Design and Realization of Integrated Management Expert System for Drip Irrigation for Cotton in Arid Area
To avoid encountering 504 error when downloading big files,please take CAJ download or PDF download by chapter as your first priority.
【Author in Chinese】 王乐；
【Author's Information】 石河子大学， 作物栽培学与耕作学， 2017， 硕士
【Abstract in Chinese】 近年来,随着科技与经济的发展,信息技术得到广泛应用,其在农业领域的应用已掀起一次新的技术革命。作为农业信息技术的分支之一,以专家系统为代表的智能化农业信息技术的作用尤为突出。新疆,作为我国最大的植棉区,由于生产管理水平较低,棉花产量和品质潜力未得到发挥,建立干旱区滴灌棉花综合管理专家系统,对于解决当前棉花栽培管理中的水、肥、药利用效率低等问题,实现棉花生长过程的动态调控及栽培技术的咨询服务等方面,具有十分重要的意义。本文开展了 29个品种、3个氮素的试验,定期调查棉花生育期、株高、叶龄、蕾铃数等数据。通过数据分析,构建棉花株高模拟模型、叶龄模拟模型,并进行模型的验证,结合新疆地区棉花栽培管理特性,建立了干旱区滴灌棉花综合管理专家系统。主要研究结果如下:1.构建了棉花株高动态模拟模型利用归一化和聚类分析法探明了棉花品种间相对有效积温与相对株高的关系,建立了基于相对有效积温的棉花相对株高模拟模型y = a/(1 + exp（b-cx）1/d。根据模拟结果,将所有品种类型分为三大类,第Ⅰ类（棉花三叶期（相对株高值小于0.14）至十一叶期（相对株高小于0.7）,株高生长速率较慢）y = 0.997/（1 + exp（26.08-33.62x））1/8.66（r=0.9976）;第 Ⅱ类（棉花三叶期（相对株高值在0.14～0.18之间）至十一叶期（相对株高在0.7～0.8之间）,株高生长速率较快）:（r=0.9967）;第Ⅲ类（棉花三叶期（相对株高值大于0.18）至十一叶期（相对株高大于0.8）,株高生长速率最快）:y = 1.02/（1 + exp（8.55—12.68x））1/3.25（r=0.9973）。对模型验证表明,RMSE= 1.6998cm,模拟值与观测值误差小,能够较好的反映干旱区滴灌条件下棉花株高的动态变化。2.构建了棉花叶龄动态模拟模型通过引入叶片生理发育因子,利用归一化处理及聚类分析法初步探明棉花品种间相对有效积温与主茎叶龄的关系,以8叶龄为界,对叶龄发育进程进行分段、分类模拟,分别建立了基于有理函数的棉花1-8叶龄动态模型和基于二次多项式、有理函数的8-13叶龄模拟模型,决定系数分别为0.9719、0.9964、0.9743、0.9733。模型验证表明,不同类型品种叶龄模拟值与观测值RMSE=0.3505,R2为0.9977,模拟值与观测值误差小,有理函数和二次多项式可以有效地预测棉花1-8叶龄及8-13叶龄的动态变化,可通过观测棉花叶龄生长状况为棉花精准管理提供依据。3.干旱区滴灌棉花综合管理专家系统的建立依据干旱区棉花栽培管理的主要措施和领域内专家知识,结合本研究所建立的模型和前人研究成果,建立了干旱区滴灌棉花综合管理专家系统,包括数据库、知识库、模型库、推理机、知识获取和人机接口等六部分,该系统实现了数据管理及系统维护、专家咨询、种植方案设计、模型模拟预测、实时调控等五大功能。系统可以根据用户输入决策地点的气候环境、土壤等基础数据,综合运用推理、预测、解释等机制帮助用户设计适宜的栽培管理方案。
【Abstract】 In recent years,with the development of science and technology,information technology has been widely used,and its application in the field of agriculture has led to a new technological revolution.As one of the branches of agricultural information technology,the role of intelligent agricultural information technology,which is represented by expert systerm,is particularly prominent.Xinjiang,as China’s largest cotton plantation,due to the low level of production management,cotton production and quality potential has not been developed,It is very important to establish the expert system for integrated management of drip irrigation for cotton in arid area,which is of great significance to solve the problem of low efficiency of water,fertilizer and medicine in cotton cultivation management,and to realize the dynamic regulation of cotton growth process and the consulting services about cultivation technology.In this study,29 varieties which are widely planted in Xinjiang and 3 nitrogen（second year）were designed.The data of cotton growth period,plant height,leaf age,buding number and so on were investigated regularly,through the data analysis,building various types of models,and model validation,based on the cotton cultivation and management measures in Xinjiang,an expert system for integrated management of drip irrigation for cotton in arid zone was established.The main results were as follows:1.A dynamic model of cotton plant height was constructedThe relationship between relative GDD and relative plant height of different cotton varieties was studied by using normalization and cluster analysis.The results showed that the relative height of the cotton plant changed consistently with relative GDD in spite of different variety.Here we established three categories of simulation model about plant height based on relative GDD:y= a/（1+exp（b-cx））1/d.Among these,the class I（from 3-leaf stage（relative plant height value is less than 0.14）to 11-leaf stage（relative plant height value is less than 0.7）,the growth rate of plant height is slow）:y = 0.997/（1 + exp（26.08-33.62x））1/8.66（r=0.9976）;The class Ⅱ（from 3-leaf stage（relative plant height value between 0.14-0.18）to 11-leaf stage（relative plant height value between 0.7 to 0.8）,the growth rate of plant height is fast）:y = 0.997/（1 + exp（22.09-28.65x））1/8.41（r=0.9967）.The class III（from 3-leaf stage（relative plant height values greater than 0.18）to 11-leaf stage（relative plant height greater than 0.8）,the growth rate of plant height is superbly fast）:y = 1.02/（1 + exp（8.55-12.68x））1/3.25（r=0.9973）.The verification of the model shows that the value of RMSE was 1.6998cm,indicating that the error between simulated and observed values was relatively small.Therefore,Richards function can effectively predict the changes of plant height in cotton.2.A dynamic model on leaf age in cotton was constructedThe relationship between relative GDD and leaf age of main stem of different cotton varieties was studied by using physiological developmental factor,normalization and cluster analysis.Based on the 8 leaf age,the developmental process of leaf age was divided into different stages.Respectively established dynamic model of cotton leaf 1-8 based on national function and dynamic model of cotton leaf 8-13 based on two times polynomial and rational function,determination coefficients were 0.9719,0.9964,0.9743,0.9733.The results of model test showed that the simulated value and observed value of leaf age of different varieties were RMSE = 0.3505,R2 were 0.9977,indicating that the error between the simulated and observed values was relatively small.Therefore,Rational functions and quadratic polynomials could effectively predict the changes of leaf age 1-8 and leaf age 8-13 in cotton,the model could be used for precision management of cotton by observing the growth condition of cotton leaf age.3.Establishment of expert system for integrated management of drip irrigation for cotton in arid areaBased on the cotton cultivation management measures and the expert knowledge in the field,combined with the model established in this research and other cotton simulation models established by others,the expert system for integrated management of drip irrigation for cotton in arid area was established,including six parts:database,knowledge base,model library,inference engine,knowledge acquisition and man-machine interface,The system realized the five functions,such as:data management and system maintenance,expert consultation,planting scheme design,model simulation and prediction,real-time control and so on.According to the user’s input decision-making location of the climate environment,soil and other basic data,the system help users to design a reasonable cultivation management program.