Clean air switch equipment fault diagnosis method and system
By adopting the CNN-LSTM-Attention model for multimodal feature fusion and lightweight deployment in clean air switchgear, the problems of insufficient monitoring accuracy and poor adaptability in existing technologies are solved, and efficient fault diagnosis and real-time prediction are achieved.
Patent Information
- Application Number
- CN202510869586.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-26
- Publication Date
- 2025-09-05
AI Technical Summary
The existing technology in clean air switchgear insulation fault diagnosis has problems such as insufficient monitoring accuracy of a single sensor, weak modeling ability of traditional machine learning models for time series data and spatial correlation, and poor model adaptability due to feature drift.
The CNN-LSTM-Attention model is used for multimodal feature fusion, combining gas concentration signals, partial discharge signals, and environmental parameters. A bidirectional long short-term memory network is used to capture long-term temporal dependencies, a convolutional neural network is used to extract spatial correlations, and multi-head self-attention is used to focus on key fault features. Simultaneously, model pruning and quantization are performed to achieve lightweight deployment, and an incremental learning mechanism is introduced.
It improves the fault detection rate and the timeliness of diagnosis and prediction, solves the problem that traditional methods cannot capture the multi-dimensional characteristics of complex faults, realizes the real-time and adaptability of the model, and reduces maintenance costs.