Decomposition of a Determinist Flux Observer for the Induction Machine: Cartesian and Reduced Order Structures 1
In order to precisely control the torque of an induction actuator, its flux must also be controlled. Unfortunately, its direct measure is tricky and relatively expensive. It is rebuilt through current and/or voltage measures.
There are two methods for flux reconstruction. The first one, estimators, uses an analytical model of the machine connecting the flux and measurable variables [VER 88]. This reconstruction is sensitive to disruptions and model errors. To offset this lack of precision, we must add complex control techniques [WAN 97]. Another solution is to use closed loop models, called observers [VER 88]. The loop gain then minimizes the estimation error. We can note that there are deterministic and stochastic strengths of observers [PIE 00] (Chapter 4). The second considers measurement and calculation of noises (e.g. Kalman filter). But this advantage is offset by a very delicate determination of gains and a very calculation time-intensive implementation [DU 95].
Two specific structures of flux observers are presented in this chapter to facilitate the real-time implementation of the estimation algorithm. The first structure, the Cartesian observer [BOU 95a], proposes a breakdown into two sub-observers based on the axes of the two-phase reference frame. These coupled subsystems facilitate the synthesis of observation gains and lead to ...