融合叶绿素含量的黄瓜幼苗光合速率预测模型  

Photosynthetic Rate Prediction Model of Cucumber Seedlings Fused Chlorophyll Content

作　　者：张海辉[1] 陶彦蓉[1] 胡瑾[1] 

机构地区：[1]西北农林科技大学机械与电子工程学院,陕西杨凌712100

出　　处：《农业机械学报》2015年第8期259-263,307,共6页Transactions of the Chinese Society for Agricultural Machinery

基　　金：'十二五'国家科技支撑计划资助项目(2012BAH29B04);陕西省科学技术研究发展计划资助项目(2013K02-03;2014K02-08-03)

摘　　要：现有的基于神经网络的光合速率模型仅考虑环境因素,且收敛速度慢。在考虑温度、CO2浓度、光照强度、相对湿度等环境因子的基础上,加入生理因子叶绿素含量,建立融合叶绿素含量的黄瓜幼苗光合速率预测模型。首先利用多因子嵌套试验获得黄瓜幼苗光合速率测试数据825组,然后采用LM训练法进行模型训练,并分析加入叶绿素含量对模型训练结果的影响,最后建立黄瓜幼苗光合速率预测模型并对其采用异校验方式进行验证。试验结果表明,在考虑叶绿素影响的条件下,其训练效果与模型拟合度均优于只考虑环境因子的训练模型,加入叶绿素含量作为输入的LM训练法可有效越过局部平坦区,具有明显的优越性,满足误差小于0.000 1的训练要求,模型预测值与实测值间的决定系数为0.987,误差小于4.68%。As the environmental factors were only considered in the existing photosyntlaetlc rate prealcuon models based on neural network, slow convergence speed was still the existing problem. The temperature, CO2 concentration, photon flux density and relative humidity, especially the chlorophyll content were considered. Photosynthetic rate prediction model of cucumber seedlings fused chlorophyll content was proposed. Firstly, 825 experimental data of cucumber seedlings photosynthetic rate were obtained by multi-factor coupling test. The temperature gradients were set at 16, 20, 24, 28, 32℃, respeetively, CO2 concentration gradients were set at 300,600,900, 1 200, 1 500 μL/L and the photon flux density gradients were set at 0, 20, 50, 100,200, 300, 500,700, 1 000, 1 200, 1 500 txmol/（m^2·s）, respectively. Secondly, Levenberg- Marquardt （LM） training method was used. Meanwhile, the effect of chlorophyll content on the training results was analyzed. Then different calibrations were used to validate the multi-factor coupling photosynthetic rate prediction model. The results showed that the training results of the training method considered chlorophyll content and the model fitting degree were superior to the training model ordy considered the environmental factors. Because of the local area, LM training method considering chlorophyll content can effectively fiat over the local area and meet the training requirement. The error rate was less than 0. 000 1 and the determination coefficient between actual measured and calculated values was 0. 987. It indicated that these two values had good correlation and similarity. Besides, the error was less than 4. 68%, which proved that the proposed model has a high accuracy.

关 键 词：黄瓜 叶绿素含量 光合速率 预测模型 LM 训练法 

分 类 号：S126[农业科学—农业基础科学]