Fuzzy systems for simulation and prediction of the residual life of insulating materials for electrical machines windings

Nowadays, the perspective diagnostic methods not only evaluate the current insulation condition, but also forecast the certain prognoses of the life of insulating material under the specified operational conditions. This contribution undertakes both simulation and prediction of the residual life of the Relanex insulating material for electrical rotating machine windings. In case of simulation, the current condition of insulating material based on the input data is determined. In case of prediction, the insulation condition in the future time k+1 is predicted based on the input quantities in the times k, k-1, k-2, etc. Prediction means forecasting an output quantity in future based on the N values of this quantity and/or the other quantities measured in the past. By means of the diagnostic quantities Ba (the quantity is proportional to the effective activation energy of polarization effects), Bv (the quantity is proportional to the effective activation energy of conductive effects) and Uk (the quantity is proportional to critical voltage) and by using a fuzzy method in the form of fuzzy models and predictors, we can determine breakdown voltage. This approach determines either the current condition of insulating material, taking into account electric strength, or the life prognosis of this material. To simulate and predict a magnitude of breakdown voltage, the Sugeno type fuzzy system was applied, especially for easy setting input parameters. According to the input quantity matrix (i.e., Ba, Bv, and Uk) measured at the Relanex insulating material samples, and according to a consecutive training phase, it is possible to simulate and predict the values of breakdown voltage Ups. The values can be compared to the real measured values of breakdown voltage Up. The evaluation of the fuzzy system quality is performed on the basis of relative and absolute errors.