基于对抗性判别域自适应的未知新增设备非侵入式负荷识别方法

打开文本图片集

何胜（1978—），男，工程师，主要从事智能电器和工控领域的技术研究、产品研发及行业应用。

关键词：非侵入式负荷监测；未知新增设备；对抗性判别域自适应；迁移学习中图分类号：TM714 文献标志码：A 文章编号：2095-8188（2025）06-0022-10DOI： 10.16628/j . cnki.2095-8188.2025.06.004

Non-Intrusive Load Identification Method for Unknown Added Devices Based on Adversarial Discriminative Domain Adaptation

HE Sheng1²， YANG Haowen³，ZHAO Jingbing³ [1. Shanghai Chint Intelligent Technology Co.，Ltd.，Shanghai 201616， China; 2. Chint Low Voltage Intelligent Electrical Appliances Research Institute， Shanghai 2O1616， China; 3.State Key Laboratory of Electrical Materials and Electrical Insulation （Xi'an Jiaotong University），Xi'an 710049，China]

Abstract：Aiming attheproblem that thediffrences insignal characteristics between different electrical devices lead to the decreasein accuracyand insuficient generalization abilityof traditional non-intrusive load recognition methods when facingunknownaditional devices，a migrationlearning method basedon theadversarial Discriminativedomainadaptation（ADDA） isproposed.First，the multi-dimensional feature parameters ofdifferent devices are extracted bycombining feature selection methods，and thesource domainmodels aretrained using the source device feature data.Then，an adversarial training strategy is employed，with a discriminator introduced to optimizethefeature extractorofthetargetdevice.Finally，themodel isfine-tunedandoptimizedtocompletethe load recognition task on the target device.Experimentalresults show thatthe optimized recognition model achieves an average accuracy of 98.90% ，an improvement of 18.41% compared to the traditional migration learning methods.

Keywords：non-intrusive load monitoring；unknown added devices；adversarial discriminative lomain adaptation （ADDA）； transfer learning

0 引言

非侵入式负荷监测技术作为智能电网和能源管理系统的重要组成部分，已被广泛应用于家庭、商业及工业电力监测中[1]。（剩余11336字）