摘要
针对土壤悬液组分复杂以及单输入变量时电极预测精准度有限的问题,以提高离子选择电极预测土壤硝态氮含量精准度为目标,建立基于多参数融合的支持向量机(SVM)土壤硝态氮预测模型。采用灰色关联分析法对影响电极法测定土壤硝态氮的主要干扰因素进行排序,建立以主干扰因素及硝酸根电极检测电势的多参数融合SVM预测模型,并与传统Nernst模型和干扰因素全输入下的SVM模型作对比验证算法可行性。实验结果表明,土壤电导率、温度与Cl^-电极检测电势为影响电极预测硝态氮精准度的主要干扰因素;输入参数为硝态氮电极检测电势、土壤电导率、温度与Cl^-电极检测电势时,SVM土壤硝态氮预测模型效果最优,与光学法测定结果回归方程的调整决定系数为0.98,平均绝对偏差为3.38 mg/L,均方根误差为4.51 mg/L,基于多参数融合的SVM预测模型可显著提高电极法硝态氮检测精准度。
The conventional method of soil nitrate-nitrogen prediction based on ion-selective electrode had the problem of complex soil suspension components and the limited prediction accuracy and precision in single input variables. To improve the prediction accuracy and precision of soil NO_3^--N concentration employing ion-selective electrodes( ISEs),the support vector machine( SVM) model of soil NO_3^--N prediction based on sensor fusion was built. Grey relational analysis was applied to screen the major interference factors,which had a great impact on the soil NO_3^--N detection employing ISEs,and the support vector machine( SVM) model based on sensor fusion was built with the major factors. Then,the classical Nernst model and the SVM model with major factors and all considered factors were compared with the conventional method. According to the testing results,EC values,temperature and Cl-were the three major interference factors which had great influence on the prediction accuracy and precision of soil NO_3^--N concentration employing ISEs. With the optimized input parameters of NO_3^--N ISE potentials,EC,temperature and Cl-ISE potentials,the adjusted R2,average absolute error and root mean square error of the SVM model were 0. 98,3. 38 mg / L and 4. 51 mg / L,respectively. The SVM model based on sensor fusion showed more advantages than the Nernst model and it could successfully achieve theprediction purpose of NO_3^--N with high prediction accuracy and precision of the ISEs in soil extracted solution.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2015年第S1期96-101,共6页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金资助项目(61134011
31201136)
中央高校基本科研业务费专项资金资助项目(2015XD001)
关键词
土壤
硝态氮检测
离子选择电极
多参数融合
支持向量机
Soil
Nitrate-nitrogen detection
Ion-selective electrode
Sensor fusion
Support vector
machine
