Research On 3D Power Predictive Control Method Of Reactor Core Based On Neural Network

With the development of society and the rapid growth of economy,people are gradually aware of the limitations and the harm of traditional energy,the search for clean and efficient energy has been put on the agenda,and the direction of energy research expands to nuclear energy.Reactor core control system,which is an important part of nuclear power plant,largely determines the safety of the reactor.But nowadays,the research of reactor core control system is based on the point reactor core model,which can not reflect the real effect of the controller accurately.Therefore,the controller based on the point reactor model is quite different from the actual controller.In this paper,the three-dimensional power distribution model of the reactor core is established,and three popular intelligent control methods are used to improve the traditional reactor control system.Firstly,for Qinshan 300 MW reactor,the three-dimensional power distribution simulation program of reactor core is established by using the three-dimensional two-group diffusion equation with six groups of delayed neutrons,based on the REMARK core physical calculation program developed by GSE.The neutron flux density is decomposed into shape function and amplitude function.In addition,an intelligent three-dimensional core power distribution model is established by using ex-core neutron detectors and neural network,in order to reconstruct three-dimensional distribution of reactor core power.Secondly,aiming at the disadvantage that the traditional PID controller can not adjust parameters on-line in real time,the artificial neural network algorithm and genetic algorithm are selected to optimize the PID controller.By combining the two intelligent algorithms with the traditional PID controller,the neural network PID intelligent controller and the genetic algorithm PID intelligent controller are designed.In addition,predictive control method is also selected to improve the reactor power control system.Rolling finite-time optimization method is used in predictive controller,which results in better dynamic control performance compared with traditional PID control method.Neural network is used as the predictive model of the controller,and a neural network predictive controller is designed.After compiling the simulation programs of three intelligent controllers,each controller is applied to the core threedimensional power distribution simulation model to verify their availability.Finally,the performance and the characteristics of traditional PID controller,neural network PID controller,genetic algorithm PID controller and neural network predictive controller are compared under two working conditions of increasing power and decreasing power.The simulation results show that the performance of three intelligent controllers is better than that of traditional PID controller whether in overshoot or in adjustment time.Among three intelligent controllers,the neural network predictive controller has the shortest adjustment time and the best performance.In this paper,a useful attempt is made to reconstruct the power of the smart reactor core and to control the power distribution of the reactor intelligently.