Toward smart energy user: Real time non-intrusive load monitoring with simultaneous switching operations
Toward smart energy user: Real time non-intrusive load monitoring with simultaneous switching operations
复制标题
迈向智能能源用户：实时非侵入式负载监控和同步开关操作
DOI：
10.1016/j.apenergy.2021.116616
复制
发表时间：
2021-02-13
期刊：
影响因子：
11.2
通讯作者：
Gao, Shan
中科院分区：
文献类型：
--
作者：
Liu, Yu;Liu, Wei;Gao, Shan
非侵入式负荷监测是智能能耗管理中的一项很有前途的技术，它可以为用户的用电模式和生活习惯提供洞察，从而为多种智能用电应用提供技术支持。作为该领域最具挑战性的问题，同时切换操作的实时负荷分解问题由于其复杂性和不确定性一直没有得到解决。本文针对实时非侵入式负荷监测框架下的同时作用问题，提出了一种实用可行的解决方案，填补了这方面的研究空白。在第一阶段，为了处理不同类型的重叠组合，采用了基于自适应窗口的检测方法来捕获状态变化。其次，在实时负载监测系统中提出了一种深度字典学习模型，该模型以半监督模块化的方式生成字典原子，并扩展了家电的高阶动态特征。在此基础上，提出了一种稀疏编码公式化算法来解决该问题，该算法兼容处理多个切换事件的同时发生。所提出的模型和算法在仿真平台和现场测量中得到了验证。结果表明，该方法除了具有较高的监控精度外，还能有效地解决作业重叠问题。
Non-intrusive load monitoring is a promising technology in intelligent energy consumption management, which can provide insights for electricity use patterns and customer living habits, leading to the technical support of multiple smart energy use applications. As the most challenging problem in this specific field, the real time load disaggregation with simultaneous switching operations still remains unsolved due to its complexity and uncertainty. In this paper, a practical and feasible solution for the simultaneous action problems under real time non intrusive load monitoring framework is proposed and investigated, which is capable of filling such research gap. At the first stage, to tackle the different kinds of overlapping combinations, an adaptive-window based detection approach is applied to capture the state changes. Secondly, a deep dictionary learning model is proposed in the real time load monitoring architecture, where the dictionary atom is generated in a semi-supervised modularization way and extended with high order dynamic features of appliances. Then, sparse coding formulized algorithm is presented to solve the problem, which is compatible of handling simultaneous occurrence of multiple switching events. The proposed model and algorithm are tested and verified in both simulation platform and field measurements. Results show that the operation-overlapping issue can be effectively addressed by the presented method, in addition to the high accuracy monitoring performance.