Due to the complex and highly nonlinear physical process of magnetic confinement fusion, the traditional numerical simulation methods have gradually failed to meet the high requirements of computational accuracy and speed in the research. Therefore, the adoption of high-performance computing technology has become the key to improve the simulation efficiency and accuracy. Plasma simulation plays a key role in optimizing the design of magnetic confinement fusion devices. In this paper, we propose a general heterogeneous parallel optimization strategy for the orbital simulation program of plasma guiding center (ORBIT), which covers the triple optimization methods of heterogeneous computing porting, dynamic load balancing and I/O storage optimization. In order to evaluate the effectiveness of the optimization strategy, this paper conducts comprehensive tests for different data sizes, and the test results show that the triple parallel optimization program achieves speedup ratios ranging from 16.45 to 25.6 times under a single process. In addition, when scaling up to 256 processes, the program performs very well in terms of weak scalability, up to 87%. This result shows that the optimization scheme proposed in this paper can effectively improve the computational efficiency of large-scale plasma simulation and maintain a significant performance improvement on heterogeneous platform, and also provides a reference value for more high-performance computing applications in the future.