融合Transformer和LSTM的蓝莓根区土壤含水量预测模型  

A prediction model for soil moisture content in blueberry root zone by integrating transformer and LSTM

作　　者：王亿 曹姗姗 孙伟[2,3] 胡博[5] 古丽米拉·克孜尔别克 孔繁涛[4] WANG Yi;CAO Shan-shan;SUN Wei;HU Bo;Gulimi Kizilbek;KONG Fan-tao(College of Computer and Information Engineering/Engineering Research Center of Intelligent Agriculture,Ministry of Education/Xinjiang Engineering Research Center for Agricultural Informatization,Xinjiang Agricultural University,Urumqi 830052,China;Agricultural Information Institute,Chinese Academy of Agricultural Sciences,Beijing 100081,China;National Agriculture Science Data Center,Chinese Academy of Agricultural Sciences,Beijing 100081,China;Institute of Agricultural Economics and Development,Chinese Academy of Agricultural Sciences,Beijing 100081,China;Qingdao Wolin Blueberry Industry Co.,Ltd.,Qingdao 266400,Shandong,China)

机构地区：[1]新疆农业大学计算机与信息工程学院/智能农业教育部工程研究中心/新疆农业信息化工程技术研究中心,乌鲁木齐830052 [2]中国农业科学院农业信息研究所,北京100081 [3]中国农业科学院国家农业科学数据中心,北京100081 [4]中国农业科学院农业经济与发展研究所,北京100081 [5]青岛沃林蓝莓果业有限公司,山东青岛266400

出　　处：《湖北农业科学》2024年第8期78-84,共7页Hubei Agricultural Sciences

基　　金：新疆维吾尔自治区重点研发任务专项(2022B02049-1-3);中国农业科学院创新工程任务项目(HT20220570)。

摘　　要：针对土壤含水量预测模型存在难以解决非线性复杂特征、易陷入局部极小值等问题,构建融合Transformer和LSTM的土壤含水量深度学习预测模型(Transformer-LSTM)。采集山东省青岛市黄岛区丁家寨村蓝莓(Vaccinium spp.)生产区冷棚与露天2个站点的蓝莓根区土壤和气象数据作为建模数据,根据皮尔逊相关性和偏自相关性分析选择模型的数据输入特征与输入长度,与单一的Transformer模型和LSTM模型进行对比分析,评估模型对土壤含水量的预测性能。结果表明,Transformer-LSTM模型在预测精度上均优于单一的Transformer模型和LSTM模型,Transformer-LSTM模型的平均绝对误差(MAE)、均方根误差(RMSE)、平均绝对百分比误差(MAPE)、决定系数(R^(2))分别为0.245 9、0.572 0、0.012 1、0.960 6。Transformer-LSTM模型可以更全面地提取蓝莓种植环境因子输入序列中的特征信息,有效提升土壤含水量因子预测精度和水平。A deep learning prediction model for soil moisture content(transformer LSTM)was constructed,which integrated transform⁃er and LSTM,to address the difficulties in solving nonlinear and complex features,as well as the tendency to fall into local minima in the soil moisture prediction model.Soil and meteorological data from the blueberry(Vaccinium spp.)root zone of two stations,cold shed and outdoor,in the blueberry production area of Dingjiazhai Village,Huangdao District,Qingdao City,Shandong Province,were col⁃lected as modeling data,based on Pearson correlation and partial autocorrelation analysis,the data input characteristics and input length of the selected model were compared and analyzed with a single transformer model and LSTM model to evaluate the predictive performance of the model on soil moisture content.The results showed that the transformer LSTM model outperformed both the single transformer model and the LSTM model in prediction accuracy.The mean absolute error(MAE),root mean square error(RMSE),mean absolute percentage error(MAPE),and coefficient of determination(R^(2))of the transformer LSTM model were 0.2459,0.5720,0.0121,and 0.9606,respectively.The transformer LSTM model could more comprehensively extract feature information from the in⁃put sequence of blueberry planting environmental factors,effectively improving the accuracy and level of soil moisture factor prediction.

关 键 词：蓝莓(Vaccinium spp.) 根区土壤 含水量 TRANSFORMER LSTM 预测模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]