作为九五国家攀登B项目——“利用热力参数进行热力系统故障诊断研究’的一个子课题 “循环流化床锅炉状态监测与故障诊断专家系统的研究”利用专家系统和故障诊断技术, 以清华大学及其国内外合作研究伙伴在循环流化床研究领域内的现有经验为知识基础，为我国开发热动力系统故障诊断专家系统提出了一套通用的、开放性的分析思想和设计方法，并针对清华大学改进型75t/h循环流化床锅炉将上述思想和方法具体实现，建立了实际的75t/h循环流化床锅炉状态监测与故障诊断专家系统CFBBEXPTS，它包括分层次状态监测系统和深浅知识混合故障诊系统两部分。 为克服单纯浅知识规则库的缺陷，CFBBEXPTS还运用面向对象的深知识表示方法建立了循环流化床锅炉系统的深知识库（包括基于对象的等级网络，关于对象的描述集合和故障对象——异常参数因果依赖网络三方面）。利用CFBBEXPTS的深浅混合知识库，诊断系统不仅能对循环流化床锅炉的常见故障进行诊断，还能对其内部大量不常见的热力对象功能恶化型故障进行预报和诊断。 为了克服利用热力参数进行故障诊断时常常遇到的对测点参数数量“过分需求”的问题，本论文设计并实现了适合大型热力系统故障诊断的深浅知识混合诊断推理控制机制，提出了一系列适用于候选故障对象的矛盾检测原则。在CFBBEXPTS的具体诊断过程中，本论文首次成功地将五种高级诊断技术：候选生成、矛盾检测、局部诊断，约束隔离和约束传播可操作地综合运用于其中。该方法从理论分析和技术实现两方面初步解决了大型热力系统故障诊断对热力参数可测性的不现实依赖难点，同时又保持了对已有测点参数的高效利用。本论文最后针对模拟故障算例，进行了专家系统CFBBEXPS故障诊断能力的分析。结果表明，该系统具有良好的故障捕获能力。该系统中包含的利用热力参数进行热力系统运行健康水平监测与诊断的分析思想和设计方法,丰富和发展了热力系统状态监测与故障诊断技术，最后，为了说明上述分析思想和设计方法的通用性和开放性行，本论文据此为山东潍坊发电总厂30万机组常规煤粉锅炉远程集中监测与诊断系统BMADS提出了解决方案。
As a sub-project of the 95 National Climbing B Project "New Technologyand New Theory of Modern Power System", "Monitoring and Fault DiagnosisExpert System for the Circulating Fluidized Bed Combustion (CFBC) Boilers"is based on the experience of Tsinghua University and its cooperate partners athome and abroad in the area of CFBC. This dissertation uses the expert systemand the fault diagnosis technology, puts forward a new versatile and openanalyzing and devising method in developing the monitoring and faultdiagnosis expert system for the thermal systems named Circulating EluidizedBed Boiler Expert System(CFBBEXPTS). This system is targeted for TsinghuaUniversity's improved 75 t/h CFBC boiler in particular. CFBBEXPTS includestwo parts: a layered monitoring system and a mixed deep-shallow knowledgefault diagnosis system. With the use of the object-oriented deep knowledge representing method,CFBBEXPTS has established a deep knowledge library for the CFBC boilercomposed of various thermal objects, which includes the object-basedhierarchical networks and the descriptive sets about the objects. Further more,the causal dependency networks among the fault objects and the abnormalparameters is derived in order to avoid the disadvantages of the pure shallowknowledge rule-of-thumb. Utilizing the mixed deep-shallow knowledge library,CFBBEXPTS can not only diagnose the ordinary faults but also forecast anddiagnose a great number of the extra ordinary functional deteriorating faults ofthe CFBC boiler. To surmount the difficulty over the excessive requirements ofthe measurepoints, which was often encountered during diagnosing the faults on the basisof the thermal parameters, a mixed deep-shallow knowledge inference controlmechanism has been devised and realized. Moreover, a series of discrepancydetection principles which face the candidate objects is proposed in this paper.Within the concrete diagnosis process of CFBBEXPTS, in this paper, a newdiagnostic mechanism based on the five advanced diagnosis techniques isproposed, which are candidate generation, discrepancy detection, localdiagnosis, constraint suspension diagnosis and constraint propagationdiagnosis. From the aspects of the theoretical analysis and the technicalrealization, this mechanism could overcorne the difficulty that the faultdiagnosis ofthe large-scale thermal power systems is irrealizablely dependenton the measurability of the themal parameters, while it maintains the efficientutilization ofthe data ofthe measure points. In the last part of this dissertation, the diagnostic ability of CFBBEXPTShas been checked and analyzed by some emulative faults. The results show thatthis expert system has very good fault-catching ability. Thus, CFBBEXPTSmade a significant stride forward in the monitoring and fault diagnosistechnology of the thermal systems. Lastly, in order to illustrate the generalityand openability of the above-mentioned analyzing and devising method, apractical proposal of the remote centralized monitoring and fault diagnosissystem for the conventional 300MW pulverized coal boiler in Weifang PowerPlant, Shangdong Province is worked out in this paper as well.
