Abstract: Physics-Informed Neural Network (PINN) is a deep learning method that incorporates physical knowledge. It has recently been applied to solve core neutron problems in computational nuclear engineering. However, the non-reusability of PINNs when solving different problems greatly hampers its value and efficiency. To address this issue, a method based on parameterized PINNs has been proposed, leading to the development of a new surrogate model for the neutron diffusion physical equation. This surrogate model is highly flexible and efficient, capable of quickly providing solutions to any given parameter neutron diffusion problem without the need for sample data. Furthermore, the predictive accuracy of the parameterized PINN surrogate model under soft and hard boundary constraints was compared. The results show that the PINN surrogate model under hard constraints has higher predictive accuracy, and the acceleration ratio is over 1 000. This work primarily focuses on the application of PINNs to core neutron problems in computational nuclear engineering. The goal is to improve the reusability of PINNs, thereby enhancing their value and efficiency in computational nuclear engineering. Based on parameterized PINNs, this work has drawn several conclusions through the study of single-region, double-region, and triple-region examples. The parameterized PINN neutron diffusion surrogate model proposed in this paper has high accuracy. When compared with the reference solution, the relative error of the single-region three-physical-parameter example can reach 0.07%, the double-region five-physical-parameter example can reach 1%, and the triple-region three-physical-parameter example can reach 2%. The predictive accuracy of the parameterized PINN neutron diffusion surrogate model under hard boundary constraints is significantly higher than that of the parameterized PINN under soft boundary constraints. The accuracy of the two differs by 1-2 orders of magnitude. The parameterized PINN neutron diffusion surrogate model under hard boundary constraints has a very obvious acceleration effect compared to the reference program COMSOL. The acceleration ratios for the three examples are 1 923.4, 1 396.1, and 1 385.1, respectively. This work can be applied to the rapid solution of complex neutron diffusion problems. Future work will focus on further optimizing this method based on parameterized PINNs to improve its performance in handling more complex problems. In addition, the possibility of applying this method to other physical problems will be explored to fully utilize its advantages in handling complex problems. At the same time, this work has also found some problems in the parameterized PINN method, such as the obvious spatial distribution characteristics of errors in the triple-region example. In the future, research will be conducted on how to better handle potential outliers to further improve the robustness of the model.