engleski

Inference Model for Automatic Diagnosis of Power Transformers Based on Dissolved Gas Analysis

Reliability of electricity delivery is highly related to condition of power system equipment, like power transformers. Lot of real-time data are being collected using existing solutions for transformer condition monitoring, but interpretation of these data and condition assessment are challenging. Therefore, R&D efforts are directed toward solutions for automatic diagnosis inference, but experience shows that besides diagnosis (condition), users need also diagnosis probability and tools for validation of obtained diagnosis to be able to make reliable decisions. This paper presents a new model of automatic diagnosis for power transformers on basis of dissolved gas analysis. The model uses eight interpretation methods as voters, and developed voting algorithm estimates probability of each diagnosis and proposes final diagnosis. Using variable diagnosis resolution of model it is possible to validate obtained diagnosis. Besides usual tests on sets of DGA data from different transformers, the model in this research was subjected to tests on historical data from transformers to verify consistency of diagnostic model on DGA data set from certain transformer. All tests showed promising results.

Fault diagnosis; Inference Mechanisms; Power Transformers

nije evidentirano