基于原位EIS参数与改进BiGRU算法的侧柏苗木生长预测研究

Research on growth prediction of Platycladus orientalis seedlings based on in situ EIS parameter and improved BiGRU algorithm

刘扬 ¹钱稷 ²陈竹筠 ³刘旋 ³任振辉³

扫码查看

作者信息

1. 河北农业大学机电工程学院,河北保定 071001;河北软件职业技术学院计算机应用工程系,河北保定 071000
2. 河北农业大学园艺学院,河北保定 071001
3. 河北农业大学机电工程学院,河北保定 071001
折叠

摘要

侧柏苗木的生长预测对于苗木未来的出圃时间、调整现有苗木的种植方案等方面意义重大.本文运用原位电阻抗图谱技术发现4 kHz下的阻抗相位角、400 kHz下阻抗正切值可以表示侧柏根系的健康程度,与茎粗、株高、分枝数等指标共同构成代表侧柏苗木质量的特征向量;同时,设计了基于多维双向的时空融合模型(CNN-DBiGRU-WBLS),该模型可以有效解决多因素输入下未来4周、8周、16周的侧柏生长数据预测问题,Willmott一致性指数(WI)值达到 0.993、0.975 和 0.948,比 CNN-BiGRU、CNN-LSTM、CNN-BiLSTM和CNN-GRU等现有模型WI最优值提升了 0.6％、1.1％和1.5％以上.试验结果表明,本文算法对侧柏苗木生长指标具有较好的预测性能.

Abstract

The growth prediction of Platycladus orientalis seedlings is of great significance for the future nursery time and the adjustment of the existing planting plan.In this paper,we used in situ electrical impedance spectroscopy technology and found that the phase Angle of impedance at 4 kHz and the tangent value of electrical impedance at 400 kHz was able to represent the root health degree of Platycladus orientalis seedlings.They constituted the feature vector together with stem diameter,plant height and branch number,they constituted the feature vector representing the quality of Platycladus orientalis seedlings.And the same time,a multi-dimensional bidirectional space-time fusion model(CNN-DBiGRU-WBLS)was designed,which can effectively solve the problem of growth data prediction at future 4-week,8-week and 16-week under multi-factor input.The Willmott Index of Agreement(WI)value reached 0.993,0.975 and 0.948,which were 0.6％ ,1.1％ and 1.5％ higher than the WI optimal values of existing models such as CNN-BiGRU,CNN-LSTM,CNN-BiLSTM and CNN-GRU.The experimental results showed that the algorithms used in this paper performed well in predicting the growth of Platycladus orientalis seedlings.

关键词

侧柏苗木/EIS/生长预测/BiGRU/CNN

Key words

Platycladus orientalis seeding/EIS/growth prediction/BiGRU/CNN

引用本文复制引用

基金项目

河北省重点研发计划项目(22327203D)

出版年

2024

河北农业大学学报

河北农业大学

河北农业大学学报

CSTPCDCSCD北大核心

影响因子：0.475

ISSN：1000-1573

参考文献量18

段落导航