Font Size: a A A

Research On On-Line And Comprehensive Diagnosis System Of Turbogenerator Faults

Posted on:2002-03-28Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y PuFull Text:PDF
GTID:1118360092970059Subject:Power system and its automation
Abstract/Summary:PDF Full Text Request
Fault diagnosis for large turbogenerators plays a very important role for their safe, stable and economic operation. Its aim is to analyze and judge occurring fault on the basis of information acquired by monitoring, to determine property, type, degree, reason and position of fault, point out developing trend and result, propose corresponding measures of regulation and overhaul to bring it under control and remove it, and finally restore the system to its normal state. This paper makes its research incorporating with development of the project, 'On-line and comprehensive fault diagnosis system for 300MW turbogenerator' for Shanghai Waigaoqiao Power Plant. It is a system with multiple functions containing fault diagnosis, diagnostic processing, simulation and training, system service and management. It can realize comprehensive diagnosis for over one hundred fault occurring in turbogenerator and its auxiliary system of hydrogen, oil and water, and propose suggestion concerning diagnostic methods, developing trend, processing methods and preventives, using multiple data sources including on-line and real-time, on-line and non real-time, off-line data and historical diagnostic information. Considering that the expert system is a large software system used in a modern large plant, and there are no precedents in our domestic country and abroad, it is very necessary to undertake such research work. Main work this paper has completed is as follows: 1. Participating in the whole procedure from general design, design in detail, programming to installment and test on site. Emphasis is put on division of function modules and their coordination, system diagnostic data flow and operation control.2. Based on physical model for all fault of turbogenerators proposed by experts, studying diagnostic strategy together with group members. Putting forward searching and control strategy in process of reasoning to solve such problems as multiple reasoning information, confrontation of matched nodes, and real-time diagnosis, etc. Besides, taking part in establishment of knowledge base, including representation of diagnostic knowledge and control knowledge, knowledge storage and knowledge check. 3. Design of function and structure of the subsystem of diagnostic informationpre-processing. Emphasis is put on diagnostic information acquisition and symptom extraction. Diagnostic information of this system includes on-line and real-time data, on-line non real-time data, off-line data and historical diagnosis information. According to data source of each kind and their features, symptoms are extracted respectively. Threshold of symptoms consists of set values and dynamic values calculated by mathematical models. 4. Study on overheating fault diagnosis of stator windings. To provide criterion for early detection of fault concerning stator temperature, computation method of standard temperature (stator bar coolant outlet water temperature and temperature detected by detectors embedded in individual slot) of stator windings for turbogenerators cooled by water and hydrogen is studied, considering influence of operating condition and position, structure and cooling condition of different bars. Method of experimental modeling is used. First, general models are set up for temperature of each measuring position theoretically, taking account of their relationships with turbogenerator structure and operating parameters. Then, on the basis of large quantity of data at different steady and healthy operating points, artificial neural network(ANN)and the least squares method are used to identify these models. These models are very accurate and adaptable for engineering practice.
Keywords/Search Tags:turbogenerator, on-line, fault diagnosis, expert system, diagnostic model, pre-processing, symptom extraction, knowledge base, identification, artificial neural network, the least squares method
PDF Full Text Request
Related items