Power transformers fault diagnosis using AI techniques

Abstract

Artificial Intelligence (AI) is a novel branch in science and engineering. AI techniques constitute the most cutting-edge method in Power Transformers Fault Diagnosis. When a transformer fails, some gases are produced and dissolved in the insulating oil, and Gas Chromatography detects them. It is a technique of separation, identification, and quantification of mixtures of gases. The analysis of these gases helps to identify the incipient fault types. The conventional method widely adopted is the Dissolved Gas Analysis (DGA). All the conventional methods have limitations because they cannot analyze all faults accurately. It usually happens when more than one fault occurs in a transformer or when the concentration of gases is near the threshold. To deal with this problem and to improve the reliability and the accuracy of fault diagnosis, various Artificial Intelligence techniques are proposed. In this paper, three AI methods are employed, a Fuzzy Inference System (FIS), an Artificial Neural Network (ANN), and an Adaptive Neuro-Fuzzy Inference System (ANFIS) in order to enhance the accuracy of conventional Rogers Ratio method, that evaluates the DGA. All these techniques are simulated using MATLAB software. Real samples of dissolved gases that have been generated in failure transformers and have been obtained from the HEDNO (Hellenic Electricity Distribution Network Operator ) are used. Finally, a comparison of the FIS, ANN, ANFIS, and the conventional Rogers Ratio method is presented.