考虑辨识结果连续性的Type-Ⅲ型工商业负荷辨识方法  

Identification Method for Type-ⅢIndustrial and Commercial Load Considering Identification Result Continuity

作　　者：段晶 李勇[1] 张振宇 李巍巍 蒋林[3] 李磊 DUAN Jing;LI Yong;ZHANG Zhenyu;LI Weiwei;JIANG Lin;LI Lei(College of Electrical and Information Engineering,Hunan University,Changsha 410082,China;Changsha Xinao Changran Energy Development Co.,Ltd.,Changsha 410005,China;Department of Electrical Engineering and Electronics,University of Liverpool,Liverpool L693BX,UK;Changsha Xinao Xiangjiang New Energy Development Co.,Ltd.,Changsha 410005,China)

机构地区：[1]湖南大学电气与信息工程学院,湖南省长沙市410082 [2]长沙新奥长燃能源发展有限公司,湖南省长沙市410005 [3]利物浦大学电气工程与电子系,利物浦L693BX,英国 [4]长沙新奥湘江新能源发展有限公司,湖南省长沙市410005

出　　处：《电力系统自动化》2021年第24期65-72,共8页Automation of Electric Power Systems

基　　金：国家重点研发计划政府间国际科技创新合作重点专项资助项目(2018YFE0125300);国家自然科学基金资助项目(52061130217);湖湘高层次人才聚集工程资助项目(2019RS1016)。

摘　　要：非侵入式负荷监测技术可以引导用户合理安排用电时间,从而减少电量消耗。其中,由于状态的连续可变性,连续变化(Type-Ⅲ)型负荷的辨识一直是非侵入式负荷监测中难以解决的问题之一。针对Type-Ⅲ型负荷的辨识难题,提出了基于深度卷积神经网络(CNN)和隐马尔可夫模型(HMM)的非侵入式负荷辨识算法。首先,根据互信息理论进行负荷特征选择;然后,利用残差神经网络作为深度CNN的基本架构,提取负荷多维特征并实现Type-Ⅲ型负荷的初辨识;最后,为了解决CNN辨识结果中存在的状态断点问题,采用HMM完成负荷辨识结果的连续性优化。在复杂的工商业运行环境中,对具有代表性的Type-Ⅲ型负荷数据进行了算法训练和验证,结果表明所提算法能有效辨识Type-Ⅲ型工商业负荷的运行状态。Non-intrusive load monitoring technology can guide users to arrange power consumption time reasonably,thereby reducing power consumption.Among them,due to the continuous variability of the state,the identification of continuously varying(Type-Ⅲ)load has always been one of the difficult problems in non-intrusive load monitoring.Aiming at the problem of Type-Ⅲload identification,a non-intrusive load identification algorithm based on deep convolutional neural network(CNN)and hidden Markov model(HMM)is proposed.Firstly,the load characteristics are selected according to the mutual information theory.Then,the residual neural network is used as the basic structure of deep CNN to extract multi-dimensional features of the load and realize the initial identification of Type-Ⅲloads.Finally,in order to solve the problem of state breakpoint in CNN identification results,the HMM is used to complete the continuous optimization of load identification results.In the complex industrial and commercial operation environment,the algorithm is trained and verified on the representative Type-Ⅲload data.The results show that the proposed algorithm can effectively identify the operation state of Type-Ⅲindustrial and commercial load.

关 键 词：非侵入式负荷辨识 互信息 残差神经网络 隐马尔可夫模型 

分 类 号：TM714[电气工程—电力系统及自动化]