MULTI-CLASS CLASSIFICATION OF TRANSFORMER FAULT FROM DISSOLVED GAS ANALYSIS

Abstract

Power transformers are a critical part of the power system. The early-stage fault detection of Power transformer is essential for the protection and prevention of further technical and financial losses. Dissolved Gas Analysis (DGA) is a commonly used diagnosis tool for keeping track of transformer status,but the existing DGA methods are based on expertise and personal experience,so their reliability can never be guaranteed,which can lead to unreliable diagnosis.Nowadays,artificial intelligence-based techniques are widely used to enhance DGA fault detection accuracy.Here,the machine learning model tries to overcome the deficiency of conventional DGA by converting DGA into a pattern recognition problem by establishing a connection between gas concentration and incipient faults. In this thesis, a new deep learning based Bidirectional Long short-term Memory(BLSTM) is introduced for multi-class classification of transformer fault and the model’s performance is compared with that of other deep learning models.