A NEW DYNAMIC GENETIC SELECTION ALGORITHM: APPLICATION TO INDUCTION MACHINE IDENTIFICATION

Premature convergence is known as a serious failure mode for genetic algorithms (GAs). This paper presents a new dynamic selection based on power ranking by varying gradually the selection pressure versus generations, in order to maintain a trade-off between exploitation and exploration in genetic algorithm and avoid premature convergence. The proposed dynamic genetic selection algorithm's performance was proven by identifying an induction machine's (IM) parameters, both electrical and mechanical, using only the starting current and the corresponding phase voltage. A comparison is established between the proposed dynamic genetic selection algorithms with other genetic selections algorithms. The evaluation is carried out on IM's (1.5 kW) parameters estimation by measured data. The matching in the transient and in steady state of computed currents with the measured ones confirms the accuracy of the identified parameters. The experimental results obtained indicate the superiority of the proposed dynamic genetic selection algorithm versus the other algorithms in terms of computing time and convergence speed.