The Research On Intelligent Automatic Start-up And Monitoring Technology Of Reactor

The start-up process of current reactors is primarily carried out manually by operators.Compared to full-power operation,the start-up process involves a greater number of systems,complex procedures,and lengthy execution times,leading to a higher potential for humaninduced accidents and posing a significant threat to reactor safety.In addition,the core power distribution changes dramatically during the start-up phase,requiring continuous monitoring of the core power distribution during this process.However,for most of the reactors in operation,their fixed in-core neutron detectors do not provide a continuous neutron signal.Small modular reactors with integrated core arrangement cannot have too many neutron detectors inside them.The ex-core neutron detector becomes most reactors’ only real-time neutron probes.Therefore,studying the intelligent automatic reactor start-up method and core power distribution online monitoring method is important.Research on automatic reactor startups has primarily been based on sequential control techniques,which are insufficient due to the complexity and nonlinearity of reactors.Accurately converting numerous tasks during startup into precise sequential logic is difficult.On the research of core power distribution monitoring method using ex-core detectors.Previous studies have relied too much on the accuracy of neutron transport calculations or considered only simple linear and nonlinear relationships between the individual physical nodules within the reactor and the ex-core detectors,without considering the relationships between the individual physical nodules in space.Based on this,this paper refers to manual startup operation strategies and applies artificial intelligence algorithms to the field of reactor automatic startups.Employing the framework of a Deep Double Q-Network(DDQN)with prioritized experience replay,an intelligent automatic reactor start-up control algorithm is developed.The online data interaction platform between the reactor simulator and the intelligent automatic reactor start-up control algorithm is built using the UDP protocol,and the intelligent automatic reactor start-up control algorithm is trained using online data interaction.Considering the limited number of ex-core detectors,this paper proposes a novel Proper Orthogonal Decomposition-Extreme Learning Machine(PODELM)as a method for online monitoring of core power distribution.The POD is used to downscale the core power distribution model,and then the ELM is used to fit the relationship between the ex-core detector and the POD orthogonal basis coefficients to obtain the new POD orthogonal basis coefficients,which in turn enables online monitoring of the core power distribution.The simulation results show that the DDQN intelligent automatic reactor start-up algorithm can realize the intelligent automatic reactor start-up and control all safety parameters within the range specified by the start-up protocol.Compared with the traditional reactor automatic start-up method,the DDQN intelligent start-up algorithm can select a reasonable action according to the state of the core,without the operation task being fully converted into an accurate sequential logic.The POD-ELM method has achieved good results in the online monitoring of power distribution.Compared with the traditional core power online monitoring methods,the POD-ELM method does not rely too much on the solution accuracy of neutron transport calculation and is a good online monitoring method for power distribution.The research in this paper has some reference value to improve the automation level of the reactor start-up phase and improve the safety and efficiency of the start-up phase.