Abstract: In the high-fidelity numerical calculation method of reactor physics, the pin-by-pin two-step method can achieve a balance between computational accuracy and computational cost, and is currently a feasible and realistic high fidelity numerical calculation method. However, compared to the traditional two-step method of assembly homogenization, pin-by-pin calculation requires even greater storage requirements of homogenization parameters. In order to reduce the number and storage scale of pin-by-pin homogenization parameters in high fidelity numerical calculations of reactor physics, and improve the practicality of high-fidelity numerical calculation methods, this paper was based on clustering methods and studied the number and clustering characteristics of typical assemblies in pressurized water reactors. Firstly, this paper conducted a study on the number and characteristics of clustering, focusing on typical fuel assemblies of pressurized water reactors, multiple clustering features and numbers were studied, fast and thermal flux ratio, fission sigma and thermal flux were chosen as clustering features, and it was determined that using the fast and heat flux ratio as the clustering feature and 5 as the clustering number were appropriate. In response to the phenomenon that using only one clustering feature may lead to clustering pins of different material types into one cluster, based on the fact that the material of the pin-cell is the most important feature of clustering, a significant deviation was introduced, therefore, this paper proposed a multi-level clustering method of “classification first and then clustering”, and this method can significantly improve the stability of clustering; Secondly, based on the traditional homogenization theory, derived three different homogenization methods of clustering, and compared the accuracy and data compressing efficacy of the three methods, this paper derived the methods of clustering region super homogenization and clustering pin-by-pin super homogenization. The clustering pin-by-pin super homogenization method can ensure the conservation of reaction rate before and after clustering, while reducing storage scale; Once again, this paper established a clustering-based method for compressing the pin-by-pin homogenization parameters of pressurized water reactor. This method establishes a whole process from assembly calculation to compressing the pin-by-pin homogenization parameters; Finally, the clustering effect of this method was verified on the HPR1000 core model, including the typical burnup of the HPR1000 core. The research results indicate that for the core problem of HPR1000 reactor, this method can ensure high accuracy and significantly reduce the scale of pin-by-pin homogenization parameters. Therefore, the pin-by-pin homogenization parameter compression method established in this study can be used to compress the pin-by-pin homogenization parameters of pressurized water reactors, reducing their storage scale.