基于混合深度学习的电力用户窃电检测方法研究

吴栩峰; 区彦黛; 张思路; 庄婉铃; 魏华杰

吴栩峰,区彦黛,张思路,庄婉铃,魏华杰.基于混合深度学习的电力用户窃电检测方法研究[J].电测与仪表,2026,63(2):148-156.
WU Xufeng,OU Yandai,ZHANG Silu,ZHUANG Wanling,WEI Huajie.Research on Electricity User Theft Detection Method Based on Hybrid Deep Learning[J].Electrical Measurement & Instrumentation,2026,63(2):148-156.

基于混合深度学习的电力用户窃电检测方法研究

Research on Electricity User Theft Detection Method Based on Hybrid Deep Learning

DOI：10.19753/j.issn1001-1390.2026.02.016

中文关键词: 分布式光伏用户窃电检测半监督卷积神经网络模型 Transformer模型鲸鱼算法

英文关键词:Distributed photovoltaic users, Electricity theft detection, Semi supervision, Convolutional neural network model, Transformer model, Whale Optimization Algorithm

基金项目:深圳供电局科技项目（项目编号：090000KK52222030）

作者中文名	作者英文名	单位
吴栩峰	WU Xufeng	深圳供电局有限公司
区彦黛	OU Yandai	深圳供电局有限公司
张思路	ZHANG Silu	云南电网有限责任公司昆明供电局
庄婉铃	ZHUANG Wanling	深圳供电局有限公司
魏华杰	WEI Huajie	深圳供电局有限公司

摘要点击次数: 89

中文摘要:

分布式光伏系统并网连接方式多样和运行模式灵活等导致窃电行为难以被察觉,有效的窃电检测方法对于遏制窃电十分关键。针对有监督分类和无监督回归窃电检测方法存在的模型泛化能力差和检测精度低等问题,本文在分析光伏窃电原理的基础上,提出了一种结合卷积神经网络和Transformer的混合半监督模型进行窃电检测。卷积神经网络模型融合双向长短时记忆网络和注意力机制,并利用Tent混沌映射、非线性收敛因子、平均位置向量、自适应惯性权重优化的鲸鱼算法寻优超参数。Transformer模型结合多尺度感知层和线性映射层对其进行优化。通过实验对所提模型的优越性进行验证。实验结果表明,所提半监督学习的光伏窃电检测模型在少量标注数据和大量未标注数据的情况下,能够准确判断分布式光伏用户是否存在窃电行为,具有较高的检测准确率和曲线下面积,准确率为96.67%,曲线下面积为99.05%。与常规相比,所提方法的准确率分别提高了4.98%和0.44%,曲线下面积分别提高了3.12%和0.74%。研究为新型电力系统下的反窃电技术提供了新的思路,有助于推动清洁能源的可持续发展。

英文摘要:

The diverse connection methods and flexible operation modes of distributed photovoltaic systems make it difficult to detect electricity theft, effective electricity theft detection methods are crucial for curbing electricity theft. In response to the problems of poor model generalization ability and low detection accuracy in supervised classification and unsupervised regression methods for electricity theft detection, this paper proposes a hybrid semi supervised model combining convolutional neural networks and Transformers for electricity theft detection based on the analysis of the principle of photovoltaic electricity theft. The convolutional neural network model integrates bidirectional long short-term memory networks and attention mechanisms, and uses the whale algorithm with Tent chaotic mapping, nonlinear convergence factor, average position vector, and adaptive inertia weight optimization to optimize hyperparameters. The Transformer model is optimized by combining multi-scale perception layers and linear mapping layers. Verify the superiority of the proposed model through experiments.The experimental results indicate that, proposed semi supervised learning photovoltaic electricity theft detection model can accurately determine whether distributed photovoltaic users have electricity theft behavior with a small amount of labeled data and a large amount of unlabeled data. It has high detection accuracy and area under the curve, with an accuracy of 96.67% and an area under the curve of 99.05%. Compared to conventional methods, the accuracy of the proposed method has increased by 4.98% and 0.44%, respectively, and the area under the curve has increased by 3.12% and 0.74%, respectively. The research provides new ideas for anti electricity theft technology in the new power system, which helps promote the sustainable development of clean energy.

查看全文查看/发表评论下载PDF阅读器

关闭