Modeling And Control Strategy Research Of Helical Tube Once-Through Steam Generator

With the rapid development of modern control theory and artificial intelligence technology,advanced control algorithms including optimal control are gradually being applied to the corresponding control systems of nuclear power plants.Once-through steam generator is one of the key heat exchange equipment in integrated pressurized water reactor,because its outlet steam parameters are easily disturbed by external factors,which will directly affect the safe and stable operation of the secondary circuit and even the entire nuclear power plant.The research on mathematical modeling and pressure control strategy of once-through steam generator is very important,which will help reduce the operating burden of operators and ensure the normal operation of related equipment and systems.In this paper,taking the helical tube once-through steam generator as the research object,a lumped parameter model is established,and the pressure control system is optimized and improved based on predictive control algorithm and deep reinforcement learning.In this paper,the physical structure and working mechanism of the equipment are deeply analyzed after fully investigating the domestic and foreign research status in the field of once-through steam generator modeling and pressure control,as well as the application of advanced control algorithms in nuclear power systems.According to the different thermal states of the secondary side working fluid,the heat transfer tube is divided into several control bodies,and the input and output parameters of the model to be established are completely sorted out through the assumed conditions and the selected modeling method.Based on the three conservation laws,the differential equations are deduced for the primary and secondary sides of the steam generator and the metal tube wall respectively,and the nonlinear lumped parameter mathematical model of the once-through steam generator is established.According to the corresponding design inlet parameters,the steady-state output value obtained by the equation group at full power level is basically consistent with the actual operating data,and the relative error is less than 1%,which shows that the steady-state accuracy of the model is high.At the same time,a single disturbance is set for the input variable of the model,and the response curve of the output variable is obtained.The results show that the dynamic change process is also consistent with the actual operating characteristics.The above model is linearized at the system equilibrium point to obtain the state space at the 100% steady-state power point,combined with the PI controller and the positive feedback signal of the steam flow,the design of the pressure control system is completed,and carry out preliminary tests to prove the effectiveness of the control system under variable working conditions,and follow-up research work can be carried out on this basis.Aiming at the problem that conventional PI controller is difficult to achieve better control effects,the two algorithms of model predictive control and deep reinforcement learning are applied to the design of controllers respectively to explore the method with the best control performance under various working conditions,and comprehensively compared the performance indicators of the four controllers,the simulation results show that the model predictive controller and the deep reinforcement learning controller have different degrees of improvement in the control effect,and the deep reinforcement learning controller performs better in most working conditions.However,there is still room for further improvement in the overshoot of some working conditions.