Data Mining of Online Diagnosed Waveforms for Probabilistic Condition Assessment of SF $_{6}$ Circuit Breakers

To establish a circuit-breaker (CB) online monitoring system, the development of an effective diagnosis framework is essential. The purpose of this online monitoring system is to accurately assess its condition while under operation, and to foresee any risk of failure. Consequently, it would be possible to plan for the required maintenance on the CB sufficiently ahead of a failure occurrence. To fulfill this, a framework is proposed to precisely assess the SF 6 CBs´ condition using its control circuit signals´ waveforms as easy-to-access, easy-to-measure, and noninvasive diagnosis parameters. The features found in these waveforms match the characteristics essential for diagnostic purposes, and could cover 50%-60% of CB failures. A data-mining process is employed to cluster the captured data against past-recorded data (in terms of faulty or healthy condition) of circuit breakers (CBs) diagnosis. To determine the probability distribution function of a CB condition cluster, a classifier is developed. In this model, the CB condition (based on its failure risk) is qualitatively classified into normal, alarm, and emergency states. In addition, to avoid any unnecessary maintenance activities during an alarm state, the condition of CBs in each state is quantified in terms of some probabilistic-based indices. The feasibility and applicability of the proposed methodology are verified using recorded data.