Enhanced situation awareness and decision making for an intelligent reconfigurable reactor power controller

A Learning Automata based intelligent reconfigurable controller has been adapted for use as a reactor power controller to achieve improved reactor temperature performance. The intelligent reconfigurable controller is capable of enforcing either a classical or an optimal reactor power controller based on control performance feedback. Four control performance evaluation measures: dynamically estimated average quadratic temperature error, power, rod reactivity and rod reactivity rate were developed to provide feedback to the control decision component of the intelligent reconfigurable controller. Fuzzy Logic and Neural Network controllers have been studied for inclusion in the bank of controllers that form the intermediate level of an enhanced intelligent reconfigurable reactor power controller (IRRPC). The increased number of alternatives available to the supervisory level of the IRRPC requires enhanced situation awareness. Additional performance measures have been designed and a method for synthesizing them into a single indication of the overall performance of the currently enforced reactor power controller has been conceptualized. Modification of the reward/penalty scheme implemented in the existing IRRPC to increase the quality of the supervisory level decision process has been studied. The logogen model of human memory (Morton, 1969) and individual controller design information could be used to allocate reward to the most appropriate controller. Methods for allocating supervisory level attention were also studied with the goal of maximizing learning rate