Application of artificial intelligence techniques in modeling and control of a nuclear power plant pressurizer system

In pressurized water reactor (PWR) nuclear power plants (NPPs) pressure control in the primary loops is fundamental for keeping the reactor in a safety condition and improve the generation process efficiency. The main component responsible for this task is the pressurizer. The pressurizer pressure control system (PPCS) utilizes heaters and spray valves to maintain the pressure within an operating band during steady state conditions, and limits the pressure changes during transient conditions. Relief and safety valves provide overpressure protection for the reactor coolant system (RCS) to ensure system integrity. Various protective reactor trips are generated if the system parameters exceed safe bounds. Historically, a proportional-integral-derivative (PID) controller is used in PWRs to keep the pressure in the set point, during those operation conditions. The purpose of this study is two-fold: first, to develop a pressurizer model based on artificial neural networks (ANNs); secondly, to develop fuzzy controllers for the PWR pressurizer modeled by the ANN and compare their performance with conventional ones. Data from a 2785 MWth Westinghouse 3-loop PWR simulator was used to test both the pressurizer ANN model and the fuzzy controllers. The simulation results show that the pressurizer ANN model responses agree reasonably well with those of the simulated power plant pressurizer, and that the fuzzy controllers have better performance compared with conventional ones.

► We model a nuclear power plant pressurizer system using artificial neural network.
► We develop fuzzy P, PI and PID controllers for the modeled pressurizer.
► The fuzzy P, PI and PID controllers were tuned by genetic algorithms.
► We compare the response of the fuzzy controllers with the conventional ones.