基于SSA-DELM算法的黑土地保护性耕作土壤温度预测

刘斌; 王柏; 司振江; 黄彦; 蒋骁童; 郝利

doi:10.13928/j.cnki.wrahe.2025.10.020

水利水电技术（中英文） ›› 2025, Vol. 56 ›› Issue (10) : 257 -276. DOI: 10.13928/j.cnki.wrahe.2025.10.020

农村水利

基于SSA-DELM算法的黑土地保护性耕作土壤温度预测

刘斌 ¹^,² ,
王柏 ¹^,²^,³ ,
司振江 ¹ ,
黄彦 ¹^,²^,³ ,
蒋骁童 ¹^,² ,
郝利 ¹^,²

作者信息 +

Prediction of soil temperature based on SSA-DELM algorithm for conservation tillage in black soil

Bin LIU ¹^,² ,
Bai WANG ¹^,²^,³ ,
Zhenjiang SI ¹ ,
Yan HUANG ¹^,²^,³ ,
Xiaotong JIANG ¹^,² ,
Li HAO ¹^,²

Author information +

文章历史 +

PDF

摘要

【目的】农田耕作层土壤温度的精准预测分析对农业生产具有重要现实意义。为有效、准确分析缺乏实测条件的黑土地保护性耕作农田土壤温度的变化特征，【方法】基于极限训练机算法(ELM),引入自编码器算法(AE)构建深度极限训练机(DELM),利用麻雀搜索算法(SSA)进行改进，构建了SSA-DELM混合模型，利用气象因子数据，对免耕无秸秆覆盖(NT0)和免耕秸秆全覆盖(NTS)两种耕作条件下不同深度土壤温度进行预测，并与ELM、RF和SSA-RF模型进行对比。【结果】结果显示，SSA-DELM模型在两种耕作条件下土壤温度预测结果决定系数(R²)分别为0.996和0.998,平均绝对误差(MAE)分别为0.16和0.1,均方根误差(RMSE)分别为0.29和0.16,纳什效率系数(NSE)为0.999和0.999,性能指数(PI)为0.051和0.056,最大残差(MaxE)均小于0.25,平均运行时间分别为17.2 s和17.6 s。【结论】结果表明，相较于其他模型，SSA-DELM模型的预测精度、泛化能力和预测效率更优，预测误差较小，在两种不同耕作条件下均能满足土壤温度预测需求，具有良好的稳定性和抗干扰能力，可为农业生产决策提供可靠的数据支持。

Abstract

[Objective] Accurate prediction and analysis of soil temperature in the tillage layer of farmland is of great practical significance for agricultural production, in order to effectively and accurately analyze the characteristics of soil temperature change in the farmland of conservation tillage of black soil lacking in real measurement conditions. [Methods] Based on the Extreme Limit Training Machine(ELM) algorithm, the Self-Encoder Algorithm(AE) was introduced to form the Depth Extreme Limit Training Machine(DELM), which was improved by using the Sparrow Search Algorithm(SSA), and a hybrid SSA-DELM model was constructed to predict soil temperatures at different depths under two types of no-tillage no-straw cover(NT0) and no-tillage full straw cover(NTS) tillage conditions by utilizing the data of meteorological factors, and then predict soil temperatures at different depths under no-tillage no-straw cover(NTS). Soil temperature was predicted and compared with ELM, RF and SSA-RF models. [Results] The results showed that the coefficient of determination(R²) of the SSA-DELM model was 0.996 and 0.998, the mean absolute error(MAE) was 0.16 and 0.1, the root mean square error(RMSE) was 0.29 and 0.16, and the coefficients of Nash′s efficiency(NSE) were 0.999 and 0.999 for the prediction of soil temperatures under two types of tillage conditions, respectively. Performance Index(PI) was 0.051 and 0.056, the maximum residual error(MaxE) was less than 0.25, and the average running time was 17.2 s and 17.6 s, respectively. [Conclusion] Compared with other models, the prediction accuracy, generalization ability and prediction efficiency of SSA-DELM model were better than other models, and the prediction error was very low, and it could satisfy the two different tillage conditions under the soil temperature prediction needs, has good stability and anti-interference ability, and can provide certain data support for agricultural production decision-making.