Recursive parameter identification of an induction machine using a non-linear programming method

Concerns the parameter identification of asynchronous machines. Extended Kalman filter is an optimal recursive estimation algorithm which allows combined state and parameter identification. In return, a good a priori knowledge on the parameters and on the stochastic properties of the noise disturbing both state and output are needed to initialize the nonlinear filter. Other methods use a linearized parametric model. Drawback of this approach is that the validity of this approximate model depends on assumptions about speed drive, which must remain constant. The estimation method described in this paper differs from other approaches in the following way: it is a recursive parameter identification method, based on a nonlinear continuous time model of the machine. A continuous time nonlinear dynamic model of the induction motor is presented. An output model where derivatives signals are not required is deduced from the reinitialized partial moment technique. After a description of the identification method, some considerations about the sensitivity functions are presented. Experimental results are also presented.