Improved Gas Ratios Models for DGA Interpretation Using Artificial Neural Networks

Abstract—Dissolved gas-in-oil analysis (DGA) has proved to be a valuable and reliable diagnostic technique for the detection of incipient fault conditions within liquid-immersed transformers this reliability comes from the effectiveness of interpretation means. Ratio methods including (IEC and Roger) is one of these interpretation means that fail often to provide diagnosis because the codes combination of certain cases full outside standard specific codes, due to this problem a back-propagation neural network is applied and verified to overcome the drawbacks of conventional methods. The proposed neural network model was verified using practical data collected from transformers in the Setif region, one of five areas where 20 % of the power transformers of Algeria are located and simulated. The ANN simulation results are presented and compared with the conventional IEC and Rogers’s detection