| 刘婧然,武海霞,刘心,刘真,王鹏宇,张有强,李玉琼.基于深度学习人工神经网络的青椒调亏灌溉水量预测[J].干旱地区农业研究,2021,39(6):105~112 |
| 基于深度学习人工神经网络的青椒调亏灌溉水量预测 |
| Water quantity prediction of regulated deficit irrigation for green peppers based on deep learning artificial neural network |
| |
| DOI：10.7606/j.issn.1000-7601.2021.06.12 |
| 中文关键词: 调亏灌溉 滴灌 人工神经网络 深度学习 青椒 预测 |
| 英文关键词:regulated deficit irrigation drip irrigation artificial neural network deep learning green peppers prediction |
| 基金项目:河北省研究生创新资助项目（CXZZBS2018004)；河北省自然科学基金资助项目（D2019402151）；河北省科技支撑计划项目（17226914D）；河北省创新能力提升计划科技研发平台建设专项（18965307H） |
|
| 摘要点击次数: 508 |
| 全文下载次数: 81 |
| 中文摘要: |
| 在2014—2018年，采用垄沟集雨覆盖种植滴灌技术与调亏灌溉技术相结合（MFR-RDI）对青椒进行试验研究，选取灌溉水利用效率最高的试验处理（即青椒结果后期重度亏水）进行灌水量预测。根据试验期间搜集的各项资料，在MFR-RDI种植方式下，以作物需水量、青椒生育期天数、作物生育期内的降水量、土壤含水率、前一天的灌水量作为模型输入因子，构建青椒作物灌水量的深度学习人工神经网络（DNN）预测模型。通过模型试验得到最佳DNN预测模型，该模型的隐含层包括4层，各隐含层神经元个数分别为：32、16、8、4。模型的激活函数采用“ReLU”，优化函数为“adam”，迭代次数为300。模型使用2018年的数据进行了测试。测试结果表明DNN模型的RMSE为0.898 mm，MAE为0.257 mm，NS为0.758，R2为0.7635，说明该预测模型具有较高的精度性能。通过预测结果可以得到此种植方式下青椒的灌溉制度，为实现高效智能节水灌溉提供参考。 |
| 英文摘要: |
| A combination of ridge and furrow rain\|collecting and rain\|covering planting drip\|irrigation technology and regulated deficit irrigation technology (MFR-RDI) was adopted from 2014 to 2018 to carry out an experimental study on green peppers. The experimental treatment with the highest utilization efficiency of irrigation water (severe deficit water in the later period of green peppers results) was selected to predict the irrigation water amount. Based on the data collected during the experiment, the deep learning artificial neural network (DNN) prediction model of green pepper crop irrigation water amount was established under the MFR-RDI planting method. The model took crop water requirement, growth period of green pepper, precipitation, soil water content and irrigation amount of the previous day as input factors. Through model test, the best DNN prediction model was obtained. The hidden layer of the model included 4 layers, and the number of neurons in each hidden layer was 32, 16, 8, and 4, respectively. The activation function of the model was “ReLU”, the optimization function was “adam”, and the number of iterations was 300. The model was tested using data from 2018. The test results showed that the RMSE of DNN model was 0.898 mm, MAE was 0.257 mm, NS was 0.758, and R2 was 0.7635, indicatinga high accuracy performance of the prediction model. Through the prediction results, the irrigation system of green peppers under this planting method can be obtained, providing reference for realization of efficient and intelligent water\|saving irrigation. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| | |