Controle preditivo baseado em modelo com conjunto de controle finito aplicado a conversores CC-CC

Abstract

The generation of energy through renewable and distributed sources has numerous advantages. However, due to intermittence and volatility, it is necessary that these systems be associated

with energy storage systems or connected to the electrical grid. This work proposes the applica- tion and performance comparison of the FCS-MPC control in face of the choice of different cost

functions applied to a scenario of Buck converter and to two different scenarios of Boost con- verter and Voltage Regulator - Battery Energy Storage System (VR-BESS). The first scenario

consists on the use of a classical cost function (voltage) and proposes the use of two others: mul- tivariable voltage and electrical current control; And indirect voltage control through a current

cost function. In this scenario there are no corrections for the delays caused by non-minimum phase (NMP). In the second scenario, a new minimum phase output is replaced on the voltage

cost function, both classical and multivariable. The converters are simulated in Matlab/Simu- link and the results obtained for the Buck converter demonstrate that the indirect control of the

output voltage reaches steady-state values very close to the desired value for voltage. In addi- tion, the multivariable cost function presents a balance between the other functions, reducing

the error values and keeping the average values closer to the desired ones, when compared to the scenarios where the control of the variables is done indirectly. For the Boost and VR-BESS converters, the results show that the use of the multivariable cost function alone does not solve the optimization problem caused by the non-minimum phase characteristic. However, when the system is indirectly controlled by current, or when operating with NMP corrections, the FCS-MPC is able to optimize the output with low steady-state errors. In addition, the use of the multivariable cost function with correction reduces steady-state oscillations and mean errors.