Data reconciliation and hypothesis testing based multi-fault diagnosis method for power plant system

Abstract

The invention provides a data reconciliation and hypothesis testing based multi-fault diagnosis method for power plant system, which belong to the technical field of power plant performance monitoring and fault diagnosing. The technical steps of the invention comprise 1) selecting quasi-static measuring data; establishing a system constraint equation and carrying out data coordination calculation; 2) calculating the global hypothesis testing statistic amount; and comparing with the allowable critical value to determine whether the parameters have significant errors; 3) if there are significant errors, recognizing the parameter with errors and removing it from proceeding to the next step; 4) calculating the testing statistic amount after rank reduction and comparing with a new critical value to determine whether the parameters have significant errors; 5) if there are no significant errors, then completing the calculation; and otherwise, repeating steps 3 and 4 until all significant errors are identified through tests. The method uses the idea of sequence elimination to realize fast and accurate fault location one by one. The significant error thresholds can be used as power plant alarm thresholds to prompt operation men at early stage so as to ensure the security of machines and reduce the production costs.

Description

technical field [0001] The invention relates to a multi-fault diagnosis method of a power station system based on data coordination and hypothesis testing, and belongs to the field of power plant performance monitoring and fault diagnosis. Background technique [0002] Accurate measurement data is an important basis for on-line performance monitoring, fault diagnosis and operation optimization of power plants. However, due to human misoperation or aging of measuring instruments, the original data obtained by direct measurement of power plants always contain measurement errors. Measurement errors are usually categorized as random errors and significant errors (gross errors). Under a large number of measurement samples, random errors obey certain statistical laws, while significant errors do not obey the statistical distribution of a large number of samples. The existence of significant errors usually reflects the failure of equipment or instruments. Therefore, detecting and...

AI Technical Summary

Problems solved by technology

[0003] Using the redundant flow measurement information in the steam turbine system, the detection and identification of flowmeter faults can be realized, but this method only considers the balance relationship of the steam turbine system, and does not consider other systems in the power plant (such as boiler systems, condensate water systems, etc.) ), and the characteristic equation of the equipment is not considered, and the data coordination objective function adopted is less robust, resulting in poor adaptability of the method, which needs to be improved urgently

Images