Optimization of Energy Saving and Consumption Reduction of Polysilicon Siemens Method Based on Improved K-means Algorithm

The core technology of polysilicon production by improved Siemens method is still controlled and monopolized by developed countries. Firstly, it is proposed to improve the algorithm instability caused by the random determination of clustering center of K-means algorithm. The initial clustering center is determined by the method of distance to improve the performance of the algorithm, and it is applied to the energy saving and consumption reduction of polysilicon Siemens method. It can be graded from the quality point of view to ensure the quality of supply. Secondly, the convective heat transfer model in Siemens reactor was established, and the total energy consumption of laboratory-scale Siemens reactor predicted by K-means model was compared with the experimental data in the public literature. The relative errors were all within 1%, indicating that the convective heat transfer model established was effective. At the same time, by adding dichlorodihydrosilane to the feed of the traditional process, it is obtained that at 100 kPa, when the molar ratio of dichlorodisilane, trichlorosilane and hydrogen is 1:1:5, the polysilicon yield is increased by 9.6%, the silicon tetrachlorosilane yield is reduced by 8.7%, and the energy consumption is reduced by 35% compared with the traditional process. Finally, combined with the production practice and other factors unchanged, the ratio of mixture flow rate to silicon rod power is taken as the research object, and the furnace times with different feed and power ratio are compared, and the operation results are analyzed and summarized.