This work compares the use of direct torque and flux control (DTFC) and model predictive control (MPC) for induction motor (IM) control. These two strategies are fundamentally different in operation since (i) DTFC decides the current control action based on a switching table constructed using a simplified model of the IM, whereas (ii)MPC decides the current control action by on-line minimization of a cost function that uses the available inverter output voltages as optimization variables. Emphasis is given in this work to the reconfiguration of the control action after voltage source inverter faults. We assume that the fault can be suitably detected and isolated and that the inverter can be reconfigured after the specific fault to continue operation, albeit with a reduced set of achievable output vectors. Based on this reduced set of vectors, we propose to reconfigure the induction motor control algorithm by (i) instructing DTFC to use a reconfigured switching table or (ii) providing the reduced set of inverter vectors as the reconfigured constraint set of optimization variables for MPC. Simulation results show that MPC considerably outperforms DTFC at a modest increment of computational cost. Moreover, this increment is less pronounced under fault since the number of optimization variables is reduced.
2010 Conference on Control and Fault Tolerant Systems (SysTol'10). Proceedings of the 2010 Conference on Control and Fault Tolerant Systems (Nice, France 6-8 October, 2010) p. 759-764