Research On Intelligent Irrigation Decision Mechanism Based On Crop Water Field Data Mining

Aagricultural irrigation decision is an important factor of affecting crop yield and soil water use efficiency.With the development of science and technology in recent years,intelligent irrigation is gradually changing the traditional irrigation mode.Intelligent irrigation control is achieved through intelligent detection equipment and control system.However,most intelligent irrigation systems only use current field meteorological data to make irrigation decisions without considering the impact of weather changes on intelligent irrigation decisions in the future.In this paper,based on the network meteorological data and data collected by in-situ meteorological stations,the rainfall prediction model of the farm in the coming day was established by using data mining technology,and the crop water demand dynamic prediction model for winter wheat was established.Based on the soil moisture content data of each layer in the water field test area,the correlation between the soil moisture of each layer in the planned wetting layer of winter wheat was studied,and the soil moisture of each layer in the planned wetting layer of winter wheat was simulated by HYDRUS-3D.An intelligent soil moisture content detection model was established.Finally,based on the principle of water balance,an intelligent irrigation decision mechanism was established to determine the appropriate when and how to irrigation for winter wheat.The HYDRUS-3D model was used to simulate the adaptability and accuracy of this decision.The results of model prediction and simulation were evaluated by the expression of determination coefficient(R~2),root mean square error(RMSE)and relative analysis error(RPD).Based on the data and numerical simulation results of the water field test area,the following conclusions can be drawn:(1)The meteorological data from Internet to forecast the rainfall is good,on the network directly to collect rainfall data to predict the results R~2,RMSE and RPD were 0.895,2.1%and 3.108,network meteorological rainfall prediction model was optimized by LM algorithm,the optimized prediction results R~2,RMSE and RPD were 0.938,1.6%and 4.045,and the optimized using the algorithm of BP algorithm and DT not optimization effect,showed that rainfall prediction model based on LM algorithm to optimize the optimal.(2)The data from network meteorological and in-situ meteorological stations were used.The in-situ meteorological stations data were used as the input to build a forecast model of crop water demand.The data of network meteorological data were used as the input to build a forecast model of crop water demand.Firstly,a forecasting model of water demand for reference crops was established based on the meteorological data of 11,7 and 4meteorological parameters.The R~2 for predicted accuracy was 0.959,0.943 and 0.905;RMSE was 0.6%,0.8%and 1.0%;RPD was 4.982,4.214 and 3.256.Based on the network meteorological data 7 and 4 meteorological parameters,the prediction model of reference crop water demand was established,and the predicted results R~2 were 0.892 and 0.850,respectively;RMSE was 1.0%and 1.2%,RPD was 3.056 and 2.598.Prediction results show that the more input meteorological parameters,the more accuracy for the forecast model,and the data of in-situ weather stations meteorological of reference crop water requirement forecasting is accurate,and the reference crop water demand prediction results from network meteorological weather stations data is poor.The effect of the dynamic forecast of reference crop water requirement and model predicted results 7 parameters and 4 parameters R~2 more than 0.85,can better predict the reference crop water requirement.Then,based on the reference crop water demand forecasting model,a dynamic crop water demand forecasting model was established based on the single crop coefficient method.The results show that 4parameters(maximum temperature,minimum temperature,average temperature and average atmospheric pressure)can be used to predict crop water demand when meteorological data are scarce.Therefore,the dynamic forecasting model of crop water demand based on network meteorological data is effective and can be used to predict crop water demand in the future.(3)Based on soil moisture data at soil depths of 30,40,50 and 60 cm,crop water demand and rainfall data in the whole growth period of winter wheat,the relationship between soil moisture of each layer was analyzed,and a prediction model was established based on soil moisture data at 30,40,50 and 60 cm,respectively,to predict soil moisture of each layer in the range of 30cm.The results showed that it was best to establish an intelligent soil moisture detection model with soil moisture content of 30 cm.The predicted results of R~2,RMSE and RPD of 40 cm were 0.774,1.1%and 2.125.The predicted results of R~2,RMSE and RPD of 50 cm were 0.824,1.1%and 2.405.The predicted results of R~2,RMSE and RPD of 60 cm were 0.734,0.9%and 1.960.Based on the precise rate of soil hydraulic parameters,application HYDRUS–3D during winter wheat growth period 30,40,50,60 cm soil moisture change simulation,predicted soil moisture at 30 cm depth,the R~2,RMSE and RPD is 0.705,1.6%and 1.968.The predicted soil moisture at 40 cm depth,the R~2,RMSE and RPD is 0.690,1.1%and 1.801.The predicted soil moisture at 50 cm depth,the R~2,RMSE and RPD is 0.758 and 1.2%,2.035.The predicted soil moisture at 60 cm depth,the R~2,RMSE and RPD is 0.620,1.0%and 1.625.The results show that the prediction model based on 30 cm soil moisture content is more accurate.HYDRUS-3D has a good effect on the simulation of soil moisture content changes in various layers,and the simulation of water transport rules during the growth period of crops has certain applicability.(4)Based on the prediction model above and combining with the principle of field water balance,an intelligent irrigation decision-making mechanism was established,so as to obtain the irrigation time and amount during the growth period of winter wheat.HYDRUS-3D was used to simulate and verify the irrigation time and irrigation amount.The simulation results showed that the soil moisture of the wetting layer during the growth period was kept between the upper and lower limits of the soil moisture,and the soil moisture could be well maintained in the appropriate moisture content state.The results show that the intelligent irrigation decision mechanism based on the data of water field test area is effective and can be used in the design of intelligent irrigation system.