【作者】 谢家兴；

【导师】 王卫星；

【作者基本信息】 华南农业大学 ， 农业电气化与自动化， 2016， 博士

【摘要】 中国荔枝种植面积在55.33万hm~2左右,年总产量约220万t,分别约占世界荔枝种植总面积和年总产量的80%和75%。而根据果园调查分析,中国的果园管理技术普遍比较落后,中小规模的果园管理设备缺乏先进性,相应设施也不完善。果树的灌溉绝大部分运用喷灌或漫灌,水分补充时间选择依据仅仅是果农自身种植经验或天气情况。果树生长需要供给养分的确定通常是按固定时间以农业中常用的肥料进行施肥,既耗费了大量的资源和人力,又得不到很好的效果。无线传感器网络(Wireless Sensor Network,WSN)综合了现代传感器技术、微电子技术、通信技术和嵌入式计算机技术等多学科技术,被认为是21世纪最重要的技术之一,其巨大的应用价值已经使其成为世界备受关注的新兴前沿研究热点领域。WSN具有功耗低、成本低、可靠性高等特点,这种新型的信息获取和处理技术可应用于农业领域,在国内外已有很多成功应用实例。应用无线传感技术,监控荔枝生长过程,对加快农业生产自动化和信息化、预测荔枝旱情及病虫害有着重要意义。本论文以荔枝为应用对象,研究了专家系统决策模型,构建了基于无线传感网的荔枝园智能节水灌溉和专家系统决策模型,进行了系统软、硬件的设计和荔枝园环境下组网和系统试验。主要研究内容和成果如下:(1)系统分析无线传感器网络系统设计原则及关键技术,研究基于无线传感器网络的果园智能灌溉技术,研究专家系统灌溉理论模型,为构建基于无线传感网的荔枝园智能节水灌溉和专家系统提供理论基础。(2)根据荔枝园环境智能监控的应用需求,进行果园智能节水灌溉无线监控系统设计,包括硬件设计、软件设计和模糊控制算法的设计和优化。硬件设计是进行系统总体结构和节点硬件结构的设计,在此基础上,完成传感器和电子元器件的选型,研究硬件各模块功能的实现方法,进行电路原理图、PCB版图、电路板及实验调试等设计研究。软件设计是设计节点嵌入式软件结构,研究嵌入式软件设计和编程方法,完成包括主程序、系统初始化、数据采集、时间同步算法等模块程序的开发;设计远程服务器,将荔枝园实时环境数据保存到数据库,为专家系统和监控网站提供数据支持;设计专家系统,提供不同的灌溉方式;设计监控网站,构建形象直观、实时性强的荔枝园环境信息实时监控平台,方便管理人员进行监控、查询和管理。设计手机客户端,方便管理人员在户外的情况下实时监控荔枝园智能灌溉系统。模糊控制算法的设计和优化是通过研究模糊控制理论和系统,设计模糊控制算法,包括确定输入输出变量论域和模糊语言变量、设计模糊规则和进行反模糊化等,并使用Matlab对该算法进行仿真,最后根据仿真结果进行模块控制器的优化。(3)研究设计荔枝园专家系统的灌溉模型,包括灌溉预报模型、定量灌溉模型、根据灌溉制度灌溉和智能灌溉模型等,解决在实际荔枝生产管理中的果农如何确定灌溉间隔,每次灌溉的用水量和如何提高荔枝产量等问题。根据灌溉模型开发专家系统软件,用户可以根据荔枝园的实际情况选择不同的灌溉方法,每种灌溉方法都有各自的优劣性。(4)设计完成不同节点的功耗测试,验证节点是否为低功耗。设计完成节点生命周期测试,验证节点能否长期应用荔枝园实地试验。设计完成室内组网测试,验证节点的网络性能是否满足荔枝园实地试验需要。设计完成传感器标定试验,以实现对荔枝园环境参数的实时准确测量。(5)在基础实验和测试的基础上对系统进行集成和性能测试,在荔枝园环境下部署网络,进行节点有效通信距离试验、网络丢包率试验、系统智能灌溉性能试验和专家系统决策准确性试验,分析和总结系统试验效果:(1)节点有效通信距离试验中,当节点处于空旷地带时,六组节点所处的高度不同,分别1.5、1.25、1.0、0.75、0.5和0.25 m时,有效通信距离均为1 205m。当节点处于试验荔枝园时,六组节点分别处于1.5、1.25、1.0、0.75、0.5和0.25 m的高度,结果是1.0m高度处的节点通信距离最长,为81.5m,其他高度的次之。(2)网络丢包率试验中,传感器节点设定为土壤含水率采集模式,即采集周期为30min,采用休眠唤醒策略,监测时间为15天,结果为单个节点最低丢包率为2.63%,最高丢包率为4.03%,整个网络的平均丢包率仅为3.87%。所有节点丢包率都比较低,网络传输稳定可靠,满足系统通信的基本要求。(3)系统智能灌溉性能试验中,分别比较试验期内的A区和B区的土壤含水率情况,其中,A区采用智能灌溉方法,结果是土壤含水率平均值为17.85%,均高于荔枝生长的最佳土壤含水率的下限,即15.55%,达到预期灌溉效果。而B区不灌溉,其土壤含水率平均值为7.75%,远远低于荔枝生长的最佳土壤含水率的下限,即15.55%。因此,系统的智能灌溉性能较好。(4)专家系统决策准确性试验系统在单个参数作为输入量进行决策得到的决策结果为:在下次灌溉时荔枝树的干旱级别为4或5,可知此时荔枝树处于比较湿润或者湿润状态;测得的土壤含水率仍然在荔枝树生长的最佳含水率范围内,即15.55%~19.14%,说明决策系统得出的荔枝树的干旱状态与实际情况不一致。系统在多参数作为输入量进行决策得到的决策结果为:在下次灌溉时荔枝树干旱级别为3,荔枝树处于适宜生长的状态;测得的土壤含水率在荔枝树生长的最佳含水率范围内,说明决策系统得出的荔枝树的干旱状态与实际情况一致。因而,多参数作为输入量会提升灌溉决策的准确性。更多还原

【Abstract】 The cultivated area of litchi orchard is about 555.3 thousand hm² and the annual output of litchi is about 2.2 million tons,respectively accounting for 80%of the cultivated area of litchi orchard and 75%of the annual output of litchi in the world approximately.And according to past investigations,the level of orchard management level in China is still very low.Lacking relevant infrastructures,middle and small orchards suffer a low level of modernization in facilities.Flooding or canal irrigation are main methods of fruit trees irrigation,which water supplement time are decided by personal experience or meteorological conditions.Nutrients needed in the fruit trees growth process usually are fertilized at a fixed time with agricultural fertilizer,consuming a lot of resources and manpower but having poor effect.Wireless Sensor Network combines multidisciplinary technologies including modern Sensor technology,microelectronics technology,communication technology and embedded computer technology and so on,which is considered as one of the most important technology in the 21st century.With great values,it does become a concerned emerging frontier hot research field around the world.WSN,a new kind of information acquisition and processing technology applied in agriculture,enjoys the benefits such as low power consumption,low cost,high reliability,and already has many successful cases at home and abroad.Applying wireless sensor technology to monitor litchi growth process has important significance in both boosting the degree of agricultural production automation as well as informatization,and predicting drought,pests and other diseases of litchi.Taking litchi as an example,this paper studied the decision model of expert system,constructed a decision-making model of intelligent water-saving irrigation and expert system in litchi orchard based on wireless sensor network,designed hardware and software for system,built the networks and tested the system in litchi orchard environment.The main contents and achievements show as follows:（1）Set up designing principles and key technologies of wireless sensor network system,investigated intelligent water-saving irrigation technology based on wireless sensor network in orchard and irrigation model of expert system,which provided theoretical basis to build a suitable decision-making model.（2）According to requirements of monitoring the environment in litchi orchard,the system was designed to monitor water-saving irrigation in orchard intelligently,with special designs of hardware and software and optimization of fuzzy control algorithm.Hardware design consists of general structure and the node hardware structure of system,on which based to complete selection of sensors and electronic components,research method to realize the function of each module hardware and design circuit schematic diagram,PCB layout,circuit boards,experimental debugging and so on.Software design consists of node embedded software structure.Firstly,embedded software design and programming methods were researched and components of program including the main program,system initialization,data acquisition,time synchronization algorithm and so on are developed gradually.Secondly,a remote database server was designed to save real-time environmental data of the litchi orchard,providing source data for the expert system and monitoring sites.Thirdly,an expert system was designed to provide different irrigation methods.Fourthly,monitoring sites was designed and real-time environmental information about litchi orchard was displayed,helping administrator monitoring,querying and managing data.Lastly,a mobile phone client was designed for administrator to monitoring real-time situation of litchi orchard even outdoors.Optimization of fuzzy control algorithm was completed after studying the fuzzy control theory and system,including choosing input and output variable domain and fuzzy language variables,creating the fuzzy rules,making defuzzification,etc.This paper using Matlab to simulate the algorithm,and then optimizing the module of the controller.（3）Through the research and design of irrigation model of litchi orchard expert system,including irrigation forecast model,quantitative irrigation model,irrigation rules and the most water-saving irrigation model,to solve the actual litchi production management to determine the irrigation interval,irrigation water per irrigation and how to improve the yield of litchi and other issues.With help of a model of expert system software,the users can choose different irrigation methods according to the actual situation of litchi orchards.Each method has their own advantages and disadvantages.（4）Tested power consumption of different nodes to verify nodes is low power consumption.Designed node life cycle test to verify the nodes can work in litchi orchard field for a long time.Tested interior network to verify the network performance can meet the needs of litchi orchard field monitoring.Designed sensor calibration experiment to verify the monitoring accuracy of environmental parameters in litchi orchard.（5）Furthermore,integration and performance tests was carried out and the network was deployed in litchi orchard environment.Effective communication distance of the nodes,network packet loss rate,intelligent irrigation performance of the system and decision accuracy of expert system was measured and assessed.The analysis and summarizes of the test results show as follows:a)In the test of effective communication distance of the nodes,the six different height--0.25m,0.5 m,0.75m,1.0 m,1.25m,and 1.5 m--where nodes installed was compared.When the nodes were in the open field,all of their effective communication were 1205 m.When the nodes were in the litchi orchard,the longest effective communication distance was 81.5m which belongs to 1.0m,larger than distance of 0.25m,0.5m 0.75m,1.25m and1.5m.b)In the test of network packet loss rate,soil moisture had been collected by sensors in nodes every 30 minutes using the strategy of dormancy awakening for 15 days.The result showed that a single node packet loss rate was 2.63%at least,loss rate of whole network 3.87%on average and 4.03%at most.Packet loss rates were low enough to satisfy stable and reliable requirement of communication.c)During the performance test period,the soil moisture content in A zone and B zone was compared.On one hand,the average of the soil moisture content of A zone was17.85%,higher than the lower threshold of litchi growth--15.55%and achieved expected effect using intelligent irrigation.On the other hand,the average of the soil moisture content in B zone without been irrigated was 7.75%,much lower than 15.55%.In conclusion,the performance of intelligent irrigation system was better.d)The results of system test of decision accuracy was measured in two scenarios.When only a parameter was input to system,it predicted litchi drought level would be 4 or5 at the next irrigation time,which meant litchi trees were in wet or humid condition now.Meanwhile,soil moisture content was still in the best level of litchi trees growth--15.55%¹9.14%,demonstrating that prediction ofdecision-making system was conflict with actual situation.When multiple parameters was input,the prediction of drought level would be 3 at the next irrigation time,litchi,which meant litchi trees were in a suitable growth state now.At the same time,soil moisture content was in the best level of litchi trees growth,proving the litchi drought condition predicted by decision-making system matched the actual situation.Therefore the accuracy of the irrigation decision could be improved by inputting multiple parameters.更多还原