一种变压器的告警方法、装置、设备及存储介质
By extracting features and processing signals from real-time transformer data, and using principal component analysis and trend prediction models, the problem of low-frequency oscillation signals being masked by noise was solved, enabling accurate fault diagnosis and dynamic alarm for transformers.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ELECTRIC POWER RES INST OF STATE GRID ZHEJIANG ELECTRIC POWER COMAPNY
- Filing Date
- 2026-06-18
- Publication Date
- 2026-07-17
AI Technical Summary
In the existing technology, the low-frequency oscillation signal of a single-phase transformer circuit is easily masked by operating noise, and the waveform shape is affected by load fluctuations and changes in ambient temperature, making it difficult to accurately identify early fault signs from complex operating data, resulting in difficulties in condition assessment.
By acquiring the real-time voltage sequence, real-time current sequence, load fluctuation sequence, and ambient temperature sequence of the transformer, abnormal frequency band signals and abnormal signal feature sets are generated. Principal component analysis algorithm is used to extract principal component feature vectors. Combined with LSTM and ARMA models, signal smoothing and trend prediction are performed, and thresholds are dynamically adjusted for alarms.
It enables accurate capture of low-frequency oscillation signals and early risk assessment, improves the accuracy of fault diagnosis, reduces the false alarm rate, dynamically adapts to changes in operating conditions, and enhances the pre-diagnosis capability of transformers.
Smart Images
Figure CN122412928A_ABST