<p>The ultra-large scale refined numerical simulation based on supercomputers is the key to promoting the development of scientific and engineering computing and other application fields. The new generation of exascale heterogeneous supercomputer in China provide strong computing power support for ultra-large scale refined numerical simulation. Parallel optimization can significantly improve numerical simulation efficiency. However, limited by the complexity of the different Chinese supercomputer architectures and heterogeneous multi-layer characteristics, the parallel optimization of numerical simulation faces technical challenges. Therefore, this paper reviews the development trend of supercomputers in China, and analyzes the common optimization direction of ultra-large scale refined numerical simulation. This paper also illustrates the challenges faced by parallel optimization technology by combing the architectural characteristics of the mainstream supercomputers in China. Finally, four key numerical simulation parallel optimization technologies are summarized. The above parallel optimization technologies and the future development trend of Chinese supercomputers are summarized and prospected.</p>