文章摘要
基于知识嵌入和深度神经网络的异常用电检测
Non-technical loss detection based on energy measurement knowledge and deep neural network
投稿时间:2019-10-18  修订日期:2019-11-20
DOI:
中文关键词: 智能电网  深度学习  神经网络  异常用电
英文关键词: smart grid  deep learning  neural network  non-technical loss detection
基金项目:国家自然科学基金项目(面上项目,重点项目,重大项目)
作者单位E-mail
李江腾 华中科技大学电信学院 lijiangteng@hust.edu.cn 
王非 华中科技大学电信学院 wangfei@hust.edu.cn 
摘要点击次数: 73
全文下载次数: 97
中文摘要:
      目前的异常用电检测研究主要以居民用户为对象,相关方法并不适用于工商业用户。针对此问题,本文提出一种融合了电能计量原理的基于深度学习的异常用电检测方法。首先,分析了各类异常用电的数据现象,结果说明单纯采用智能电表数据不足以准确检测异常用电。本文遵循电能计量原理,将描述电气参量内部逻辑关系的指标作为知识嵌入智能电表数据,构建深度学习样本模型。然后,本文提出了一个改进的深度混合残差神经网络,确保从海量智能电表数据中学习用于识别异常用电的高级特征。实验结果表明,相比多个基准算法,本文方法在所有评估指标上均取得了明显的提升。
英文摘要:
      The current research on non-technical loss(NTL) detection is mainly for residential customers, and the related methods are not suit industrial and commercial customers. According to this problem, joint with the principle of electricity measurement, this paper proposes a deep learning based NTL detection method. Firstly, we analyzed various NTL. The data phenomena show that only smart meter data is not enough to detect NTL. Hence, this paper embeds the electricity measurement knowledge which reflects the internal relationship among electricity magnitudes into smart meter data, and construct as 1D vector style sample for deep neural network analyzing. Secondary, this paper proposes an improved hybrid residual neural network to extract advanced features of NTL from massive smart meter data. The experiment results show that our approach can achieve satisfactory performance. Compared with the baselines, our method has achieved significant improvement on all metrics.#$NLKeywords: smart grid; deep learning; neural network; non-technical loss detection
查看全文   查看/发表评论  下载PDF阅读器
关闭