1978 年 20 巻 6 号 p. 433-439
An optimal control problem of large-scale reactors composed of a number of coupled subcores is solved by a hierarchical two-level scheme based on the Decomposition Principle, On the infemal level, local optimums are explored by optimization method applied to one-point reactor models, with the interaction variables among the subcores tentatively set at arbitrary values. On the supremal level, the interaction variables are searched by an iterative gradient-type procedure that aims at equalizing the local optimums with the overall optimum.
Compared with straightforward optimization using the Generalized Newton Method, the twolevel scheme has the advantage of reducing both computer time and computer storage needed for the optimization, as demonstrated by examples of actual computation covering coupled core reactors with either two or six cores. This merit further tends to as the number of subsystems.