Wheat Yield Forecast And Visualization System Based On HP Filter And BP Neural Network

Wheat is one of the three major food crops in the world,and its yield accounts for about one third of the total grain yield in China.As the granary of Central Plains,it is of great significance for Henan to maintain the stability of planting area and yield to ensure national food security.Scientific and accurate yield prediction can provide the basis for the relevant departments to set grain production goals,guide agricultural production,promote farmers’ income,and provide technical support for carrying the important task of stabilizing grain production.Based on the temperature and rainfall data of 15 meteorological stations in Henan Province from 1984 to 2018,this paper divided the climate year type by the combination of accumulated temperature year type and rainfall year type,divided the yield year type according to the wheat yield data,and then summarized and analyzed the general law between the climate year type and yield year type.Based on the five major wheat regions in Henan Province,including North,East,West,South and Nanyang Basin,the meteorological yield and trend yield in the process of wheat production were separated by using HP filtering method.The meteorological yield prediction model was constructed by using BP neural network driven by meteorological factors,and the trend yield of wheat was predicted by using linear regression model Finally,a wheat yield prediction and visualization system was developed by using visualization technology.The main results are as follows.1.Through the division of accumulated temperature,rainfall and yield year types in Henan Province,the differences of different climate year types and their effects on yield year types were analyzed.The results show that the 35 year accumulated temperature years of 15 meteorological stations in Henan Province from 1984 to 2018 include 99 years of moderate warm years,314 years of normal temperature years and 112 years of cold temperature years;159 years of moderate wet years,183 years of normal water years and 183 years of dry years;145 years of high yield years,248 years of normal yield years and 132 years of low yield years.The results show that the accumulated temperature years in Henan Province are mainly warm years and normal years,and the distribution of rainfall years is relatively uniform.The climate years are mainly normal years and dry years.In warm and wet years,wheat is easy to obtain high yield,and the frequency of high yield years is 76.9%.In cold and wet years,wheat is not easy to obtain high yield,and the frequency of low yield years is 67.9%.In meteorological data,accumulated temperature is the main factor affecting the fluctuation of wheat yield,and the frequency of warm and wet years is 76.9% It is easier for wheat to have high yield under annual type,and the probability of low yield is higher in cold year.2.Based on the separation of meteorological yield and trend yield of wheat in Henan Province,a comprehensive prediction model of wheat yield was established,and the wheat yield in five major wheat regions of Henan Province was predicted.The results showed that the prediction effect of wheat yield was different with different rainfall and water management.Compared with the actual yield,the prediction error of wheat yield in the dry wheat region of western Henan was the largest,with a relative error of 0.58%,and that of wheat yield in Northern Henan was the highest The results show that the prediction and error of the three other wheat regions are 0.36%,0.48% and 0.38% respectively,which indicates that it is feasible to use HP filter and BP neural network technology to predict wheat yield.3.The wheat yield prediction and visualization system platform was established.That is to say,the Map Reduce technology of parallel computing is used to calculate accumulated temperature and rainfall,and the judgment and analysis of climate year type are realized by combining with the national standard of crop climate year type;the HP filtering method is programmed by python,and a high pass filter is designed by the principle of data moving average method to realize the function of wheat yield separation;the encapsulation technology of Java programming is used to realize the function of yield prediction Through the restful interface of springboot server,the packaged model is called to realize the output prediction in the system.Based on ecarts framework,map(),bar(),pie(),line()methods are constructed to realize the visual display of accumulated temperature and rainfall,yield separation and prediction function results.Users can carry out relevant function operation and real-time visual view.