首页|基于ME-BiLSTM模型的苜蓿叶面积指数预测方法

基于ME-BiLSTM模型的苜蓿叶面积指数预测方法

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
连续时序的叶面积指数(Leaf Area Index，LAI)可反映苜蓿长势的变化情况，预测苜蓿未来时段的LAI对指导田间管理决策具有重要作用。针对LAI数据采集困难，导致苜蓿时序LAI存在训练数据不足的问题，该文以生长天数为自变量，采用修正的Logistic模型对实测苜蓿LAI变化的动态过程进行建模，根据LAI模拟曲线进行数据插补，从而构建宁夏引黄灌区试验区3 年的逐日苜蓿LAI数据集。在插补数据集的基础上，为解决苜蓿刈割后数据突变问题，提出了一种ME-BiLSTM模型。该模型集成移动累计和检验方法(MOSUM)以及基于双向长短期记忆网络(BiLSTM)的编码器-解码器神经网络。MOSUM方法可以实现LAI数据集中突变点检测，并剔除包含突变点训练批次，同时应用改进的BiLSTM模型进行预测。结果表明:ME-BiLSTM模型能较好地进行苜蓿LAI未来曲线变化的预测，其决定系数(R2)、均方根误差(RMSE)值分别为0。998 5 和0。072 2。对于苜蓿生长的各个茬次，预测模型对于第1 茬、第4 茬的预测精度最高，第2 茬和第3 茬的预测精度稍有降低。
Prediction Method of Alfalfa Leaf Area Index Based on ME-BiLSTM Model
The continuous temporal Leaf Area Index(LAI)reflects the changes in alfalfa growth.Predicting future LAI in alfalfa plays a crucial role in guiding field management decisions.Aiming at the problem of insufficient training data for alfalfa temporal LAI due to dif-ficulties in LAI data collection,we employ the growth days as independent variables and utilize a modified Logistic model to dynamically model the observed changes in alfalfa LAI.By interpolating data based on the simulated LAI curve,a three-year daily alfalfa LAI dataset for the Ningxia Yellow River Irrigation District experimental area is constructed.To address abrupt data changes after alfalfa cutting,we introduce the ME-BiLSTM model which integrates the Moving Sum and Moving Average(MOSUM)method with a Bidirectional Long Short-Term Memory(BiLSTM)encoder-decoder neural network.The MOSUM detects mutation points in the LAI dataset and eliminates training batches containing these points,followed by predictions using the improved BiLSTM model.It is demonstrated that the ME-BiLSTM model predicts future alfalfa LAI curve changes effectively,with coefficient of determination(R2)and root mean square error(RMSE)values of 0.998 5 and 0.072 2,respectively.The first and fourth alfalfa growth cycles have the best predictive model ac-curacy,whereas the second and third cycles have slightly lower accuracy.

alfalfaleaf area indexLogistic modelMOSUMbidirectional long short-term memory network

杨松涛、葛永琪、王静、刘瑞

展开 >

宁夏大学 信息工程学院,宁夏 银川 750021

苜蓿 叶面积指数 Logistic模型 MOSUM 双向长短期记忆网络

国家自然科学基金地区科学基金项目国家自然科学基金地区科学基金项目宁夏回族自治区重点研发计划宁夏回族自治区重点研发计划宁夏自然科学基金宁夏自然科学基金

62262052621620522021BEB040162022BDE030072021AAC030412022AAC03004

2024

10.20165/j.cnki.ISSN1673-629X.2024.0058

计算机技术与发展
陕西省计算机学会

计算机技术与发展

CSTPCD
影响因子：0.621
ISSN：1673-629X
年,卷(期)：2024.34(5)
  • 25