张合兵, 潘怡莎, 聂小军,等.基于神经网络模型的县域尺度农业碳排放研究[J].河南理工大学学报（自然科学版）,2025,44(5):111-120.

ZHANG H B, PAN Y S, NIE X J,et al.Study on agricultural carbon emission at county scale based on neural network model[J].Journal of Henan Polytechnic University(Natural Science) ,2025,44(5):111-120.

基于神经网络模型的县域尺度农业碳排放研究

张合兵1, 潘怡莎1, 聂小军1, 王重洋2, 张慧芳1

1.河南理工大学 测绘与国土信息学院，河南 焦作  454000；2.重庆大学 资源与安全学院，重庆  400044

摘要: 目的 为测算平顶山市各县区2010—2020年农业碳排放，开展基于神经网络模型的县域尺度农业碳排放研究。  方法 从县域角度出发，从投入与产出角度对各影响因子进行分析，并在此基础上建立农业碳排放预测模型。采用灰色关联分析和Robust回归分析，得出各影响因素的关联程度及对农业碳排放的影响，初步确定各影响因素权重，建立神经网络预测模型，并将预测结果与实际值进行检验评价。  结果 结果表明：（1）平顶山市受农业生产分布区域影响，环中心城区县市承担主要农业生产活动，农业碳排放量较高；（2）灰色关联分析结果显示，农资投入要素对平顶山农业碳排放量影响显著，其中化肥与碳排放量相关度最高，产出因素相关度存在一定差异；（3）Robust回归分析结果给出了各影响因素的影响方向，指出玉米种植对农业碳排放的产生呈负相关关系，油料，瓜果，农业劳动力与农业碳排放关系不明显；（4）预测模型结果与实际值相关系数R₂为0.99，拟合度较好。 结论 研究结果可为区域农业高质量发展和农业碳减排政策的制定提供一定理论支持与技术支撑。

关键词:农业碳排放;灰色关联;神经网络;Robust回归分析;农业碳排放影响因素

DOI:10.16186/j.cnki.1673-9787.2023060001

基金项目:国家自然科学基金资助项目（41977284）

收稿日期:2023/06/01

修回日期:2023/08/22

出版日期:2025/07/23

Study on agricultural carbon emission at county scale based on neural network model

Zhang Hebing1, Pan Yisha1, Nie Xiaojun1, Wang Chongyang2, Zhang Huifang1

1.School of Surveying and Land Information Engineering， Henan Polytechnic University， Jiaozuo  454000， Henan， China；2.School of Resources and Safety Engineering， Chongqing University， Chongqing  400044， China

Abstract: Objectives To calculate the agricultural carbon emissions in Pingdingshan City from 2010 to 2020， Conduct research on agricultural carbon emissions at the county scale based on neural network models. Methods from the perspective of county level， the influencing factors were analyzed from the perspective of input and output. On this basis， the agricultural carbon emission prediction model was established. Grey correlation analysis and Robust regression analysis were used to obtain the correlation degree of each influencing factor and the direction of influence on agricultural carbon emissions. The weight of each influencing factor was preliminarily determined to establish a neural network prediction model， and the prediction results were tested and evaluated with the actual values.  Results The results showed that： （1） Pingdingshan City was affected by the distribution of agricultural production activities， the counties and cities around the central city were the main agricultural production activities， and their agricultural carbon emissions were higher. （2） The results of grey correlation analysis showed that agricultural input factors had a significant impact on agricultural carbon emissions in Pingdingshan， among which fertilizer had the highest correlation with carbon emissions， while output factors had different correlation degrees. （3） The Robust regression analysis results gave the influence direction of each influencing factor， pointing out that maize planting had a negative correlation with the production of agricultural carbon emissions， and identified oil crops， fruits and farm workers as factors with insignificant influence on agricultural carbon emissions. （4） The correlation coefficient R₂ between the prediction model results and the actual values reached 0.99， which had a good degree of fit.  Conclusions The study results could provide certain support for the high-quality development of regional agriculture and the formulation of agricultural carbon emission reduction policies.

Key words:agricultural carbon emission;grey relation;neural network;Robust regression analysis;factor influencing agricultural carbon emission