| Establishing a proper irrigation system is a vital task of reliving China's water crisis and constructing water-saving society.It is also the basis of optimizing irrigation scale,determining the optimized solution of water project operation and setting the optimum water supply system dispatch in irrigation management. It is of great importance to forecast the irrigation water requirement and to set a proper irrigation system especially in Xinjiang with obvious water shortage.Film-hole irrigation is a new water-saving technology; It is widely applied in various crops because of its distinguished water conserving and calefacient effect and obtains the double effect in water conservation and an increase in production.Computer artificial neural network is non-linear science booming in the 80s,20th century.With its fast development in last 20 years, this technology has made outstanding achievement in artificial intelligence,automation and pattern recognition,etc.Applying the neural network in modeling and forecasting in nonlinear system could not restrained by linear models thus could effectively present the complex nonlinear characteristics of uncertain, multi-input problems.The paper combine Computer artificial neural network with film-hole irrigation of cotton's developing status,which could not only attain water conservation and an increase in production, but also of great significance in establishing an proper irrigation system.The paper established the model based on the RBF neural network system to forecast the film-hole irrigation on cotton's developing status in Xinjiang Shihezi areas,and examined the model by field experiment.The results are as followed:1.With the meteorology data provided by 148 Regiment and MATLAB toolbox,the paper established the neural network system model to forecast the water consumption affected by the average temperature, the sunshine hours and vapor pressure.Through the water equilibrium equation calculated crop water consumption compared with predictive value of RBF network model,we come to the figures that the maximum value of absolute error is 0.0277mm/d;the minimum is 0.0001mm/d and 0.0089mm/d on average;while the relative error is 4.789%,the minimum is 0.004%,and 0.614%on average,which in all proves a preferable prediction.The statistics reveal a relatively high accuracy that could be well applied in the productive practice.Predicting the crop water consumption is the vital problem of irrigation prediction in irrigated areas.Solving this problem would lay the foundation of irrigation prediction in irrigated areas.2.The paper applied the water equilibrium equation as the prediction method of cotton real time irrigation,and make thorough introduction of the determination of parameters in the water equilibrium equation.Crop water requirement is predicted by the RBF neural network model.The irrigation prediction in irrigated areas by water equilibrium equation could take advantage of the short term weather forecast data thus make the irrigation system accord with the practical situation of the irrigated areas.3.The paper applied the trial method to calculate the days to irrigate,and applied the formula I_i=10γH_iβ_fp(1-φll) to forecast the irrigation amount in the first experimental field.4.The prediction table 4-3 shows 11 times of real irrigation while the model predicted 8 times of irrigation,and the real irrigation amount has reached 320.04mm while the model predicted 246.68mm of irrigation amount.Through the analysis,the predicted times and amount of irrigation take better advantages of the rainfall and is accord with the practical situation and achieve the purpose of water-saving.It further suggested that the real time irrigation forecast model could not only reduce waste of water and could better use the meteorology data to make the irrigation system well correspond with the real situation.
|
PDF Full Text Request