【作者】 任勃；

【导师】 黄璜；

【作者基本信息】 湖南农业大学 ， 生态学， 2004， 硕士

【摘要】 本文针对目前农业专家系统开发研究中存在的问题和未来发展趋势，研究了构建生产类农业专家系统的理论与技术，并选用目前国内最新的专家系统开发平台PAID4.0，结合湖南示范区(常德、岳阳等市)的水稻生产实际，开发了“基于Web的水稻生产专家系统(WRPES)”。WRPES采用目前网络应用的主流技术“Web浏览器／Web服务器／数据库系统”，集成本地区农业(水稻生产)领域专家们的理论知识、经验知识和最新的科研成果，能够提供模糊水稻产量预测、智能化病虫草害诊断与跟踪指导防治、不同生育期管理决策咨询和优化调控稻田生态系统等功能，能够服务于水稻生产的产前、产中和产后全过程。 本文的研究内容和结果有： (1) 系统知识库的构建。知识库是专家系统的核心组成部分，是决定专家系统性能的关键。一个专家系统的优劣在很大程度上取决于知识库中知识的质量和数量。系统知识库的构建包括知识获取和知识的表示。 知识获取是研究如何从各个知识源收集特定的知识，并将其转变成计算机可以存储和使用的形式。系统知识来源主要有：公开出版物、专家经验和补充试验研究。 知识表示是把从各种知识源获取的领域专家的知识、经验和思考、推理方式等以一定的方法或技术转变成计算机可以存储和使用的信息符号的过程。考虑到农业(水稻)领域中实际问题的复杂性和模糊性及农业专家表达领域知识和求解领域问题的特点，本文提出一种由传统的“产生式规则的知识表示法”改进的“模糊加权产生式规则+模型”的知识表示法，实现了确定性知识和模糊知识、定性知识和定量知识的归一化表示，使得各种知识能够适合开发平台的推理机制和解释机制。 本文在本领域专家的指导下，通过长时间对湖南示范区(岳阳、常德等地区)水稻生产领域专家知识和经验的搜集、挖掘和整理，最终获得各类知识规则2858条，建立了系统知识库。 (2) 系统基础数据库的构建。由于土壤数据、气象数据和地理数据等基础数据是推理过程或模型求解过程中必须的证据和参数，为专家系统决策的各个环节提供必须的基本信息，所以基础数据库是知识库和模型库等的支撑库。本文通过表结构设计和数据录入，用SQL Server 2000数据库软件组建了土壤资源数据库、农业气象数据库、农业基本情况数据库、水稻品种资源数据库等多个基础数据库。 (3)系统模型库的构建。本文在系统模型库的构建中，着重研究了基于模糊聚类的水稻产量预测模型和基于人工智能的水稻病虫草害诊断、识别和防治模型。由于采用“模糊加权产生式规则十模型”的归一化的知识表示法，各种模型除在模型库中显式出现外，还分散隐式出现在系统不同功能模块的知识规则中。 (4)系统的推广应用。系统的推广应用是本研究的最终目标。系统通过国家863计划项目专家组的鉴定验收后，在湖南示范区进行了大面积推广应用，并在应用过程中总结出了“三向带动、两网并行、人机合一、立体推进”的321立体推广模式和“公司+基地+农户”的具体应用模式。更多还原

【Abstract】 According to the problems and trends in studying Agricultural Expert System (AES), this paper discusses principles and techniques of developing expert systems for crop production, and development of Web-based Rice Production Expert System (WRPES). WRPES uses the structure of Web Browser/Web Server/Database, and is built based on PAID4.0 which is a updated tool for Expert System (ES) building. It integrates theories, experiences and updated research fruits of experts which are authoritative in localagricultural (rice) field. It can be used in all phases of rice production--before ricesowing, when rice growing and after rice harvesting. It can give managing suggestions in different phases of rice growing and measures for optimizing and controlling the ecosystem of rice field. It also provide some other functions, such as, doing fuzzy prediction on rice yield, diagnosing diseases of rice, recognizing pests and weeds in rice fields, and tracking prevention and cure of the diseases, pest and weeds.The research contents and results of this paper presents as following: 1. Development of knowledge base.Knowledge base is a hardcore of WRPES, and it is a key to the function of WRPES. The performance of an ES mainly depends on the quality and quantity of knowledge in knowledge base. Development of knowledge base includes knowledge acquisition and knowledge expression.Knowledge acquisition is on how to collect the wanted knowledge from all kinds of knowledge sources. The sources where the knowledge of WRPES comes from are publications, experiences of experts and some complementary experiments etc.Knowledge expression is on how to formalize the knowledge collected to be stored and used by computers. Considering the complexity and fuzziness of knowledge, the characteristic of practical questions solved by agricultural experts, this paper puts forward a knowledge expressing method called "weighted fuzzy logic producing rules + model" which derives from traditional "producing rules". The method translates all kinds of knowledge, such as, certain knowledge and fuzzy knowledge, qualitative knowledge andquantitative knowledge, into one format, which makes all kinds of knowledge fit the reasoning and interpreting mechanism of PAID4.0 better.Directed by experts of agricultural (rice) field, we take long time to collect the knowledge and experience of experts in agricultural (rice) field, and finally obtain 2858 items of knowledge rules for the knowledge base of WRPES.2. Development of foundational data base.Because soil data, meteorology data, geography data etc, are reasoning proofs or model parameters which can be primary information of ES decision-making, foundation data base is a support base of knowledge base and model base. After designing structures of data tables of foundation data base and then inputting data, we build soil database, agro-meteorology database, agricultural primary information database and rice variety database etc with SQL Server 2000.3. Development of model base.In this part, we discuss fuzzy clustering prediction models on rice yield, diagnosing models of rice diseases, recognizing models of pests and weeds in rice fields. All models present not only in model base but also in knowledge rules of different function modules.4. Spreading application of WRPES.Spreading application of WRPES is the final end. After appraisement, check and accept, WRPES is applied in large demonstrating areas of Hunan (including Changde, Yueyang etc). During the application, we summarize a "321" three-dimensional spreading mode which means "three aspect drive, two network parallel, man and computer tied in, three-dimensional advance", and an application mode called "company + base + famer".更多还原