支持向量机的盐碱土壤Philip入渗模型参数预测研究  

Saline-alkali Soil Philip Infiltration Model Parameters Prediction Based on Support Vector Machine

作　　者：程诗念 樊贵盛[1] 

机构地区：[1]太原理工大学水利科学与工程学院,太原030024

出　　处：《节水灌溉》2017年第9期47-50,共4页Water Saving Irrigation

基　　金：国家自然科学基金项目"区域尺度上土壤入渗参数多元非线性传输函数研究"(40671081);山西省科技攻关项目"节约淡水型盐碱荒地开发利用技术研究"(2007031070)

摘　　要：以大田原生盐碱荒地土壤入渗试验数据为样本,应用支持向量机回归算法,建立了盐碱土含水率、容重、有机质含量、黏粒含量、粉粒含量、全盐量以及pH值与Philip入渗模型参数间的预测模型。预测结果表明,训练样本中吸渗率S的相对误差平均值为4.05%,稳渗率A的相对误差平均值为5.49%,90 min累积入渗量I_(90)的相对误差平均值为4.28%;检验样本中S、A和I_(90)的相对误差平均值分别为4.22%、3.58%和4.48%。可以看出,不论训练样本还是检验样本,入渗参数预测值与实测值基本吻合,所建立的预测模型精度较高,表明基于支持向量机的盐碱土壤Philip入渗模型参数的预测是可行的,可为改良盐碱土壤提供入渗参数的技术支撑。With the test data of soil infiltration in field native saline wasteland as sample, using the regression algorithm of the support vector machine, the prediction model was established among the soil moisture content, bulk density, organic matter content, clay content, silt content, soluble, pH and the Philip infiltration parameters under the condition of saline. The prediction results showed that the average relative error of the sorptivity was 4.05%, the steady infiltration rate was 5.49%, and the ninety minutes cumulative infiltration amount was 4. 28% under the training samples; the average relative error of the sorptivity, the steady infiltration rate and the ninety minutes cumulative infiltration amount were 4.22%, 3.58% and 4.48%, respectively, under the testing samples. It could be seen that beth the training samples and testing samples, the predictive values of the two infiltration parameters were well-coincident with the actual values and the accuracy of the established forecast was high. It shows that the prediction model of the Philip infiltration parameters based on the SVM is feasible under the condition of saline and it can provide technical support of infiltration parameters for improving the saline soil.

关 键 词：支持向量机 盐碱土壤 土壤入渗参 Philip入渗模型 土壤理化参数 

分 类 号：S152.7[农业科学—土壤学]