Automatic identification, clustering and reporting of recurrent faults in electric distribution feeders

Latent power line conditions, such as vegetation intrusion and apparatus that have failed or are in the process of failing can cause recurring fault events. Many such conditions are influenced by other factors such as wind and moisture, and therefore cause fault events only intermittently. These conditions are difficult to detect and locate with conventional technologies. Fault current and arcing from recurrent faults can cause further damage to already weak apparatus, ultimately leading to a catastrophic failure, at which time there may be more consequential damage to apparatus, including burned-down lines. For more than a decade, Texas A&M researchers have instrumented dozens of feeders using sensitive, high-fidelity waveform recorders to document numerous apparatus failure conditions, including multiple instances in which failing apparatus and other factors have caused recurring faults and momentary interruptions, spread over significant periods of time, without causing sustained outages. A series of related faults can escape notice when an unmonitored, pole-mount recloser is the interrupting device, unless customers report momentary interruptions, and experience indicates this often does not happen. Even if customers report individual momentary interruptions, the utility may not recognize that the multiple interruptions are related to each other, particularly if time intervals between operations are sufficiently long for operator memories to fade. Awareness of recurrent fault conditions would enable utilities to make timely, proactive repairs, thus avoiding additional faults and interruptions, as well as potentially preventing more catastrophic failures (e.g., equipment damage, downed conductors, fires). This paper describes an on-line, automated method to mine, cluster and report recurrent faults to utility operators in a near real-time fashion. This paper also documents one of multiple real-world examples where the methodology described in this paper was successfully used by utilities to locate and fix problematic components and prevent further faults.