Prediction Model And Application Of Grain Yield In The Middle And Lower Reaches Of The Yellow River

Food is the cornstone of social stability,economic growth,and national security,which also serves as the backbone of agriculture.The downstream area of the Yellw River is fertile and receives a lot of precipitation.It is a significant grain producing region in our nation,with an annual grain output that exceeds 30% of China’s total.To secure our nation’s grain supply and food security,it is crucial to forecast grain output in this area.Thus,it is a pressing issue to employ a method that is more effective and logical in order to increase the precision of grain output forecasts in the Yellow River’s middle and lower reaches.The majority of current research objects on grain yield prediction in the middle and lower sections of the Yellow River concentrate on provinces and cities in the area,with relatively few studies being undertaken on regional grain yield prediction as a whole.Moreover,a poor fitting effect of grain yield data is caused by an insufficient selection of affecting elements,and a time series model with a simple structure has poor short-term prediction accuracy.For a neural network model,predicting the medium-and long-term development trend of grain yield is challenging.The primary research focus of this paper is as follows,and it is based on the existing status and characteristics of grain yield forecasts in the middle and lower portions of the Yellow River:First,in order to analyze the influencing factors of grain yield in this area,a combination analysis method based on GRA,correlation coefficient analysis,and machine learning was proposed after a detailed analysis and summary of the development trend and current situation of various factors affecting grain yield in the middle and lower reaches of the Yellow River.Eventually,the primary influencing elements of grain yield in this region were identified as being the total power of agricultural machinery,effective irrigation area,conversion quantity of fertilizer application,and rural electricity consumption.Second,a time series model is created by using time series to describe historical grain yield data.In view of the gray property,non-stationary and dynamic randomness of grain yield data,the idea of recursion was used to improve the VMD.Data on grain yield was divided into several IMFs,and these functions’ properties were examined.To forecast the IMFs,the GM(1,1)and ARIMA models were employed.In order to actualize the short-term forecast of grain output in the middle and lower portions of the Yellow River,IMF was then rebuilt as the final projected outcome.Thirdly,the mapping relationship between influencing factors and grain yield was established,and a prediction model based on CNN,Bi LSTM and Bi GRU was proposed.The main influencing factors of grain yield were further extracted through one-dimensional dilatation convolution and attention mechanism,and then two hidden layers established by Bi LSTM and Bi GRU were established to mine the timing features inside the data and forecast,and finally the prediction results were output.Based on time series and affecting factors,the model’s prediction findings and actual values had errors of 1.98% and 1.16%,respectively.The two models presented in this study have relatively high prediction accuracy when compared to other conventional models,significantly enhancing prediction accuracy and offering fresh ideas for forecasting grain yield in the Yellow River’s middle and lower reaches.