THE USE OF REFERENCE OBJECTIVES FOR SELECTING POLYOPTIMAL CONTROL IN A MULTISTAGE PROCESS

The paper is devoted to an algorithm for selecting polyoptimal control for multistage processes. The processes are characterized by the possibility of distinguishing the time subsequent parts which differ in mathematical description i.e. state equations, control constraints, performance indices. The problem of selecting polyoptimal control is solved as follows: at each stage of the multistage process we have many performance indices; they constitute a base for a new one - the stage performance index. For each stage a local optimization problem is solved with the corresponding stage performance index and with a state equation for this stage. These local optimizations are governed by a global optimization problem which consists in coordinating the solutions for individual stages. Thus we obtain one polyoptimal solution. By modyfying the problem we can get other polyoptimal solutions and choose one considered the best. The method is based on the decomposition principle.