Advanced reactor simulations require the inversion of key parameters of the whole reactor based on less actual detection data.To meet this need,this paper constructs a computational model based on variational residual physics-informed neural network(hp-VPINN)with high-order polynomial domain decomposition function,which is used to solve the neutron diffusion equation forward and backward.The model uses neural network as trial function,and substitutes it into the neutron diffusion equation to form variational residual as loss function for gradient descent.In order to improve the accuracy and efficiency of the solution,this paper also proposes some innovative key technologies such as effective multiplication factor intelligent search and inversion based on the physical characteristics of neutron diffusion equation,and realizes the self-optimization of neural network hyperparameters based on whale optimization algorithm(WOA).Finally,several examples are verified,and the results show that the method has high accuracy and low dependence on training data,providing a solution with less training data and higher accuracy output for advanced reactor simulations.
维普
万方数据
