Fault Diagnosis In Circulating Fluidized Bed Boiler Coking Based On Mutiple PCA Models

As an indispensable and important part of unit in power, the performance and rel-iability of circulating fluidized bed boiler has become the important factor to ensure its safe production. Coking is a common combustion fault of boiler system, and it has great harm to the safe and economic operation, even can reduce the output power of boiler and the thermal efficiency.With the distributed control system is widely used in the complex of large power plants, A large number of data in it may provide the info-rmation of process variables that relate to the coking fault. Since accurate mathemati-cal model is hard to build in most industrial production process, the fault diagnosis method based on the mathematical model meets the great difficulties in practical application, and the multivariate statistical process control method which does not depend on the mathematical model has currently become one of the hot topics.In this paper, with the circulating fluidized bed boiler in Shenhua ZhunGeEr energy gangue power plant as the research object, it makes an overview based on the summary of domestic and international statistical process control theory and fault diagnosis, then, the principal component analysis theory and coking fault in the boiler process flow and the burning process is detailed introduced, also, the related process variables is analyzed and pretreated. Finally, A detailed analysis and simulation about PCA under stable operation in fault diagnosis is metioned. Due to the variety of operation conditions in the power plant boiler, Firstly, this paper discusses the limitations of traditional principal component analysis method. At the same time, an overall scheme about mutiple PCA models fault diagnosis which is used in the variety of operation process is given here. According to the variety of operation conditions in the power plant, the fuzzy c-means clustering method is used to classify the process monitoring data to multiple main component analysis statistical models. Finally, multiple process variables are selected to complete the fault diagnosis of boiler coking phenomenon, also, a lot of simulation experiments are made. This method increases the sensitivity of fault diagnosis, thus it lays a good foundation for practice.