A fault early warning method, device and computer readable storage medium of a wind turbine generator
By collecting multi-source data and using an attention mechanism for feature weighted fusion and time-series prediction, the problem of inaccurate identification of early hidden faults in wind turbine fault diagnosis is solved, and high-precision fault warning and early warning are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- POWER DISPATCHING CONTROL CENT OF GUANGDONG POWER GRID CO LTD
- Filing Date
- 2026-04-30
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, fault diagnosis of wind turbines relies on analysis of a single data source, which makes it difficult to capture early hidden faults. Furthermore, multi-data source fusion technology lacks a dynamic adjustment mechanism for feature weights, resulting in features being unable to effectively distinguish key fault information from background noise, thus affecting the accuracy of time series prediction models.
Real-time acquisition of CMS vibration monitoring data, SCADA system operating parameter data, and acoustic signature monitoring data; weighted fusion of features calculated using an attention mechanism; input of the data into a time-series prediction model using a sliding window mechanism; and continuous monitoring of residual vectors to generate fault warning signals.
It enables accurate early warning of hidden faults in wind turbine units, improves the accuracy of multi-source data fusion and the precision of time-series prediction, reduces the false alarm rate, and improves the reliability of wind turbine unit operation and maintenance.
Smart Images
Figure CN122407474A_ABST