A Fuzzy Nearest Neighbor Fusion Diagnosis Method for Thermal System Faults

Abstract

The invention discloses a fuzzy nearest neighbor fusion diagnosis method for thermal system faults, which belongs to the technical field of thermal system fault diagnosis. Obtain the characteristic parameters of the thermal system faults under the rated working conditions of the unit, and determine the normal values ​​of the characteristic parameters under the rated working conditions; use the symptom calculation method to standardize the characteristic parameters, and use the gravity search algorithm to obtain typical fault prototypes; according to the characteristic parameters The real-time fault symptoms are calculated by the measured value and the normal value; the similarity between the real-time fault symptoms and the typical fault prototype is compared, and the fault membership degree is given by the fuzzy nearest neighbor classifier; Diagnosis, give the final diagnosis result. It has the advantages of fast diagnosis speed and high diagnostic accuracy; it can be used for thermal system fault diagnosis under rated working conditions and different steady-state working conditions; through real-time prediction of fault characteristic parameters, it can be applied to dynamic process thermal system fault diagnosis under variable working conditions .

Description

technical field [0001] The invention belongs to the technical field of thermal system fault diagnosis, in particular to a fuzzy nearest neighbor fusion diagnosis method for thermal system faults. Background technique [0002] The thermal system is an important system of a thermal power plant, which has an important impact on the safety and economy of the overall operation of the unit. Carrying out fault diagnosis research on the thermal system is crucial to improving the overall availability of thermal power units, and the safety, reliability and economy of unit operation. It has important theoretical and practical value. The thermal system is a complex series-parallel system composed of multiple strongly coupled subsystems. There are many parameters and mutual influences. The manifestations of faults are diverse. The fault characteristics vary greatly with random group loads and working conditions, resulting in an excessively large sample set of thermal system faults. The ...

AI Technical Summary

Problems solved by technology

The existing methods do not consider the effective reduction of the fault sample set, and only study the diagnosis method. The existence of a large number of redundant and noisy samples increases the storage requirements and calculation costs of the diagnosis system, which seriously affects the actual application effect of the diagnosis system in the field.

In addition, most of the existing methods are researched on a certain stable load condition (such as the rated load condition), and only a few literatures use neural networks to diagnose faults under different steady-state conditions.

At present, large-scale thermal power units need to participate in peak-shaving operation, and the random group load adjustment of the thermal system is often operated under variable conditions. The existing methods cannot adapt to this new situation, which greatly limits the practicability of the fault diagnosis system.

Images