Sugarcane Yield Prediction Based On Field Iots Combined With Meteorological Data

As China’s largest sugarcane producing area,Guangxi ensures the supply of sugar in China.The quality of sugarcane is closely related to environmental factors like temperature,humidity,light intensity and rainfall.With the rapid development and extensive use of the agricultural Internet of Things,a large amount of agricultural data has been generated in the agricultural production process.How to use the data of the agricultural Internet of Things monitoring to better predict the sugarcane production is the main research pur:pose of this paper.This paper firstly designed a Field Internet of Things monitoring system based on ZigBee technology.The data collected by the sugarcane field sensor was uploaded to the gateway node through the built wireless sensor network,and then the data was uploaded to the server through the GPRS module and published in the form of a webpage to realize the viewing and sharing of data.Further,in order to effectively analyze the data collected by the Internet of things and better guide the planting and management of sugarcarne,this paper carried out in-depth mining and analysis on the environmental data of four key periods of germination,seedling,tillering and elongation of sugarcane crops.Including:preprocessing,hierarchical clustering,evaluation grading.The results of data clustering showed that the CPCC values of each growth period reached 0.8491,0.8355,0.8239 and 0.8175,respectively,indicating that the hierarchical clustering method had a good classification effect on environmental data and can provide guidance for optimizing sugarcane planting.Finally,this paper useed the eight factors of precipitation,average temperature,maximum temperature,minimum temperature,soil moisture mean,soil temperature minimum,soil temperature mean and soil temperature maximum as the input factors,sugar cane yield as the output,using stepwise regression,BP neural network and genetic algorithm optimized BP neural network to predict the sugarcane yield of Anji experimental irrigation station,and by comparing the prediction results of these three models,it was found that the BP neural network model based on genetic algorithm was the smallest average relative error(0.64%)in the prediction set for predicting sugarcane yield,lower than BP neural network(5.66%)and stepwise regression model(8.97%),indicating that the BP neural network with genetic algorithm had the greatest predictive potential.