长短期记忆神经网络在多时次土壤水分动态预测中的应用

基于长沙站2016—2019年10 cm深度土壤水分自动观测小时数据集,利用长短期记忆神经网络(LSTM)模型结合随机采样学习方法,开展了土壤水分多时次预测,结果表明:LSTM模型对6、12、24、48h后的土壤体积含水量预测均方根误差(RMSE)分别为0.22%、0.28%、0.38%、0.54%,决定系数(R2)分别为0.99、0.99、0.98、0.96,除6 h预测步长外,准确率均优于自回归整合滑动平均(ARIMA)模型,且误差稳定、无异常值出现,预测准确率远优于相关研究。该结果证实了基于LSTM模型精准预测土壤水分动态的可行性,为精准灌溉和干旱预警提供了计算机技术及手段支撑,为政府及科研部门水资源管理政策的制定提供了数据支持。

Application of Long/Short Term Memory Neural Network in Soil Moisture Multi-time Dynamic Prediction

Accurate prediction of soil moisture will help improve the utilization efficiency of water resources for agricultural production. Based on the data set of hourly soil moisture automatic observation at 10cm depths from 2016 to 2019 in Changsha station, this study used the neural network of Long Short-Term Memory(LSTM) combined with random sampling learning to carry out multi-time prediction of soil moisture. The results showed that the Root Mean Square Error(RMSE) of prediction in 6, 12, 24, 48 hours was 0.22, 0.28, 0.38, 0.54%, and Coefficient of Determination(R2) was 0.99, 0.99, 0.98, 0.96, respectively. The prediction accuracy was better than Autoregressive Integrated Moving Average model(ARIMA) except 6 time steps, the deviation was stable and no abnormal value appeared, the prediction accuracy was far better than relevant studies. The results prove that the feasibility of accurately predicting in soil moisture dynamics based on LSTM model, make contribution to computer technology and methods support for precise irrigation and drought warning, and providing data support for the formulation of water resource management policies by government and research institutions.