MULTIOBJETIVE FINITE CONTROL SET MODEL PREDICTIVE TORQUE AND STATOR FLUX CONTROL OF AN INDUCTION MACHINE
Abstract
Resumen extenso, ver multimedia o impreso During the last decades, basically two control strategies for electrical drives have
dominated high-performance industrial applications: field-oriented control (FOC) and
direct torque control (DTC). Nowadays, these control strategies are implemented on digital
platforms. Digital signal processors (DSP) allows high flexibility, the integration of more
functionality, and the implementation of more complex control schemes. Due to the
development of powerful and fast microprocessors, increasing attention has been dedicated
to the use of model predictive control (MPC) in power electronics. The first ideas about
this strategy applied to power converters started in the 1980s. The main concept is based
on the calculation of the future system behavior to compute optimal actuation variables.
Due to the wide range ofMPC methods, the MPC techniques applied to power electronics
have been classified into two main categories: classical MPC and finite control set MPC
(FCS-MPC). In the first type, the control variable is the converter output voltage, in the
form of a continuous duty cycle, while an open-loop receding horizon optimization problem
is solved at every sampling step to calculate the best actuation. This actuation is applied
usually using pulsewidth modulation (PWM) or space vector modulation (SVM). The second
type, uses the inherent discrete nature of the power converter to solve the optimization
problem using a single cost function. Here, the input is restricted to a finite set of discrete
values. The discrete system model is evaluated for every possible actuation sequence and
then compared with the reference in order to select the best voltage vector.
The research done up to now has revealed that a key issue in FCS-MPC implementations
is the selection of the weighting factors used in the cost function. Weighting factors are used
to give more importance to one or another variable and to normalize the different control
objectives. These scalar factors are parameters to adjust, and its selection is an important
task because it is more complex than the tuning of proportional-integral (PI) coefficients or
hysteresis bands of traditional controllers. Several methods using offline and online search
procedures have been implemented at the present state of the art, but they are strongly
dependent on the system parameters and they are formulated for two control objectives in
a specific application only. When more objectives are considered, the weighting factors are
usually obtained using trial and error procedures and running time-consuming simulations.
The use of a single cost function to solve the optimization problem at each sampling
time is not the only possible alternative. The possibility of the use of a different optimizer
is the origin of this work. Different simple multiobjective optimization methods in order
to eliminate the requirement of weighting factors in the predictive torque and flux control (PTC) scheme are presented. The optimization problem is solved using a multiobjective
approach, giving rise to a multiobjective predictive torque and flux control. The scheme is
then applied to an induction machine drive fed by a commercial two-level voltage source
inverter (2L-VSI).