Identification et commande de la machine asynchrone par les techniques du softcomputing

Abstract

In this work a new philosophy for the off line identification of electrical and mechanical parameters of three phase asynchronous machine is presented. The novelty lies in introducing an approach based on the principle of causality for locating the best algorithm rather than the best value of objective function. When the stage of biographic search we have reported that the works concerning the identification are based on two classical criteria: values of objective function and execution time. Taking in consideration the dynamic change of machine parameters, according to the conditions of operation of the machine, the minimum obtained is only local. Hence, we conclude that the comparative study has located the best value but not the best algorithm. This has induced us to propose a new approach based on the principle of causality. In this context, the behaviour of optimisation techniques, belonging to four different classes, towards the deficiency or abundance of the information of the acquired quantities is studied. The method of the reference model has showed the superiority of the ant colony algorithm over the other used techniques. For the permanent regime, the ant colony algorithm is the greedier. The transient regime, where the information of the experimental quantities is richer, permits more fecund results. Indeed, the lower value of objective function is obtained with analytical techniques, but the behaviour of the ant colony algorithm is more flexible. Genetic algorithm has exhibited an adaptive behaviour towards the information of the acquired quantities. The obtained parameters are useful for simulation or the initialisation of the algorithm of on line adaptation of machine parameters. The characteristics of the indirect vector control are greatly dependent on the values of machine parameters. The simulation of the indirect vector control has showed that modified rotor field and electromagnetic torque are decoupled.