A multi-criterion hybrid expert model for health assessment and maintenance of power transformer insulation

In the present paper, a novel multi-criterion-based adaptive neuro-fuzzy inference system (ANFIS) is proposed to determine the overall health index (OHI) of power transformer insulation. The proposed model concentrates on 12 diagnostic attributes related to dissolved gases, oil and paper insulation. On-field data from 150 transformers collected from Himachal Pradesh State Electricity Board, India, are utilised. Initially, the collected data are normalised and correlation among the attributes is established using a multi-criterion approach (MCA) according to IEEE standards C57.104-2019 and C57.106-2015. Closely related attributes are grouped into three grades. The obtained three grades are applied as input to the ANFIS model and the transformer health index (HI) is output. Among the collected data, 80% are used for training and 20% are used for testing. The performance of the proposed model is evaluated using two crucial error metrics: root mean square error (RMSE) and correlation coefficient (R ² ). Moreover, the proposed model is validated using 110 data samples collected from the literature and compared with existing expert models, achieving an accuracy of 98.18%. The integration of the MCA with the ANFIS model overcomes the shortcomings of the previous ANFIS model, requiring huge training time, larger rule formation and a high computational burden on the network. The proposed MCA-based ANFIS model is easy to implement and accurately predicts the health indices. The present work is beneficial for diagnostic experts looking to take appropriate remedial actions on the current health status of transformer insulation.