CN-122020353-A - Power generation equipment fault diagnosis method and system based on multi-mode data fusion

CN122020353ACN 122020353 ACN122020353 ACN 122020353ACN-122020353-A

Abstract

The invention discloses a power generation equipment fault diagnosis method and system based on multi-mode data fusion, and relates to the technical field of equipment diagnosis. The method comprises the steps of carrying out operation dynamic collection on power generation equipment to obtain multi-mode heterogeneous data, carrying out space-time alignment on the multi-mode heterogeneous data to construct multi-mode data streams, constructing a combined dual-channel network, carrying out feature analysis on the multi-mode data streams through the combined dual-channel network to generate two-dimensional feature parameters, activating a multi-head self-attention mechanism to fuse the two-dimensional feature parameters, calculating multi-mode feature association weights to generate multi-mode fusion feature vectors, backtracking the multi-mode fusion feature vectors to the power generation equipment to carry out full-connection classification based on the multi-mode fusion feature vectors, carrying out fault identification according to classification results, and generating fault diagnosis results. The technical problem that the failure diagnosis accuracy is low due to the fact that the multi-mode data are difficult to integrate effectively in the prior art is solved, and the technical effect of improving the failure diagnosis accuracy is achieved through deep fusion of the multi-mode data.

Inventors

Dou Cai
JIA WEIJIE
FAN HAO
XIAO XIANGWU
ZHANG BO
ZHOU YU
CHEN XU
YANG SHAOJUN
FAN ZIWEI
MA HONGGUANG
DONG ZHANJIANG

Assignees

辽宁大唐国际新能源有限公司
中国大唐集团数字科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251215

Claims (10)

1. The power generation equipment fault diagnosis method based on multi-mode data fusion is characterized by comprising the following steps of: Carrying out operation dynamic acquisition on power generation equipment to obtain multi-mode heterogeneous data, carrying out space-time alignment on the multi-mode heterogeneous data, and constructing a multi-mode data stream; Constructing a joint dual-channel network, and performing feature analysis on the multi-mode data stream through the joint dual-channel network to generate a two-dimensional feature parameter; activating a multi-head self-attention mechanism to fuse the two-dimensional characteristic parameters, calculating multi-mode characteristic association weights, and generating a multi-mode fusion characteristic vector; and backtracking to the power generation equipment based on the multi-mode fusion feature vector to perform full-connection classification, and performing fault identification according to the classification result to generate a fault diagnosis result.
2. The method for diagnosing a power generating device fault based on multi-modal data fusion as claimed in claim 1, wherein the method for operating and dynamically collecting power generating device to obtain multi-modal heterogeneous data, performing space-time alignment on the multi-modal heterogeneous data to construct a multi-modal data stream comprises: Introducing a clock signal, taking the clock signal as a reference parameter, and decomposing the multi-mode heterogeneous data by the power generation equipment according to a plurality of sampling frequencies based on the reference parameter to obtain a first signal set, a second signal set and a third signal set; extracting vibration signals based on the first signal set to perform short-time Fourier transform, and generating a time-frequency spectrogram as first-mode time sequence data; Extracting a thermal image sequence based on the second signal set, traversing key components of the power generation equipment according to the thermal image sequence to extract region interests, and obtaining a local thermal image sequence as second-mode time sequence data; extracting acoustic spectrum data based on the third signal set, performing mutual information analysis on the acoustic spectrum data and the first-mode time sequence data, and performing time offset compensation on the acoustic spectrum data according to mutual information parameters to obtain a fourth signal set; and recombining the first-modality time sequence data, the second-modality time sequence data, the third signal set and the fourth signal set according to a time axis to generate the multi-modality data stream.
3. The method for diagnosing a power generating apparatus fault based on multi-modal data fusion as claimed in claim 2, wherein a joint dual-channel network is constructed, the multi-modal data stream is subjected to feature analysis through the joint dual-channel network, and a two-dimensional feature parameter is generated, the method comprising: constructing a combined two-channel network by utilizing a convolutional neural network and a long-term memory network, wherein the convolutional neural network comprises a first CNN network and a second CNN network, and the long-term memory network comprises a first LSTM network and a second LSTM network; Synchronizing a time-frequency spectrogram of the first modal time-sequence data to the first CNN network, and extracting deep space spectrum features of vibration signals to serve as a first feature vector; synchronizing the second mode time sequence data to the second CNN network, and extracting the abnormal hot spot characteristics of the temperature distribution of the infrared thermal image as a second characteristic vector; Synchronizing the acoustic spectrum data of the third signal set to the first LSTM network for acoustic feature learning, and obtaining an acoustic feature short-time variation parameter as a third feature vector; Extracting multidimensional operation working condition parameters based on the fourth signal set, synchronizing the multidimensional operation working condition parameters to the second LSTM network, and carrying out working condition learning analysis to obtain a long-period time sequence dependency relationship of the learning working condition parameters as a fourth feature vector; And carrying out dimension division on the first feature vector, the second feature vector, the third feature vector and the fourth feature vector to construct the two-dimensional feature parameter.
4. The power generation equipment fault diagnosis method based on multi-modal data fusion as claimed in claim 3, wherein the construction process of the first CNN network, the second CNN network, the first LSTM network, the second LSTM network, the method includes: The first CNN network is composed of at least two convolution blocks and is used for extracting frequency domain space features from local to global from a time-frequency spectrogram of the first modal time-sequence data; the second CNN network is composed of at least two convolution blocks and is used for extracting spatial distribution morphological characteristics of a temperature field from an image sequence of the second modality time sequence data; The first LSTM network at least comprises a layer of LSTM units and is used for processing the acoustic spectrum data of the third signal set to capture short-time dynamic changes; the second LSTM network at least comprises a layer of LSTM unit, and is used for processing the multidimensional operation condition parameters of the fourth signal set to conduct long-period evolution analysis.
5. The method for diagnosing a power generating device fault based on multi-modal data fusion as claimed in claim 3, wherein the multi-headed self-attention mechanism is activated to fuse the two-dimensional feature parameters, the multi-modal feature association weight is calculated, and a multi-modal fusion feature vector is generated, the method comprising: splicing the first feature vector, the second feature vector, the third feature vector and the fourth feature vector to form an initial fusion feature matrix; synchronizing the initial fusion feature matrix to a multi-head self-attention layer for linear transformation calculation, and constructing a multi-head self-attention mechanism which comprises a plurality of attention heads; Traversing the plurality of attention heads for normalization analysis, and determining attention weight distribution data; weighting and summing the attention heads based on the attention weight distribution data to obtain multi-mode feature association weights of the attention heads; And linearly projecting the first feature vector, the second feature vector, the third feature vector and the fourth feature vector based on the multi-modal feature association weight to obtain the multi-modal fusion feature vector.
6. The method for diagnosing a power generating equipment fault based on multi-modal data fusion as claimed in claim 5, wherein the first feature vector, the second feature vector, the third feature vector and the fourth feature vector are spliced to form an initial fusion feature matrix, the method comprising: sequentially splicing the first feature vector, the second feature vector, the third feature vector and the fourth feature vector along the feature dimension direction to determine a target row vector; And expanding based on the target row vector to generate a two-dimensional matrix, and taking the two-dimensional matrix as the initial fusion feature matrix, wherein each row or each column of the initial fusion feature matrix represents an independent feature vector.
7. The method for diagnosing a power generating apparatus fault based on multi-modal data fusion as claimed in claim 1, wherein the multi-modal fusion feature vector is traced back to the power generating apparatus for full connection classification, fault recognition is performed according to the classification result, and a fault diagnosis result is generated, the method comprising: constructing a deep neural network classifier, wherein the deep neural network classifier comprises a plurality of full connection layers; Synchronizing the multi-mode fusion feature vector to the plurality of full-connection layers for classification to obtain a classification result; carrying out state analysis on the power generation equipment according to the classification result to obtain equipment state parameters, wherein the equipment state parameters comprise fault state parameters and health state parameters; Carrying out distribution calculation based on the fault state parameters and the health state parameters to obtain probability distribution data of the fault state parameters and probability distribution data of the health state parameters; and carrying out fault identification based on the probability distribution data of the fault state parameters and the probability distribution data of the health state parameters, and generating the fault diagnosis result.
8. The method for diagnosing a power generating apparatus fault based on multi-modal data fusion as claimed in claim 7, wherein synchronizing the multi-modal fusion feature vector to the plurality of fully connected layers for classification, the method comprising: The plurality of full-connection layers at least comprise a first full-connection layer, a second full-connection layer and an output layer; The first full-connection layer is used for receiving the multi-mode fusion feature vector to perform linear transformation and generating a first output result; The second full-connection layer receives the first output result of the first full-connection layer and performs characteristic compression to generate a second output result; And the output layer maps the second output result generated by the second full-connection layer to the same dimension as the number of the fault categories to obtain the classification result.
9. The power generation equipment fault diagnosis method based on multi-modal data fusion as set forth in claim 7, wherein fault recognition is performed based on probability distribution data of the fault state parameters, probability distribution data of the health state parameters, and the fault diagnosis result is generated, the method including: Setting a probability threshold value based on the probability distribution data of the health state parameters, and comparing and judging the probability distribution data of the fault state parameters with the probability threshold value; when the probability distribution data of the fault state parameters exceeds the probability threshold value, judging that the power generation equipment has faults to perform fault identification, and determining a fault mode; Triggering an abnormal detection alarm when the fault state parameters are all lower than a probability threshold and the health state probability is lower than a preset health threshold; Performing confidence analysis based on the fault mode to generate a first confidence coefficient; performing confidence analysis based on the anomaly detection alarm to generate a second confidence level; and carrying out fault identification based on the first confidence coefficient and the second confidence coefficient in combination with the fault mode and the abnormal detection alarm, and generating a fault diagnosis result.
10. A power generation equipment fault diagnosis system based on multi-modal data fusion, characterized in that it is used for implementing the power generation equipment fault diagnosis method based on multi-modal data fusion according to any one of claims 1-9, the system comprising: The data processing module is used for carrying out operation dynamic acquisition on the power generation equipment to obtain multi-mode heterogeneous data, carrying out space-time alignment on the multi-mode heterogeneous data and constructing a multi-mode data stream; The characteristic analysis module is used for constructing a combined dual-channel network, carrying out characteristic analysis on the multi-mode data stream through the combined dual-channel network and generating a two-dimensional characteristic parameter; the feature fusion module activates a multi-head self-attention mechanism to fuse the two-dimensional feature parameters, calculates multi-modal feature association weights and generates multi-modal fusion feature vectors; And the fault identification module is used for backtracking to the power generation equipment based on the multi-mode fusion feature vector to carry out full-connection classification, carrying out fault identification according to the classification result and generating a fault diagnosis result.

Description

Power generation equipment fault diagnosis method and system based on multi-mode data fusion Technical Field The invention relates to the technical field of equipment diagnosis, in particular to a power generation equipment fault diagnosis method and system based on multi-mode data fusion. Background During the operation of the power generation equipment, various monitoring data such as electric quantity, mechanical quantity, heat, environmental quantity, images, acoustics and the like can be generated, and the monitoring information reflects the operation state and potential fault characteristics of the equipment. However, the existing monitoring system generally collects data of different modes independently, and obvious differences exist among modes in sampling frequency, data format, space-time resolution and noise characteristics, so that fusion analysis is difficult to directly perform. Most of traditional fault diagnosis methods are based on single-mode or simply spliced data to perform feature extraction, and cannot effectively capture the association relation between multi-mode data, so that feature expression capacity is insufficient, and further equipment faults under complex working conditions are difficult to accurately identify. Disclosure of Invention The application provides a power generation equipment fault diagnosis method and system based on multi-mode data fusion, which solve the technical problem that the accuracy of fault diagnosis is low because multi-mode data are difficult to integrate effectively in the prior art. In a first aspect of the present application, there is provided a power generation equipment fault diagnosis method based on multi-modal data fusion, the method comprising: The method comprises the steps of carrying out operation dynamic collection on power generation equipment to obtain multi-mode heterogeneous data, carrying out space-time alignment on the multi-mode heterogeneous data to construct multi-mode data streams, constructing a combined dual-channel network, carrying out feature analysis on the multi-mode data streams through the combined dual-channel network to generate two-dimensional feature parameters, activating a multi-head self-attention mechanism to fuse the two-dimensional feature parameters, calculating multi-mode feature association weights to generate multi-mode fusion feature vectors, carrying out full-connection classification on the multi-mode fusion feature vectors to the power generation equipment based on the multi-mode fusion feature vectors, carrying out fault recognition according to classification results, and generating fault diagnosis results. In a second aspect of the present application, there is provided a power generation equipment fault diagnosis system based on multi-modal data fusion, the system comprising: The system comprises a data processing module, a feature analysis module, a feature fusion module, a fault identification module and a fault diagnosis module, wherein the data processing module is used for carrying out operation dynamic acquisition on power generation equipment to obtain multi-mode heterogeneous data, carrying out space-time alignment on the multi-mode heterogeneous data to construct multi-mode data flow, the feature analysis module is used for constructing a joint dual-channel network, carrying out feature analysis on the multi-mode data flow through the joint dual-channel network to generate two-dimensional feature parameters, the feature fusion module is used for activating a multi-head self-attention mechanism to fuse the two-dimensional feature parameters, calculating multi-mode feature association weights to generate multi-mode fusion feature vectors, and the fault identification module is used for carrying out full-connection classification on the multi-mode fusion feature vectors back to the power generation equipment and carrying out fault identification according to classification results to generate fault diagnosis results. One or more technical schemes provided by the application have at least the following technical effects or advantages: Firstly, running dynamic acquisition is carried out on power generation equipment to obtain multi-mode heterogeneous data, space-time alignment is carried out on the multi-mode heterogeneous data, and a multi-mode data stream is constructed. Then, constructing a joint dual-channel network, and carrying out feature analysis on the multi-mode data stream through the joint dual-channel network to generate a two-dimensional feature parameter. And then, activating a multi-head self-attention mechanism to fuse the two-dimensional feature parameters, calculating the multi-mode feature association weight, and generating a multi-mode fusion feature vector. And finally, backtracking to the power generation equipment based on the multi-mode fusion feature vector to perform full-connection classification, and performing fault identification according to the classification r