| 摘要 |
PROBLEM TO BE SOLVED: To shorten the processing time of an optimization calculation, and to suppress a deterioration in the calculation accuracy in model prediction control. SOLUTION: A model prediction control method is provided for determining the optimal manipulated variables by applying numerical differentiation to each of manipulated variables using the dynamic model of a control object to find out a differentiated value, by moving evaluation points based on a Jacobian matrix with the differentiated values as elements, and by evaluating optimization for each moved evaluation point, wherein an evaluation saving flag is turned on (S14) when the differentiated values are compared with a predetermined value (S13) and the differentiated value is smaller than the predetermined value. In the next Jacobian matrix preparation process, the differentiated value is set to 0, and the numerical differentiation can be saved. In this case, when the numerical differentiation step with the differentiated value as 0 ends, the evaluation saving flag is turned on (S18). In the next numerical differentiation step, numerical differentiation is executed, so that calculation accuracy can be improved. COPYRIGHT: (C)2006,JPO&NCIPI
|
