Adaptive Nuclear Reactor Control Based on Optimal Low-Order Linear Models

The problem of optimally controlling the power level changes of a nuclear reactor for a wide variety of cost functions is considered. Admissible performance indexes are arbitrary functions of the power level and its derivative as well as of the reactivity input and its derivative. The proposed method is based on the use of optimal second-order models, the parameters of which are computed on-line to produce piece-wise approximations to the response of the plant. While the controller parameters, which give optimal performance for the model, result in only a suboptimal system response, the deviations between the two converge rapidly to zero. The proposed technique is illustrated by computing the responses of a power reactor for integral quadratic cost functions, as well as for least pth and minimax criteria. The results indicate that near-optimal performance may be achieved under conditions encountered in practice.