论文标题:基于知识模型的棉花管理决策支持系统的研究
Knowledge Model-Based Decesion Support System for Cotton Management
论文作者 张怀志
论文导师 曹卫星,论文学位 博士,论文专业 作物栽培学与耕作学
论文单位 南京农业大学,点击次数 140,论文页数 112页File Size5578k
2003-05-01论文网 http://www.lw23.com/lunwen_5173417/ 棉花;知识模型;播前方案设计;生育指标动态;诊断调控;决策支持系统
Cotton;Knowledge model;Pre-sowing plan design;Dynamic development indices;Diagnosis and regulation;Decision support system
本研究着重利用系统分析原理和数学建模技术来研究作物栽培管理的知识表达体系，在广泛收集及充分理解和分析棉花栽培管理专家知识、经验和资料的基础上，利用棉花栽培理论与技术方面的现有研究成果，并结合必要的试验支持，解析、提炼和综合棉花生育及管理指标与品种类型、生态环境及生产水平之间的定量化关系，首次构建了棉花栽培管理动态知识模型，并进一步结合基于知识规则的棉花栽培管理知识库表达系统，设计和实现了综合性、智能化和构件化的基于知识模型的棉花管理决策支持系统(KMDSSCM)。 具有时空适应性的棉花栽培管理动态知识模型主要包括播前方案设计、生育指标动态和苗情诊断调控知识模型三个部分。其中，播前方案设计知识模型包括产量目标、品种选择、播栽日期、种植密度与播种量、肥料和水分运筹等。生育指标动态知识模型包含有适宜生育期、株高、叶面积指数、干物质积累量、果枝和蕾铃数目，营养物质积累量等。苗情诊断调控知识模型包括生育指标差异、调控可能性分析以及调控强度推荐等。 产量目标知识模型是以棉花光温水生产潜力的估算为基础，通过量化土壤肥力、历史产量水平和生产技术水平等多因子对皮棉产量的影响而建立的。品种选择基于品种特性与生态环境之间的定量化关系，计算棉花抗病抗虫性及产量和品质指标的综合影响。根据种植制度(种植方式)决定品种熟性，由品种熟性与用户要求决定品种，依据品种特性和产量目标确定播种日期的原理建立播期确定知识模型。基于“以产定铃、以铃定节、以节定枝、以枝定苗”的原理，同时考虑播种日期、打顶日期、≥12℃的有效积温、品种株型特性、肥水管理水平差异等因素来确定适宜的种植密度；在此基础上，应用相对权重法计算土壤含水量、含盐量以及整地质量和播种方式等多个生态环境因子对出苗率的综合影响，进而确定播种量。根据平衡施肥原理和棉花水分需求规律，在综合考虑土壤理化特性、品种遗传特征、产量水平等因子影响的基础上，建立了具有系统性和普适性的棉花肥料及水分运筹动态知识模型，模型可以完成肥料总量需求，有机肥与无机肥的比例、基肥与追肥的比例、追肥施用时间、水分需求总量及其在各个生育时期的分配等。 以≥12℃的有效积温和品种熟性因子为基础，建立了一种统一的生理时间尺度，可预测不同环境条件下不同品种从播种到吐絮的全过程。进一步根据产量和品质目标构建了株高、叶面积指数、干物质积累量、果枝和蕾铃数目变化与生理时间之间的动态关系知识模型，初步量化了棉株养分积累量与干物质积累量之间的动态关系，从而为定量化的苗情诊断和管理调控提供了参考标准。 根据田间实际棉花生长发育指标与生育动态指标知识模型生成的适宜生育指标之间的差异及差异出现的生理时间，苗情诊断调控知识模型利用余弦函数曲线来表示调控目标实现的可能性，并进一步判断苗情特征对产量目标的可能影响，最后推荐适宜的调控管理措施。 在构建棉花栽培管理动态知识模型的基础上，进一步结合基于知识规则的棉花知识库表达系统，在 Vsual＋＋6对平台上构建了综合性、智能化和构件化的基于知识模型的棉花管理决策支持系统，实现了预测功能和决策功能的有倾合与集成，从而为建立其它作物的栽培管理决策支持系统提供了开发框架和思路，为精确农作和数字化农作奠定了基础． 利用南京、安阳、太原和石河子等生态点的不同品种、不同土壤类型和不同产量目标等资料对所建系统的可靠性进行了实例测试与检验。结果表明，系统具有较好的决策性和适应性。
This research focused on applying the system analysis principle and mathematical modeling technique to study knowledge expression system for crop cultivation management. Based on extensively collecting, understanding, analysis, and integration of expert"s knowledge and experience, literature and experiment data for cotton cultivation management, the dynamic relationships of cotton growth and management indices to variety types, ecological environments and production levels were quantified, and a dynamic knowledge model for cotton management (CottonKnow) was developed. By further incorporating the rule-based knowledge system for cotton management, a comprehensive and intelligent knowledge model-based decision support system for cotton management (KMDSSCM) was established with component design.The dynamic knowledge model with temporal and spatial characters for cotton management includes three modules as pre-sowing plan design, dynamic development indices, diagnosis and regulation. The knowledge model for pre-sowing plan design includes submodels of target yield calculation, variety selection, sowing or transplanting date, population density and sowing rate, fertilization and water management strategy. The knowledge model for the dynamics of main development indices includes submodels of suitable development stages, plant height, leaf area index, dry matter accumulation, numbers of fruit branch, square and boll, plant nutrient accumulation. The knowledge model for diagnosis and regulation includes calculation of differences in growth indices, possibility of regulation and intensity of regulation practices.The submodel for target yield prediction was developed through integrating the effects of cotton radiation-temperature-water yield potential, soil fertility, average yield of last three years and production technique levels. The sub-model for variety selection was established by qualifying the relationships of variety characters to eco-environments through the combined effects of disease and insect resistances, yield and quality traits. The sub-model for suitable sowing or transplanting date was developed according to the principle of determining variety maturity characters from planting style, variety type fromvariety maturity characters and user"s requirement, and sowing date from variety and target yields. The sub-model for design of population density was developed according to the principle of determining boll number from target yield, fruit node from boll number, fruit branch from fruit node and population density from fruit branch by integrating the effects of sowing date, cutout date, effective temperature accumulation above 12 , variety type, and fertilizer and water management levels. Sowing rate was then decided by integrating the effects of different soil water and salt contents, pH, temperature and sowing style on seedling emergence rate with relative weight method. According to the principle of nutrient balance and water requirement in cotton, the sub-model for fertilization and water management was developed by integrating the effects of soil characters, variety traits and yield target. The submodel can make decisions on the suitable total nutrient and water rates and distributions among main growth stages, ratio of organic to inorganic nitrogen, and the ratio of base to topdressing fertilizer.Based on the effective temperature accumulation above 12 and variety maturity characters, physiological time for predicting the development processes from sowing to boil opening under different environments was determined. The knowledge model for the dynamics of main development indices as plant height, leaf area index, dry matter accumulation, numbers of fruit branch, square and boll was developed based on the physiological time and target yield and quality. In addition, the dynamic relationships between plant nutrients and dry matter accumulation was quantified. All these sub-models provide the reference standards for quantitative and dynamic growth diagnosis and management regulation.
|
