PURPOSE: To attain the forecasting control of a process having integration characteristics and to improve the operation efficiency of the process by finding out a manipulated variable based upon the forecasting of disturbance in the process.

CONSTITUTION: A disturbance forecasting model 21 finds out a disturbance forecasting model from disturbance (d) and steam pressure (y) A forecasting model setting part 22 finds out a forecasting model from the relation between an input to the process 10 and an output from the process 10. On the other hand, a normative locus setting part 23 finds out a normatively set locus based upon a set value inputted by a setter. A control part 24 finds out the manipulated variable (u) of the process 10, i.e., the quantity of carbon feed to a boiler, based upon an output signal from the setting part 23. In order to obtain a control result matched with the locus, it is necessary to make the normative value coincide with the forecasting model of the process at a certain time. The optimum control of the process 10 can be executed by repeating said algorithm.