Service life prediction method of cable insulation faults based on multiple kernel learning model

Predicting the insulation defect status of power cables is challenging, making it difficult to accurately assess the cable's operational condition. This paper presents a service life predictive model of cable defects utilizing a multiple kernel learning model. The approach includes extracting features from multi-parameter data, fusing these features, and predicting defect service life. Initially, aggregate features are derived from cable equipment's multi-parameter data. Next, the multiple kernel learning optimization algorithm determines kernel weights for different data perspectives. Lastly, the SVM algorithm projects these weights into the target space to develop a service life predictive model of cable insulation failure. The results of experiment and case studies show that this method can effectively integrates diverse cable detection data, and tracks the progression of insulation faults accurately, maintains robustness in evaluations, and achieves high accuracy. It is suitable for predicting insulation defect trends and assessing insulation conditions in different environments.