0.4873 | 0.6269 | 0.8837 | 0.4692 | 0.6130 |
HD | 0.9013 | 0.6816 | 0.7762 | 0.9452 | 0.5937 | 0.7293 | 0.9109 | 0.6808 | 0.7792 |
HD-S | 0.9010 | 0.6809 | 0.7756 | 0.9446 | 0.5928 | 0.7284 | 0.9102 | 0.6814 | 0.7794 |
WD | 0.8327 | 0.9158 | 0.8723 | 0.8378 | 0.9041 | 0.8697 | 0.8280 | 0.8982 | 0.8617 |
CNCP-L | 0.9121 | 0.9248 | 0.9184 | 0.9357 | 0.9322 | 0.9339 | 0.9472 | 0.9821 | 0.9643 |
CNCP | 0.9125 | 0.9253 | 0.9189 | 0.9392 | 0.9385 | 0.9388 | 0.9508 | 0.987 | 0.9685 |
Training = 60% | Training = 70% | Training = 80% | |||||||
---|---|---|---|---|---|---|---|---|---|
Precision | Recall | F1-Score | Precision | Recall | F1-Score | Precision | Recall | F1-Score | |
1 | CNCP | CNCP | CNCP | CNCP | CNCP | CNCP | CNCP | CNCP | CNCP |
2 | CNCP-L | CNCP-L | CNCP-L | CNCP-L | CNCP-L | CNCP-L | CNCP-L | CNCP-L | CNCP-L |
3 | HD | WD | WD | HD | WD | WD | HD | WD | WD |
4 | HD-S | SVM | HD | HD-S | SVM | HD | HD-S | SVM | HD |
5 | XGB | HD | HD-S | XGB | HD | HD-S | XGB | HD | HD-S |
6 | RF | HD-S | SVM | RF | HD-S | SVM | RF | HD-S | SVM |
7 | WD | LR | XGB | WD | LR | LR | WD | LR | LR |
8 | LR | XGB | LR | LR | XGB | XGB | LR | XGB | XGB |
9 | SVM | RF | RF | SVM | RF | RF | SVM | RF | RF |
Convolution Branch | Non-Convolution Branch | ||
---|---|---|---|
Layer | FLOPs | Layer | FLOPs |
Conv_1 | 2 × (3 × 3 × 37 × 28) × (1 × 64) | FC256 | (37 × 28) × 256 |
Conv_2 | 2 × (3 × 3 × 37 × 28) × (64 × 64) | FC128 | 256 × 128 |
Max_pool | 37 × 28 × 64 | FC64 | 128 × 64 |
Conv_3 | 2 × (3 × 3 × 19 × 14) × (64 × 128) | ||
Conv_4 | 2 × (3 × 3 × 19 × 14) × (128 × 128) | ||
Max_pool | 19 × 14 × 128 | ||
Conv_5 | 2 × (3 × 3 × 10 × 7) × (128 × 256) | ||
Conv_6 | 2 × (3 × 3 × 10 × 7) × (256 × 256) | ||
Conv_7 | 2 × (3 × 3 × 10 × 7) × (256 × 256) | ||
Gmpool | 256 × 10 × 7 | ||
FC60 | 256 × 60 |
Size of the Kernel | FLOPs |
---|---|
3 × 3 | 4.02 × 10 |
5 × 5 | 1.116 × 10 |
7 × 7 | 2.187 × 10 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Wang, Y.; Jin, S.; Cheng, M. A Convolution–Non-Convolution Parallel Deep Network for Electricity Theft Detection. Sustainability 2023, 15, 10127. https://doi.org/10.3390/su151310127
Wang Y, Jin S, Cheng M. A Convolution–Non-Convolution Parallel Deep Network for Electricity Theft Detection. Sustainability. 2023; 15(13):10127. https://doi.org/10.3390/su151310127
Chicago/Turabian StyleWang, Yiran, Shuowei Jin, and Ming Cheng. 2023. "A Convolution–Non-Convolution Parallel Deep Network for Electricity Theft Detection" Sustainability 15, no. 13: 10127. https://doi.org/10.3390/su151310127