Abstract: To enhance the perception accuracy and early identification capability of vacuum degradation in high-voltage vacuum circuit breakers, this paper proposes a vacuum monitoring technique based on multi-source data fusion and support vector machine (SVM) modeling. A comprehensive multi-source data acquisition system is developed, incorporating electrical signals, mechanical characteristics, and thermal parameters, to enable holistic perception of the circuit breaker’s operating status. Feature extraction and information fusion are performed using methods such as normalization, noise suppression, and principal component analysis, thereby constructing a high-dimensional feature space to improve the sensitivity to vacuum level changes. An SVM model is then employed to intelligently classify vacuum states and assess degradation trends. The results demonstrate that the proposed method outperforms traditional vacuum detection techniques in terms of recognition accuracy, generalization ability, and real-time responsiveness.

Key words: high-voltage vacuum circuit breaker, vacuum monitoring, multi-source data fusion, SVM