基于BP神经网络的土壤水热动态预测模型研究

发布时间：2018-08-18 17:57

【摘要】：以土壤施氮量、灌水量、土壤各层初始温度、各层初始含水率以及时间作为神经网络的输入因子,以土壤温度、含水率作为输出因子,基于BP神经网络,建立了拓扑结构为17-21-14的BP-W-T预测模型,利用Matlab软件对BP神经网络进行训练,并结合实测值对温度分布和含水率分布进行检验。结果表明:土壤温度预测模型模拟值与实测值的平均相对误差、相关系数和决定系数分别为2.28%~3.36%、0.954~0.972和0.910 2~0.944 7;含水率预测模型模拟值与实测值的平均相对误差、相关系数和决定系数分别为1.87%~3.09%、0.974 3~0.992 6和0.949 3~0.985 2;该预测模型具有较高的预测精度和良好的稳定性,可以较好地描述温度梯度下土壤水热动态变化情况,为预测土壤温度与含水率提供了一条新的途径。
[Abstract]:Soil nitrogen application amount, irrigation amount, initial temperature of each layer, initial moisture content of each layer and time were taken as input factors of neural network, and soil temperature and water content were taken as output factors based on BP neural network. A BP-W-T prediction model with a topological structure of 17-21-14 is established. BP neural network is trained by Matlab software, and the temperature distribution and moisture content distribution are tested with the measured data. The results showed that the average relative error, correlation coefficient and determination coefficient of the simulated and measured values of soil temperature prediction model were 2.283.36 and 0.910 20.97, respectively, and the average relative error between the simulated value and the measured value of the water content prediction model was obtained. The correlation coefficient and the determining coefficient were 1.874.3996 and 0.949 30.9852, respectively. The prediction model had high prediction precision and good stability, and could describe the dynamic changes of soil water and heat under the temperature gradient, and the correlation coefficient and the determination coefficient were 3.0974 30.992 6 and 0.949 30.985 2, respectively. It provides a new way to predict soil temperature and moisture content.
【作者单位】：太原理工大学水利科学与工程学院;
【基金】：国家自然科学基金项目(51579168,51249002) 山西省自然科学基金项目(201601D011053) 山西省科技攻关项目(20140311016-6)
【分类号】：S152;TP183

【相似文献】