Research On Thermal-hydraulic Parameter Prediction Method Of The Small Lead–bismuth Fast Reactor Core Based On Adaptive RBF Neural Network

As the most important reactor of the fourth generation nuclear power system,the lead-bismuth fast reactor has remarkable advantages,such as:high thermal efficiency,safe operation,and stable performance.In order to obtain higher thermal efficiency,the outlet coolant temperature of a lead-bismuth fast reactor core is allowed to exceed 500℃,and the center temperature of fuel pellets may reach 800 ℃.However,in some transient conditions,spontaneous variations in the heat transfer capacity of the cycle significantly affect the core thermal parameters,thereby affecting the thermal safety of the entirereactor.Therefore,it is of significant importance to quickly and accurately predict the thermal parameters of the lead bismuth fast reactor core under steady-state and transient condition.In the present study,the sub-channel analysis program SUBCHANFLOW is applied to study the operation of the SPALLER-100 reactor,which is a long-life small natural circulation lead-based fast reactor.To this end,the core sub-channel was modeled and the steady-state th ermal parameters of the reactor were studied.The obtained results indicate that the Pu enrichment of the assembly is an accurate indicator to raflect the radial power distribution of the core,assembly temperature,and the coolant flow.It is found that these parameters have M-shaped distributions.Moreover,the axial distribution of the core coolant temperature and cladding temperature shows a monotonic upward trend,and a thermal spike appears at the outlet of the reactor core.Compared with the core,the axial temperature distribution in a single assembly is more sensitive to axial power.Moreover,the larger the equivalent diameter of the flow channel,the larger the flow area,the better the cooling effect,and the lower the temperature.The performed analyses reveal that the core center temperature of the hottest fuel rod is717.22°C,which is lower than the melting point of nitride fuel.Based on the performed analyses,a new method is proposed to rapidly predict the thermal and hydraulic parameters of the reactor core based on an adaptive RBF neural network.Firstly,SUBCHANFLOW program and Latin hypercube sampling method were used to determine the sample data to ensure the correctness and uniformity of the sample data,and the sample set and prediction set required for the training of the prediction model were constructed.Secondly,the adaptive RBF neural network model is applied to predict the thermal parameters of the SPALLER-100 reactor under steady-state and transient conditions.The obtained results show that the maximum prediction error of the adaptive RBF neural network for the maximum temperature of the fuel cladding of the reactor core is less than 5%,and the average relative error is less than1%.In the case of flow response and power response under transient conditions,the maximum prediction error of the adaptive RBF neural network for the maximum temperature of the fuel cladding of the reactor core does not exceed 3%.It is concluded that the adaptive RBF neural network can rapidly and accurately predict the maximum fuel cladding temperature of a small lead bismuth fast reactor core.