Linear least squares and genetic algorithm based induction motor identification and classification by performing a locked rotor test at variable frequency

Abstract EN

Induction motors (IM) has different rotor types : single cage, double cage or deep bar cage.

In major applications, any induction motor is represented by a single cage model which is not suitable to characterize the dynamic behavior of all IMs.

In fact, IM robust control depends mainly on the IM model and on the accuracy of the parameters identification.

This paper presents the modeling of the single cage, double cage and deep bar induction motor by using a Model of Invariant Parameters.

An offline IM identification based upon steady state electric quantities (voltage, stator current and active power) is developed by performing a locked rotor test for different frequencies.

The linear Least Squares Technique (LST) and the Genetic Algorithm (GA) are used.

The sensitivity to measurement errors is evaluated for each method.

Thus the GA identification is used to classify the motor according to its rotor type.

The approach is simulated using Matlab 7.1.0, and applied to twenty IM with known manufacturer parameters and rotor types.

These parameters are used to establish a simulation database so as to validate the classification and identification procedure.

GA is judged more efficient since it persists and converges for a measurement noise of 5 %.