| 引用本文: | 周步祥,赵雯雯,臧天磊,陈阳,闵昕玮.基于低频功率差量特征与双长短期记忆网络的非侵入式负荷监测方法[J].电力自动化设备,2023,43(8):167-173,209 |
| ZHOU Buxiang,ZHAO Wenwen,ZANG Tianlei,CHEN Yang,MIN Xinwei.Non-intrusive load monitoring method based on low-frequency power difference characteristic and dual long short-term memory network[J].Electric Power Automation Equipment,2023,43(8):167-173,209 |
|
| 摘要: |
| 为提升负荷监测中事件检测与负荷识别的准确性与适应性，提出一种基于低频功率差量特征与双长短期记忆网络的非侵入式负荷监测方法。基于低频数据，根据电器正常运行造成的功率波动与事件启停造成的功率跳变之间的特性差异，提出一种事件检测算法，该算法通过滑动窗内功率波动的差量特征排除波动干扰，实现事件准确定位并获取相关功率数据；建立一种双长短期记忆网络，对不同电器构建专一电器判别单元并进行训练；建立由各判别单元组成的事件识别网络，根据各判别单元输出的概率对事件进行综合判别，实现非侵入式负荷监测。基于测试数据集的仿真结果验证了所提方法的有效性与准确性。 |
| 关键词: 事件检测 非侵入式负荷监测 长短期记忆网络 负荷识别 低频特征 |
| DOI：10.16081/j.epae.202212011 |
| 分类号:TM73 |
| 基金项目:国家自然科学基金资助项目(51907097)；国家重点研发计划项目(2021YFB4000500)；四川省科技计划项目(2020JDRC0049) |
|
| Non-intrusive load monitoring method based on low-frequency power difference characteristic and dual long short-term memory network |
|
ZHOU Buxiang, ZHAO Wenwen, ZANG Tianlei, CHEN Yang, MIN Xinwei
|
|
College of Electrical Engineering, Sichuan University, Chengdu 610065, China
|
| Abstract: |
| In order to improve the accuracy and adaptability of event detection and load identification in load monitoring, a non-intrusive load monitoring method based on low-frequency power difference characteristic and dual long short-term memory network is proposed. Based on the low-frequency data, an event detection algorithm is proposed according to the characteristic difference between power fluctuation caused by normal operation of appliances and power jumping caused by event start/stop, which eliminates fluctuation interference through the difference characteristic of power fluctuation in the sliding window, realizes accurate event location and obtains relevant power data. A dual long short-term memory network is established, and the specific discrimination units of different appliances are constructed and trained. An event identification network consisting of each discrimination unit is established, and the events are comprehensively discriminated according to the probabilities outputted by each discrimination unit, and the non-intrusive load monitoring is realized. The effectiveness and accuracy of the proposed method are verified based on the test data sets. |
| Key words: event detection non-intrusive load monitoring long short-term memory network load identification low-frequency characteristic |
