Study On Intelligent Water Requirements Perception And Water Saving Irrigation Decision-making Of Green Peppers Drip Irrigation With Rainwater Harvesting And Regulated Deficit

In recent years,with the development of economy,the problem of water shortage is becoming more and more serious.Irrigation water accounts for about 70%of the world's water resources.Developing smart agriculture,forecasting crop water demand and realizing smart irrigation are particularly important to save water and solve the problem of water shortage.In this paper,the precision irrigation test site(the original site)of Hebei University of Engineering was taken as the test site,and green peppers was taken as the test object.The experiments of covered rainwater harvesting and regulated deficit drip irrigation(MFR-RDI)and traditional plain full irrigation were carried out from 2014 to 2018.Collect all the relevant data of meteorological crops,soil for the suitable water saving irrigation methods,on the basis of crop water requirement forecasting model,establish water-saving irrigation decision system as the goal,the integrated use of multidisciplinary technology such as the ministry of artificial intelligence and the Internet of things,perception and our irrigation water requirement for regional agricultural intelligence decision system to study related issues.The paper selects the planting mode with the highest utilization efficiency of irrigation water(IWUE)under the MFR-RDI planting mode to carry out the intelligent prediction of the water demand of green pepper,and on this basis,establishes the irrigation decision system,and finally builds the decision system platform.The research results have important guiding significance for the water-saving irrigation of green pepper planting in Handan area.The main research contents and results are as follows:(1)Combined with rainwater harvesting technology and regulated deficit drip irrigation technology,field experiments were conducted to collect test data and make statistical analysis.it was concluded that under sufficient irrigation conditions,rainwater harvesting drip irrigation can significantly improve fruit yield,V_c content,and IWUE compared with traditional plain cropping.In the covered rainwater harvesting and drip irrigation,regulated deficit irrigation can significantly improve fruit V_c content than full irrigation(CK_1R).Among them,the IWUE of T_8R was the highest in 2014-2018,and there was no significant difference between T_8R and CK_1R in green pepper yield from 2015 to 2018,and the fruit V_ccontent was higher.Therefore,based on the test data obtained from T_8R of the highest IWUE,an irrigation decision-making system is established to save irrigation water to the maximum extent.(2)The intelligent forecasting models of green peppers crop water requirements was established based on support vector(SVM)and Elman neural network optimized by Genetic Algorithm(GA)and Elman neural network optimized by Mind Evolutionary Algorithm(MEA).The results show that under the same input factors,the results of GA-Elman was better than GA-SVM,and the performance of MEA-Elman was better than GA-Elman.The accuracy of the optimized artificial intelligence prediction model can be improved by introducing canopy temperature into the model input factors.In addition,selecting different input factors to predict crop water demand at different growth stages can further improve the accuracy of the prediction model.Root Mean Square Error(RMSE),Mean Absolute Error(MAE)and Nash-Sutcliffe coefficient(NS)values of the prediction model were 0.359 mm/d,0.294 mm/d and 0.941,respectively.(3)Based on intelligent prediction model of water requirements of green peppers,an irrigation decision system based on Depth Neural Network(DNN)was built.Crop factors,meteorological factors and soil factors were taken as the input factors of the model,and the output of the crop model of irrigation water was taken as the output factors.The data used for training was from 2014 to 2018,the test data was 2018.The hidden layers of the optimal DNN irrigation decision system included 4 layers,and 32,16,8 and 4 were the number of hidden layer neurons respectively.The activation function of the system was"Re LU",and the optimization function was"adam".The decision-making system can obtain the irrigation schedule of T_8R under MFR-RDI planting mode.Compared with the actual value calculated by using,the equation of water balance,the RMSE,MAE,NS and water-saving rate of the decision model were 0.898 mm,0.257 mm,0.758 and 1.3%,respectively.In 2018,the yield of green pepper was 12886.2 kg·hm^-2,the content of V_c was 51.1 mg·100g,and the IWUE was 32.6 kg·hm^-2·mm^-1.Compared with CK_1R,the water-saving rate of green pepper was about 26.4%.(4)The platform of intelligent irrigation decision system based on Lo Ra technology was built.The platform can monitor the data of agricultural meteorology and soil moisture and make irrigation decision.