Multivariable Non-minimum State Space Predictive Control Based on the Disturbance Observer

To suppress the effect of lumped disturbances (i. e., external and internal disturbances caused by model mismatch and coupling among variables), the existing methods are usually based on feedback control and feedforward compensation, which cannot guarantee the optimal output of the system. A disturbance-observer-based multivariable non-minimum state space predictive control (D-MNMSSPC) method is proposed in this study. In this method, first, lumped disturbances are estimated by the disturbance observer. Then, the estimated disturbance and output variables are simultaneously introduced into the state variables to form a composite multivariable non-minimum state space prediction model, which can avoid the design of a state observer and reduce the burden of design. Furthermore, the disturbance is directly involved in rolling optimization of predictive control to obtain the optimal control performance of the system. The simulation results verify the validity of the D-MNMSSPC method.