An algorithm for the selection of a method for the modelling of direct-on-line starting of cage induction motors from finite supplies

The objective of the study was the derivation of an algorithm for the selection of a numerical simulation method for the direct-on-line starting of induction motors from power supplies whose voltage and frequency may vary during the simulation period. A major part of the work consisted of an evaluation of methods for treating the variation of motor parameters under different conditions and the relative effect of these variations on predicted output. Two new methods were introduced for predicting the variation of leakage reactance with current and a new method was developed for deriving the rotor parameter variation with slip for bars of arbitrary, but known, cross-section. An existing method was modified to derive model parameters from manufacturer's quoted performance data. The results are given of an investigation into the effect on the derived parameters and consequently on the predicted performance, of allowing the quoted performance data to vary within the tolerances permitted by Australian Standard 1359. A set of descriptive measures for simulation output were defined. This enabled the motor's starting performance to be quantified in terms of a small number of variables. The factorial method of experimental design was used to calculate relative coefficients of performance. These related the effect of changes in program input data to the resultant variation in program output and thus allowed numerical measures to be placed on the significance of data uncertainty and model complexity as they affected the simulated performance. PC-based data acquisition and processing techniques were used to take measurements from two laboratory motors which confirmed the results of numerical simulation work. The broad general conclusion of the thesis was that in most situations, the total system data needs to be included in the model if the uncertainty is to be improved beyond that obtainable with simple non-recursive calculations. An expert system shell was used to present an algorithm for the selection of a method of modelling appropriate to the quality of the data, the type of system, the purpose of the study and the availability of computational tools.