Research On Prediction Model Of Greenhouse Environment And Group Physiological Parameters Based On Deep Learning

Modern intelligent greenhouse faces greater challenges in order to meet the requirement of precise regulation with the rapid development of facility agriculture.In this paper,SPA technology is introduced to improve the regulation efficiency of modern intelligent greenhouse.Taking the net photosynthetic rate and transpiration rate of greenhouse crops as the research object,a prediction model based on deep learning was built to obtain the key physiological parameters of greenhouse crops accurately and quickly.The main research is as follows:Taking the net photosynthetic rate of tomato leaf in greenhouse as the research material,a prediction model of net photosynthetic rate of tomato based on WDNN was established.The performance of the traditional DNN prediction model was improved by the introduction of ADAM algorithm and the parallel add of Wide Network.The net photosynthetic rate prediction model of greenhouse tomato based on DCN was established by comprehensively considering environmental factors and crop physiological factors.The model takes environmental sensor data and physiological parameters of tomato that are easy to obtain as input,and the net photosynthetic rate of tomato leaf that is difficult to obtain as output.By replacing the Wide network with the Cross network,the DCN model automatically constructs the low-order Cross features,which reduces the work of manual construction of the Cross features.At the same time,by comparing with the traditional net photosynthetic rate prediction model,the advantages of DCN model in precision and convergence speed are verified.The results indicate that the DCN model could be utilized to characterize the key physiological index--net photosynthetic rate of greenhouse crops.A multi-region transpiration rate prediction model of greenhouse watermelon was established based on DCN to predict the transpiration rate of greenhouse watermelon.Firstly,the proposed feature selection method based on τ Kendall-Entropy by considering the feature correlation and the amount of information is verified.The experimental results indict that the method is reliable.Then,the accuracy and convergence speed of the model under the whole greenhouse environment and different regions were analyzed.Meanwhile,the advantages of the deep learning model in predicting the transpiration rate of greenhouse crops were verified by comparing with other types of models.Finally,Py Qt5 and QTDesigner were applied to implement the monitoring software of greenhouse environment and key physiological parameters of crops.By importing characteristic information and using the deep learning model that has been trained,the net photosynthetic rate and transpiration rate of greenhouse crops were predicted.According to the set rules,corresponding auxiliary decision-making suggestions were given.The test of the software shows that the software can accomplish the above basic functions.Environmental sensors were utilized to get the input data and then deep learning model was utilized to constract the prediction model for key physiological indexes of greenhouse crops.Through experiments and data analysis,it is verified that deep learning model has advantages in greenhouse crop parameter prediction.The results show that the established deep learning model can be used as the characterization means of the key physiological indicators,which provides a certain data support for the precise regulation of facility agriculture.