农业植保预测和决策支持系统的若干关键问题研究

【作者】 黎敬涛；

【导师】 邹平；

【作者基本信息】 昆明理工大学 ， 管理科学与工程， 2013， 博士

【摘要】 中国作为人口大国,农业植保系统是决定全社会吃饭问题的关键系统,在农业生产中起着非常重要的作用。植保预测和决策是植保系统的核心,涉及到生产组织和管理、病虫害防治、预报等农业生产全过程。植保系统是一个非常复杂的系统,其中不明确、不确定的因素非常多,导致了系统呈现灰色系统特性,因此在农业生产中植保系统的一些问题把握不好,将导致农业生产受到较大影响,严重影响社会农业系统内人员的收入,甚至影响全国的粮食供应。在农业植保预测预报系统建设过程中,存在很多不确定因素和技术难点,而这些问题却又是农业植保系统的关键问题,也是农业生产精准化、可控化的关键问题。论文作者在项目研究过程中,参加了云南农业大学植保学院和昆明市植保站,以及云南省内8个县区植保站试验田的生产和种植试验,进行了历时数年大量的试验数据采集和试验调研,并与大量的农业植保专家进行合作研究,充分听取农民种植户、基层农技人员、县区植保技术人员、省市级植保专家、院校植保专家的建议和意见,对于植保系统中存在的一些难点问题进行深入仔细的分析研究,并选择其中农业植保预测和决策支持系统进行研究,以期解决一些难点问题。论文的部分成果已经在云南8县一市应用,最终取得了较高的学术、社会、经济效益。论文从农业植保系统中一些传统经验和方法遇到的困境出发,积极利用现代科技的优势,引入管理工程、计算机、信息采集、信息处理、数据库以及人工智能知识管理等方法和技术,对传统的以手工和经验为主的植保预测和预报进行了深入研究,对其原理和方法的科学性进行了讨论和改进,充分利用管理工程和信息技术的方法和技术,建立了一些应用模型,用计算机进行开发实现,并已经进行了一定时间段的生产试验。论文主要解决了如下问题：1、阈期的预测是植保系统广泛存在的一个难题,传统以来大多以经验模型进行阈期的预测,由此开发的预测系统通用性很差。论文将传统的经验模型进行科学化处理,并以除草阈期预测作为例子,提出了除草阈期模型自适应性问题,即模型参数的动态自适应解决方法,彻底改变了小样本特性的经验模型在不同作物生长环境适应性差的难题。2、以模糊评判等方法建立了田间杂草生长的优势评估以及除草剂效果评估模型,为植保决策支持系统中的优势因素确定和效果评估提供了方法。3、利用矢量合成等方法建立病害传播模型,解决了传统植保预测预报系统中传播趋势难以预测的问题,为病害传播的多因素趋势自动预测提供了方法。4、将植保系统中极其重要的农业专家的评价进行了量化处理,传统方法在系统里专家的支持度一般都是固定不变的,论文提出了动态支持度概念,建立了支持度动态调整模型,为专家参与植保决策支持系统的自动运行,以及客观描述专家在系统中的重要性提供了一套方法。5、精准化农业的关键是数据采集的密度和精确度问题,也是预测决策的关键,通常采集密度越高预测越准确,而采集密度越高采集成本就越高,甚至无法实施。论文提出了2套低成本、高密度、易实施的采集系统方案,彻底解决大田数据采集问题,为精准农业的实施打好了基础。上述这些理论和方法的建立,虽然是建立在传统的管理工程思想和方法基础上,但这些理论和方法在解决农业预测和决策问题上有一定的创新性,为农业生产的精确化、科学化管理提供了新的思路和方法,一定程度改变农业生产以经验性决策为主的模式,对农业植保领域的预测系统和决策支持系统建设起到了一定的示范作用。论文工作数据采集等历时数年,工作量巨大,技术难度较高,也获得了较多成果。更多还原

【Abstract】 Agriculture plant protection system of China is the key system that decides the whole country’s eating problem. This system is very important for agriculture production. Forecast and decision are the core in the plant protection system, which involve the whole production process about production organization and management, pest control and forecasting etc. Agriculture plant protection system is a complex system. It includes many of undefined and uncertain factors that lead to this system showing grey system characteristic. If we can not deal with the factors well in the agriculture plant protection system; it may cause many huge negative results about the income of agriculture system member and even the Chinese provisionment.Establishing process of Agriculture plant protection forecasting system has many uncertain factors and technical difficulties. These problems are the key factors in this system and are the key in the production precision and control process. During this project research, the author participated in the product and plant test in test field organized by the plant protection institute of Yunnan Agricultural University, Kunming plant protection unit, and other eight counties plant protection units in Yunnan province. In the last few years the author got a large number of test data collection and test research, and cooperative studied with a large number of agricultural plant protection experts, as well as collected advices and opinions from the peasant farmers, basic technique personnel, technical personnel, plant protection experts who came from counties, cities and province, and considerately analysed research on the difficult problems in plant protection system. At last the author selected the plant protection forecasting and decision support system as his research direction. At the same time, parts of achievements from this paper have been used in Yunnan province which achieved high academic, social and economic benefits.The paper based on the difficulties from traditional experience and methods in agricultural plant protection system, and actively used modern science and technology advantages, and introduced knowledge management methods and techniques, such as management engineering, computer, information collection, information processing, database and artificial intelligence etc to deeply research on traditional plant protect and forecast system which was primary with manual and experience. Discussing and improving the science of the principle and method, using the method and technology from management engineering and information technology established a series of application models, and developed them by computer. The application models have been tested for a long time.The paper mainly to solve the following problems:1, The prediction of threshold period is a widespread problem in plant protection system, and people always uses the experience model to judge the traditional threshold period, thus the universality of the prediction system is very poor. This paper deals with the traditional experience model scientifically, and put forward the self-adaptive problem of the weeding threshold period by weeding threshold period prediction as an example. The solution of the parameters dynamic self-adaptive method thoroughly changes the difficult problems of the poor self-adaptive of small sample experience model in different crop growth environment.2, Using fuzzy evaluation method establishes the superiority evaluation model of weeds in the field and the herbicides effect evaluation model, and provides the method for advantage factors determine and effect evaluation in plant protection decision support system.3, Using vector synthesis method to establish the disease transmission model, solves the problem that spread trend is difficult to predict in traditional plant protection forecast system, and provides a method to the automatic prediction of the factors trend for the spread of the disease.4, This paper quantitatively processes the most important factor of the assessment on agricultural expert, but it’s usually keep changeless in traditional method. It puts forward the concept of dynamic adjustment degree, and establishes the model of support degree dynamic adjustment, that provides a set of methods for the automatic operation of experts to participate the plant protection decision support system and the objective description the importance of experts in system.5, Density and accuracy of the data acquisition is the key to precision agriculture, which is the key to decision. Generally speaking, the higher density acquisition gets the more accurate prediction, however the higher collection density the cost of collection is more expensive, finally you can’t afford to it. This paper puts forward two sets of low-cost acquisition system structure scheme and implementation technologies, and thoroughly resolves field data collection problem, which lay the foundation for the implementation of precision agriculture.Although these theories and methods, are based on traditional management engineering ideas and methods, have certain innovation in solving agricultural forecast and decision. They provide a new train and method for the precision of agricultural production and scientific management, change the model of mainly empirical decision in agricultural production a certain degree, and have some demonstration effects on forecast and decision support system in agricultural production protection field. The paper’s data acquisition experienced several years. Although workload is huge, and technical difficulty is higher, also have the more achievements.更多还原