Optimization of the number of fuzzy rules towards a better temperature control of nuclear reactors

A nuclear power plant must be capable to face the fine evolution of the energy demand. Several parameters of the reactor are mainly concerned by operations of follow-up of load. We propose a fuzzy regulation of the average temperature of the primary circuit. The base is set to few rules at the beginning and an optimization algorithm fills the base. The principle consists by adding rules if the output is not satisfactory. The rules are added according to the areas where the control is fairly good. It is also possible to end the optimization and set the number of rules without filling the entire table. The possibility of free filling the base aims at refining the regulation and compensating the effects of oscillations due to the instability of the system. The results are satisfactory.