电力信息物理系统开关及拓扑状态感知方法、设备和介质
By constructing a multi-source detection network for switch states consisting of a multi-source temporal feature encoding layer and a state-space temporal detection layer, the problems of insufficient identification of weak transient features and non-compliance of multi-source signal fusion with physical laws in the detection of switch states in distribution networks are solved, thus achieving efficient and accurate switch state identification and real-time detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HEFEI UNIV OF TECH
- Filing Date
- 2026-05-12
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for detecting the status of distribution network switches lack sufficient feature recognition under weak transient conditions such as high-resistance grounding and light-load closing. Furthermore, the fusion of multi-source signals does not conform to the physical laws of the power system, resulting in decreased detection accuracy and poor model generalization ability, making it difficult to meet real-time detection requirements.
A multi-source temporal feature encoding layer is used to enhance learnable topological transient energy and drive multi-source attention fusion driven by topological coupling complex correlation. Combined with a state-space temporal detection layer with adaptive discrete length, a multi-source detection network for switch states is constructed. Transient derivative features are extracted by first-order difference, and topological analytic signals are constructed by Hilbert transform. This achieves enhancement of weak transient features and drives the physical coupling characteristics of multi-source signals, thereby optimizing the recursive process of the state-space model.
It improves the detection performance and model generalization ability under weak transient conditions, enhances detection accuracy and real-time performance, solves the problems of insufficient feature recognition and overfitting in existing technologies, and achieves efficient and accurate identification of the switch status of distribution networks.
Smart Images
Figure CN122196700B_ABST