An ANFIS Based Comprehensive Correlation Between Diagnostic and Destructive Parameters of Transformer’s Paper Insulation

摘要

In this work, a fine correlation among destructive and diagnostic parameters of transformer’s paper insulation is presented.The degree of polymerization (DP) directly assesses the condition of the paper insulation strength and is measured throughdestructive procedure. Generally, 2-FAL, CO_2and CO are the aging products of paper decomposition and referred as diagnosticparameters. An Adaptive Neuro Fuzzy Inference System (ANFIS) is proposed and developed to estimate the value of DPas the function of amount of diagnostic parameters. The proposed system has an advantage of diagnosing the health of solidinsulation without performing destructive tests. The diagnostic parameters are taken as inputs to the system to determine thevalue of DP. The system uses 630 data points for training the ANFIS model and follows a ten-fold cross-validation approach.The average validation error has been determined to be. 0.0029. Further, the model’s performance has been assessed usingexperimental data. The optimal ANFIS model has been achieved by suitably selecting the number and type of membershipfunction. The estimated value of DP has been found to conform to the experimental measurements for every case under test.The performance of this model has also been compared with a fuzzy inference system (FIS) model in the reported literature.The comparison shows that the ANFIS model determines the DP values with a greater degree of accuracy than the FIS model.