On-Line Health Condition Monitoring of Power Connectors Focused on Predictive Maintenance

Abstract

Electrical power connectors are critical points of electrical networks. Failure in high-voltage connectors may result in major power outages, safety risks and important economic consequences. Therefore, there is an imperious need to tackle such issue by developing suitable on-line condition monitoring strategies to minimize the aforementioned problems and to ease the application of predictive maintenance tasks. This work develops an on-line condition monitoring method to predict early failures in power connectors from data acquired on-line (electric current and voltage drop across the connector, and temperature) to determine the instantaneous value of the connector resistance, since it is used as a signature or indicator of its health condition. The proposed approach combines a parametric degradation model of the resistance of the connector, whose parameters are identified by means of the Markov chain Monte Carlo stochastic method, which also provides the confidence intervals of the electrical resistance. This fast approach allows an on-line diagnosis of the health condition of the connector, anticipating its failure and thus, easing the application of predictive maintenance plans. Laboratory results emulating the ageing conditions of the connectors prove the suitability and feasibility of the proposed approach, which could be applied to other power products and apparatus.