Development of knowledge-based regulator for a PWR-type nuclear power plant

Abstract

In thic study, a rule-based fuzzy logic controller for a FUR nuclear power plant has been oeveloped in order to reguiate the power around z full power setpoint. In this artificial intelligence applicatlor knowledge acquisition wan performed through numerical simuiation using a validated linear model of the H.B. Robinson power plant and production rules were used for knowledge representation. For comparison purposes broken-line and S-shaped fuzzy sets were investigated and broken-line fuzzy sets were preferred. The regulator was implentecd on an IBM-compatible PC using the PASCAL language. The performance of the rule-based controller was compared to that of an optimal controller and was found to be better in the sense that the overshoots were less. Also, the eftect of noise in sensor data and variation in reactor parameters were investigated and their effect on the performance of the controller was found to be sufficiently robust.