IOT based classification of transformer faults using emerging techniques of E-nose and ANFIS

E-Nose finds its use in a wide range of applications such as quality assessment in food processing to toxic gas identification in chemical industry either in the offline or online mode. Their usage can be extended to transformer condition monitoring in the online mode. Considering the importance of transformers in power system and the impact it could create if faults in them are unidentified or left unattended, their functioning should be monitored on a real time basis. This work, describes the realization of a prospective E-Nose for online transformer incipient fault identification. The resistive gas sensor array has been simulated in real time using variable resistances forming one arm of a Wheatstone bridges. Separate variable resistances have been calibrated using characteristics of different fault gas sensors. The sensor array of the E-Nose helps to identify the transformer fault gases resulting from an incipient fault condition at the nascent stage itself and prompts for the necessary corrective action well before a catastrophic situation arises. Furthermore, ANFIS model of the Duval’s Triangle (DT) method have been developed to facilitate the online classification of incipient faults. The ANFIS models of other popularly used incipient fault interpretation methods, reported in earlier works, have also been used for a comparative analysis on their diagnostic capabilities. The developed model has been tested using the fault cases of IEC-TC10 fault database and the results thus obtained have been found to be very promising.