CONTRIBUTION TO THE NEURO-FUZZY CONTROL OF VOLTAGE INDUCTION MACHINE INVERTER FED

This article presents a modern strategy to improve the performance of electrical drives control. The approach used consists of combining both fuzzy and neural control algorithms to make better use of the knowledge resulting from both techniques. The numerical simulation tests carried have shown that the techniques robustness are very good and the obtained results validate the proposed neuro-fuzzy control diagram taking into account the induction motor model complexity due to high non linearity presence. The advance of power electronics and industrial computing technologies have allowed induction machine to be widely used in variable speed applications. However, the requirements of better performances and the complex model of the induction motor require powerful control algorithms. The estimators enabling to train the non measurable parameters present a good identification of the machine. The recent artificial intelligent algorithms have helped in solving most of non linear control problems. Amongst these techniques, neuro-fuzzy approaches are very interesting not only due to their intrinseque non linear proprieties, but also due to the learning process; which offers them certain ability for adaptative control. The neural networks are mainly characterised by their fast operation and the great ability to approximate non linear dynamics. The fuzzy logic is a technique used to deal with imprecise knowledge based on analogues linguistics to those used in communication. Hence, in a perspective to be able to deal with all types of information came the idea to combine both approaches in order to design neuro-fuzzy control algorithms. This consists of linking known knowledge from fuzzy techniques with that knowledge learnt from neural techniques. In this article, contribution on the study and analysis of voltage fed inverter induction motor robust control is proposed. We first present a squirrel cage induction motor model to evaluate the model complexity of the machine. We then briefly present reasoning modes related both to fuzzy sets and to neural networks, and also we deal with the adaptative neuro-fuzzy controllers (ANFIS). The obtained simulation results and their interpretations will enable us to conclude that the proposed neuro-fuzzy control applied to inverter fed induction motor for speed variation is a powerful and effective tool that give the induction motor performances the same merits as those obtained from a direct current machines (1), (2).