Partial Discharge Monitoring Data: Statistical Processing to Assessment Transformer Condition

The paper dwells upon developing a method for assessing the condition of a high-voltage electric-furnace transformer. The method is based on probabilistic processing of partial discharge (PD) monitoring data. The paper summarizes monitoring systems as implemented for the transformers of electric steelmaking units at iron and steel works. Statistical processing of online readings is found to be of benefit. The article further substantiates the method of means as the most effective mathematical method for estimate random sample differences. The authors hereof discuss the basic theory and present the structure of an algorithm capable of detecting changes in the transformer condition. Emphasis is made on presenting the statistical processing output of PD diagnostic parameters represented as time-stamped samples. Means and standard deviations of PD power and amplitude in ladle-furnace transformer phases are analyzed on pre-repair data. This results in diagnosing a progressing failure while also detecting where it has likely occurred. Repairs taken are detailed. The paper concludes with statistical processing of PD parameters as sampled over a similarly long post-repair timeframe. The finding is that the source of increased discharging has been eliminated, and that the PD parameters are within acceptable limits. This proves the PD intensity and amplitude analysis method efficient; it also proves the developed condition diagnosis algorithm feasible as implemented in online HV transformer monitoring software.