Research On Forecasting Model Of Winter Wheat Yield In Anhui Province Based On Meteorological Data

As a major agricultural province,the grain output of Anhui Province plays an important role in the whole country.In recent years,due to the changeable climate,the grain production in many areas of Anhui Province has been reduced,which brings a lot of uncertainty to agricultural production.Therefore,it is necessary to use meteorological data to predict the grain production.In this paper,Anhui Province winter wheat as the research object,temperature,precipitation and sunshine hours as the three main meteorological factors to carry out winter wheat yield prediction model research.The specific work is as follows:1)The selection and preprocessing of meteorological and winter wheat yield data were completed.Based on the database of"monitoring and early warning platform for crop diseases and insect pests in Anhui Province"and"Anhui Statistical Yearbook",Fuyang City in Anhui Province was selected as the representative area,and the winter wheat yield data,average temperature,precipitation and sunshine hours in the growth period from 2006to 2018 in this area were selected as the sample data,and the meteorological sample data were divided into 18 meteorological factor variations Quantity.2)Time trend yield and meteorological yield were separated.Generally,winter wheat yield can be divided into two parts:time trend yield and meteorological yield.First of all,the meteorological yield and time trend yield in the total winter wheat yield of the experimental year are separated by time regression analysis,and the time trend yield prediction model is established;then,the meteorological yield prediction model is constructed by using the meteorological data and the separated meteorological yield data,using the stepwise regression analysis method and BP neural network algorithm respectively.3)The prediction model of winter wheat yield based on meteorological data was studied.First,six main meteorological factors which are significantly related to meteorological output are selected from 18 meteorological factor variables by stepwise regression analysis as independent variables,and then a prediction model of meteorological output is constructed according to the parameters in the results of stepwise regression analysis;then,the six main meteorological factors selected by stepwise regression analysis are used as the input of neural network by BP neural network algorithm,and the gas As the output of the neural network,the output data is used to set the parameters of the neural network and train the prediction model.4)The comparative analysis between the two prediction models is carried out.Based on the prediction results of meteorological production,the advantages and disadvantages of the two prediction models are compared from the overall prediction accuracy,average relative error and root mean square error.The results show that the overall prediction accuracy of the model established by BP neural network algorithm is 84%,which is 4%higher than that of the model established by stepwise regression analysis(80%),the average relative error is 0.19,and the root mean square error is 44.37,which are lower than the average relative error(0.20)and root mean square error(74.59)of the model established by stepwise regression analysis The model established by the algorithm has better effect on the prediction of meteorological output.Taking 2019 as the trial year,combining the time trend yield prediction model and BP neural network algorithm prediction model,the prediction value of winter wheat yield in 2019 is 7157.726kg/hm~2,compared with the actual yield of 6355kg/hm~2,the prediction accuracy is 88%.