Contribution à la commande sans capteur mécanique de la machine synchrone à réluctance variable

This thesis focuses on variable speed synchronous machine drives, highlighting their advantages such as robustness and high reliability. Previous research has explored variable speed control of these drives, particularly for applications such as rail traction and marine propulsion.
Many problems require the estimation of the state of a system using an observer.
However, modelling and synthesising the observer becomes a difficult task for non-linear systems.
Synchronous Variable Reluctance Machines (SynRMs) have advantages, but their control can be complex due to model non-linearity and parameter fluctuations.
The research presented in this thesis focuses on the state estimation of non-linear systems represented by an Mras-Mpc controller which is proposed as the newest strategy, it combines all cascade structures into one control law to obtain the switching vector by minimising the predefined cost function.
After an introduction to predictive control (MPC) and the optimisation criterion, the Mras is presented in detail while developing its stability in order to propose two approaches with two different cost functions combining the two observers into a single one to consolidate the quantities to be estimated. The first approach consists of feeding the input quantities of the predictive control model with the output quantities of the model's reference adaptive system, in order to achieve the optimum speed.
In the second approach, we propose a controller that estimates the speed by replacing traditional PI controllers with a predictive control model that calculates a new speed estimate based on the MRAS scheme using only stator current and voltage measurements and then optimises the vector voltage that minimises the current error.