基于神经网络和遗传算法的铸锭晶体硅质量控制及工艺优化

摘要/Abstract

摘要： 铸锭晶体硅是太阳能级晶硅材料的重要来源之一,为了进一步降低硅片成本,需要在保证晶体质量的同时发展大尺寸铸锭晶硅。影响铸造晶体硅质量的热场控制核心参数包括晶体生长速度与生长界面温度梯度之比V/G、壁面热流q、生长界面高度差Δh和硅熔体内部温差ΔT等。针对铸锭晶体硅生长过程中的质量控制问题,本研究基于人工神经网络(ANN)模型对晶体生长过程建立了工艺控制优化方法,利用实验测量数据和数值仿真模拟结果构建铸锭晶体硅生长过程的工艺控制数据集,以底部隔热笼开口和侧、顶加热器功率比作为主要工艺控制参数,V/G、|q|、|Δh|和ΔT为优化目标,建立用于研究晶体生长工艺控制参数和热场参数之间映射关系的神经网络模型。使用训练完成的模型分析底部隔热笼开口及侧、顶加热器功率比对晶体生长过程热场的影响规律,并采用遗传算法(GA)对铸锭晶体硅生长过程的工艺控制参数以提高晶体质量为目标进行优化,最后结合实际生产中的检测图像讨论了V/G对晶体质量的影响。研究表明晶体生长中期的V/G沿横向变化较平缓,对应缺陷较少且分布均匀,因此增大V/G在横向上的均匀度也是提高晶体质量的一个重要因素。

关键词: 铸锭晶体硅, 人工神经网络, 遗传算法, V/G, 定向凝固, 晶体质量

Abstract: Ingot crystalline silicon is one of the important sources for solar-grade crystalline silicon materials. For reducing the cost of silicon wafers, it is necessary to develop large-size ingot crystalline silicon while ensuring the crystal quality. The core parameters that affect the quality of cast crystalline silicon include the ratio of crystal growth rate to temperature gradient at the solidification interface V/G, wall heat flow q, height difference of solidification interface Δh, and temperature difference of silicon melt ΔT. Aiming at the quality control during the growth of ingot crystalline silicon, this study established a process control optimization method based on the artificial neural network (ANN) model for the crystal growth process. A process control data set for the growth process of ingot crystalline silicon was constructed based on the experimental measurement data and numerical simulation results. The opening of the bottom insulation cage and the power ratio of the side and top heaters were used as the main process control parameters, and V/G, |q|, |Δh|, ΔT are the optimization goals. A neural network model for mapping the relationship between process control parameters and thermal field characteristics was established. The influence of the bottom heat insulation cage opening and the power ratio of the side to top heaters on the thermal field during the crystal growth process were analyzed using the trained neural network model. The process control parameters for the ingot crystalline silicon growth process are optimized using genetic algorithm (GA), improving the quality of grown crystal and reducing energy consumption. Finally based on the experimental test images, the influences of V/G on the crystal quality were discussed. Research shows that V/G in the middle stage of crystal growth changes smoothly along the horizontal direction, and the corresponding defects are less and evenly distributed. Therefore, increasing the uniformity of V/G in the horizontal direction is also an important factor to improve the quality of crystal.

Key words: ingot crystalline silicon, artificial neural network, genetic algorithm, V/G, directional solidification, crystal quality

中图分类号:

O469
TQ127.2

郝佩瑶, 朱金伟, 廖继龙, 郑丽丽, 张辉. 基于神经网络和遗传算法的铸锭晶体硅质量控制及工艺优化[J]. 人工晶体学报, 2022, 51(3): 385-397.

HAO Peiyao, ZHU Jinwei, LIAO Jilong, ZHENG Lili, ZHANG Hui. Process Control and Optimization of Ingot Crystalline Silicon Growth Using Neural Network and Genetic Algorithm[J]. JOURNAL OF SYNTHETIC CRYSTALS, 2022, 51(3): 385-397.

参考文献

[1] 曾卉洁.PERC电池或将在2028年成为主流技术[J].高科技与产业化,2019(7):28-30.
ZENG H J. PERC solar cells may become a mainstream technology in 2028[J]. High-Technology & Commercialization, 2019(7): 28-30(in Chinese).
[2] MIYAZAWA H, LIU L J, KAKIMOTO K. Numerical analysis of influence of crucible shape on interface shape in a unidirectional solidification process[J]. Journal of Crystal Growth, 2008, 310(6): 1142-1147.
[3] MA X, ZHENG L L, ZHANG H, et al. Thermal system design and optimization of an industrial silicon directional solidification system[J]. Journal of Crystal Growth, 2011, 318(1): 288-292.
[4] MA W C, ZHONG G X, SUN L, et al. Influence of an insulation partition on a seeded directional solidification process for quasi-single crystalline silicon ingot for high-efficiency solar cells[J]. Solar Energy Materials and Solar Cells, 2012, 100: 231-238.
[5] QI X F, ZHAO W H, LIU L J, et al. Optimization via simulation of a seeded directional solidification process for quasi-single crystalline silicon ingots by insulation partition design[J]. Journal of Crystal Growth, 2014, 398: 5-12.
[6] WEI J A, ZHANG H, ZHENG L L, et al. Modeling and improvement of silicon ingot directional solidification for industrial production systems[J]. Solar Energy Materials and Solar Cells, 2009, 93(9): 1531-1539.
[7] LI Z Y, LIU L J, ZHANG Y F, et al. Preservation of seed crystals in feedstock melting for cast quasi-single crystalline silicon ingots[J]. International Journal of Photoenergy, 2013, 2013: 670315.
[8] FHNER T, JUNG T. Use of genetic algorithms for the development and optimization of crystal growth processes[J]. Journal of Crystal Growth, 2004, 266(1/2/3): 229-238.
[9] SU J, CHEN X J, LI Y, et al. A niching genetic algorithm applied to optimize a SiC-bulk crystal growth system[J]. Journal of Crystal Growth, 2017, 468: 914-918.
[10] DANG Y F, LIU L J, LI Z Y. Optimization of the controlling recipe in quasi-single crystalline silicon growth using artificial neural network and genetic algorithm[J]. Journal of Crystal Growth, 2019, 522: 195-203.
[11] ASADIAN M, SEYEDEIN S H, ABOUTALEBI M R, et al. Optimization of the parameters affecting the shape and position of crystal-melt interface in YAG single crystal growth[J]. Journal of Crystal Growth, 2009, 311(2): 342-348.
[12] DROPKA N, HOLENA M. Optimization of magnetically driven directional solidification of silicon using artificial neural networks and Gaussian process models[J]. Journal of Crystal Growth, 2017, 471: 53-61.
[13] TSUNOOKA Y, KOKUBO N, HATASA G, et al. High-speed prediction of computational fluid dynamics simulation in crystal growth[J]. CrystEngComm, 2018, 20(41): 6546-6550.
[14] QI X F, MA W C, DANG Y F, et al. Optimization of the melt/crystal interface shape and oxygen concentration during the Czochralski silicon crystal growth process using an artificial neural network and a genetic algorithm[J]. Journal of Crystal Growth, 2020, 548: 125828.
[15] DANG Y F, ZHU C, IKUMI M, et al. Adaptive process control for crystal growth using machine learning for high-speed prediction: application to SiC solution growth[J]. CrystEngComm, 2021, 23(9): 1982-1990.
[16] YU W C, ZHU C, TSUNOOKA Y, et al. Geometrical design of a crystal growth system guided by a machine learning algorithm[J]. CrystEngComm, 2021, 23(14): 2695-2702.
[17] 闵乃本.晶体生长的物理基础[M].上海:上海科学技术出版社,1982.
MIN N B. Physical basis of crystal growth[M]. Shanghai: Shanghai Science and Technology Press, 1982(in Chinese).
[18] 梅向阳,马文会,戴永年,等.定向凝固技术的发展及其在制备太阳能级硅材料中的应用[J].轻金属,2008(9):64-71.
MEI X Y, MA W H, DAI Y N, et al. Development and application in preparating solar grade silicon material of directional solidification technology[J]. Light Metals, 2008(9): 64-71(in Chinese).
[19] 斋藤康毅.深度学习入门:基于Python的理论与实现[M]. 陆宇杰译.北京:人民邮电出版社,2018.
SAITO K. Introduction to deep learning: theory and implementation based on Python[M].Translated: LU Y J. Beijing: Posts and Telecommunications Press, 2018(in Chinese).
[20] KINGMA D, BA J. Adam: A method for stochastic optimization[J]. Computer Science, 2014.
[21] IOFFE S, SZEGEDY C. Batch Normalization: Accelerating deep network training by reducing internal covariate shift[J]. JMLR.org, 2015.
[22] 玄光男,林林,著. 网络模型与多目标遗传算法[M]. 梁承姬,于歆杰译.北京:清华大学出版社,2017:151.
GEN M, HAYASHI L. Network model and multi-objective genetic algorithm[M]. Translated: LIANG C J, YU X J. Beijing: Tsinghua University Press, 2017: 151(in Chinese).
[23] 马旭.晶硅定向凝固过程中的热输运与组分输运研究[D].北京:清华大学,2012.
MA X. Thermal and species transport during directional solidification of silicon for solar cells[D]. Beijing: Tsinghua University, 2012(in Chinese).