Modeling And Water Level Reconstruction Of Natural Circulation Steam Generator And Control Strategy Research

The safety of nuclear energy is the top priority in the development of nuclear energy.With the development of artificial intelligence,the nuclear power system is undergoing intelligent transformation step by step.The collection of data generated during the operation of nuclear power plant and the establishment of a data-driven prediction and analysis system can realize automatic control under some scenarios.The artificial intelligence can be used to evaluate the system status and assist operators to make appropriate decisions.As the complex equipment in the nuclear power system,the monitoring and control of the water level parameters of the steam generator is an important topic.To reconstruct the water level signal of the steam generator and study the control strategy is conducive to reducing the burden of operators,detecting sensor faults and ensuring the safe operation of nuclear facilities.Due to the difficulty in obtaining actual data,it is necessary to establish a model of the natural circulation steam generator.There are three means to model the steam generator,including ready-made simulation software,classical transfer function model and differential equations based on the conservation law.Because the simulation software to establish the model can comprehensively reflect the change of thermodynamic parameters,but not easy to do research on control,and the classical transfer function model limited to control research,this paper is concentrated on studying the dynamic mathematical model of the steam generator.The water level reconstruction technology and water level control strategy are studied based on the established steam generator model and the feasibility of using the reconstructed water level signal for automatic control is verified.Firstly,the thermal process of the natural cycle steam generator is analyzed,and the dynamic characteristics of the water level are studied based on E.Irving model.It is found that the false water level is an important feature of the dynamic characteristics of the steam generator.The steam generator is divided into modules,using the lumped parameter method and the homogeneous flow model,and combining the three factors that affect water level changes,considering the changes in the volume of bubbles under the water surface,and establishing a set of ordinary differential equations that can describe the operating behavior of the steam generator This model can not only reflect the false water level characteristics of the water level,but also the dynamic changes of various thermal parameters.The false water level characteristics of the model are verified with the E.Irving model.Then,the established dynamic model is based on the input of random step signals to stimulate the dynamic performance of the model,obtain the input and output data of the water level reconstruction model,and use BP neural network,NARX dynamic neural network,parameter optimization of support vector regression model to reconstruct water level.It is found that both the support vector regression model and BP neural network can be used to reconstruct the water level signal step by step.And they are suitable for fault detection of the water level sensor.NARX dynamic neural network can perform short-term rolling reconstruction of water level signal.Finally,in the building of the steam generator model on single PID control and cascade PID control and PID control based on genetic algorithm optimization simulation,explore the controller for the steam generator water level control.The research found that genetic algorithm can perform PID parameters better tuning,the cascade PID controller is beneficial to resist steam flow disturbances.Then,based on the cascade PID control system established,the signal reconstructed by NARX dynamic neural network is used to replace the real water level signal.And the research finds that the reconstructed water level signal can guarantee the normal operation of the automatic control system in short time.