| 摘 要: 作物模型参数的筛选与优化对于提升其适应能力具有重要意义。为提高参数优化效率,采用随机森林和梯度提升树两种基于集成学习的机器学习算法,对 APSIM NG 旱地春小麦生长过程中的敏感参数进行筛选,并利用 Nelder-MeadSimplex算法和 DREAM-zs算法对这些敏感参数进行优化,实现了在 R语言中一键完成作物模型敏感性分析及优化全过程。实验结果表明,优化后产量实测值与模拟值的均方根误差(RMSE)从215.12kg·hm-2降至100.26kg·hm-2,标准均方根误差(NRMSE)从14.63%降至6.82%。优化后的参数符合旱地春小麦的生长发育过程,适用性较高。机器学习算法为参数选择提供了新方法,使模型优化过程更加便捷。 |
| 关键词: 参数筛选 APSIM NG 机器学习 旱地春小麦 参数优化 |
|
中图分类号: TP391.7
文献标识码: A
|
|
| Research on Parameter Screening of Crop Models Based on Machine Learning Algorithms |
|
YU Qi, NIE Zhigang
|
(College of Information Science and Technology, Gansu Agricultural University, Lanzhou 730070, China)
yu2213964993@163.com; niezg@gsau.edu.cn
|
| Abstract: The screening and optimization of crop model parameters are of great significance for enhancing their adaptability. To improve the efficiency of parameter optimization, this study employed two ensemble learning-based machine learning algorithms—random forest and gradient boosting decision trees—to screen sensitive parameters in the growth process of dryland spring wheat in APSIM NG (Agricultural Production Systems sIMulator Next Generation).The Nelde-r Mead Simplex algorithm and the DREAM-zs algorithm were used to optimize these sensitive parameters,
achieving a one-click process for sensitivity analysis and optimization of the crop model in the R language. Experimental results showed that the Root Mean Square Error (RMSE) between the measured and simulated yields decreased from 215.12 kg·hm-2 to 100.26 kg·hm-2 after optimization, while the Normalized Root Mean Square Error (NRMSE) decreased from 14.63% to 6.82% . The optimized parameters align with the growth and development process of dryland spring wheat, demonstrating high applicability. Machine learning algorithms provide a new method for parameter selection, making the model optimization process more convenient. |
| Keywords: parameter screening APSIM NG machine learning dryland spring wheat parameter optimization |
