Prediction Of Rice Water Demand Based On Time Series Convolutional Network And System Development

Rice is one of the main food crops in my country.With the increase of rice planting area,the demand for water resources is expanding,and the contradiction between supply and demand is becoming more and more serious.Accurately predicting the water demand of rice can save agricultural irrigation water,determine the irrigation cycle in advance,and ensure agricultural production and production.Food security,water demand prediction has become an important research content of rice water-saving irrigation.Due to the need to obtain more meteorological parameters in the process of rice water demand calculation,it is difficult to calculate rice water demand,and the prediction accuracy of traditional machine learning methods is low.Therefore,in order to make rice water demand prediction more convenient and have higher prediction accuracy,it is necessary to design and develop a convenient,intelligent and accurate rice water demand prediction system.This paper takes the water demand of rice in Qingan and Xiangyang,Heilongjiang Province as the research object,uses deep learning technology to predict the water demand of rice during the growth period,and designs and develops a rice water demand prediction system.The main work is as follows:First,in view of the difficulty in calculating water demand,the modified Penman formula and coefficient method were used to calculate the water demand during the growth period of rice.First,download meteorological data such as the average temperature and sunshine hours in Qingan and Xiangyang from 2010 to 2020 from the China Meteorological Science Data Network.Second,use the modified Penman formula to calculate the ET₀of the two regions.Third,the coefficient method was used to calculate the water demand of rice in the growth period,and the variation law of rice water demand in the two regions and the difference of water demand in different growth periods were analyzed.The analysis results showed that the water demand of rice in the two areas changed periodically,and the water demand was the largest in the tillering stage,and the total water demand in Xiangyang area was lower than that in Qingan area.Secondly,in view of the low accuracy of rice water demand prediction,a rice water demand prediction model based on time series convolutional network is proposed.First,the zero-mean normalization method was used to normalize the water demand of rice.The data from 2010 to 2018in Qingan and Xiangyang were used as the training set,and the data from 2019 to 2020 was used as the test set.Error,mean square error and root mean square error are three evaluation indicators to evaluate the performance of the model.Second,a rice water demand prediction model based on time series convolutional network is proposed,which adjusts the model hyperparameters（time step,batch size and number of iterations）to improve the prediction accuracy of the model.Third,compare the prediction accuracy of time series convolutional network and support vector regression,random forest,long and short-term memory network,and gated unit network.The experimental results show that among the five rice water demand prediction models,when the time step of the time series convolutional network is set to 14,the batch size is set to 16,and the number of iterations is set to 500,it has the best performance.The water demand forecast has the best effect,with the mean absolute error,mean square error and root mean square error of 0.07,0.33,and0.42,respectively.Finally,aiming at the inconvenient problem of rice water demand forecasting,a rice water demand forecasting system is designed and implemented.First,analyze the requirements that the system should meet from the perspective of users and business,and design the overall framework of the system on this basis.Second,use Pyqt5 to build a rice water demand forecasting system.The system is divided into six functional modules,namely user login and registration module,planting area overview module,meteorological data acquisition and trend visualization module,rice water demand calculation and trend visualization module,Deep learning model training module,rice water demand prediction module.Third,perform functional and systematic tests on the system.The functional test results show that the system is fully functional and can meet various user needs.The performance test results show that the average response time of each module is 3 seconds,and the system can provide a relatively good user experience.The TCN-based rice water demand prediction system designed and developed in this paper provides a solution for the inconvenient acquisition of meteorological data,the complexity of rice water demand calculation and the difficulty in establishing a prediction model,and provides technical support for rice water demand prediction and water-saving irrigation in Heilongjiang Province It can also provide ideas and references for the development of water demand forecasting systems for crops such as corn and wheat.