Multi-objective Optimization To Winter Wheat Irrigation Schedule Based On The Water-fertilizer Coupling

Water and fertilizer are two very important aspects in agricultural production, andthe two have a synergistic effect, increase water can increase the stimulation effect offertilizer, add fertilizer can increase the stimulation effect of the water. Nowadays thecrop fertilization in field irrigation schedule still is given priority to extensive and ex-perience, not only waste a lot of water, but also large quantities of fertilizer will causethe more agriculture production cost, pollutes soil and water, destroy the ecological en-vironment, harm to human body health. So optimization of water and fertilizer irrigationschedule reasonable has become a key problem. In agricultural production, water andfertilizer are mutual restriction and coordination to promote each other, Only matchingwater and fertilizer factors reasonable,can rise to regulate water fertilizers, and give fullplay to the water and fertilizer double factors coupling production function. So Researchthe effect of water and fertilizer coupling, to improve the utilization efficiency of waterand fertilizer, to avoid the destruction of the ecological environment and deterioration,Improve the comprehensive benefit of agricultural production, ensure the sustainabledevelopment of agriculture, has an important significance. This paper builds a irrigationschedule Multi-objective optimization model based on the water-fertilizer coupling, Themain research of this paper includes the following two aspects:At first building the relationship between irrigation-fertilization-production basedon modern design method.Building the model of relationship between irriga-tion-fertilization-production by SVM,that has the advantage in solving pattern recogni-tion problems which are small samples, nonlinear and high dimensional.The C andσare key parameters to determine the regression accuracy in the model, and seted inadvance, and no fixed rule.So this paper optimize the key parameters of the SVM modelby using the PSO, avoid the "learning" or "owe learning" phenomenon because of theparameters selected improperly.On the other hand, The BP neural network is widelyused in recent years,because of the outstanding in self learning and adaptive ability,strong generalization ability. This paper also set up the water and fertilizer productionfunction model based on BP neural network.Comparing the model of PSO-SVM withthe BP neural network, determine the best water fertilizer production function model. For another, in order to determine the reasonable irrigation schedule based on wa-ter and fertilizer coupling. This paper builds a multi-objective optimal model based onthe PSO-SVM water-fertilizer and production relationship by using GNSGA-II,andtakes water and fertilizer as decision variables for synchronous optimization in order toachieve a goal of minimum water, minimum fertilizer but maximum production in thewhole growth period.Determining the water and fertilizer reasonably makes the controlof water and fertilizer management more detailed. At the same time, considering themodern agricultural production conditions, with the goal of maximizing the economicbenefit of crops per unit area, and the environment benefit of fertilizer and irrigationwater of synchronous optimization. Its optimal method and the results can provide ref-erence for modern agricultural production.This paper according to the relationship between water-fertilizer and productionbuilds the model of PSO-SVM and BP neural network respectively, and builds a mul-ti-objective optimal model based on the PSO-SVM water-fertilizer and production rela-tionship, to determine a reasonable water and fertilizer coupling irrigation schedule.And take the model being applied to the instance, get the following conclusion:(1)This paper optimized the key parameters of SVM by using PSO, avoids thedisadvantages of SVM in parameter choice effectively, ensures the learning and predic-tion ability of SVM, improves the prediction accuracy(2)The model of water-fertilizer and production relationship based on PSO-SVM isbetter than BP neural network in the precision of forecast and the choice of parame-ters, and has a better robustness.(3)This paper builds a multi-objective optimal model based on the PSO-SVM wa-ter-fertilizer and production relationship by using GNSGA-II, and takes water and fer-tilizer as decision variables, production and economic benefit as goal for synchronousoptimization,part of the pareto solutions are obtained then decided by user. The optimalmodel can provide reference for water and fertilizer to the field of fine management.