CN-121997503-A - Method for identifying vibration abnormality of pipeline of thermal energy storage system

CN121997503ACN 121997503 ACN121997503 ACN 121997503ACN-121997503-A

Abstract

The invention provides a method for identifying vibration abnormality of a pipeline of a thermal energy storage system, which belongs to the technical field of thermal energy storage systems, and comprises the steps of installing vibration sensors at key nodes of a pipeline network of the thermal energy storage system, constructing a topological model based on graph theory, extracting multidimensional vibration feature vectors by wavelet packet decomposition and empirical mode decomposition, establishing a working condition identification code system, utilizing a longest public subsequence algorithm to match historical working condition references, the method comprises the steps of constructing an upper game model with maximum detection precision as a target and a lower game model with minimum false alarm rate as a target, realizing parameter collaborative optimization, combining finite element transfer function calculation to realize accurate positioning of an abnormal source, carrying out hierarchical abnormal judgment according to a deviation metric value, and adopting a self-adaptive monitoring frequency adjustment and sliding window database updating mechanism to solve the technical problems of insufficient detection precision and higher false alarm rate of pipeline vibration of a thermal energy storage system.

Inventors

ZHANG LEI
XU MEI
ZHANG CHENXI
WANG ZEZHONG
ZHU CHANG
WEI FEI
BAI DINGRONG

Assignees

鄂尔多斯实验室
清华大学

Dates

Publication Date: 20260508
Application Date: 20251203

Claims (10)

1. A method for identifying vibration abnormality of a pipeline of a thermal energy storage system includes the steps of installing vibration sensors at key nodes of each pipeline section of the thermal energy storage system, establishing a pipe network topology model based on graph theory structures, collecting original vibration signals of each vibration sensor, conducting noise reduction processing through wavelet packet decomposition technology and spectral kurtosis self-adaptive selection, adopting empirical mode decomposition to extract intrinsic mode functions of the vibration signals after noise reduction to form multidimensional vibration characteristic vectors, establishing working condition identification codes according to working condition parameters of the current system, searching historical working condition references in a multi-working condition vibration fingerprint database through a longest common subsequence algorithm, establishing an upper game model and a lower game model, determining optimal parameters of the vibration abnormality identification model through double-layer game optimization, inputting the current multidimensional vibration characteristic vectors into the vibration abnormality identification model, calculating to obtain abnormal probability distribution and abnormal type prediction results of each node of the pipe network, locating abnormal source positions based on a finite element transfer function and a shortest path algorithm, calculating deviation metric value of the current abnormal probability distribution and the historical working condition references to conduct abnormal judgment, adjusting vibration monitoring frequencies according to abnormal type and severity, and updating the multi-working condition vibration fingerprint database.
2. The method for identifying abnormal vibration of a pipeline of a thermal energy storage system according to claim 1, wherein the pipe network topology model based on graph theory structure is characterized in that the thermal energy storage system pipe network is abstracted into a graph structure, pipe sections are used as edges, connection points are used as nodes, and connection relations among the pipe sections and pipeline geometric parameters are recorded.
3. The method for identifying vibration anomalies of a thermal energy storage system pipeline according to claim 2, wherein the upper game model aims at maximizing vibration anomaly detection precision, the lower game model aims at minimizing false alarm rate, objective functions are a detection precision function and a false alarm rate function, and the two objective functions are coupled through an anomaly discrimination threshold.
4. The method for identifying abnormal vibration of a pipeline of a thermal energy storage system according to claim 3, wherein the spectral kurtosis is adaptively selected, in particular to a method for automatically determining an optimal filtering frequency band based on statistical characteristics of signals, and a frequency range containing the most fault information is selected by calculating the spectral kurtosis value of each frequency band.
5. The method for recognizing the vibration abnormality of the heat energy storage system pipeline according to claim 4, wherein the vibration abnormality recognition model is specifically an abnormality detection model constructed based on a deep learning technology, wherein multidimensional vibration feature vectors are input, and an abnormality probability distribution and an abnormality type prediction result are output.
6. The method for identifying abnormal vibration of a thermal energy storage system according to claim 5, wherein the multi-dimensional vibration feature vector is a multi-dimensional array including vibration signal frequency domain, time domain and energy features, and is used for describing the comprehensive characteristics of the vibration signal.
7. The method for identifying abnormal vibration of a pipeline of a thermal energy storage system according to claim 6, wherein the condition identification code is specifically a character sequence generated according to a fixed coding rule according to current load level, flow rate and temperature operation parameters of the thermal energy storage system, and is used for identifying different operation conditions.
8. The method for identifying abnormal vibration of a pipeline of a thermal energy storage system according to claim 7, wherein the multi-working-condition vibration fingerprint database, in particular a knowledge base storing normal vibration characteristics under different working conditions, comprises vibration frequency spectrums and amplitude distribution standard modes of the working conditions.
9. The method according to claim 8, wherein the finite element transfer function, in particular describing the propagation characteristics of the vibration signal in the pipe structure, is used for calculating the propagation path and attenuation coefficient of the vibration in the pipe network.
10. The method for identifying abnormal vibration of a thermal energy storage system pipeline according to claim 9, wherein the deviation metric, in particular, a quantization index of a difference between a current abnormal probability distribution and a reference of a historical working condition is obtained by calculating a euclidean distance, the method is judged to be in a normal state when the deviation metric is epsilon [0,0.25 ], a secondary detection flow is triggered when the deviation metric is epsilon [0.25,0.7 ], and the abnormal vibration is confirmed when the deviation metric is epsilon [0.7,1 ].

Description

Method for identifying vibration abnormality of pipeline of thermal energy storage system Technical Field The invention belongs to the technical field of heat energy storage systems, and particularly relates to a method for identifying abnormal vibration of a heat energy storage system pipeline. Background The vibration monitoring of the heat energy storage system pipeline is a key technical link for guaranteeing the safe and stable operation of the system, the traditional vibration abnormality detection method mainly adopts a single-dimension signal processing technology such as fixed threshold comparison, spectrum analysis and the like, vibration signals are acquired in real time and compared with a preset threshold by arranging vibration sensors on a key pipeline section, and when the vibration amplitude exceeds the set threshold, the abnormal state is judged. The method is widely applied to the fields of industrial pipeline monitoring, petrochemical device vibration diagnosis and the like, and has the advantages of simplicity in implementation and lower cost. However, the traditional vibration detection technology has obvious defects that the normal property of the same vibration signal under different working conditions is obviously different due to the complex and changeable operation working conditions of the thermal energy storage system, the fixed threshold method is difficult to adapt to the working condition change and is easy to generate false alarm, meanwhile, the traditional method mostly adopts single characteristic parameters to carry out abnormal judgment, the comprehensive analysis on the multidimensional characteristics of the vibration signal is lacking, the detection precision is limited, and in addition, the prior art lacks an effective abnormal source positioning mechanism, and the abnormal occurrence position is difficult to quickly and accurately determine. Therefore, in the prior art, due to the lack of a self-adaptive abnormality detection mechanism aiming at the change of multiple working conditions and a comprehensive analysis method of multidimensional vibration characteristics, the technical problems of insufficient precision and high false alarm rate exist in the detection of the vibration abnormality of the pipeline of the heat energy storage system. Disclosure of Invention In view of the above, the invention provides a method for identifying vibration abnormality of a thermal energy storage system pipeline, which can solve the technical problems of insufficient detection precision and higher false alarm rate of the vibration abnormality of the thermal energy storage system pipeline in the prior art. The invention provides a method for identifying vibration abnormality of a pipeline of a heat energy storage system, which comprises the steps of installing vibration sensors at key nodes of each pipeline section of the heat energy storage system, establishing a pipe network topology model based on a graph theory structure, collecting original vibration signals of each vibration sensor, carrying out noise reduction processing through wavelet packet decomposition technology and spectral kurtosis self-adaptive selection, adopting empirical mode decomposition to extract intrinsic mode functions of the noise-reduced vibration signals to form multidimensional vibration characteristic vectors, establishing working condition identification codes according to working condition parameters of the current system, searching historical working condition references in a multi-working condition vibration fingerprint database by utilizing a longest common subsequence algorithm, establishing an upper game model and a lower game model, determining optimal parameters of the vibration abnormality identification model through double-layer game optimization, inputting the current multidimensional vibration characteristic vectors into the vibration abnormality identification model, calculating to obtain abnormal probability distribution and abnormal type prediction results of each node of the pipe network, positioning abnormal source positions based on a finite element transfer function and a shortest path algorithm, calculating deviation metric value of the current abnormal probability distribution and the historical working condition references, carrying out abnormal judgment, adjusting vibration frequency according to the abnormal type and severity, and updating the multi-working condition fingerprint database. The pipe network topology model based on the graph theory structure specifically abstracts a pipe network of the thermal energy storage system into a graph structure, takes pipe sections as edges, takes connection points as nodes, and records connection relations among the pipe sections and geometric parameters of the pipeline. The upper game model aims at maximizing vibration anomaly detection precision, the lower game model aims at minimizing false alarm rate, objective functions are a det