Detection of Transformer Defects in Smart Environment Using Frequency Response Analysis and Artificial Neural Network Based on Data-Driven Systems

Data analysis is becoming increasingly important in modern industrial systems. Widespread use of sensors and smart meters has introduced an information layer to the electricity distribution network for data collection, storage, and analysis. Distribution power transformers have a crucial role in smart data-driven power networks, so diagnosing the faults that occurred in transformers is a vital issue. In this paper, a 1600kVA distribution power transformer was utilized to collect datasets by Frequency Response Analysis (FRA) method. Artificial Neural Network (ANN), as a powerful technique in pattern recognition, has been investigated to detect various mechanical and electrical defects. For extracting feature sets from the FRA database, a novel statistical indicator called Fitting Percentage (FP) has been applied to FRA signatures that led to obtaining the best accuracy in the fault classification procedure.