CN115392370A - Fault diagnosis method and system for thermal power generation equipment - Google Patents

Fault diagnosis method and system for thermal power generation equipment Download PDF

Info

Publication number
CN115392370A
CN115392370A CN202211017981.1A CN202211017981A CN115392370A CN 115392370 A CN115392370 A CN 115392370A CN 202211017981 A CN202211017981 A CN 202211017981A CN 115392370 A CN115392370 A CN 115392370A
Authority
CN
China
Prior art keywords
thermal power
boiler
feature vector
vector
working parameters
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211017981.1A
Other languages
Chinese (zh)
Inventor
罗恒
方果
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhuji Yingduo Trading Co ltd
Original Assignee
Zhuji Yingduo Trading Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhuji Yingduo Trading Co ltd filed Critical Zhuji Yingduo Trading Co ltd
Priority to CN202211017981.1A priority Critical patent/CN115392370A/en
Publication of CN115392370A publication Critical patent/CN115392370A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The embodiment of the application provides a fault diagnosis method and a fault diagnosis system for thermal power generation equipment, wherein a context encoder is used for encoding multiple working parameters of a boiler of a thermal power generating unit at multiple preset time points, a first convolutional neural network is used for processing the working parameters to obtain a second characteristic vector, high-dimensional local implicit associated features in a time sequence dimension are extracted, a time sequence encoder is used for encoding sequences of the working parameters in the multiple preset time points, a second convolutional neural network is used for processing the working parameters to obtain a fourth characteristic vector, high-dimensional local implicit features of high-dimensional change features in the time sequence dimension among different parameter samples are extracted, then, derived information hyper-convexity factors between the second characteristic vector and the fourth characteristic vector are used for weighted fusion of weighting coefficients, and the condition that the classification precision is influenced due to the fact that the derived information is lost between the second characteristic vector and the fourth characteristic vector is avoided, and therefore the fault diagnosis precision of the boiler is improved.

Description

Fault diagnosis method and system for thermal power generation equipment
Technical Field
The present disclosure relates to the field of fault diagnosis of thermal power plants, and more particularly, to a fault diagnosis method and system for a thermal power plant.
Background
With the continuous development of the economic level of China, the demand of the society on the power industry is increasingly improved. In the electric power structure of China, thermal power generation still occupies a very important position. In a thermal generator set, a boiler is one of key devices, and has the characteristics of complex system, high coupling, multiple operation parameters and the like; and because the boiler equipment works in a high-temperature, high-pressure and high-vibration environment for a long time, the frequency of faults is relatively high, and the faults of the boiler equipment in one day can cause the unplanned shutdown of a power plant, influence the stability of the output power of the power transmission to a power grid, waste energy and improve the power generation cost.
Therefore, a technical solution for diagnosing a fault in a boiler of a thermal power generating unit is desired.
Disclosure of Invention
The application provides a fault diagnosis method and a fault diagnosis system for thermal power generation equipment, wherein a context encoder is used for encoding multiple working parameters of a boiler of a thermal power generating unit at multiple preset time points, a first convolutional neural network is used for processing to obtain a second characteristic vector, high-dimensional local implicit correlation characteristics in a time sequence dimension are extracted, a time sequence encoder is used for encoding sequences of the working parameters in the multiple preset time points, a second convolutional neural network is used for processing to obtain a fourth characteristic vector, high-dimensional local implicit characteristics of high-dimensional change characteristics in the time sequence dimension among different parameter samples are extracted, then, derived information super-convexity factors between the second characteristic vector and the fourth characteristic vector are used as weighting fusion of weighting coefficients to obtain classification characteristic vectors, the influence on classification accuracy caused by information loss of the second characteristic vector and the fourth characteristic vector is avoided, and therefore the fault diagnosis accuracy of the boiler is improved.
In a first aspect, the present application provides a fault diagnosis system for a thermal power plant, the system comprising:
the system comprises a working parameter data acquisition unit, a data processing unit and a data processing unit, wherein the working parameter data acquisition unit is used for acquiring a plurality of working parameters of a boiler of the thermal power generating unit at a plurality of preset time points including a current time point, and the plurality of working parameters comprise superheated steam flow, superheated steam pressure, superheated steam temperature, reheated steam flow, reheated steam inlet pressure, reheated steam outlet pressure, reheated steam inlet temperature, reheated steam outlet temperature, normal boiler water volume, feed water temperature, primary air volume, secondary air volume, coal consumption, flue gas volume and corrected exhaust smoke temperature;
the context parameter coding unit is used for enabling a plurality of working parameters of the boiler of the thermal power generating unit at each preset time point to pass through a context coder to obtain a plurality of eigenvectors, and cascading the plurality of eigenvectors to obtain first eigenvectors of the plurality of working parameters of the boiler of the thermal power generating unit corresponding to each preset time point;
the parameter time sequence correlation coding unit is used for performing time dimension two-dimensional arrangement on first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit at each preset time point to form a first characteristic matrix and then generating second eigenvectors by using a first convolution neural network of a first convolution kernel;
the parameter time sequence coding unit is used for arranging various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points into input vectors according to the time dimension respectively and then generating third feature vectors corresponding to various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points through a time sequence coder comprising a one-dimensional convolution layer and a full connection layer;
the parameter sample correlation coding unit is used for two-dimensionally arranging third eigenvectors of each working parameter in a plurality of working parameters of a boiler of the thermal power generating unit at a plurality of preset time points into a second eigenvector matrix according to a sample dimension and then generating a fourth eigenvector by using a second convolution neural network of a second convolution kernel, wherein the size of the second convolution kernel of the second convolution neural network is larger than that of the first convolution kernel of the first convolution neural network;
a weighted fusion unit, configured to perform weighted fusion on the second feature vector and the fourth feature vector based on a derived information hyper-convex metric between the second feature vector and the fourth feature vector as a weighting coefficient to obtain a classified feature vector, where the derived information hyper-convex metric between the second feature vector and the fourth feature vector is generated based on a weighted sum of absolute values of differences between feature values of corresponding positions in the second feature vector and the fourth feature vector; and
and the early warning unit is used for enabling the classified characteristic vectors to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether fault early warning is generated or not.
In one possible implementation manner, in the above fault diagnosis system, the context parameter encoding unit is further configured to: mapping each parameter in a plurality of working parameters of a boiler of the thermal power generating unit at the same preset time point into an embedded vector by using an embedded layer of the context encoder to obtain a sequence of the embedded vectors; and global semantic encoding the sequence of embedded vectors based on the upper and lower bits using a converter of the context encoder to obtain the plurality of feature vectors.
In one possible implementation manner, in the fault diagnosis system, the parameter timing-related encoding unit is further configured to: performing, using layers of the first convolutional neural network, in a layer forward pass, respectively input data: performing convolution processing on the input data by using a first convolution kernel to generate a convolution characteristic map; performing global pooling processing based on a feature matrix on the convolution feature map to generate a pooled feature vector; and performing non-linear activation based on the pooled feature vectors to generate activation feature vectors; wherein the activation feature vector output by the last layer of the first convolutional neural network is the second feature vector.
In one possible implementation manner, in the fault diagnosis system, the parameter time-series encoding unit is further configured to: arranging various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the preset time points into one-dimensional input vectors corresponding to the boiler of the thermal power generating unit on a daily basis according to the time dimension; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000031
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
Figure BDA0003813051010000032
represents a matrix multiplication;
performing one-dimensional convolutional coding on the input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000033
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
In one possible implementation manner, in the fault diagnosis system, the weighted fusion unit is further configured to:
calculating a derivative information super-convexity coefficient between the second feature vector and the fourth feature vector as a weighting coefficient according to the following formula;
wherein the formula is:
Figure BDA0003813051010000034
wherein v is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Representing the second eigenvector sum V 4 Denotes a fourth feature vector and w denotes a weighting coefficient. In one possible implementation manner, in the fault diagnosis system, the early warning unit is further configured to:
processing the classification feature vector using the classifier to generate a classification result with the formula: softmax { (W) n ,B n ):…:(W 1 ,B 1 )|X)}，W 1 To W n As a weight matrix for each fully connected layer, B 1 To B n A bias matrix representing the layers of the fully connected layer.
In a second aspect, the present application provides a fault diagnosis method for a thermal power plant, including:
acquiring multiple working parameters of a boiler of the thermal power generating unit at multiple preset time points including the current time point, wherein the multiple working parameters comprise superheated steam flow, superheated steam pressure, superheated steam temperature, reheated steam flow, reheated steam inlet pressure, reheated steam outlet pressure, reheated steam inlet temperature, reheated steam outlet temperature, normal water volume of the boiler, feedwater temperature, primary air volume, secondary air volume, coal consumption, flue gas volume and corrected exhaust smoke temperature;
obtaining a plurality of eigenvectors by passing a plurality of working parameters of the boiler of the thermal power generating unit at each preset time point through a context encoder, and cascading the plurality of eigenvectors to obtain a first eigenvector corresponding to the plurality of working parameters of the boiler of the thermal power generating unit at each preset time point;
performing time dimension two-dimensional arrangement on first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit at each preset time point to form a first eigenvector matrix, and generating a second eigenvector by using a first convolution neural network of a first convolution kernel;
after all working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points are respectively arranged as input vectors according to the time dimension, a time sequence encoder comprising a one-dimensional convolution layer and a full connection layer is used for generating third eigenvectors corresponding to all working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points;
arranging third eigenvectors of each working parameter in multiple working parameters of a boiler of the thermal power generating unit at multiple preset time points in a second eigenvector matrix according to a sample dimension two-dimensional way, and generating a fourth eigenvector by using a second convolution neural network of a second convolution kernel, wherein the size of the second convolution kernel of the second convolution neural network is larger than that of the first convolution kernel of the first convolution neural network;
performing weighted fusion on the second feature vector and the fourth feature vector based on a derivative information super-convexity metric factor between the second feature vector and the fourth feature vector as a weighting coefficient to obtain a classified feature vector, wherein the derivative information super-convexity metric factor between the second feature vector and the fourth feature vector is generated based on the weighted sum of absolute values of differences between feature values of corresponding positions in the second feature vector and the fourth feature vector; and
and passing the classified feature vectors through a classifier to obtain a classification result, wherein the classification result is used for indicating whether fault early warning is generated or not.
In one possible implementation manner, in the fault diagnosis method, the passing multiple operating parameters of the boiler of the thermal power generating unit at each predetermined time point through a context encoder to obtain multiple eigenvectors includes:
mapping each parameter in a plurality of working parameters of a boiler of the thermal power generating unit at the same preset time point into an embedded vector by using an embedded layer of the context encoder to obtain a sequence of the embedded vectors; and
global semantic encoding, using a converter of the context encoder, the sequence of embedded vectors based on upper and lower bits to obtain the plurality of feature vectors.
In one possible implementation manner, in the fault diagnosis method, after arranging the operating parameters of the boilers of the thermal power generating units at the predetermined time points as input vectors according to the time dimension, generating a third eigenvector corresponding to each of the operating parameters of the boilers of the thermal power generating units at the predetermined time points by using a time sequence encoder including a one-dimensional convolution layer and a fully-connected layer, respectively, includes:
arranging various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the preset time points into one-dimensional input vectors corresponding to the boiler of the thermal power generating unit on a daily basis according to the time dimension;
using a full-connection layer of the time sequence encoder to perform full-connection encoding on the input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000051
Figure BDA0003813051010000052
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
Figure BDA0003813051010000053
represents a matrix multiplication;
performing one-dimensional convolutional coding on the input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000054
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
In one possible implementation manner, in the fault diagnosis method, the performing weighted fusion on the second feature vector and the fourth feature vector based on a hyperconvexity metric factor of derived information therebetween as a weighting coefficient to obtain a classified feature vector includes:
calculating a derivative information super-convexity coefficient between the second feature vector and the fourth feature vector as a weighting coefficient according to the following formula;
wherein the formula is:
Figure BDA0003813051010000055
wherein v is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Representing the sum of the second feature vectors V 4 Denotes a fourth feature vector and w denotes a weighting coefficient.
In a third aspect, the present application provides an electronic device, comprising:
one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, the one or more computer programs comprising instructions which, when executed by the apparatus, cause the apparatus to perform the method of the second aspect.
In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program which, when run on a computer, causes the computer to perform the method according to the second aspect.
In a fifth aspect, the present application provides a computer program for performing the method of the second aspect when the computer program is executed by a computer.
In a possible design, the program of the fifth aspect may be stored in whole or in part on a storage medium packaged with the processor, or in part or in whole on a memory not packaged with the processor.
The fault diagnosis method and the fault diagnosis system for the thermal power generation equipment at least achieve the following beneficial effects: the method comprises the steps of coding multiple working parameters of a boiler of a thermal power generating unit at multiple preset time points by using a context coder, processing the working parameters through a first convolutional neural network to obtain a second characteristic vector so as to extract high-dimensional local implicit associated features in a time sequence dimension, coding a sequence of each working parameter in the multiple working parameters at the multiple preset time points by using the time sequence coder, processing the sequence through a second convolutional neural network to obtain a fourth characteristic vector so as to extract high-dimensional local implicit features of high-dimensional change features in the time sequence dimension among different parameter samples, and then conducting weighted fusion on the second characteristic vector and the fourth characteristic vector based on a derivative information hyperconvexity factor between the second characteristic vector and the fourth characteristic vector as a weighting coefficient to avoid causing information loss of the second characteristic vector and the fourth characteristic vector to influence classification precision, so that the precision of fault diagnosis of the boiler is improved.
Drawings
Fig. 1 is a schematic view of an application scenario of an embodiment of a fault diagnosis system for a thermal power plant according to the present application;
FIG. 2 is a schematic structural diagram of an embodiment of a fault diagnosis system for a thermal power plant according to the present application;
FIG. 3 is a schematic flow diagram illustrating an embodiment of a fault diagnosis system for a thermal power plant according to the present application;
FIG. 4 is a schematic diagram of a fault diagnosis method for a thermal power plant according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of an embodiment of an electronic device according to the present application.
Detailed Description
The terminology used in the description of the embodiments section of the present application is for the purpose of describing particular embodiments of the present application only and is not intended to be limiting of the present application.
Summary of the application
In recent years, deep learning and neural networks have been widely used in the fields of computer vision, natural language processing, text signal processing, and the like. In addition, deep learning and neural networks also exhibit a level close to or even exceeding that of humans in the fields of image classification, object detection, semantic segmentation, text translation and the like. Therefore, the applicant provides a new solution for diagnosing the fault of the boiler of the thermal power generating unit in consideration of deep learning and development of a neural network.
Essentially, fault diagnosis or fault early warning of a boiler is a classification problem, that is, whether a fault exists in the boiler is judged based on external parameter characterization of the boiler. Specifically, in the embodiment of the present application, a plurality of operating parameters of a boiler of a thermal power generating unit at a plurality of predetermined time points including a current time point are obtained. In order to prevent data dimension explosion and take into account that sufficient information for classification judgment can be extracted, in the embodiment of the application, the selected working parameters comprise superheated steam flow, superheated steam pressure, superheated steam temperature, reheated steam flow, reheated steam inlet pressure, reheated steam outlet pressure, reheated steam inlet temperature, reheated steam outlet temperature, normal boiler water volume, feedwater temperature, primary air volume, secondary air volume, coal consumption, flue gas volume and corrected exhaust smoke temperature.
It should be understood that during operation, the boiler is associated with parameters at the same point in time, the boiler is associated with parameters at different points in time, and the boiler is also associated with parameters at different points in time. In order to fully utilize the information, in the embodiment of the present application, a context encoder is first used to encode a plurality of operating parameters of the boiler of the thermal power generating unit at each of the predetermined time points. In one embodiment, the context encoder is a converter-based Bert model, and it should be understood that the converter-based Bert model can perform global context-based semantic encoding on each parameter of a plurality of operating parameters of a boiler of a thermal power generating unit at the same predetermined time point to obtain a plurality of feature vectors. Here, each of the plurality of feature vectors corresponds to one of the parameters. And then, cascading the eigenvectors to obtain first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit corresponding to each preset time point.
In order to extract the correlation of the global feature vectors of the multiple parameters at each preset time point in the time sequence dimension, a convolutional neural network model is further used for carrying out explicit spatial coding on a first feature matrix formed by two-dimensional arrangement of the first feature vectors of the multiple working parameters of the boiler of the thermal power generating unit at each preset time point along the time dimension so as to extract high-dimensional implicit local features in the first feature matrix to obtain a second feature vector, namely the high-dimensional implicit local correlation features of the global feature vectors of the multiple parameters at each preset time point in the time sequence dimension.
Further, in the embodiment of the present application, a time-series encoder is used to encode the sequence of the plurality of working parameters at the plurality of predetermined time points to extract locally associated features of each of the plurality of working parameters at the sequence of the plurality of predetermined time points to obtain a third feature vector, that is, variation feature information of the same parameter of the boiler in a time-series dimension. Similarly, a feature matrix composed of a third feature vector for representing variation feature information of the same parameter in a time sequence dimension is subjected to explicit spatial coding by using a convolutional neural network model to extract implicit associated features in the feature matrix to obtain a fourth feature vector, namely, high-dimensional local implicit features of high-dimensional variation features of the same parameter of the boiler in the time sequence dimension among different parameter samples are extracted.
And then, fusing the second feature vector and the fourth feature vector to judge the boiler by fault classification or perform fault early warning. However, since the feature vector obtained by the context encoder is larger in scale than the feature vector obtained by the time-series encoder, the convolution kernel of the convolution neural network model corresponding to the second feature vector is larger than that of the convolution kernel of the convolution neural network model corresponding to the fourth feature vector. That is, due to the difference in the dimension between the samples of the context semantic feature of the parameter corresponding to the second feature vector and the time-series associated feature of the parameter corresponding to the fourth feature vector, the convex monotonicity deviation of the high-dimensional manifold is caused in the case of direct point-and-add fusion.
In particular, in the solution of the present application, a second eigenvector V is calculated 2 And a fourth feature vector V 4 The derivative information among them is super convex coefficient as weighting coefficient, namely:
Figure BDA0003813051010000081
wherein v is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Representing the second eigenvector sum V 4 A fourth feature vector is represented that represents a fourth feature vector,w represents a weighting coefficient.
By taking the derivative information hyperconvexity measurement factor as a weighting coefficient, the hyperconvexity consistency derivative expression of manifold can be carried out through the information measurement between the internal element sub-dimensions of the feature vector in a high-dimensional space, so that the manifold difference between the feature vectors can adapt to the convex monotonicity on the projection of each sub-dimension, and the classification effect of the fused feature vectors is improved.
Based on this, the application provides a fault diagnosis method, a system and an electronic device for thermal power generation equipment, wherein a context encoder is used for encoding multiple working parameters of a boiler of a thermal power generating unit at multiple preset time points, a first convolutional neural network is used for processing the working parameters to obtain a second feature vector, so as to extract high-dimensional local implicit associated features in a time sequence dimension, a time sequence encoder is used for encoding sequences of the working parameters in the multiple preset time points in the multiple working parameters, a second convolutional neural network is used for processing the working parameters to obtain a fourth feature vector, so as to extract high-dimensional local implicit features of high-dimensional change features in the time sequence dimension among different parameter samples, and then the second feature vector and the fourth feature vector are subjected to weighted fusion based on a derivative information hyperconvexity factor between the two as a weighting coefficient to obtain a classification feature vector, so that the influence on the classification accuracy caused by information loss of the two is avoided, and the fault diagnosis accuracy of the boiler is favorably improved.
The fault diagnosis method for the thermal power generation equipment can be used for judging whether a boiler of the thermal power generation equipment has a fault or not. As shown in fig. 1, in an application scenario of the present application, first, a plurality of operating parameters of a boiler (e.g., T2 in fig. 1) of a thermal power generating unit at a plurality of predetermined time points including a current time point are acquired through a sensor group (e.g., T1 in fig. 1) disposed on a thermal power generating device (e.g., P in fig. 1), and then, the acquired plurality of operating parameters of the boiler of the thermal power generating unit at the plurality of predetermined time points including the current time point are input into a server (e.g., S in fig. 1) disposed with a fault diagnosis algorithm for the thermal power generating device, wherein the server can process the acquired plurality of operating parameters of the boiler of the thermal power generating unit at the plurality of predetermined time points including the current time point by using the fault diagnosis algorithm for the thermal power generating device to output a classification result indicating whether fault early warning is generated.
Fig. 2 is a schematic diagram of a method of an embodiment of the fault diagnosis system for a thermal power plant according to the present application. As shown in fig. 2 and 3, the fault diagnosis system 100 for a thermal power plant may include: the operating parameter data acquiring unit 110 is configured to acquire a plurality of operating parameters of a boiler of the thermal power generating unit at a plurality of predetermined time points including a current time point; a context parameter encoding unit 120, configured to obtain a plurality of eigenvectors by passing through a context encoder the plurality of working parameters of the boiler of the thermal power generating unit at each predetermined time point, and cascade the plurality of eigenvectors to obtain a first eigenvector corresponding to the plurality of working parameters of the boiler of the thermal power generating unit at each predetermined time point; the parameter time sequence association coding unit 130 is configured to perform time-dimension two-dimensional arrangement on first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit at each predetermined time point to obtain a first eigenvector, and then generate a second eigenvector by using a first convolution neural network of a first convolution kernel; the parameter time sequence encoding unit 140 is configured to arrange, according to the time dimension, each working parameter of the multiple working parameters of the boiler of the thermal power generating unit at the multiple predetermined time points as an input vector, and then generate, by using a time sequence encoder including a one-dimensional convolution layer and a full connection layer, a third feature vector corresponding to each working parameter of the multiple working parameters of the boiler of the thermal power generating unit at the multiple predetermined time points; the parameter sample correlation encoding unit 150 is configured to arrange third eigenvectors of each of the plurality of working parameters of the boiler of the thermal power generating unit at the plurality of predetermined time points in a second eigenvector matrix according to a two-dimensional arrangement of sample dimensions, and then generate a fourth eigenvector by using a second convolutional neural network of a second convolutional kernel; a weighted fusion unit 160, configured to perform weighted fusion on the second feature vector and the fourth feature vector based on a super-convex metric of derived information therebetween as a weighting coefficient to obtain a classified feature vector; and an early warning unit 170, configured to pass the classified feature vectors through a classifier to obtain a classification result, where the classification result is used to indicate whether a fault early warning is generated.
Specifically, in the embodiment of the present application, the operating parameter data acquiring unit 110 is configured to acquire a plurality of operating parameters of a boiler of a thermal power generating unit at a plurality of predetermined time points including a current time point. Considering that in a thermal generator set, a boiler is one of key devices, and has the characteristics of complex system, high coupling, multiple operation parameters and the like; and because the boiler equipment works in the high-temperature, high-pressure and high-vibration environment for a long time, the frequency of fault occurrence is relatively high, so that the timely detection of faults existing in the operation of the boiler is particularly important, and the problem is essentially a classification problem. Namely, whether the thermal power generation equipment has faults or not is comprehensively and accurately classified and judged on the basis of a plurality of working parameters of the boiler of the thermal power generating unit, so that the equipment fault rate is reduced, and the thermal power generation benefit is improved.
In order to prevent data dimension explosion and allow for extraction of sufficient information for classifying and judging whether the boiler is in fault, in the embodiment of the present application, the selected operating parameters include, but are not limited to, superheated steam flow, superheated steam pressure, superheated steam temperature, reheat steam flow, reheat steam inlet pressure, reheat steam outlet pressure, reheat steam inlet temperature, reheat steam outlet temperature, boiler normal water volume, feedwater temperature, primary air volume, secondary air volume, coal consumption, flue gas volume, and flue gas corrected temperature. Specifically, each operating parameter may be acquired by a plurality of sensors deployed on the thermal power generation equipment, and the type and the installation position of each sensor are aimed at acquiring one or more of the operating parameters, which is not limited in the embodiment of the present application.
The context parameter encoding unit 120 is configured to obtain a plurality of eigenvectors by passing through a context encoder the plurality of working parameters of the boiler of the thermal power generating unit at each predetermined time point, and cascade the plurality of eigenvectors to obtain a first eigenvector corresponding to the plurality of working parameters of the boiler of the thermal power generating unit at each predetermined time point. It should be understood that during operation, the boiler is associated with parameters at the same point in time, the boiler is associated with parameters at different points in time, and the boiler is also associated with parameters at different points in time. Therefore, in order to fully utilize the above information, in the embodiment of the present application, a context encoder is first used to encode a plurality of operating parameters of the boiler of the thermal power generating unit at each of the predetermined time points to obtain a plurality of feature vectors. In an embodiment of the present application, the context encoder is a converter-based Bert model, and it should be understood that the converter-based Bert model can perform global context semantic coding on each parameter of multiple operating parameters of a boiler of a thermal power generating unit at the same predetermined time point to obtain multiple feature vectors. Here, each of the plurality of feature vectors corresponds to one of the parameters. And then, cascading the eigenvectors to obtain first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit corresponding to each preset time point.
More specifically, in the embodiment of the present application, the context parameter encoding unit 120 is further configured to: mapping each parameter in a plurality of working parameters of a boiler of the thermal power generating unit at the same preset time point into an embedded vector by using an embedded layer of the context encoder to obtain a sequence of the embedded vectors; and global semantic encoding the sequence of embedded vectors based on the upper and lower bits using a converter of the context encoder to obtain the plurality of feature vectors.
That is, the context parameter encoding unit 120 performs a word segmentation process on each of the plurality of working parameters to convert each of the plurality of working parameters into a word sequence composed of a plurality of words. Then, each word in the sequence of words is mapped to an embedded vector using an embedding layer of the context encoder to obtain a sequence of embedded vectors. Then, a global context-based semantic encoding is performed on the sequence of embedded vectors using a converter of the context encoder to obtain the plurality of feature vectors. It should be appreciated that encoding the embedded vector using a converter-based context encoder may result in a plurality of feature vectors being obtained with global textual feature association information. And finally, cascading the plurality of eigenvectors to obtain first eigenvectors of a plurality of working parameters of the boiler of the thermal power generating unit corresponding to each preset time point.
It is noted that other models, such as the bi-directional LSTM model, may be used for processing in other examples. Specifically, after word segmentation processing is performed on each parameter in the multiple working parameters to convert each parameter in the multiple working parameters into a word sequence composed of multiple words, the word sequence is input into a bidirectional LSTM model to obtain first feature vectors of the multiple working parameters of the boiler of the thermal power generating unit corresponding to each predetermined time point.
That is, in the embodiment of the present application, each of the plurality of operating parameters is mapped into a uniform embedding vector space using an embedding layer. And then, carrying out context semantic coding on the sequence of the embedded vectors by using a semantic coding model based on a converter so as to extract a global semantic implicit distribution characteristic representation of the plurality of working parameters of the boiler, and fully mining the global characteristics of the plurality of working parameters of the boiler.
The parameter time sequence association coding unit 130 is configured to perform time-dimension two-dimensional arrangement on the first eigenvectors of the multiple working parameters of the boiler of the thermal power generating unit at each predetermined time point to obtain a first eigenvector, and then generate a second eigenvector by using a first convolution neural network of a first convolution kernel. In other words, in order to extract the correlation of the global feature vectors of the multiple parameters at each predetermined time point in the time sequence dimension, a convolutional neural network model is further used to perform explicit spatial coding on a first feature matrix formed by two-dimensionally arranging the first feature vectors of the multiple working parameters of the boiler of the thermal power generating unit at each predetermined time point along the time dimension so as to extract high-dimensional implicit local features in the first feature matrix to obtain a second feature vector, that is, the high-dimensional implicit local correlation features of the global feature vectors of the multiple parameters at each predetermined time point in the time sequence dimension.
Specifically, the parameter timing correlation encoding unit 130 is further configured to: performing, using layers of the first convolutional neural network, in a layer forward pass, respectively input data: performing convolution processing on the input data by using a first convolution core to generate a convolution characteristic diagram; performing global pooling processing based on a feature matrix on the convolution feature map to generate a pooled feature vector; and performing non-linear activation based on the pooled feature vectors to generate activation feature vectors; wherein the activation feature vector output by the last layer of the first convolutional neural network is the second feature vector.
That is to say, the input data of the first layer of the first convolutional neural network is a first feature matrix obtained by performing time-dimension two-dimensional arrangement on first feature vectors of a plurality of operating parameters of a boiler of the thermal power generating unit at each predetermined time point, and each layer of the first convolutional neural network performs convolution processing based on a first convolution kernel, global pooling processing based on the feature matrix, and activation processing based on nonlinear activation on the input data in forward transmission of the layer respectively to output the activation feature vectors from the last layer of the first convolutional neural network, where the activation feature vectors output from the last layer of the first convolutional neural network are the second feature vectors. Therefore, in the embodiment of the present application, the first convolution neural network with the first convolution kernel is used to encode the multiple operating parameter feature representations of the boiler of the thermal power generating unit at the respective predetermined time points so as to extract a high-dimensional implicit association between the multiple operating parameter feature representations of the boiler at the respective predetermined time points, that is, the first convolution neural network with the first convolution kernel is used as a feature extractor to extract multiple operating parameter feature distribution representations of the boiler of the thermal power generating unit at the respective predetermined time points, that is, the operating parameter distribution features at the respective predetermined time points.
The parameter time sequence encoding unit 140 is configured to arrange, according to a time dimension, each working parameter of the multiple working parameters of the boiler of the thermal power generating unit at the multiple predetermined time points as an input vector, and then generate, by using a time sequence encoder including a one-dimensional convolution layer and a full connection layer, a third feature vector corresponding to each working parameter of the multiple working parameters of the boiler of the thermal power generating unit at the multiple predetermined time points. That is, a time sequence encoder is used for encoding the sequence of the plurality of the working parameters at the plurality of the predetermined time points so as to extract the local associated features of the working parameters at the sequence of the plurality of the predetermined time points to obtain a third feature vector, namely, the variation feature information of the same working parameter of the boiler in the time sequence dimension.
More specifically, the parameter temporal coding unit 140 is further configured to: arranging various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the preset time points into one-dimensional input vectors corresponding to the days or the times of the boiler of the thermal power generating unit according to the time dimension by taking the days or the times as a unit; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000121
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
Figure BDA0003813051010000122
represents a matrix multiplication; performing one-dimensional convolutional coding on the input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000123
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
The parameter sample association coding unit 150 is configured to arrange third eigenvectors of each of multiple working parameters of the boiler of the thermal power generating unit at the multiple predetermined time points into a second eigenvector according to a two-dimensional sample dimension, and then generate a fourth eigenvector by using a second convolutional neural network of a second convolutional kernel. Similarly, a feature matrix composed of a third feature vector for representing variation feature information of the same parameter in a time sequence dimension is subjected to explicit spatial coding by using a convolutional neural network model to extract implicit associated features in the feature matrix to obtain a fourth feature vector, namely, high-dimensional local implicit features of high-dimensional variation features of the same parameter of the boiler in the time sequence dimension among different parameter samples are extracted.
It should be noted that, because the feature vector (such as the first feature vector) obtained by the context encoder is larger in scale than the feature vector (such as the third feature vector) obtained by the time sequence encoder, the first convolution neural network with the larger convolution kernel performs feature extraction on a first feature matrix formed by two-dimensionally arranging the first feature vectors of the multiple working parameters of the boiler of the thermal power generating unit at each predetermined time point along the time dimension to obtain a second feature vector, and the second convolution neural network with the smaller convolution kernel performs explicit spatial coding on a feature matrix composed of the third feature vectors for representing variation feature information of the same parameter in the time sequence dimension to extract implicit relevant features in the feature matrix to obtain a fourth feature vector, that is, the size of the second convolution kernel of the second convolution neural network is larger than the size of the first convolution kernel of the first convolution neural network. For example, the first convolution kernel of the first convolution neural network is a 1 × 1 point convolution kernel, and the second convolution kernel of the second convolution neural network is a 3 × 3 convolution kernel, which is not limited herein.
It should be understood that the large-size convolution kernel has a relatively large receptive field, the small-size convolution kernel has a relatively small receptive field, the large-size convolution kernel is more suitable for capturing high-dimensional local implicit features of high-dimensional variation features of the same parameter of the boiler in time sequence dimensions among different parameter samples from the third feature vector obtained by the time sequence encoder, and the small-size convolution kernel is more suitable for capturing high-dimensional implicit associated features among various working parameter feature representations of the boiler at various preset time points from the first feature vector obtained by the context encoder, so that the extraction of the local features is enhanced, and the information loss is avoided.
In an optional embodiment of the present application, after obtaining the second feature vector and the fourth feature vector, the fault diagnosis system may perform fault diagnosis by fusing the second feature vector and the fourth feature vector, so as to obtain a classification result indicating whether fault early warning is generated. For example, the second feature vector and the fourth feature vector are adjusted in a linear scaling manner, so that the adjusted second feature vector and the adjusted fourth feature vector are kept at the same scale, and therefore the fusion can be performed in a position point addition manner. Of course, the fusion method of the first feature map and the second feature map may also adopt other alternative fusion methods, and is not limited herein.
However, the applicant considers that due to the scale difference between the context semantic features of the parameters corresponding to the second feature vector and the time-series associated features of the parameters corresponding to the fourth feature vector in the dimension between samples, the convex monotonicity deviation of the high-dimensional manifold is caused in the case of directly passing through the point-and-add fusion, and if the second feature vector and the fourth feature vector are adjusted in a linear scaling manner, the information loss of the second feature vector and the fourth feature vector can be caused to affect the classification precision. Therefore, in the embodiment of the present application, the weighted fusion unit 160 performs weighted fusion on the second feature vector and the fourth feature vector in consideration of using a super-convex coefficient of derived information between the second feature vector and the fourth feature vector as a weighting coefficient when performing fusion on the second feature vector and the fourth feature vector, so as to avoid information loss between the second feature vector and the fourth feature vector from affecting classification accuracy.
The weighted fusion unit 160 is configured to perform weighted fusion on the second feature vector and the fourth feature vector based on a derived information hyper-convex metric between the two as a weighting coefficient to obtain a classified feature vector. Wherein the derivative information hyper-convexity metric between the second eigenvector and the fourth eigenvector is generated based on a weighted sum of absolute values of differences between eigenvalues of corresponding locations in the second eigenvector and the fourth eigenvector.
That is, the difference in scale in the inter-sample dimension of the context semantic features of the parameters corresponding to the second feature vector and the time-dependent features of the parameters corresponding to the fourth feature vector is taken into account. Therefore, in order to solve the scale difference between the two in the high-dimensional feature space, the weighted fusion unit 160 performs weighted fusion on the second feature vector and the fourth feature vector based on a super-convex scale factor of derivative information between the two as a weighting coefficient to obtain a classified feature vector. That is to say, by using the derived information hyperconvexity metric factor as a weighting coefficient, the hyperconvexity consistency derivation representation of the manifold can be performed through the information measurement between the internal element sub-dimensions of the feature vector in the high-dimensional space, so that the manifold difference between the feature vectors can adapt to the convex monotonicity on the projection of each sub-dimension, and the classification effect of the fused feature vectors is improved. By the mode, the accuracy of judging whether the boiler fails is improved.
The weighted fusion unit 160 is further configured to: calculating a derivative information super-convexity metric between the second feature vector and the fourth feature vector as a weighting coefficient by the following formula;
wherein the formula is:
Figure BDA0003813051010000141
wherein b is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Represents the second featureSum of eigenvector V 4 Denotes a fourth feature vector and w denotes a weighting coefficient.
Then, the weighted fusion unit 160 fuses the second feature vector and the fourth feature vector by the weighting coefficient to obtain a classified feature vector. Specifically, the weighted fusion unit 160 is further configured to: calculating a weighted sum of the second feature vector and the fourth feature vector to obtain a classification feature vector with the following formula:
M s ＝wY 2 +V 4
wherein M is s For the classification feature vector, V 2 Is the second feature vector, V 4 For the fourth feature vector, "+" indicates the addition of elements at corresponding positions of the second feature vector and the fourth feature vector, and w is a weighting coefficient for controlling the balance between the second feature vector and the fourth feature vector.
The early warning unit 170 is configured to pass the classification feature vector through a classifier to obtain a classification result, where the classification result is used to indicate whether a fault early warning is generated.
That is to say, if the classification result indicates that the boiler fails or has a hidden trouble, a failure early warning is sent out to prompt the user to process in time, and if the classification result indicates that the boiler does not fail, the failure early warning is not sent out.
Specifically, the early warning unit 170 is further configured to: processing the classification feature vector using the classifier to generate a classification result with the formula: softmax { (W) n ,B n ):…:(W 1 ,B 1 )|X}，W 1 To W n As a weight matrix for each fully connected layer, B 1 To B n A bias matrix representing the layers of the fully connected layer.
To sum up, the fault diagnosis system 100 for a thermal power plant provided in the embodiment of the present application uses a context encoder to encode multiple working parameters of a boiler of a thermal power plant at multiple predetermined time points, and processes the working parameters through a first convolutional neural network to obtain a second eigenvector, so as to extract a high-dimensional local implicit correlation feature in a time sequence dimension, and uses a time sequence encoder to encode a sequence of each working parameter in the multiple working parameters at the multiple predetermined time points, and processes the working parameters through a second convolutional neural network to obtain a fourth eigenvector, so as to extract a high-dimensional local implicit feature of a high-dimensional change feature in the time sequence dimension between different parameter samples, and then performs weighted fusion on the second eigenvector and the fourth eigenvector based on a derivative information hyperconvexity factor between the two as a weighting coefficient, thereby avoiding information loss between the two from affecting classification accuracy, which is beneficial to improving accuracy of fault diagnosis of the boiler.
As described above, the fault diagnosis system 100 for a thermal power generation apparatus according to the embodiment of the present application can be implemented in various terminal apparatuses, such as a server or the like for a fault diagnosis algorithm of the thermal power generation apparatus. In one example, the fault diagnosis system 100 for a thermal power plant according to the embodiment of the present application may be integrated into a terminal device as one software module and/or hardware module. For example, the fault diagnosis system 100 for a thermal power generation apparatus may be a software module in the operating system of the terminal apparatus, or may be an application developed for the terminal apparatus; of course, the fault diagnosis system 100 for a thermal power plant may also be one of many hardware modules of the terminal device.
Alternatively, in another example, the fault diagnosis system 100 for thermal power generation equipment and the terminal equipment may also be separate equipment, and the fault diagnosis system 100 for thermal power generation equipment may be connected to the terminal equipment through a wired and/or wireless network and transmit mutual information in an agreed data format.
Fig. 4 is a schematic diagram illustrating an embodiment of the fault diagnosis method for a thermal power plant according to the present application. As shown in fig. 4, the fault diagnosis method may include:
acquiring multiple working parameters of a boiler of the thermal power generating unit at multiple preset time points including the current time point, wherein the multiple working parameters comprise superheated steam flow, superheated steam pressure, superheated steam temperature, reheated steam flow, reheated steam inlet pressure, reheated steam outlet pressure, reheated steam inlet temperature, reheated steam outlet temperature, normal water volume of the boiler, feed water temperature, primary air volume, secondary air volume, coal consumption, flue gas volume and corrected exhaust smoke temperature;
obtaining a plurality of eigenvectors by passing a plurality of working parameters of the boiler of the thermal power generating unit at each preset time point through a context encoder, and cascading the plurality of eigenvectors to obtain a first eigenvector corresponding to the plurality of working parameters of the boiler of the thermal power generating unit at each preset time point;
performing time dimension two-dimensional arrangement on first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit at each preset time point to form a first eigenvector matrix, and generating a second eigenvector by using a first convolution neural network of a first convolution kernel;
after all working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points are respectively arranged as input vectors according to time dimensions, a time sequence encoder comprising a one-dimensional convolution layer and a full connection layer is used for generating third eigenvectors of all working parameters in the multiple working parameters of the boiler of the thermal power generating unit corresponding to the multiple preset time points;
arranging third eigenvectors of each working parameter in multiple working parameters of a boiler of the thermal power generating unit at multiple preset time points in a second eigenvector matrix according to a sample dimension two-dimensional way, and generating a fourth eigenvector by using a second convolution neural network of a second convolution kernel, wherein the size of the second convolution kernel of the second convolution neural network is larger than that of the first convolution kernel of the first convolution neural network;
performing weighted fusion on the second feature vector and the fourth feature vector based on a derivative information hyper-convexity metric between the second feature vector and the fourth feature vector as a weighting coefficient to obtain a classified feature vector, wherein the derivative information hyper-convexity metric between the second feature vector and the fourth feature vector is generated based on the weighted sum of absolute values of differences between feature values of corresponding positions in the second feature vector and the fourth feature vector; and
and passing the classified feature vectors through a classifier to obtain a classification result, wherein the classification result is used for indicating whether fault early warning is generated or not.
In one possible implementation manner, in the fault diagnosis method, the passing multiple operating parameters of the boiler of the thermal power generating unit at each predetermined time point through a context encoder to obtain multiple feature vectors includes:
mapping each parameter in a plurality of working parameters of a boiler of the thermal power generating unit at the same preset time point into an embedded vector by using an embedded layer of the context encoder to obtain a sequence of the embedded vectors; and
global semantic encoding, using a converter of the context encoder, the sequence of embedded vectors based on upper and lower bits to obtain the plurality of feature vectors.
In a possible implementation manner, in the fault diagnosis method, after the time-dimension two-dimensional arrangement of the first eigenvectors of the multiple operating parameters of the boiler of the thermal power generating unit at each predetermined time point is performed as the first eigenvector, the first convolution neural network of the first convolution kernel is used to generate the second eigenvector: performing, using layers of the first convolutional neural network, in a layer forward pass, respectively input data:
performing convolution processing on the input data by using a first convolution kernel to generate a convolution characteristic map;
performing global pooling processing based on a feature matrix on the convolution feature map to generate a pooled feature vector; and
performing a non-linear activation based on the pooled feature vectors to generate activation feature vectors;
wherein the activation feature vector output by the last layer of the first convolutional neural network is the second feature vector.
In one possible implementation manner, in the fault diagnosis method, after arranging the operating parameters of the boilers of the thermal power generating units at the predetermined time points as input vectors according to the time dimension, generating a third eigenvector corresponding to each of the operating parameters of the boilers of the thermal power generating units at the predetermined time points by using a time sequence encoder including a one-dimensional convolution layer and a fully-connected layer, respectively, includes:
arranging various working parameters in the plurality of working parameters of the boiler of the thermal power generating unit at the preset time points into one-dimensional input vectors corresponding to the boiler of the thermal power generating unit in a day-by-day mode according to the time dimension;
using a full-connection layer of the time sequence encoder to perform full-connection encoding on the input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000171
Figure BDA0003813051010000172
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
Figure BDA0003813051010000173
represents a matrix multiplication;
performing one-dimensional convolutional coding on the input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
Figure BDA0003813051010000174
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
In one possible implementation manner, in the fault diagnosis method, the performing weighted fusion on the second feature vector and the fourth feature vector based on a super-convex metric of derived information therebetween as a weighting coefficient to obtain a classified feature vector includes:
calculating a derivative information super-convexity coefficient between the second feature vector and the fourth feature vector as a weighting coefficient according to the following formula;
wherein the formula is:
Figure BDA0003813051010000175
wherein b is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Representing the second eigenvector sum V 4 Denotes a fourth feature vector, and w denotes a weighting coefficient.
In one possible implementation manner, in the fault diagnosis method, the passing the classification feature vector through a classifier to obtain a classification result includes:
processing the classification feature vector using the classifier to generate a classification result with the formula: softmax { (W) n ,B n ):…:(W 1 ,B 1 )|X}，W 1 To W n As a weight matrix for each fully connected layer, B 1 To B n A bias matrix representing the layers of the fully connected layer.
It can be understood that, the fault diagnosis method provided by the embodiment shown in fig. 4 may refer to the technical solution of the embodiment of the fault diagnosis system shown in fig. 2 of the present application, and the implementation principle and the technical effect thereof may further refer to the related description in the embodiment of the fault diagnosis system.
It is to be understood that some or all of the steps or operations in the above-described embodiments are merely examples, and other operations or variations of various operations may be performed by the embodiments of the present application. Further, the various steps may be performed in a different order presented in the above-described embodiments, and it is possible that not all of the operations in the above-described embodiments are performed.
Fig. 5 is a schematic structural diagram of an embodiment of an electronic device according to the present application, and as shown in fig. 5, the electronic device may include: one or more processors; a memory; and one or more computer programs.
The electronic device may be a computer, a server, a thermal power generation device, or the like.
Wherein the one or more computer programs are stored in the memory, and the one or more computer programs comprise instructions which, when executed by the apparatus, cause the apparatus to perform the functions/steps of the fault diagnosis method for a thermal power plant provided by the method embodiment shown in fig. 4 of the present application.
As shown in fig. 5, electronic device 900 includes a processor 910 and a memory 920. Wherein, the processor 910 and the memory 920 can communicate with each other through the internal connection path to transmit control and/or data signals, the memory 920 is used for storing computer programs, and the processor 910 is used for calling and running the computer programs from the memory 920.
The memory 920 may be a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a Random Access Memory (RAM), or other types of dynamic storage devices that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, blu-ray disc, etc.), magnetic disc storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, etc.
The processor 910 and the memory 920 may be combined into a processing device, and more generally, independent components, and the processor 910 is configured to execute the program codes stored in the memory 920 to realize the functions. In particular implementations, the memory 920 may be integrated with the processor 910 or may be separate from the processor 910.
In addition, in order to make the functions of the electronic device 900 more complete, the electronic device 900 may further include one or more of a sensor 930, a power source 940, an input unit 950, and the like.
Optionally, power supply 940 is used to provide power to various devices or circuits in the electronic device.
It should be understood that the electronic device 900 shown in fig. 5 is capable of implementing the processes of the methods provided by the embodiments shown in fig. 4 of the present application. The operations and/or functions of the respective modules in the electronic device 900 are respectively for implementing the corresponding flows in the above-described method embodiments. Reference may be made specifically to the description of the embodiment of the method illustrated in fig. 4 of the present application, and a detailed description is appropriately omitted herein to avoid redundancy.
It should be understood that the processor 910 in the electronic device 900 shown in fig. 5 may be a system on chip SOC, and the processor 910 may include a Central Processing Unit (CPU), and may further include other types of processors, such as: an image Processing Unit (hereinafter, referred to as GPU), and the like.
In summary, various parts of the processors or processing units inside the processor 910 may cooperate to implement the foregoing method flows, and corresponding software programs of the various parts of the processors or processing units may be stored in the memory 920.
The application also provides an electronic device, the device includes a storage medium and a central processing unit, the storage medium may be a non-volatile storage medium, a computer executable program is stored in the storage medium, and the central processing unit is connected with the non-volatile storage medium and executes the computer executable program to implement the method provided by the embodiment shown in fig. 4 of the application.
In the above embodiments, the processors may include, for example, a CPU, a DSP, a microcontroller, or a digital Signal processor, and may further include a GPU, an embedded Neural Network Processor (NPU), and an Image Signal Processing (ISP), and the processors may further include necessary hardware accelerators or logic Processing hardware circuits, such as an ASIC, or one or more integrated circuits for controlling the execution of the program according to the technical solution of the present application. Further, the processor may have the functionality to operate one or more software programs, which may be stored in the storage medium.
Embodiments of the present application further provide a computer-readable storage medium, in which a computer program is stored, and when the computer program runs on a computer, the computer is caused to execute the method provided by the embodiment shown in fig. 4 of the present application.
Embodiments of the present application also provide a computer program product, which includes a computer program, when the computer program runs on a computer, causing the computer to execute the method provided by the embodiment shown in fig. 4 of the present application.
In the embodiments of the present application, "at least one" means one or more, "a plurality" means two or more. "and/or" describes the association relationship of the associated objects, and means that there may be three relationships, for example, a and/or B, and may mean that a exists alone, a and B exist simultaneously, and B exists alone. Wherein A and B can be singular or plural. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c may represent: a, b, c, a and b, a and c, b and c or a and b and c, wherein a, b and c can be single or multiple.
Those of ordinary skill in the art will appreciate that the various elements and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, any function, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U disk, a portable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other media capable of storing program codes.
The above description is only for the specific embodiments of the present application, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present disclosure, and all the changes or substitutions should be covered by the protection scope of the present application. The protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. A fault diagnosis system for a thermal power plant, characterized by comprising: the system comprises a working parameter data acquisition unit, a data processing unit and a data processing unit, wherein the working parameter data acquisition unit is used for acquiring a plurality of working parameters of a boiler of the thermal power generating unit at a plurality of preset time points including a current time point, and the plurality of working parameters comprise superheated steam flow, superheated steam pressure, superheated steam temperature, reheated steam flow, reheated steam inlet pressure, reheated steam outlet pressure, reheated steam inlet temperature, reheated steam outlet temperature, normal boiler water volume, feed water temperature, primary air volume, secondary air volume, coal consumption, flue gas volume and corrected exhaust smoke temperature; the context parameter coding unit is used for enabling a plurality of working parameters of the boiler of the thermal power generating unit at each preset time point to pass through a context coder to obtain a plurality of eigenvectors, and cascading the plurality of eigenvectors to obtain first eigenvectors of the plurality of working parameters of the boiler of the thermal power generating unit corresponding to each preset time point; the parameter time sequence correlation coding unit is used for performing time dimension two-dimensional arrangement on first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit at each preset time point to form a first characteristic matrix and then generating second eigenvectors by using a first convolution neural network of a first convolution kernel; the parameter time sequence coding unit is used for respectively arranging various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points into input vectors according to time dimensions, and then generating third feature vectors of the various working parameters in the multiple working parameters of the boiler of the thermal power generating unit corresponding to the multiple preset time points through a time sequence coder comprising a one-dimensional convolution layer and a full connection layer; the parameter sample correlation coding unit is used for two-dimensionally arranging third eigenvectors of each working parameter in a plurality of working parameters of a boiler of the thermal power generating unit at a plurality of preset time points into a second eigenvector matrix according to a sample dimension and then generating a fourth eigenvector by using a second convolution neural network of a second convolution kernel, wherein the size of the second convolution kernel of the second convolution neural network is larger than that of the first convolution kernel of the first convolution neural network; a weighted fusion unit, configured to perform weighted fusion on the second feature vector and the fourth feature vector based on a derived information hyper-convex metric between the second feature vector and the fourth feature vector as a weighting coefficient to obtain a classified feature vector, where the derived information hyper-convex metric between the second feature vector and the fourth feature vector is generated based on a weighted sum of absolute values of differences between feature values of corresponding positions in the second feature vector and the fourth feature vector; and the early warning unit is used for enabling the classified characteristic vectors to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether fault early warning is generated or not.
2. The fault diagnosis system for a thermal power plant according to claim 1, wherein the context parameter encoding unit is further configured to: mapping each parameter in a plurality of working parameters of a boiler of the thermal power generating unit at the same preset time point into an embedded vector by using an embedded layer of the context encoder to obtain a sequence of the embedded vectors; and global semantic encoding the sequence of embedded vectors based on the upper and lower bits using a converter of the context encoder to obtain the plurality of feature vectors.
3. The fault diagnosis system for a thermal power plant according to claim 2, wherein the parameter timing-related encoding unit is further configured to: performing, using layers of the first convolutional neural network, in a layer forward pass, respectively input data: performing convolution processing on the input data by using a first convolution kernel to generate a convolution characteristic map; performing global pooling processing based on a feature matrix on the convolution feature map to generate a pooled feature vector; and performing non-linear activation based on the pooled feature vectors to generate activation feature vectors; wherein the activation feature vector output by the last layer of the first convolutional neural network is the second feature vector.
4. The fault diagnosis system for a thermal power plant according to claim 3, wherein the parameter timing encoding unit is further configured to: arranging various working parameters in the plurality of working parameters of the boiler of the thermal power generating unit at the preset time points into one-dimensional input vectors corresponding to the boiler of the thermal power generating unit in a day-by-day mode according to the time dimension; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
Figure FDA0003813050000000021
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
Figure FDA0003813050000000022
represents a matrix multiplication; performing one-dimensional convolutional coding on the input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
Figure FDA0003813050000000023
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
5. The fault diagnosis system for a thermal power plant according to claim 4, wherein the weighted fusion unit is further configured to: calculating a derivative information super-convexity coefficient between the second feature vector and the fourth feature vector as a weighting coefficient according to the following formula; wherein the formula is:
Figure FDA0003813050000000024
wherein v is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Representing the second eigenvector sum V 4 Denotes a fourth feature vector and w denotes a weighting coefficient.
6. The fault diagnosis system for thermal power generation equipment according to claim 5, wherein the early warning unit is further configured to: processing the classification feature vector using the classifier to generate a classification result with the formula: softmax { (W) n ,B n ):…:(W 1 ,B 1 ) | X }, wherein W 1 To W n For each layer to be fully connected with the layer rightHeavy matrix, B 1 To B n A bias matrix representing the layers of the fully connected layer.
7. A fault diagnosis method for a thermal power generation apparatus, characterized by comprising: acquiring multiple working parameters of a boiler of the thermal power generating unit at multiple preset time points including the current time point, wherein the multiple working parameters comprise superheated steam flow, superheated steam pressure, superheated steam temperature, reheated steam flow, reheated steam inlet pressure, reheated steam outlet pressure, reheated steam inlet temperature, reheated steam outlet temperature, normal water volume of the boiler, feed water temperature, primary air volume, secondary air volume, coal consumption, flue gas volume and corrected exhaust smoke temperature; obtaining a plurality of eigenvectors by passing a plurality of working parameters of the boiler of the thermal power generating unit at each preset time point through a context encoder, and cascading the plurality of eigenvectors to obtain a first eigenvector corresponding to the plurality of working parameters of the boiler of the thermal power generating unit at each preset time point; performing time dimension two-dimensional arrangement on first eigenvectors of multiple working parameters of the boiler of the thermal power generating unit at each preset time point to form a first eigenvector matrix, and generating a second eigenvector by using a first convolution neural network of a first convolution kernel; after all working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points are respectively arranged as input vectors according to the time dimension, a time sequence encoder comprising a one-dimensional convolution layer and a full connection layer is used for generating third eigenvectors corresponding to all working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the multiple preset time points; arranging third eigenvectors of each working parameter in multiple working parameters of a boiler of the thermal power generating unit at multiple preset time points in a second eigenvector matrix according to a sample dimension two-dimensional way, and generating a fourth eigenvector by using a second convolution neural network of a second convolution kernel, wherein the size of the second convolution kernel of the second convolution neural network is larger than that of the first convolution kernel of the first convolution neural network; performing weighted fusion on the second feature vector and the fourth feature vector based on a derivative information super-convexity metric factor between the second feature vector and the fourth feature vector as a weighting coefficient to obtain a classified feature vector, wherein the derivative information super-convexity metric factor between the second feature vector and the fourth feature vector is generated based on the weighted sum of absolute values of differences between feature values of corresponding positions in the second feature vector and the fourth feature vector; and enabling the classified characteristic vectors to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether fault early warning is generated or not.
8. The fault diagnosis method for a thermal power plant according to claim 7, wherein the passing the plurality of operating parameters of the boiler of the thermal power generating unit at the respective predetermined time points through a context encoder to obtain a plurality of eigenvectors comprises: mapping each parameter in a plurality of working parameters of a boiler of the thermal power generating unit at the same preset time point into an embedded vector by using an embedded layer of the context encoder to obtain a sequence of the embedded vectors; and global semantic encoding the sequence of embedded vectors based on the upper and lower bits using a converter of the context encoder to obtain the plurality of feature vectors.
9. The fault diagnosis method for a thermal power plant according to claim 8, wherein the generating a third eigenvector corresponding to each of the plurality of operating parameters of the boiler of the thermal power plant at the plurality of predetermined time points by a time-sequential encoder including a one-dimensional convolution layer and a fully-connected layer after arranging each of the plurality of operating parameters of the boiler of the thermal power plant at the plurality of predetermined time points as an input vector according to the time dimension comprises: arranging various working parameters in the multiple working parameters of the boiler of the thermal power generating unit at the preset time points into one-dimensional input vectors corresponding to the boiler of the thermal power generating unit on a daily basis according to the time dimension; using a full-connection layer of the time sequence encoder to perform full-connection encoding on the input vector by using the following formula to extract high-dimensional implicit features of feature values of all positions in the input vector, wherein the formula is as follows:
Figure FDA0003813050000000031
Figure FDA0003813050000000041
where X is the input vector, Y is the output vector, W is the weight matrix, B is the offset vector,
Figure FDA0003813050000000042
represents a matrix multiplication; performing one-dimensional convolutional encoding on the input vector by using a one-dimensional convolutional layer of the time sequence encoder according to the following formula so as to extract high-dimensional implicit correlation characteristics among characteristic values of all positions in the input vector, wherein the formula is as follows:
Figure FDA0003813050000000043
wherein, a is the width of the convolution kernel in the x direction, F is the parameter vector of the convolution kernel, G is the local vector matrix operated with the convolution kernel function, and w is the size of the convolution kernel.
10. The fault diagnosis method for a thermal power plant according to claim 9, wherein the performing weighted fusion of the second feature vector and the fourth feature vector based on a derivative information hyper-convexity coefficient therebetween as a weighting coefficient to obtain a classification feature vector includes: calculating a derivative information super-convexity metric between the second feature vector and the fourth feature vector as a weighting coefficient by the following formula; wherein the formula is:
Figure FDA0003813050000000044
wherein v is 2i ∈V 2 And v is 4i ∈V 4 ，V 2 Representing the second eigenvector sum V 4 Denotes a fourth feature vector, and w denotes a weighting coefficient.
CN202211017981.1A 2022-08-24 2022-08-24 Fault diagnosis method and system for thermal power generation equipment Pending CN115392370A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211017981.1A CN115392370A (en) 2022-08-24 2022-08-24 Fault diagnosis method and system for thermal power generation equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211017981.1A CN115392370A (en) 2022-08-24 2022-08-24 Fault diagnosis method and system for thermal power generation equipment

Publications (1)

Publication Number Publication Date
CN115392370A true CN115392370A (en) 2022-11-25

Family

ID=84119859

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211017981.1A Pending CN115392370A (en) 2022-08-24 2022-08-24 Fault diagnosis method and system for thermal power generation equipment

Country Status (1)

Country Link
CN (1) CN115392370A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115797708A (en) * 2023-02-06 2023-03-14 南京博纳威电子科技有限公司 Power transmission and distribution synchronous data acquisition method
CN115796173A (en) * 2023-02-20 2023-03-14 杭银消费金融股份有限公司 Data processing method and system for supervision submission requirements
CN116130721A (en) * 2023-04-11 2023-05-16 杭州鄂达精密机电科技有限公司 Status diagnostic system and method for hydrogen fuel cell
CN117124910A (en) * 2023-09-20 2023-11-28 漳州建源电力工程有限公司 Smart city charging pile node fault alarm system and method

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115797708A (en) * 2023-02-06 2023-03-14 南京博纳威电子科技有限公司 Power transmission and distribution synchronous data acquisition method
CN115796173A (en) * 2023-02-20 2023-03-14 杭银消费金融股份有限公司 Data processing method and system for supervision submission requirements
CN116130721A (en) * 2023-04-11 2023-05-16 杭州鄂达精密机电科技有限公司 Status diagnostic system and method for hydrogen fuel cell
CN117124910A (en) * 2023-09-20 2023-11-28 漳州建源电力工程有限公司 Smart city charging pile node fault alarm system and method

Similar Documents

Publication Publication Date Title
CN115392370A (en) Fault diagnosis method and system for thermal power generation equipment
Jin et al. Anomaly detection in nuclear power plants via symbolic dynamic filtering
CN116627708A (en) Storage fault analysis system and method thereof
CN116092701B (en) Control system and method based on health data analysis management platform
CN115859437A (en) Jacket underwater stress detection system based on distributed optical fiber sensing system
CN116086790B (en) Performance detection method and system for high-pressure valve of hydrogen fuel cell
CN116336400B (en) Baseline detection method for oil and gas gathering and transportation pipeline
CN116405326B (en) Information security management method and system based on block chain
Pu et al. Sparse Bayesian learning for gas path diagnostics
CN116680557B (en) Real-time monitoring system and method for coal bed gas drilling engineering
CN115324843A (en) Wind generating set fault diagnosis system and method based on monitoring data
CN104361238A (en) Fault sensitive characteristic extraction method based on information entropy improved PCA (Principal Component Analysis)
CN116247824A (en) Control method and system for power equipment
CN117269644A (en) Line fault monitoring system and method for current transformer
CN115526202A (en) Offshore wind turbine fault diagnosis system based on data driving and diagnosis method thereof
Rasaienia et al. Feature-based fault detection of industrial gas turbines using neural networks
Li et al. Model-based monitoring and fault diagnosis of fossil power plant process units using group method of data handling
Peng et al. Fault detection and isolation for self powered neutron detectors based on Principal Component Analysis
CN117231590A (en) Fault prediction system and method for hydraulic system
Uren et al. An integrated approach to sensor FDI and signal reconstruction in HTGRs–Part II: Case studies
CN117950314A (en) Shutdown control system and method for wind generating set
CN116385733A (en) High-precision positioning system and method for bucket wheel machine
Chaibakhsh Modelling and long-term simulation of a heat recovery steam generator
Lee et al. Event diagnosis method for a nuclear power plant using meta-learning
Du Rand et al. Fault diagnosis of generation IV nuclear HTGR components–Part II: The area error enthalpy–entropy graph approach

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination