Optimization of model predictive controller parameters based on Imperialist Competitive algorithm

Model predictive control is used as an effective tool for the control of complex multivariable problems with complicated constraints of input/output in industrial processes. In design of predictive controllers, the right choice of control parameters is very important, since best output results are achieved with appropriate choice of these parameters. In this paper, a new method to obtain the parameters of the model predictive controller is presented based on Imperialist Competitive Algorithm, Thus, at first, three unknown parameters, namely the prediction horizon and control horizon and the sampling time, are entered in Imperialist Competitive Algorithm as an array of countries. After these values have been calculated offline, the controller will begin its work with these starting values. In this paper, this type of controller is implemented on a four-tank CSTR system and its nonlinear responses are plotted. The results show that the proposed model predictive controller improves the system speed.