Multi-Dimensional Prediction of Mechanical State and Adaptive Phase-Selection Optimization Control Technology for Medium-Voltage Distribution Switches

Given that the mechanical characteristics of medium-voltage distribution switches deteriorate over operation cycles through cumulative wear, action time drift, and speed attenuation, traditional inspection methods struggle to provide early identification of degradation trends and dynamic optimization of control strategies. To enhance the observability of mechanical state and extend the electrical service life of the equipment, this paper constructs a mechanical characteristic fingerprint based on non-invasive current signals, integrates a time-series deep learning model for state prediction, and proposes an adaptive phase-selection control strategy aimed at balancing operational load. Validation through physical testing on 6 kV vacuum circuit breakers with permanent magnet/spring mechanisms shows that the model achieves high prediction accuracy, and the adaptive strategy can significantly reduce uneven wear and extend the operational lifespan of the equipment.