Optimizing Nuclear Reactor Operation Using Soft Computing Techniques

The strict safety regulations for nuclear reactor control make it difficult to implement new control techniques such as fuzzy logic control (FLC). FLC however, can provide very desirable advantages over classical control, like robustness, adaptation and the capability to include human experience into the controller. Simple fuzzy logic controllers have been implemented for a few nuclear research reactors, among which the Massachusetts Institute of Technology (MIT) research reactor [1] in 1988 and the first Belgian reactor (BR1) [2] in 1998, though only on a temporal basis. The work presented here is a continuation of earlier research on adaptive fuzzy logic controllers for nuclear reactors at the SCK•CEN [2, 3, 4] and [5] (pp 65–82). A series of simulated experiments has been carried out using adaptive FLC, genetic algorithms (GAs) and neural networks (NNs) to find out which strategies are most promising for further research and future application in nuclear reactor control. Hopefully this contribution will lead to more research on advanced FLC in this domain and finally to an optimised and intrinsically safe control strategy.