An optimized, adaptative, reduced-order flux observer

The authors deal with a new, robust and adaptative state observer for the torque control of asynchronous machines used in railways. Vector control requires an accurate estimation of the flux vector through a model. The first problem is the robustness of the observer since some parameters of the model vary with temperature and speed and some noise is always present on measurements. The second problem is real time, since all computations have to be performed in a very short time (less than 100 μs for a 200 Hz machine). The authors have developed a reduced-order model that can observe a linear combination of the rotor flux vector and the stator current vector. The two coefficients of this combination give two degrees of freedom to optimize the robustness of the observer with respect to the errors on the most critical parameters and measures i.e. errors on rotor resistance, stator resistance and stator voltage