Electricity theft detection based on t-LeNet and time series classification

doi:10.3969/j.issn.1000-5641.2021.05.010

Abstract

Abstract:

Electricity theft results in significant losses in both electric energy and economic benefits for electric power enterprises. This paper proposes a method to detect electricity theft based on t-LeNet and time series classification. First, a user’s power consumption time series data is obtained, and down-sampling is used to generate a training set. A t-LeNet neural network can then be used to train and predict classification results for determining whether the user exhibits behavior reflective of electricity theft. Lastly, real user power consumption data from the state grid can be used to conduct experiments. The results show that compared with the time series classification method based on Time-CNN (Time Convolutional Neural Network) and MLP (Muti-Layer Perception), the proposed method offers improvements in the comprehensive evaluation index, accuracy rate, and recall rate index. Hence, the proposed method can successfully detect electricity theft.

Key words: time series classification, t-LeNet, electricity theft detection

CLC Number:

TP391

Xiaoqin MA, Xiaohui XUE, Hongjiao LUO, Tongyu LIU, Peisen YUAN. Electricity theft detection based on t-LeNet and time series classification[J]. Journal of East China Normal University(Natural Science), 2021, 2021(5): 104-114.

Figures/Tables 10

Fig.1

Fig.2

Fig.3

Table 1

Table 2

Table 3

Table 4

Fig.4

Fig.5

Table 5

References 23

1	谢治春, 赵兴庐, 刘媛. 金融科技发展与商业银行的数字化战略转型. 中国软科学, 2018, 184- 192.
2	YANG S, ZHANG Z Q, ZHOU J, et al. Financial risk analysis for smes with graph-based supply chain mining [C]// Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI-20), Special Track on AI in FinTech. 2020: 4661-4667.
3	乔宇锋. 智能化金融监管: 模型框架、边缘约束和实践策略. 南方金融, 2021(4), 71- 80.
4	程雪军. 人工智能深度介入消费金融: 动因、风险及防控. 深圳大学学报(人文社会科学版), 2021, 38, 67- 76.
5	PASSERINI F, TONELLO A M. Smart grid monitoring using power line modems: Effect of anomalies on signal propagation. IEEE Access, 2019, (7): 27302- 27312.
6	李炳森, 胡全贵, 陈小峰, 等. 电网企业数据中台的研究与设计. 电力信息与通信技术, 2019, 17, 29- 34.
7	王全兴, 李思韬. 基于采集系统的反窃电技术分析及防范措施. 电测与仪表, 2016, 53, 78- 83.
8	李端超, 王松, 黄太贵, 等. 基于大数据平台的电网线损与窃电预警分析关键技术. 电力系统保护与控制, 2018, 46, 143- 151.
9	张晓新, 王奇超, 林峰, 等. 窄带物联网在专变用户防窃电应用中的研究. 电子器件, 2021, 44, 178- 181. doi: 10.3969/j.issn.1005-9490.2021.01.034
10	庄池杰, 张斌, 胡军, 等. 基于无监督学习的电力用户异常用电模式检测. 中国电机工程学报, 2016, 36, 379- 387, 594.
11	黄悦华, 郭思涵, 鲍刚, 等. 基于用电特征分析的异常用电检测方法. 三峡大学学报(自然科学版), 2021, 43, 96- 101.
12	ISMAIL FAWAZ H, FORESTIER G, WEBER J, et al. Deep learning for time series classification: A review. Data Mining and Knowledge Discovery, 2019, 33 (4): 917- 963. doi: 10.1007/s10618-019-00619-1
13	EBRAHIM S A , J POSHTAN, JAMALI S M, et al. Quantitative and qualitative analysis of time-series classification using deep learning. IEEE Access, 2020, (8): 90202- 90215.
14	WANG Z G, YAN W Z, OATES T. Time series classification from scratch with deep neural networks: A strong baseline [C]// 2017 International Joint Conference on Neural Networks (IJCNN) . IEEE, 2017: 1578-1585.
15	LINES J, DAVIS L M, HILLS J, et al. A shapelet transform for time series classification [C]// Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2012: 289-297.
16	HE Q, DONG Z, ZHUANG F Z, et al. Fast time series classification based on infrequent shapelets [C]// 2012 11th International Conference on Machine Learning and Applications. IEEE, 2012: 215-219.
17	LE GUENNEC A, MALINOWSKI S, TAVENARD R. Data augmentation for time series classification using convolutional neural network [C/OL]// ECML/PKDD Workshop on Advanced Analytics and Learning on Temporal Data. (2016)[2021-07-02]. https://halshs.archives-ouvertes.fr/halshs-01357973/document.
18	CUI Z, CHEN W, CHEN Y. Multi-scale convolutional neural networks for time series classification [EB/OL]. (2016-05-11)[2021-06-15]. https://arxiv.org/pdf/1603.06995.pdf.
19	曹阳, 闫秋艳, 吴鑫. 不平衡时间序列集成分类算法. 计算机应用, 2021, 41, 651- 656.
20	李艳霞, 柴毅, 胡友强, 等. 不平衡数据分类方法综述. 控制与决策, 2019, 34, 673- 688.
21	LI J, SI Y Y, XU T, et al. Deep convolutional neural network based ECG classification system using information fusion and one-hot encoding techniques [J/OL]. Mathematical Problems in Engineering, (2018-12-02)[2021-07-02]. https://doi.org/10.1155/2018/7354081.
22	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (6): 1137- 1149.
23	CHALAPATHY R, CHAWLA S. Deep Learning for Anomaly Detection: A Survey [EB/OL]. (2019-01-23)[2020-07-02]. https://arxiv.org/pdf/1901.03407.pdf.

层数	名称	参数设置
1	Input
2	Convolution 1	filter = 5, kernel_size = 5, activation = ‘relu’, padding = ‘same’
3	Pooling 1	Max Pooling, pool size = 2
4	Convolution 2	filter = 20, kernel_size = 5, activation = ‘relu’, padding = ‘same’
5	Pooling 2	Max Pooling, pool size = 4
6	Flatten
7	Dense	uints = 500, activation = ‘relu’
8	Output(Dense)	uints = n, activation = ‘softmax’

模型	实验参数
模型	输入序列长度	批量大小	迭代次数	层数
Time-CNN	1034	16	50	4
MLP	1034	16	50	4
t-LeNet	1034	16	50	5

模型	学习率区间	单次增长幅度	最高准确率	最高准确率对应学习率
Time-CNN	[0.01, 0.10]	0.010	0.7123	0.020
Time-CNN	[0.001, 0.010]	0.001	0.5138	0.006
MLP	[0.001, 0.010]	0.001	0.6542	0.008
MLP	[0.01, 0.20]	0.01	0.6736	0.050
t-LeNet	[0.01, 0.10]	0.01	0.6971	0.010 00
	[0.001, 0, 010]	0.001	0.700 6	0.001 00
	[0.0001, 0.0010]	0.000 1	0.7013	0.00010
	[0.00001, 0.00010]	0.000 01	0.712 8	0.000 07

模型	学习率	A	${P_{ {\text{macro} } } }$	$ {R_{{\text{macro}}}} $	$F_{1{ {\text{macro} } } }$
Time-CNN	0.002	0.7123	0.7139	0.7129	0.7121
MLP	0.050	0.6763	0.7322	0.6803	0.6586
t-LeNet	0.00007	0.7128	0.7258	0.7218	0.7212

模型	A	${P_{ {\text{macro} } } }$	$ {R_{{\text{macro}}}} $	$F_{1{ {\text{macro} } } }$
Time-CNN	0.7137	0.6784	0.7020	0.6900
MLP	0.6867	0.5909	0.5941	0.5925
t-LeNet	0.7289	0.6905	0.7122	0.7012