The theory of dynamic properties evolution algorithm for polymer nanomaterials is an emerging computational method, has demonstrated significant advantages in the field of material crystallization kinetics research. The heterogeneous computing platform, by leveraging the strengths of both processors, significantly enhances computational efficiency and application scope. This paper focuses on the characteristics of the dynamic properties evolution algorithm of polymer nanomaterials, and conducts parallel porting and algorithm optimization on a heterogeneous computing platform. It proposes three optimization acceleration strategies to improve the computational efficiency and performance, including heterogeneous parallel optimization, thread-limited loop structure reconfiguration optimization, and adaptive fusion calculation optimization of redundant data. The strategies combine the characteristics of the algorithm with the unique computing methods in heterogeneous systems, optimizing thread allocation and data transmission methods. The experimental results show that, compared with the original serial program, the optimized program achieves an acceleration ratio of up to 28.67x on a single node. When the optimized program is scaled up to 256 processes, in three scales, strong scalability reaches 42.62%–58.31%, and weak scalability reaches 61.73%–72.43%.