Study On Fault Diagnosis Of Thermal Power Plants Based On Multiscale Analysis

In order to ensure the reliable and economical operation of the thermal powerplants, the study of fault diagnosis of them has been given a significant attention. It isfurther complicated by the size and complexity of modern process plants with com-puter control system. But the requirement of the reliability of the whole power plantshave been improved with the competition growing. The focus of fault diagnosis isalso transferred from the key part equipments to the whole operation in power plants.Power plants can be recognized as one kinds of complex systems in which multiscaleidea has been studied intensively, it is therefor meaningful to study the methods offault diagnosis using multiscale idea for modifying the fault diagnosis system andimproving the safety and reality in power plants.So the study of this thesis has been developed around the multiscale fault diag-nosis in power plants. We analyzed the multiscale features and complexity of faultdiagnosis for power plants as a starter, next provide fault diagnosis methods of mul-tiscale statistic according to the process historic data, then consider the informationof multiscale structure and cost obtaining feature of fault detection and analysis somemethod for fault diagnosis. The contents in this thesis ware had mainly in the follow-ing aspects.First, the multiscale theory is introduced to diagnosis method of power station.In the multscale theory, the problem is decomposed at different scale and the infor-mation is used between each scale. It can reduce the complexity of the problem. Themultiscale study is divided into three part through the analysis of fault diagnosis andmultiscale research: characteristic multiscale, measure multiscale and cognitive mul-tiscale. The point of power station diagnosis is focused on the coordination of thisthree multscale diagnosis method. The characteristic multiscale, measure multiscaleand cognitive multiscale in power station is analysis from power station structure andrun procedures. The further analysis to the complexity of power station diagnosisshow that complexity of regular classical diagnosis cause the problem can not be re-alized and the multiscale diagnosis have advantage in this aspect.Second, the multiscale research in power station is diagnosis method based on statistical techniques. By using of the correspondence between time scale of mea-surement signal and space scale of power station, association rules is applianced toanalysis the change of running state relations in different time scale. From this, thefault detection method is introduced. In order to solve the problem that the detail scalecontain little macroscopic information. The multiscale diagnosis method of utilizingentire characteristic is given by the catching and analysis multiscale characteristic ofsystem running.Again, the structure information of power station is used to improve fault diag-nosis precision accuracy. Due to the discrepancy of measuring elements in monitoringsystem, the directed graph diagnosis method concluding unmeasured node symbol isresearched to improve the application extent of scale model. Due to the performanceof variable working condition, the diagnosis method of using SDG model is study.Finally, the optimization problem of fault diagnosis strategy is study in thoughtof detecting expense diversity. Most of detection property is multiple values.Multiple-valued attribute problem is analyzed and optimal diagnostic strategy algo-rithm is proposed. Due to the common phenomena of detecting information is imper-fect. We analyze the diagnostic strategy base on imperfect information. According tothe research feature in diagnosis process, the scale function represent diagnosis pro-cess information is introduced, therefore, diagnostic strategy optimization methods ofmultiple-valued attributes and imperfect information is suggested.