Boiler fault diagnosis model generation method and device, boiler fault diagnosis method and device, equipment and medium
By establishing and training the target fault diagnosis model in the boiler, the problems of boiler fault diagnosis accuracy and low efficiency are solved, and more accurate and efficient fault diagnosis is achieved.
Patent Information
- Application Number
- CN202411761547.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-03-28
AI Technical Summary
In the prior art, some boiler failures are difficult to be determined directly based on the acquired operating data, resulting in low accuracy and diagnostic efficiency of boiler fault diagnosis.
By obtaining the sample data set of the boiler under different operating conditions, an initial boiler fault diagnosis model is established, and through transfer learning and multiple trainings, a target boiler fault diagnosis model can be generated. This model can directly diagnose faults based on the operating data under different operating conditions.
It improves the accuracy and diagnostic efficiency of boiler fault diagnosis, and can effectively solve the problem that it is difficult to directly determine boiler faults.
Smart Images
Figure CN119719892A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of fault diagnosis, and in particular to a method and device for generating a boiler fault diagnosis model, a method and device for generating a boiler fault diagnosis model, an electronic device, and a machine-readable storage medium. Background Art
[0002] The boiler system is a key equipment in a thermal power plant, but it is also prone to failure. As a pressure-bearing power equipment, the boiler is subject to long-term erosion and wear from flue gas, flames, wind pressure, dust and scale during operation, and is prone to various failures. If the failure in the boiler system is not discovered and handled in time, it may lead to more serious problems and cause unplanned downtime, which not only affects the stability of power generation, but also increases operating costs and energy consumption.
[0003] In the related technology, boiler fault diagnosis mainly relies on the operator's observation and experience judgment of some parameters during the operation of the equipment, which easily leads to low diagnostic accuracy. In addition, some fault causes are not easy to find intuitively. The diagnosis process needs to judge the fault location from a large amount of measurement data, state variables and alarm information, which further reduces the accuracy of the diagnosis and also reduces the diagnostic efficiency. Summary of the invention
[0004] The purpose of the embodiments of the present application is to provide a boiler fault diagnosis model generation method and device, a boiler fault diagnosis method and device, an electronic device and a machine-readable storage medium, so as to solve the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0005] In order to achieve the above-mentioned object, the first aspect of the present application provides a method for generating a boiler fault diagnosis model, the method comprising:
[0006] Acquire a first sample data set of the boiler under a first operating condition and a second sample data set of the boiler under a second operating condition different from the first operating condition; wherein each first sample data subset in the first sample data set includes first operating data and a corresponding fault category label, and the second sample data set includes a labeled sample data subset and an unlabeled sample data subset, each labeled sample data subset includes second operating data and a corresponding fault category label, and each unlabeled sample data subset includes third operating data;
[0007] Establishing an initial boiler fault diagnosis model, and training the initial boiler fault diagnosis model using the first sample data set to obtain a first boiler fault diagnosis model;
[0008] Input each of the first operating data and the third operating data into the first boiler fault diagnosis model, respectively, to obtain a first deep feature data set and a second deep feature data set corresponding to each target neural network layer in the first boiler fault diagnosis model; wherein the first deep feature data set includes the deep feature data corresponding to each of the first operating data, and the second deep feature data set includes the deep feature data corresponding to each of the third operating data;
[0009] Based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, transfer learning is performed on the first boiler fault diagnosis model to obtain a second boiler fault diagnosis model;
[0010] The second boiler fault diagnosis model is trained by using each labeled sample data subset to obtain a target boiler fault diagnosis model.
[0011] In an embodiment of the present application, the initial boiler fault diagnosis model adopts a multi-scale convolutional neural network structure; the initial boiler fault diagnosis model includes multiple layers of sequentially connected neural network structures, each neural network structure includes a convolutional layer and a maximum pooling layer connected in sequence, and a fully connected layer and an output layer are sequentially connected after the last neural network structure in the initial boiler fault diagnosis model, and the convolution layers in the first two neural network structures in the initial boiler fault diagnosis model both use large convolution kernels.
[0012] In an embodiment of the present application, based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, transfer learning is performed on the first boiler fault diagnosis model to obtain a second boiler fault diagnosis model, including:
[0013] Establishing a target migration model; wherein the target migration model has the same structure as the initial boiler fault diagnosis model;
[0014] For each target neural network layer, determining a feature distribution distance corresponding to the target neural network layer according to a first deep feature data set and a second deep feature data set corresponding to the target neural network layer;
[0015] For each target neural network layer, if the feature distribution distance corresponding to the target neural network layer meets the preset migration conditions, the parameters of the target neural network layer are migrated to the neural network layer corresponding to the target neural network layer in the target migration model.
[0016] In an embodiment of the present application, for each target neural network layer, determining the feature distribution distance corresponding to the target neural network layer according to the first deep feature data set and the second deep feature data set corresponding to the target neural network layer includes:
[0017] Determine the edge probability distribution distance and conditional probability distribution distance corresponding to the target neural network layer according to the first deep feature data set and the second deep feature data set corresponding to the target neural network layer;
[0018] According to the edge probability distribution distance and conditional probability distribution distance corresponding to the target neural network layer, the feature distribution distance corresponding to the target neural network layer is determined.
[0019] In an embodiment of the present application, the method further includes:
[0020] Generate a test sample data set according to the first sample data subset and the labeled sample data subset; wherein each test sample data subset in the test sample data set includes test operation data and its corresponding fault category label;
[0021] Inputting each test operation data into the target boiler fault diagnosis model respectively to obtain a prediction label corresponding to each test operation data;
[0022] Determining the prediction accuracy of the target boiler fault diagnosis model according to the fault category label and the prediction label corresponding to each test operation data;
[0023] If the prediction accuracy does not meet the preset prediction conditions, return to the step of establishing an initial boiler fault diagnosis model, and train the initial boiler fault diagnosis model through the first sample data set to obtain a first boiler fault diagnosis model, until the prediction accuracy of the target boiler fault diagnosis model meets the preset prediction conditions.
[0024] A second aspect of the present application provides a boiler fault diagnosis method, the method comprising:
[0025] Obtaining the operation data to be predicted for the boiler;
[0026] The operation data to be predicted is input into the target boiler fault diagnosis model obtained by the method described in the first aspect above, and a fault category label corresponding to the operation data to be predicted is obtained.
[0027] A third aspect of the present application provides a boiler fault diagnosis model generation device, the device comprising:
[0028] A sample data acquisition module, used to acquire a first sample data set of the boiler under a first operating condition and a second sample data set of the boiler under a second operating condition different from the first operating condition; wherein each first sample data subset in the first sample data set includes first operating data and a corresponding fault category label, and the second sample data set includes a labeled sample data subset and an unlabeled sample data subset, each labeled sample data subset includes second operating data and a corresponding fault category label, and each unlabeled sample data subset includes third operating data;
[0029] A first model training module is used to establish an initial boiler fault diagnosis model, and train the initial boiler fault diagnosis model through the first sample data set to obtain a first boiler fault diagnosis model;
[0030] A deep feature data determination module, used to input each of the first operating data and the third operating data into the first boiler fault diagnosis model, respectively, to obtain a first deep feature data set and a second deep feature data set corresponding to each target neural network layer in the first boiler fault diagnosis model; wherein the first deep feature data set includes the deep feature data corresponding to each of the first operating data, and the second deep feature data set includes the deep feature data corresponding to each of the third operating data;
[0031] A transfer learning module, configured to perform transfer learning on the first boiler fault diagnosis model based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, so as to obtain a second boiler fault diagnosis model;
[0032] The second model training module is used to train the second boiler fault diagnosis model through each labeled sample data subset to obtain a target boiler fault diagnosis model.
[0033] A fourth aspect of the present application provides a boiler fault diagnosis device, the device comprising:
[0034] A module for acquiring data to be predicted, used for acquiring operation data to be predicted for the boiler;
[0035] The fault prediction module is used to input the operation data to be predicted into the target boiler fault diagnosis model obtained by the method described in the first aspect above, and obtain the fault category label corresponding to the operation data to be predicted.
[0036] The fifth aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the boiler fault diagnosis model generation method described in the first aspect is implemented.
[0037] A sixth aspect of the present application provides a machine-readable storage medium having instructions stored thereon, which, when executed by a processor, configures the processor to execute the boiler fault diagnosis model generation method described in the first aspect above.
[0038] The embodiment of the present application performs preliminary training on the constructed model based on the operating data under the first operating condition with complete corresponding fault category labels, then performs transfer learning on the model obtained by the preliminary training based on the operating data without corresponding fault category labels under each second operating condition, and then re-trains the model obtained by transfer learning based on the operating data with corresponding fault category labels under each second operating condition, and finally obtains a target boiler fault diagnosis model applicable to the first operating condition and the second operating condition. The model can directly perform fault diagnosis based on the operating data under different operating conditions, thereby improving the accuracy and efficiency of boiler fault diagnosis, and effectively solving the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0039] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:
[0041] Figure 1 A schematic diagram of a first process flow of a boiler fault diagnosis model generation method according to an embodiment of the present application is shown;
[0042] Figure 2 A second flow chart of the boiler fault diagnosis model generation method according to an embodiment of the present application is schematically shown;
[0043] Figure 3 The flowchart of the boiler fault diagnosis method according to the embodiment of the present application is schematically shown;
[0044] Figure 4 A structural block diagram of a boiler fault diagnosis model generating device according to an embodiment of the present application is schematically shown;
[0045] Figure 5 A schematic diagram of a boiler fault diagnosis device according to an embodiment of the present application is shown;
[0046] Figure 6 The internal structure diagram of the computer device according to the embodiment of the present application is schematically shown.
[0047] Description of Reference Numerals
[0048] A01-processor; A02-network interface; A03-internal memory; A04-display screen; A05-input device; A06-non-volatile storage medium; B01-operating system; B02-computer program. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0050] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), such directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.
[0051] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.
[0052] In view of the fact that some boiler faults in the related art are difficult to determine directly based on the acquired operating data, which reduces the accuracy and efficiency of boiler fault diagnosis, the present application provides a boiler fault diagnosis model generation and diagnosis method, device, equipment and medium. The boiler fault diagnosis model generation and diagnosis method, device, equipment and medium provided in the present application are described in detail below in conjunction with the accompanying drawings through specific examples and implementation methods.
[0053] like Figure 1 and Figure 2 As shown, in one embodiment of the present application, a method for generating a boiler fault diagnosis model is provided, and the method may include the following steps.
[0054] Step A200: acquiring a first sample data set of the boiler under a first operating condition and a second sample data set of the boiler under a second operating condition different from the first operating condition.
[0055] Among them, each first sample data subset in the first sample data set includes first operating data and its corresponding fault category label, the second sample data set includes a labeled sample data subset and an unlabeled sample data subset, each labeled sample data subset includes second operating data and its corresponding fault category label, and each unlabeled sample data subset includes third operating data.
[0056] Specifically, the boiler has a variety of operating conditions. Under different operating conditions, the boiler may have different operating parameter values and performance performances. Adjusting and controlling the operating parameters of the boiler according to the specific operating conditions can ensure efficient, stable and safe operation of the boiler. In the embodiment of the present application, the first operating condition and the second operating condition are both boiler operating conditions known to those skilled in the art. Under specific operating conditions, the fault category label of the boiler includes normal operating state, thermal deviation, scaling in the tube, abnormal cooling water, secondary combustion in the flue, and ash slagging on the heating surface.
[0057] In a specific example, the number of second operating conditions is 5, that is, step A200 includes sample data subsets under 6 operating conditions (the first sample data subset, the labeled sample data subset and the unlabeled sample data subset are all sample data subsets), and the one-dimensional operating data contained in each sample data subset (the first operating data, the second operating data and the third operating data are all one-dimensional operating data) are all sets of historical values of multiple operating parameters, wherein the multiple operating parameters include main steam pressure, main steam temperature, furnace pressure, actual load, left furnace flue gas temperature, burnout wind swing angle, right furnace flue gas temperature, tail flue temperature, cooling water valve opening, tail flue oxygen content, right high steam pressure, right high steam temperature, left high steam pressure, left high steam temperature, secondary cooling water inlet temperature, ratio of feed water flow to main steam flow and secondary cooling water outlet temperature.
[0058] Step A300: establish an initial boiler fault diagnosis model, and train the initial boiler fault diagnosis model using the first sample data set to obtain a first boiler fault diagnosis model.
[0059] In the embodiment of the present application, the initial boiler fault diagnosis model adopts a multi-scale convolutional neural network structure. Specifically, the initial boiler fault diagnosis model includes multiple layers of sequentially connected neural network structures, each of which includes sequentially connected convolutional layers and maximum pooling layers, and the last neural network structure in the initial boiler fault diagnosis model is sequentially connected with a fully connected layer and an output layer.
[0060] In order to adapt to the input characteristics of one-dimensional operation data, the convolution layer and the maximum pooling layer in the initial boiler fault diagnosis model both adopt a one-dimensional structure. Since large convolution kernels can better suppress high-frequency noise, the convolution layers in the first two neural network structures in the initial boiler fault diagnosis model (i.e., the first and second convolution layers in the model structure) both adopt large convolution kernels, such as 64 and 32. This setting can also extract short-term features, obtain more data and provide more information for the deep layer of the network.
[0061] It can be understood that the convolution layer, the maximum pooling layer, the fully connected layer, and the output layer are all neural network layers of the initial boiler fault diagnosis model. In the embodiment of the present application, the convolution layer and the fully connected layer are the target neural network layers in the initial boiler fault diagnosis model. Each target neural network layer in the initial boiler fault diagnosis model has its own initial parameters (such as weights and biases), and the model has set hyperparameters, such as the number of samples processed per batch, the optimizer type, and the convolution kernel size and step size of each neural network layer.
[0062] In addition, in the embodiment of the present application, the activation function of each convolution layer in the initial boiler fault diagnosis model adopts ReLU, and the activation function of the output layer adopts Softmax. In the embodiment of the present application, batch normalization is performed after each convolution operation to increase the generalization ability of the model.
[0063] In a specific example, the initial boiler fault diagnosis model adopts a one-dimensional convolutional neural network (CNN) structure, which includes 5 neural network structures. The hyperparameter settings of each neural network layer in the initial boiler fault diagnosis model are shown in Table 1 below.
[0064] Table 1 - Hyperparameter information of neural network layers
[0065]
[0066]
[0067] In this example, the optimizer type uses the Adam optimizer, which can automatically adjust the learning rate during the learning process to avoid human interference. The number of samples processed in each batch is selected as 32, that is, each time data is input, the model will process 32 samples at the same time.
[0068] In the process of training the initial boiler fault diagnosis model, the convolution layer in the initial boiler fault diagnosis model performs local convolution operations on its input by using multiple convolution kernels. The convolution kernel slides on the input according to the set step size, and gradually calculates and generates output features. Each convolution kernel generates a feature, and the number of convolution kernels determines the depth of the output feature. Among them, the convolution operation can be expressed as:
[0069]
[0070] Among them, y l represents the output of the lth convolutional layer, f(·) represents the activation function, c l-1 represents the cth channel of the l-1th convolutional layer, represents the weight matrix of the lth convolutional layer, * represents the convolution operation, represents the output of the ith channel of the l-1th convolutional layer, Represents the bias term.
[0071] By setting a maximum pooling layer in the initial boiler fault diagnosis model, model parameters and feature dimensions can be reduced. The maximum pooling layer performs a maximum pooling operation on the local area (perception domain) in its input, selects the maximum value from each local area, and generates output features. The maximum pooling operation can be expressed as:
[0072]
[0073] in, represents the output of the i-th channel of the l+1-th convolutional layer, S represents the size of the pooling kernel, j represents the step size, Represents the output value of the tth neuron in the ith channel of the lth convolutional layer.
[0074] Based on the hyperparameter settings of the neural network layer in Table 1, the initial boiler fault diagnosis model can directly take the one-dimensional operation data as input, and then perform forward propagation, extract features layer by layer through alternating convolutional layers and maximum pooling layers, globally integrate the extracted features through the fully connected layer, and use the Softmax classifier for classification.
[0075] The initial boiler fault diagnosis model uses cross entropy as the loss function, measures the difference between the probability distribution output by the Softmax function and the true label (i.e., the fault category label) through cross entropy, and minimizes the cross entropy loss to train the model, thereby improving the accuracy of fault classification. The parameters of the target neural network layer (such as weights and biases) are updated through the back propagation algorithm to minimize the loss function and optimize the model performance. The calculation formula of cross entropy can be expressed as:
[0076]
[0077] Wherein, p(x) represents the fault category label of the one-dimensional operating data x input into the initial boiler fault diagnosis model, and q(x) represents the Softmax output classification result of the one-dimensional operating data x.
[0078] Of course, it is understandable that the initial boiler fault diagnosis model may also adopt other existing neural network models, and the present application does not specifically limit the structure of the initial boiler fault diagnosis model.
[0079] Step A400, input each of the first operating data and the third operating data into the first boiler fault diagnosis model respectively, and obtain a first deep feature data set and a second deep feature data set corresponding to each target neural network layer in the first boiler fault diagnosis model.
[0080] In the embodiment of the present application, the target neural network layer refers to a neural network layer whose parameter values can be adjusted by model training. It can be understood that since the first boiler fault diagnosis model is obtained by training based on the initial boiler fault diagnosis model, the first boiler fault diagnosis model has the same structure as the initial boiler fault diagnosis model, and the target neural network layer in the first boiler fault diagnosis model includes a convolutional layer and a fully connected layer.
[0081] In the embodiment of the present application, the first deep feature data set includes the deep feature data corresponding to each first operation data, and the second deep feature data set includes the deep feature data corresponding to each third operation data. It should be understood that for each one-dimensional operation data, after the one-dimensional operation data is input into the first boiler fault diagnosis model, the output of each target neural network layer is the deep feature data corresponding to the one-dimensional operation data.
[0082] Step A500, based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, transfer learning is performed on the first boiler fault diagnosis model to obtain a second boiler fault diagnosis model.
[0083] In the embodiment of the present application, step A500 may include the following steps:
[0084] Step A510, establishing a target migration model.
[0085] The target migration model has the same structure as the initial boiler fault diagnosis model.
[0086] Step A520, for each target neural network layer, determine the feature distribution distance corresponding to the target neural network layer based on the first deep feature data set and the second deep feature data set corresponding to the target neural network layer.
[0087] Specifically, the step A520 includes:
[0088] Step A521, based on the first deep feature data set and the second deep feature data set corresponding to the target neural network layer, determine the edge probability distribution distance and the conditional probability distribution distance corresponding to the target neural network layer.
[0089] In a specific implementation, step A521 uses the Maximum Mean Discrepancy (MMD) method to measure the distribution distance of two different but related data sets (i.e., the first deep feature data set and the second deep feature data set) in the Reproducing Kernel Hilbert Space (RKHS), i.e., the marginal probability distribution distance and the conditional probability distribution distance.
[0090] Step A522, determine the feature distribution distance corresponding to the target neural network layer based on the edge probability distribution distance and conditional probability distribution distance corresponding to the target neural network layer.
[0091] In a specific embodiment, step A522 determines the sum of the edge probability distribution distance and the conditional probability distribution distance corresponding to the target neural network layer as the feature distribution distance corresponding to the target neural network layer, and uses the feature distribution distance as the migration judgment indicator data for subsequently determining whether to perform parameter migration.
[0092] Step A530, for each target neural network layer, if the feature distribution distance corresponding to the target neural network layer meets the preset migration conditions, the parameters of the target neural network layer are migrated to the neural network layer corresponding to the target neural network layer in the target migration model.
[0093] In a specific embodiment, the preset migration condition is set as the feature distribution distance being less than or equal to the preset distance threshold. The preset distance threshold can be specifically set according to the actual situation. That is to say, for each target neural network layer, if the feature distribution distance corresponding to the target neural network layer is less than or equal to the preset distance threshold, the parameters of the target neural network layer are migrated to the neural network layer corresponding to the target neural network layer in the target migration model. On the contrary, if the feature distribution distance corresponding to the target neural network layer is greater than the preset distance threshold, the parameters of the target neural network layer are not migrated, so that the parameters of the neural network layer corresponding to the target neural network layer in the target migration model remain the initial parameters.
[0094] It can be understood that in the parameter migration process of the embodiment of the present application, by fully and comprehensively considering the distribution differences of the values of different operating parameters under the first operating condition and the second operating condition, the migration model thus obtained (i.e., the second boiler fault diagnosis model) has good feature extraction capabilities for the one-dimensional operating data under the first operating condition and each second operating condition.
[0095] Step A600: training the second boiler fault diagnosis model using each labeled sample data subset to obtain a target boiler fault diagnosis model.
[0096] Specifically, the trained target boiler fault diagnosis model can be applied to perform fault diagnosis on the operating data under the first operating condition and each second operating condition.
[0097] Optionally, the method may further include the following steps.
[0098] Step A710: Generate a test sample data set according to the first sample data subset and the labeled sample data subset.
[0099] Wherein, each test sample data subset in the test sample data set includes test running data and its corresponding fault category label.
[0100] In a specific implementation, part of the first sample data subset and the labeled sample data subset are randomly selected as the test sample data subset, and the selected first running data in the first sample data subset and the second running data in the labeled sample data subset are the test running data.
[0101] Step A720: Input each test operation data into the target boiler fault diagnosis model to obtain a prediction label corresponding to each test operation data.
[0102] Step A730, determining the prediction accuracy of the target boiler fault diagnosis model according to the fault category label and the prediction label corresponding to each test operation data.
[0103] Specifically, for each test operation data, if the corresponding fault category label and prediction label are not completely the same, it is considered that the target boiler fault diagnosis model has made a prediction error.
[0104] Step A740: If the prediction accuracy does not meet the preset prediction condition, return to step A300 until the prediction accuracy of the target boiler fault diagnosis model meets the preset prediction condition.
[0105] In a specific implementation, the preset prediction condition is set as a prediction accuracy greater than or equal to a preset accuracy threshold, wherein the preset accuracy threshold can be specifically set according to actual conditions.
[0106] The embodiment of the present application calculates the prediction accuracy of the target boiler fault diagnosis model obtained through training, and measures whether the target boiler fault diagnosis model meets the expected requirements based on the preset prediction conditions, and then determines whether model training and parameter migration are needed again, thereby further ensuring the fault diagnosis capability of the target boiler fault diagnosis model finally obtained.
[0107] In order to verify the prediction accuracy and effect of the target boiler fault diagnosis model obtained by the method provided in the embodiment of the present application, the one-dimensional test sample data in each newly acquired test sample data subset is firstly diagnosed with boiler faults by using K-Nearest Neighbor (KNN) classifier, Support Vector Machine (SVM) classifier, Random Forest (RF) classifier, one-dimensional CNN model classifier and the target boiler fault diagnosis model, that is, its fault category label is determined, and then the average diagnostic accuracy of each classifier (the target boiler fault diagnosis model provided in the embodiment of the present application is also a classifier) is determined according to the actual fault category label in each test sample data subset.
[0108] Experiments have shown that the average diagnostic accuracy of the five classifiers is 93.22%, 95.8%, 94.5%, 93.9% and 98.49% respectively. By comparison, it can be seen that the target boiler fault diagnosis model obtained by the method provided in the embodiment of the present application has better fault diagnosis capability.
[0109] It can be seen that the boiler fault diagnosis model generation method provided in the embodiment of the present application performs preliminary training on the constructed model based on the operating data under the first operating condition with complete corresponding fault category labels, and then performs transfer learning on the model obtained by the preliminary training based on the operating data without corresponding fault category labels under each second operating condition, and then re-trains the model obtained by transfer learning based on the operating data with corresponding fault category labels under each second operating condition, and finally obtains a target boiler fault diagnosis model applicable to the first operating condition and the second operating condition. The model can directly perform fault diagnosis based on the operating data under different operating conditions, thereby improving the accuracy and efficiency of boiler fault diagnosis, and effectively solving the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0110] It should be understood that although Figure 1The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.
[0111] Figure 3 The flow chart of the boiler fault diagnosis method according to the embodiment of the present application is schematically shown. Figure 3 As shown, in one embodiment of the present application, a boiler fault diagnosis method is provided, and the method may include the following steps.
[0112] Step B200, obtaining the operation data to be predicted for the boiler.
[0113] The operation data to be predicted is a one-dimensional data set, which includes the values of multiple operation parameters of the boiler.
[0114] Step B300: input the operation data to be predicted into the target boiler fault diagnosis model obtained by the method described in the above embodiment to obtain the fault category label corresponding to the operation data to be predicted.
[0115] The boiler fault diagnosis method provided in the embodiment of the present application realizes real-time boiler fault diagnosis based on the collected operating data, and effectively reduces the unplanned shutdown of the boiler and shortens the maintenance time by accurately diagnosing the fault category label corresponding to the predicted operating data.
[0116] Since the boiler fault diagnosis method provided in the embodiment of the present application includes the boiler fault diagnosis model generation method provided in the above embodiment, it can also solve the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0117] Figure 4 The structure block diagram of the boiler fault diagnosis model generation device according to the embodiment of the present application is schematically shown. Figure 4 As shown, in one embodiment of the present application, a boiler fault diagnosis model generation device is provided, and the boiler fault diagnosis model generation device may include the following functional modules.
[0118] The sample data acquisition module is used to acquire a first sample data set of the boiler under a first operating condition and a second sample data set of the boiler under a second operating condition different from the first operating condition. Each first sample data subset in the first sample data set includes first operating data and its corresponding fault category label, the second sample data set includes a labeled sample data subset and an unlabeled sample data subset, each labeled sample data subset includes second operating data and its corresponding fault category label, and each unlabeled sample data subset includes third operating data.
[0119] The first model training module is used to establish an initial boiler fault diagnosis model, and train the initial boiler fault diagnosis model through the first sample data set to obtain a first boiler fault diagnosis model.
[0120] The deep feature data determination module is used to input each of the first operating data and the third operating data into the first boiler fault diagnosis model, and obtain the first deep feature data set and the second deep feature data set corresponding to each target neural network layer in the first boiler fault diagnosis model. The first deep feature data set includes the deep feature data corresponding to each of the first operating data, and the second deep feature data set includes the deep feature data corresponding to each of the third operating data.
[0121] The transfer learning module is used to perform transfer learning on the first boiler fault diagnosis model based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer to obtain a second boiler fault diagnosis model.
[0122] The second model training module is used to train the second boiler fault diagnosis model through each labeled sample data subset to obtain a target boiler fault diagnosis model.
[0123] In an embodiment of the present application, the initial boiler fault diagnosis model adopts a multi-scale convolutional neural network structure. The initial boiler fault diagnosis model includes multiple layers of sequentially connected neural network structures, each of which includes sequentially connected convolutional layers and maximum pooling layers, and the last neural network structure in the initial boiler fault diagnosis model is sequentially connected with a fully connected layer and an output layer, and the convolutional layers in the first two neural network structures in the initial boiler fault diagnosis model both use large convolution kernels.
[0124] In an embodiment of the present application, the transfer learning module may include:
[0125] The model building unit is used to build a target migration model, wherein the target migration model has the same structure as the initial boiler fault diagnosis model.
[0126] The distance determination unit is used to determine, for each target neural network layer, the feature distribution distance corresponding to the target neural network layer based on the first deep feature data set and the second deep feature data set corresponding to the target neural network layer.
[0127] The transfer learning unit is used to migrate the parameters of each target neural network layer to the neural network layer corresponding to the target neural network layer in the target migration model if the feature distribution distance corresponding to the target neural network layer meets the preset migration conditions.
[0128] In the embodiment of the present application, the distance determination unit may include:
[0129] The first distance determination subunit is used to determine the edge probability distribution distance and the conditional probability distribution distance corresponding to the target neural network layer according to the first deep feature data set and the second deep feature data set corresponding to the target neural network layer.
[0130] The second distance determination subunit is used to determine the feature distribution distance corresponding to the target neural network layer according to the edge probability distribution distance and conditional probability distribution distance corresponding to the target neural network layer.
[0131] In the embodiment of the present application, the device may further include:
[0132] The training data generation module is used to generate a test sample data set according to the first sample data subset and the labeled sample data subset, wherein each test sample data subset in the test sample data set includes test operation data and its corresponding fault category label.
[0133] The label prediction module is used to input each test operation data into the target boiler fault diagnosis model to obtain the prediction label corresponding to each test operation data.
[0134] The accuracy determination module is used to determine the prediction accuracy of the target boiler fault diagnosis model according to the fault category label and the prediction label corresponding to each test operation data.
[0135] The condition judgment module is used to return to execute the first model training module if the prediction accuracy does not meet the preset prediction condition, until the prediction accuracy of the target boiler fault diagnosis model meets the preset prediction condition.
[0136] Since the boiler fault diagnosis model generation device provided in the embodiment of the present application is a virtual device corresponding to the boiler fault diagnosis model generation method of the above-mentioned embodiment, it can also solve the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0137] Figure 5 The structure block diagram of the boiler fault diagnosis device according to the embodiment of the present application is schematically shown. Figure 5 As shown, in one embodiment of the present application, a boiler fault diagnosis device is provided, and the boiler fault diagnosis device may include the following functional modules.
[0138] The module for acquiring data to be predicted is used to acquire the operation data to be predicted for the boiler.
[0139] The fault prediction module is used to input the operation data to be predicted into the target boiler fault diagnosis model obtained by the method described in the above embodiment to obtain the fault category label corresponding to the operation data to be predicted.
[0140] Since the boiler fault diagnosis device provided in the embodiment of the present application is a virtual device corresponding to the boiler fault diagnosis method of the above embodiment, it can also solve the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0141] An embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the boiler fault diagnosis model generation method described in the above embodiment is implemented.
[0142] The electronic device provided in the embodiment of the present application includes a processor capable of running the boiler fault diagnosis model generation method of the aforementioned embodiment, and can therefore also solve the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0143] An embodiment of the present application provides a machine-readable storage medium having instructions stored thereon. When the instructions are executed by a processor, the processor is configured to execute the boiler fault diagnosis model generation method described in the above embodiment.
[0144] The machine-readable storage medium provided in the embodiment of the present application stores instructions for enabling the machine to execute the boiler fault diagnosis model generation method of the above embodiment. Therefore, it can also solve the problem in the prior art that some boiler faults are difficult to determine directly based on the acquired operating data, thereby reducing the accuracy and efficiency of boiler fault diagnosis.
[0145] Figure 6 The internal structure diagram of the computer device of the embodiment of the present application is schematically shown. Figure 6As shown, in one embodiment of the present application, a computer device is provided, which can be a terminal. The computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05 and a memory (not shown in the figure) connected through a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A06. The network interface A02 of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor A01, a method for generating a boiler fault diagnosis model is implemented. The display screen A04 of the computer device can be a liquid crystal display or an electronic ink display, and the input device A05 of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.
[0146] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0147] In one embodiment, the boiler fault diagnosis model generation device provided in the present application can be implemented in the form of a computer program. Figure 6 The computer device is run on the computer device shown. The memory of the computer device can store various program modules constituting the boiler fault diagnosis model generation device, and the computer program composed of various program modules enables the processor to execute the steps of the boiler fault diagnosis model generation method of each embodiment of the present application described in this specification.
[0148] Figure 6 The computer device shown can be Figure 3 The sample data acquisition module in the boiler fault diagnosis model generation device shown executes step A200, the first model training module executes step A300, the deep feature data determination module executes step A400, the transfer learning module executes step A500 and the second model training module executes step A600.
[0149] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0150] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0151] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0152] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0153] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0154] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0155] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0156] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0157] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.
Claims
1. A boiler fault diagnosis model generation method, characterized in that: The method comprises: Acquire a first sample data set of the boiler under a first operating condition and a second sample data set of the boiler under a second operating condition different from the first operating condition; wherein each first sample data subset in the first sample data set includes first operating data and a corresponding fault category label, and the second sample data set includes a labeled sample data subset and an unlabeled sample data subset, each labeled sample data subset includes second operating data and a corresponding fault category label, and each unlabeled sample data subset includes third operating data; Establishing an initial boiler fault diagnosis model, and training the initial boiler fault diagnosis model using the first sample data set to obtain a first boiler fault diagnosis model; Input each of the first operating data and the third operating data into the first boiler fault diagnosis model, respectively, to obtain a first deep feature data set and a second deep feature data set corresponding to each target neural network layer in the first boiler fault diagnosis model; wherein the first deep feature data set includes the deep feature data corresponding to each of the first operating data, and the second deep feature data set includes the deep feature data corresponding to each of the third operating data; Based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, transfer learning is performed on the first boiler fault diagnosis model to obtain a second boiler fault diagnosis model; The second boiler fault diagnosis model is trained by using each labeled sample data subset to obtain a target boiler fault diagnosis model.
2. The method according to claim 1, characterized in that The initial boiler fault diagnosis model adopts a multi-scale convolutional neural network structure; the initial boiler fault diagnosis model includes multiple layers of sequentially connected neural network structures, each neural network structure includes a convolution layer and a maximum pooling layer connected in sequence, and a fully connected layer and an output layer are sequentially connected after the last neural network structure in the initial boiler fault diagnosis model, and the convolution layers in the first two neural network structures in the initial boiler fault diagnosis model both use large convolution kernels.
3. The method according to claim 1, characterized in that: Based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, transfer learning is performed on the first boiler fault diagnosis model to obtain a second boiler fault diagnosis model, including: Establishing a target migration model; wherein the target migration model has the same structure as the initial boiler fault diagnosis model; For each target neural network layer, determining a feature distribution distance corresponding to the target neural network layer according to a first deep feature data set and a second deep feature data set corresponding to the target neural network layer; For each target neural network layer, if the feature distribution distance corresponding to the target neural network layer meets the preset migration conditions, the parameters of the target neural network layer are migrated to the neural network layer corresponding to the target neural network layer in the target migration model.
4. The method according to claim 3, characterized in that For each target neural network layer, determining a feature distribution distance corresponding to the target neural network layer according to a first deep feature data set and a second deep feature data set corresponding to the target neural network layer includes: Determine the edge probability distribution distance and conditional probability distribution distance corresponding to the target neural network layer according to the first deep feature data set and the second deep feature data set corresponding to the target neural network layer; According to the edge probability distribution distance and conditional probability distribution distance corresponding to the target neural network layer, the feature distribution distance corresponding to the target neural network layer is determined.
5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Generate a test sample data set according to the first sample data subset and the labeled sample data subset; wherein each test sample data subset in the test sample data set includes test operation data and its corresponding fault category label; Inputting each test operation data into the target boiler fault diagnosis model respectively to obtain a prediction label corresponding to each test operation data; Determining the prediction accuracy of the target boiler fault diagnosis model according to the fault category label and the prediction label corresponding to each test operation data; If the prediction accuracy does not meet the preset prediction conditions, return to the step of establishing an initial boiler fault diagnosis model, and train the initial boiler fault diagnosis model through the first sample data set to obtain a first boiler fault diagnosis model, until the prediction accuracy of the target boiler fault diagnosis model meets the preset prediction conditions.
6. A boiler fault diagnosis method, characterized in that: The method comprises: Obtaining the operation data to be predicted for the boiler; The operation data to be predicted is input into a target boiler fault diagnosis model obtained by the method according to any one of claims 1 to 5 to obtain a fault category label corresponding to the operation data to be predicted.
7. A boiler fault diagnosis model generation device, characterized in that: The device comprises: A sample data acquisition module, used to acquire a first sample data set of the boiler under a first operating condition and a second sample data set of the boiler under a second operating condition different from the first operating condition; wherein each first sample data subset in the first sample data set includes first operating data and a corresponding fault category label, and the second sample data set includes a labeled sample data subset and an unlabeled sample data subset, each labeled sample data subset includes second operating data and a corresponding fault category label, and each unlabeled sample data subset includes third operating data; A first model training module is used to establish an initial boiler fault diagnosis model, and train the initial boiler fault diagnosis model through the first sample data set to obtain a first boiler fault diagnosis model; A deep feature data determination module, used to input each of the first operating data and the third operating data into the first boiler fault diagnosis model, respectively, to obtain a first deep feature data set and a second deep feature data set corresponding to each target neural network layer in the first boiler fault diagnosis model; wherein the first deep feature data set includes the deep feature data corresponding to each of the first operating data, and the second deep feature data set includes the deep feature data corresponding to each of the third operating data; A transfer learning module, configured to perform transfer learning on the first boiler fault diagnosis model based on the first deep feature data set and the second deep feature data set corresponding to each target neural network layer, so as to obtain a second boiler fault diagnosis model; The second model training module is used to train the second boiler fault diagnosis model through each labeled sample data subset to obtain a target boiler fault diagnosis model.
8. A boiler fault diagnosis device, characterized in that: The device comprises: A module for acquiring data to be predicted, used for acquiring operation data to be predicted for the boiler; A fault prediction module is used to input the operation data to be predicted into a target boiler fault diagnosis model obtained by the method according to any one of claims 1 to 5 to obtain a fault category label corresponding to the operation data to be predicted.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the boiler fault diagnosis model generating method according to any one of claims 1 to 5 is implemented.
10. A machine-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by a processor, the processor is configured to execute the boiler fault diagnosis model generation method according to any one of claims 1 to 5.