Research On Steam Generator Modeling And Predictive Control Based On The State Space Model For PWR Nuclear Power Unit

With the demand for clean,efficient and sustainable energy utilization,the trend of nuclear power development in China is unstoppable.As an important equipment for connecting nuclear islands and conventional islands in nuclear power,steam generators(SG for short)determine whether the entire nuclear power plant can operate economically and safely.However,the SG system is highly complex,non-linear,time-varying,and non-minimum-phase.It has large measurement errors under low-power operation conditions.The dynamic characteristics are unstable due to changes in operating power,and it is subject to "shrinkage" and "swelling" phenomenon,which has brought obstacles and difficulties to the SG's water level control.Excellent and effective control requires more in-depth research and accurate models.In this paper,the steam generator is analyzed by means of combination of partition and lumped parameters,a dynamic two-input and two-output mechanism model is established,and the accuracy of the model is verified by DAE micro-decomposition method.And the transfer model is obtained by simplifying and identifying the mechanism model,and a state space model predictive control method is designed.This paper deals with the mechanism modeling of objects,the derivation of transfer function models and the identification of parameters,the transformation of state-space models,and the predictive control based on the state-space model.It provides a complete,clear,and innovative research method for solving the modeling and control research of complex objects.The specific research content is as follows:(1)Establish the SG mechanism model.Through the law of conservation of mass,energy and laws of thermodynamics,the method of partition combined with lumped parameter is used to model the mechanism of SG,and the simulation and dynamic characteristics of SG water level are analyzed.The partitioning can analyze the different states of the working fluid in the whole SG at different working conditions.It is mainly divided into the primary coolant section,secondary circuit descending section,hot water section,boiling section,steam separator section,steam chamber section.A model established by using the lumped parameter method for each segment divided above.And using the special solution method of DAE differential equation in MATLAB to redistribute the quality matrix parameters of ODE solver,and transform the differential algebraic equation into nonlinear state space form to realize the exact solution of differential algebraic equation,and then verify the model's accuracy.(2)Establish the SG water level control model.Because the above model is very complex,it is not conducive to the control of the water level.Therefore,the proposed mechanism model was optimized and deduced,combined with the actual dynamic characteristics of the object,the format of transfer function model was obtained.Besides,the improved particle swarm optimization algorithm used to identify the dynamic and static parameters the transfer function model based on the field operating data,and the model which is beneficial to the water level control is obtained ultimately.The global search capability of the improved PSO is fast and efficient,which overcomes the problem that other intelligent optimization such as basic PSO which can easily fall into a local extremum.The improved PSO has good identification and optimization capabilities.(3)The tf2ss function in Matlab is used to transform the transfer function and the state space model.The Kalman filter is used to obtain the current state variable estimation value,and the augmented state space prediction model is used instead of the conventional state space prediction model,which optimizes the derivation and calculation process and improves computing efficiency greatly.A predictive controller based on the state space model is designed,and the objective function of predictive control is determined.Then and the control variable group is obtained.Only the current control variable in the control variable group is taken and the control variable is re-acquired at the next time,which is exactly online rolling optimization.Besides,the author uses the feedback form of the output error correction system model to solve the mismatch between the control model and the actual equipment.