Application of Adaptive Neuro-Fuzzy Inference System for Transformer Fault Diagnosis

Transformer is one of the key components in a power system. It can experience thermal and electrical stresses during its operation. The mineral oil and the insulation paper used in transformer can undergo chemical changes under these stresses and gases are generated. These gases dissolve in mineral oil. The gases dissolved in oil are extracted in the laboratory using gas chromatograph. The gas content and the type of gas can reveal the nature of fault present in the Transformer. Natures of faults considered in this study are partial discharges, discharges of low energy, discharges of high energy, and thermal faults. Based on the dissolved gases present in the transformer oil several diagnostic methods such as Doernburg method, Roger's ratio method etc are commonly used for fault diagnosis. In some cases, all these methods are not able to diagnose the fault. Hence, there is a need to develop new methods. In this paper an Adaptive Neuro-Fuzzy inference system is applied for diagnosis of transformer faults.