With the exponential growth of global patent application, conducting domain-specific patent analysis to uncover latent information in patent data, identify emerging productive forces, and explore technological innovation pathways has become a critical focus for national strategies, industrial sectors, and research institutions. However, current conventional patent analysis tools increasingly reveal significant limitations, including static analytical methods, insufficient interactivity, and difficulties in representing technological correlations. This study concentrates on patent relevance algorithms and D3.js visualization techniques to construct a patent analysis technical solution based on dynamic network graphs, achieving core functionalities such as deep correlation mining, dynamic real-time interaction, massive data processing, and multidimensional patent clustering analysis. Validation through a uranium mining patent visualization case demonstrates that the underlying algorithms exhibit high data accuracy, clear representation, and robust performance. The dynamic network graphs effectively visualize core patent identification, technology branch analysis. This technical solution demonstrates substantial practical value and research significance, potentially advancing patent analysis from "static statistics" to a new phase of "dynamic decision-making".