尿素对土壤水分传感器测量精度的影响  

Effect of Urea on Measurement Accuracy of Soil Moisture Sensor

作　　者：石玉娇 尹峥 王红叶 刘晓辰 石庆兰[1,3] 梅树立 SHI Yujiao;YIN Zheng;WANG Hongye;LIU Xiaochen;SHI Qinglan;MEI Shuli(College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China;Center of Cultivated Land Quality Monitoring and Protection,Ministry of Agriculture and Rural Affairs,Beijing 100125,China;National Innovation Center for Digital Fishery,Beijing 100083,China)

机构地区：[1]中国农业大学信息与电气工程学院,北京100083 [2]农业农村部耕地质量监测保护中心,北京100125 [3]国家数字渔业创新中心,北京100083

出　　处：《农业机械学报》2020年第S02期388-394,407,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(61871380)。

摘　　要：土壤水分的精准测量对节水灌溉、墒情监测、水肥一体化等领域具有重要意义,土壤氮含量会影响水分传感器的测量。为了消除这种影响,设计了不同尿素质量对不同水分含量土壤样本的监测实验,采用高灵敏度水分传感器并对尿素干扰下的输出电压进行监测,通过称重法监测土壤样本的含水率,使用LCR电桥测试仪监测土壤样本的电容和电阻。为了研究氮含量影响水分测量的机理,根据实验数据建立了三元三次多项式、BP神经网络、深度学习3种预测模型,并对预测结果进行误差分析。结果表明,相同土壤含水率条件下,尿素质量与土壤水分传感器输出值呈周期性的振荡关系。3种预测模型的平均绝对误差分别为0.77%、0.64%、0.75%,BP神经网络模型有98%误差集中在0~2%区间,误差峰值仅为2.07%,确立BP神经网络模型为最佳抗尿素干扰水分预测模型。The accurate measurement of soil moisture is of great significance to the fields of water-saving irrigation,moisture monitoring,water and fertilizer integration,and the soil nitrogen content will affect the measurement of the moisture sensor.In order to eliminate this effect,a monitoring experiment of different urea on soil samples with different moisture contents was designed,a high-sensitivity moisture sensor was used to monitor the output voltage value under the interference of urea,and the moisture content of the soil sample was monitored by weighing method,the capacitance and resistance of the soil sample was monitored by LCR bridge tester.Totally 800 sets of sample data were obtained,of which 75%were used as the training set and 25%were used as the validation set.In order to study the mechanism of the influence of nitrogen content on moisture measurement,three predictive models,including a three-element cubic polynomial model,a BP neural network model,and a deep learning model,were established based on experimental data,and error analysis was performed on the prediction results.The analysis showed that the highest errors of the three models were 2.86%,2.07%and 3.82%,and the errors were concentrated in the range of 0 to 2%,accounting for 89%,98%and 90%,respectively.The following conclusions were obtained:different urea contents had different influences on the predicted value,which was roughly in a periodic oscillation relationship.When the soil moisture content was lower,the interference of urea content on voltage was greater,and the same was true for impedance,but capacitive reactance was only sensitive to soil moisture,but not to changes in urea.Therefore,the interference of urea on soil moisture measurement was mainly caused by interference with soil resistance.The average absolute errors of the three-dimensional cubic polynomial,BP neural network,and deep learning models were 0.77%,0.64%and 0.75%,respectively,and BP neural network model was the most stable,with the most prediction results concentrated in

关 键 词：土壤水分传感器 尿素 多项式回归 BP神经网络 深度学习 

分 类 号：S151.9[农业科学—土壤学] S126[农业科学—农业基础科学]