Reserch On Modeling And Control Method Of Power Control System In Pwr Nuclear Power Plant

In recent years,the installed capacity of nuclear power in China has gradually increased,and higher requirements have been put forward on the modeling and controling of nuclear power plant systems.The pressurized water reactor nuclear power unit is a complex nonlinear system,and the subsystems are interrelated and uncertain.Neural networks can deal with nonlinear modeling problems with good results.Based on the analysis of running characteristics of each part of PWR power control system,the coolant temperature and power of nuclear reactor are selected as the research object.The improved pruning algorithm on BP neural network,the modeling of the temperature of coolant and the nonlinear model of the coordinated control system are emphatically studied,and the predictive control algorithm is used in controlling study.To improve the generalization ability of neural network,an improved pruning algorithm is proposed from the perspective of structural optimization.An influence factor is introduced to each input node and hidden node.The product of influence factor and the output of the node is transferred to output node.Under different training precision,the weight value and the influence factor are cross modified.The influence factor can represent the contribution of the node to the network output,and the value of the influence factor is used as the reference for the deletion of the nodes,and the network is continuously reduced.The simulation results show that the neural network with pruning algorithm can ensure convergence precision and convergence speed on the basis of simplifying the network structure and enhancing the network generalization ability.By analyzing the power control system,coolant temperature and reactor power in primary loop and steam turbine power in secondary loop is set as research objects.Aiming at PWR primary loop system,differential dynamic equations of point reactor neutron kinetics model,coolant temperature and fuel temperature model,and reactivity model are established.Combined with the design and operation parameters,the state space model and transfer function model with control rod position as control variables are obtained.The dynamic characteristic curve of the object shows the self-stability of the system and proves the correctness of the model.A 2 x 2 coordination control system with valve opening,control rod displacement as control quantity,main steam pressure and turbine power as the controlled quantity is established.The object model of transfer function is used to generate sample data,and the neural network with pruning ability is used for object modeling.The training and testing results are good,and the network model can accurately reflect the characteristics of the objects.In view of the nonlinear and uncertainty of the reactor power control system,the control strategy of generalized predictive control and the BP neural network as the prediction model are used in this paper,and the control simulation of the primary or secondary loop is studied.The step response curve of the object shows that the control method can achieve better control effect.Compared with the effect of PID control,the predictive control algorithm can have shorter response time,and further verify the accuracy of the model and the effectiveness of the control algorithm.