A non-invasive household electrical load identification and safety warning method and system
By employing a collaborative architecture of high-precision feature extraction at the edge and federated learning in the cloud, combined with a user feedback optimization mechanism, the problems of blind spots in the identification of new devices and device aging in non-intrusive load identification have been solved, realizing the application of intelligent identification and safety early warning technology for household electricity use.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIHUA UNIV
- Filing Date
- 2026-04-17
- Publication Date
- 2026-07-17
AI Technical Summary
Existing non-intrusive load identification technologies struggle to identify new devices in complex household electrical environments, and their accuracy declines due to device aging. Furthermore, the lack of user interaction and feedback loops leads to the accumulation of identification errors and a poor user experience.
It adopts a collaborative architecture of high-precision feature extraction at the edge and incremental update through cloud-based federated learning, combined with a user feedback-guided recognition optimization mechanism. It extracts multi-dimensional load features through a sliding window, detects load switching events, and performs incremental learning and user feedback correction in the cloud to dynamically update the model.
It enables effective identification of new equipment and adaptation to equipment aging, improves identification accuracy and user participation, prevents error accumulation, and ensures the system's identification stability and safety early warning capabilities during long-term operation.
Smart Images

Figure CN122418979A_ABST