Research And Application Of Sugar Beet Fertilization Prediction Based On GRU Neural Network

The planting of sugar beet in my country is mainly distributed in the northeast.At present,most of these planting areas have adopt artificial or semi-automatic irrigation and fertilization mode,and this planting mode has not only brought about salinization of land and waste of fresh water resources,but also has affected crop yield and labor effectiveness.Therefore,how to achieve precise irrigation and fertilization has been a difficult problem facing the improvement of sugar beet yield and labor efficiency at this stage.In view of the above problems,this paper have proposed the research and application of sugar beet fertilization forecast based on GRU neural network.This article mainly studies and applies the prediction of sugar beet fertilization from three main aspects: algorithm modeling,algorithm simulation,and application realization.Firstly,the GRU neural network is constructed as the basic prediction algorithm.In order to solve the problem that the hyperparameters of the GRU neural network were difficult to determine,the improved particle swarm algorithm and Adam gradient descent algorithm are used to optimize the hyperparameters of the GRU neural network,thereby obtaining the IPSO-Adam-GRU prediction model.Secondly,the IPSO-Adam-GRU model is compiled by Python to simulate and analyze the application of nitrogen,phosphorus,potassium fertilizers and soil moisture in multiple sugar beet plots;the predicted fertilizer application can be irrigated and fertilized by the method of fertilizer following water flow.The fuzzy PID control is modelled by MATLAB,the predicted soil moisture is took as the input value,and the sugar beet required moisture as the set value,so as to control the opening and closing time of the solenoid valve for irrigation and fertilization.Finally,the intelligent system for predicting and controlling sugar beet irrigation and fertilization is based on the Web platform has used HTML to compile relevant pages,including: data management,fertilization prediction,irrigation and fertilization control,scientific planting,and user management pages.The experimental results have showed that the prediction accuracy of the IPSO-Adam-GRU model is 35%?45% higher than that of the PSO-Adam-GRU model,GRU model,and BP model,thus precise fertilization of sugar beet is realized.Therefore,the prediction of sugar beet fertilization based on GRU neural network proposed in this paper has had certain practical significance.