A multi-hypothesis sequential probability test for partial discharges localization in power transformers

We present a new method to accurately locate partial discharge by using a sequential fault detection and identification (FDI) algorithm for detecting a bias fault in the measurements of partial discharge in transformer insulation system using acoustic signals. In this paper, a novel technique is proposed to identify the possibility of measurement errors generated from acoustic emission sensors during partial discharge localization inside a transformer tank. The technique probabilistically detects and identifies possible bias on the sensors´ measurement. This bias is possibly caused by sensor´s fault, sensor´s aging, or proximity of the PD to a certain sensor in comparison to other sensors. Correct detection of sensors´ measurement bias enables high accuracy PD localization as will be demonstrated in this paper. The accuracy and convergence characteristics of the proposed algorithm are verified in a simulation environment. This study is tremendously important for scheduling and starting maintenance/repair actions cost and time efficiently or to perform a risk analysis.