A boiler fault detection method and device, electronic equipment and storage medium

Abstract

The application discloses a boiler fault detection method and device, electronic equipment and storage medium, and relates to the technical field of machine fault. A target operation parameter corresponding to a boiler to be detected in a current operation state is acquired, wherein the target operation parameter is used to represent an attribute characteristic parameter of the boiler to be detected; the target operation parameter is input into a trained fault prediction model to determine a fault detection result of the boiler to be detected; if the fault detection result indicates that the boiler to be detected has a fault, a target maintenance strategy of the boiler to be detected is determined. The application reduces the fault rate, improves the efficiency and reliability of fault determination, and greatly reduces the labor cost.

Description

Technical Field

[0001] This application relates to the field of machine failure technology, and in particular to a method, apparatus, electronic device and storage medium for detecting boiler failures. Background Technology

[0002] Currently, in the oil and gas field industry, hot injection boilers are key equipment for converting water into steam and injecting it into oil reservoirs to reduce crude oil viscosity and improve oil production efficiency. Therefore, hot injection boilers play a vital role in improving the recovery rate of heavy oil fields and other types of oil reservoirs.

[0003] However, hot injection boilers in the oil and gas field industry often experience operational failures. For example, leakage in the valve plates and seats of plunger pumps can lead to unstable discharge, and frequency converters may shut down due to poor heat dissipation in high-temperature environments. Therefore, it is necessary to detect faults in the operation of hot injection boilers. Traditional hot injection boiler fault detection usually adopts manual inspection. However, manual inspection has limited effectiveness and cannot fully cover all fault checkpoints, which can lead to safety hazards in hot injection boilers. Manual inspection has low efficiency, high failure rate, and poor reliability in determining faults, and it also consumes a lot of manpower. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for detecting boiler faults. The embodiments provided by this application solve the technical problems of low efficiency, high failure rate, and poor reliability of manual inspection in the prior art, as well as the high labor cost of manual inspection. The embodiments provided by this application improve the efficiency and reliability of fault detection while reducing the failure rate, and greatly reduce labor costs.

[0005] In a first aspect, this application provides a method for detecting boiler malfunctions, the method comprising:

[0006] Obtain the target operating parameters corresponding to the boiler under test in its current operating state, wherein the target operating parameters are used to characterize the attribute feature parameters of the boiler under test;

[0007] The target operating parameters are input into the trained fault prediction model to determine the fault detection result of the boiler to be tested.

[0008] If the fault detection result indicates that the boiler under test has a fault, then the target maintenance strategy for the boiler under test is determined.

[0009] In one feasible implementation, the step of inputting the target operating parameters into a trained fault prediction model to determine the fault detection result of the boiler under test includes:

[0010] By inputting the target operating parameters into the trained fault prediction model, the predicted performance degradation value of the boiler under test is obtained.

[0011] Based on the predicted performance degradation value, the fault detection result of the boiler under test is determined.

[0012] In one feasible implementation, the fault detection result includes whether a fault exists, and determining the fault detection result of the boiler under test based on the performance degradation prediction value includes:

[0013] Based on the performance range within which the predicted performance degradation value falls, it is determined whether the boiler under test has a fault.

[0014] In one feasible implementation, the trained fault prediction model is determined in the following manner:

[0015] Obtain sample data corresponding to the boiler to be tested, wherein the sample data includes historical sample operating parameters and label information corresponding to the historical sample operating parameters, and the label information is used to characterize the true value of sample performance degradation corresponding to the historical sample operating parameters.

[0016] Based on the sample data, the initial fault prediction model is trained at least once until the training termination condition is met, and a trained fault prediction model is obtained.

[0017] The training operations include:

[0018] For any of the sample data, the historical sample operating parameters are input into the initial fault prediction model to obtain the sample performance degradation prediction value;

[0019] The total training loss is determined based on the difference between the predicted performance degradation value of each sample data and the label information.

[0020] The parameters of the initial prediction model are adjusted based on the total training loss, and the adjusted initial prediction model is used as the prediction model for the next training operation.

[0021] In one feasible implementation, the sample data includes positive sample data and negative sample data, and determining the total training loss based on the difference between the predicted performance degradation value of each sample data and the label information includes:

[0022] The first training loss is determined based on the difference between the predicted performance degradation value and the label information corresponding to each positive sample data.

[0023] The second training loss is determined based on the difference between the predicted performance degradation value of each negative sample and the label information.

[0024] Based on the first training loss and the second training loss, the total training loss is determined.

[0025] In one feasible implementation, the boiler fault detection method further includes:

[0026] Determine the first weight coefficient corresponding to the first training loss;

[0027] Based on the preset coefficient weighting rule and the second weight coefficient, the second weighting coefficient corresponding to the second training loss is determined, and the sum of the first weight coefficient and the second weight coefficient is 1.

[0028] In one feasible implementation, determining the second weighting coefficient corresponding to the second training loss based on a preset coefficient weighting rule and a second weighting coefficient includes:

[0029] Determine the ratio between positive and negative sample data;

[0030] Based on the preset weighting rules and the ratio, the preset weighting coefficients are determined;

[0031] The second weighting coefficient corresponding to the second training loss is determined based on the preset weighting coefficient and the second weighting coefficient, wherein the preset weighting coefficient is greater than 1.

[0032] In one feasible implementation, acquiring the sample data corresponding to the boiler to be tested includes:

[0033] Obtain the initial sample data corresponding to the boiler to be tested;

[0034] Based on the preset parameter variation range, the initial sample data is screened for fault parameters, and the initial sample data related to the fault parameters is determined as the first candidate sample data.

[0035] The first sample data is standardized to obtain the second candidate sample data corresponding to the boiler to be tested.

[0036] The second candidate sample data is expanded to obtain the sample data corresponding to the boiler to be tested.

[0037] In a second aspect, this application provides a boiler fault detection device, comprising:

[0038] The acquisition module is used to acquire the target operating parameters corresponding to the boiler under test in the current operating state, wherein the target operating parameters are used to characterize the attribute feature parameters of the boiler under test;

[0039] The first determining module is used to input the target operating parameters into the trained fault prediction model to determine the fault detection result of the boiler to be detected;

[0040] The second determining module is used to determine the target maintenance strategy for the boiler under test if the fault detection result indicates that the boiler under test has a fault.

[0041] In a third aspect, this application provides an electronic device including a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the boiler fault detection method described above.

[0042] In a sixth aspect of this application, an embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the boiler fault detection method described above.

[0043] Compared with the prior art, the boiler fault detection method, device, electronic equipment and storage medium provided in this application determine the fault detection result of the boiler under test by inputting the target operating parameters into a trained fault prediction model, and determine the target maintenance strategy of the boiler under test when the fault detection result indicates that the boiler under test has a fault. Compared with the fault determination method of manual inspection, this application improves the efficiency and reliability of fault determination while reducing the fault rate, and greatly reduces labor costs. Attached Figure Description

[0044] Figure 1 A flowchart illustrating a boiler fault detection method provided in an embodiment of this application is shown.

[0045] Figure 2 The diagram shows a loss training curve in a boiler fault detection method provided in an embodiment of this application.

[0046] Figure 3 A structural block diagram of a boiler fault detection device provided in an embodiment of this application is shown;

[0047] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown.

[0048] Figure 3 and Figure 4 The correspondence between the figure labels and figure titles in the accompanying drawings is as follows:

[0049] 300 Boiler fault detection device; 310 Acquisition module; 320 First determination module; 330 Second determination module; 400 Electronic device; 410 Processor; 420 Memory; 430 Bus. Detailed Implementation

[0050] To better understand the technical solutions provided in the embodiments of this specification, the technical solutions of the embodiments of this specification will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this specification and the specific features in the embodiments are detailed descriptions of the technical solutions of the embodiments of this specification, rather than limitations on the technical solutions of this specification. In the absence of conflict, the embodiments of this specification and the technical features in the embodiments can be combined with each other.

[0051] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The term "two or more" includes two or more cases.

[0052] First, the applicable application scenarios of this application will be introduced. The embodiments provided in this application are applicable to the field of machine fault technology.

[0053] Currently, hot injection boilers in the oil and gas field industry often experience operational failures. For example, leakage in the valve plates and seats of plunger pumps can lead to unstable discharge, and frequency converters may shut down due to poor heat dissipation in high-temperature environments. Therefore, it is necessary to detect faults in the operation of hot injection boilers. Traditional hot injection boiler fault detection usually adopts manual inspection. However, manual inspection has limited effectiveness and cannot fully cover all fault checkpoints, which can lead to safety hazards in hot injection boilers. Manual inspection has low efficiency, high failure rate, and poor reliability in determining faults, and it also consumes a lot of manpower.

[0054] Traditional manual inspection methods make it difficult to monitor equipment status in real time, and manual inspections can lead to untimely maintenance work, which in turn affects the service life of the equipment and cannot effectively deal with emergencies.

[0055] Based on this, embodiments of this application provide a method, apparatus, electronic device and storage medium for detecting boiler faults. This application reduces the failure rate, improves the efficiency and reliability of fault determination, and greatly reduces labor costs.

[0056] Figure 1 A flowchart illustrating a boiler fault detection method provided in an embodiment of this application is shown. Figure 1 As shown, an exemplary method for determining a fault identification model includes the following steps:

[0057] S101. Obtain the target operating parameters corresponding to the boiler under test in the current operating state, wherein the target operating parameters are used to characterize the attribute features of the boiler under test.

[0058] In this step, during the fault detection process of the boiler under test, this application first obtains the target operating parameters of the boiler under test, which are used to characterize the attribute feature parameters, through different types of external sensors or other types of data acquisition devices, under the current operating state.

[0059] Understandably, the attribute characteristic parameters of the boiler to be tested provided in this application can be specifically characteristic parameters whose values ​​change over time during the operation of the boiler to be tested. Here, the attribute characteristic parameters of the boiler to be tested can be specifically, but are not limited to, the flue gas temperature, outlet temperature, furnace tube temperature, fuel oil temperature, inlet temperature, flue gas oxygen content, steam temperature, superheater pressure drop, inlet pressure, inlet temperature, and fuel gas temperature. Pressure, radiant output temperature of the boiler under test, inlet water pressure of the boiler under test, radiant output pressure of the boiler under test, outlet pressure of the boiler under test, inlet pressure of the boiler under test, superheated outlet tube wall temperature of the boiler under test, pump outlet pressure of the boiler under test, superheated outlet temperature of the boiler under test, superheated inlet temperature of the boiler under test, separation drainage flow rate of the boiler under test, separator liquid level of the boiler under test, pump outlet temperature of the boiler under test, superheated outlet pressure of the boiler under test, furnace pressure of the boiler under test, superheat of the boiler under test, feedwater flow rate of the boiler under test, bearing temperature of the boiler under test, and steam pressure of the boiler under test, etc.

[0060] For example, the boiler to be tested in the embodiments provided in this application may be specifically, but is not limited to, a hot injection boiler at a target location within a target area.

[0061] S102. Input the target operating parameters into the trained fault prediction model to determine the fault detection result of the boiler to be tested.

[0062] In this step, after collecting the target operating parameters of the boiler to be tested, the target operating parameters are input into the trained fault prediction model to detect the performance of the boiler to be tested, and the fault detection result of the boiler to be tested is determined based on the detected performance prediction results.

[0063] Understandably, if the target operating parameters are input into the trained fault prediction model, and the trained fault prediction model predicts that the boiler under test has performance degradation based on the target operating parameters, and the performance degradation value predicted by the trained fault prediction model reaches a certain range, then it is determined that the boiler under test has a fault; if the performance degradation value predicted by the trained fault prediction model is within the specified degradation range, then it is determined that the boiler under test does not have a fault.

[0064] The fault detection results in the embodiments provided in this application include, but are not limited to, two prediction results: having a fault and not having a fault.

[0065] For example, the trained fault prediction model in the embodiments provided in this application can be specifically set as a type of recurrent neural network, namely the gated recurrent unit (GRU) model, according to different application scenarios.

[0066] The GRU was chosen in the embodiments provided in this application because: GRU is a variant of Recurrent Neural Network (RNN). GRU can solve the problems of long-term dependency and gradient vanishing in RNN. In addition, GRU has a special gating structure. Therefore, the structure of GRU is more concise and the number of parameters is less. Therefore, GRU has advantages in training speed and computational efficiency.

[0067] The GRU structure has two gates: the reset gate and the update gate. The reset gate determines how to combine new input information with previous memories. The update gate is used to integrate the forget gate and input gate in the Long Short-Term Memory Artificial Neural Network (LSTM) into one gate. It determines the amount of information in the historical memory to be saved to the current time step. The update gate directly acts on the hidden state of the previous time step.

[0068] The embodiments provided in this application reduce model parameters by using GRU, thereby merging the input gate and forget gate in LSTM, and effectively capturing long-term dependencies in data under limited resources.

[0069] S103. If the fault detection result indicates that the boiler under test has a fault, then determine the target maintenance strategy for the boiler under test.

[0070] In this step, when the fault detection result of the boiler under test confirms that the boiler under test has a fault, that is, the boiler under test has performance degradation, and the performance degradation value predicted by the trained fault prediction model reaches a certain range, an alert message is sent to the operator. Specifically, this is to remind the operator to formulate a corresponding target maintenance strategy for the boiler under test based on the fault detection result, and then perform routine maintenance on the boiler under test according to the corresponding target maintenance strategy, and record the routine maintenance process. At the same time, the operator can also determine the performance degradation status of the predicted equipment of the boiler under test based on the performance degradation degree determined by the trained fault prediction model, and combine the above performance degradation prediction value to determine the remaining service life of the boiler under test, thereby ensuring that the boiler under test can operate with high energy efficiency and high reliability, reducing production safety hazards, reducing labor costs, and improving utilization efficiency.

[0071] It is understood that the target maintenance strategy provided in the embodiments of this application can be a maintenance strategy set and selected by the operator based on experience values, or it can be a pre-set strategy based on an external large database. Different target maintenance strategies can be customized and judged according to specific application scenarios and applicable conditions.

[0072] The boiler fault detection method provided in this application determines the fault detection result of the boiler under test by inputting the target operating parameters into a trained fault prediction model. When the fault detection result indicates that the boiler under test has a fault, the target maintenance strategy of the boiler under test is determined. Compared with the fault determination method of manual inspection, this application improves the efficiency and reliability of fault determination while reducing the fault rate, and greatly reduces labor costs.

[0073] In one embodiment, step S102 includes the following sub-steps:

[0074] Sub-step 1021: Input the target operating parameters into the trained fault prediction model to obtain the predicted performance degradation value of the boiler to be tested.

[0075] In this step, since the target operating parameters collected by the sensors in the embodiments provided in this application are all attribute feature parameters with degradation characteristics, after inputting the collected target operating parameters of the boiler to be tested into the trained fault prediction model, the embodiments provided in this application can directly determine the performance degradation prediction value of the boiler to be tested. Then, based on the magnitude of the performance degradation prediction value, it is necessary to determine the fault detection result of the boiler to be tested, that is, whether the boiler to be tested has a fault.

[0076] Sub-step 1022: Based on the performance degradation prediction value, determine the fault detection result of the boiler to be tested.

[0077] In this step, firstly, it is determined that the fault detection result of the boiler to be tested in the embodiments provided in this application can be specific but not limited to: including whether there is a fault. At this time, after determining the performance degradation prediction value, the embodiments provided in this application need to determine whether there is a fault in the boiler to be tested based on the performance range in which the performance degradation prediction value is located.

[0078] For example, if the performance degradation prediction value of the embodiment provided in this application is determined to be within a predetermined preset performance range, then it is determined that the boiler under test has no fault; if the performance degradation prediction value of the embodiment provided in this application is determined to be outside the predetermined preset performance range, that is, the performance degradation prediction value is greater than the predetermined performance degradation threshold specified in the preset performance range, then it is determined that the boiler under test has a fault.

[0079] In one embodiment, the embodiments provided in this application specifically determine the trained fault prediction model through the following sub-steps:

[0080] Sub-step 1: Obtain sample data corresponding to the boiler to be tested. The sample data includes historical sample operating parameters and label information corresponding to the historical sample operating parameters. The label information is used to characterize the true value of sample performance degradation corresponding to the historical sample operating parameters.

[0081] In this step, before constructing the trained fault prediction model, the embodiments provided in this application first need to determine the sample data containing historical sample operating parameters. The historical sample operating parameters in the sample data contain the pre-set true value of sample performance degradation, as well as relevant information such as whether the historical sample operating parameters have historical faults during historical operation.

[0082] Sub-step 2: Based on the sample data, perform at least one training operation on the initial fault prediction model until the training termination condition is met, and obtain the trained fault prediction model.

[0083] In this step, the training operation provided in this application can be specifically as follows: for any sample data, input the historical sample running parameters into the initial fault prediction model to obtain the sample performance degradation prediction value; determine the total training loss based on the difference between the sample performance degradation prediction value corresponding to each sample data and the label information; adjust the parameters of the initial prediction model based on the total training loss, and use the initial prediction model with adjusted parameters as the prediction model for the next training operation until the training termination condition is met, and obtain the trained fault prediction model. In addition, after determining the trained fault prediction model, the embodiment provided in this application will also perform performance testing on the trained fault prediction model through threshold judgment, and then apply the sklearn.metrics module to calculate the classification performance indicators: precision, recall, F1 score, and AUC value, etc.

[0084] It should be noted that the performance test of the trained fault prediction model using threshold judgment is to transform the sample performance degradation prediction value output by the trained fault prediction model into a binary classification problem, that is, a binary classification problem with prediction values ​​of 0 and 1, which can be specifically represented as:

[0085] If the predicted value of sample performance degradation output by the trained fault prediction model is greater than 0.6, the predicted value is determined to be 1; if the predicted value of sample performance degradation output by the trained fault prediction model is less than 0.6, the predicted value is determined to be 0.

[0086] It is understood that the initial fault prediction model and the method of training total loss in the embodiments provided in this application can be customized and determined according to different application scenarios.

[0087] For example, the initial fault prediction model in the embodiments provided in this application is specifically a gated recurrent unit (GRU) model, and the method of training the total loss in the embodiments provided in this application is specifically, but not limited to, using the cross-entropy loss function to train the total loss and calculate the total training loss.

[0088] Here, the GRU structure has two gates: the reset gate and the update gate. In the actual application scenario described in this application, both the reset gate and the update gate are manifested as: combining the output of the sample performance degradation prediction value trained from the previous round of sample data with the input represented by the label of the sample data at the current moment, that is, combining the true value of the sample performance degradation of the sample data at the current moment, to obtain the candidate sample performance degradation prediction value output. This process continues until the total training loss converges, and the trained fault prediction model is determined.

[0089] It should be noted that the reset gate in the GRU structure determines how new input information is combined with previous memories, and the reset gate is mainly reflected in the cell unit it represents, and the current candidate state of the cell unit. The calculation formula is as follows:

[0090]

[0091] in, Used to characterize the output under the current candidate state; W h Used to represent the learned weight matrix; h t-1 Used to represent the output of the previous time step; tanh is used to represent a vector generated for the update; x t Used to represent the input at the current moment; h t-1 Used to characterize the output at the previous time step; r t The output used to characterize the reset gate is an output value between 0 and 1.

[0092] For ease of understanding, let's expand it as follows:

[0093]

[0094] Here, tanh is used to characterize a vector generated during the update; Used to characterize the output in the current candidate state; x t Used to represent the input at the current moment; h t-1 Used to characterize the output of the previous time step; W hx and W hh Both are used to represent the learned weight matrix; r t The output used to characterize the reset gate is an output value between 0 and 1; b h Used to characterize the bias term.

[0095] According to formula (2), if the output r of the reset gate is... t The smaller the value, that is, (r t h t-1 W hh The closer to 0, the more (r) t h t-1 W hh The smaller the value, the more likely it is that the predicted sample performance degradation value h from the output of the previous time step is... t-1 The more information that needs to be discarded, the more the gate's output r will be reset. t The larger the value, the higher the predicted sample performance degradation value h from the output of the previous time step. t-1 The more information that needs to be retained, the better.

[0096] It should be noted that the update gate of GRU is equivalent to a gate formed by integrating the forget gate and input gate of LSTM. It determines the amount of information in the historical memory that is saved to the current moment, which is used to update the memory.

[0097] For example, after the sigmoid activation function is applied, the formula for calculating the GRU update gate is as follows:

[0098] z t =σ(W z ·[h t-1 ,x t ]) (3);

[0099] Where σ() is used to characterize the sigmoid activation function; z t Used to characterize the output of the update gate.

[0100] In the above, the output z of the update gate is... t It is also determined by the input x at the current moment. t and the output h of the previous time step t-1 They jointly determine and limit their value to the range between 0 and 1 through the sigmoid activation function.

[0101] In summary, the output h at the current moment t The calculation formula is as follows:

[0102]

[0103] Among them, h t Used to characterize the output at the current time, it can be known that the output h at the current time is... t Mainly composed of the output z of the update gate t Decision: If the output z of the updated gate is to be updated... t The closer it is to 1, the better the state h of the previous time step. t-1 The less information retained in the candidate state, the better. The more information retained, the better; conversely, if the output z of the updated gate is reduced... t The closer h is to 0, the better. t-1 The more information is retained, the better. The less information is retained, the more effective the new memory can be.

[0104] In summary, the current state h of the cell unit t Includes the state h from the previous time step t-1 The cell state is formed by combining certain information from the input sample with some information from the actual performance degradation value at the current moment, and then passed on to the next cell.

[0105] In one embodiment, the sample data includes positive sample data and negative sample data. Based on the difference between the predicted performance degradation value and the label information corresponding to each sample data point, the total training loss is determined, including the following sub-steps:

[0106] Sub-step 1: Determine the first training loss based on the difference between the predicted sample performance degradation value and the label information corresponding to each positive sample data.

[0107] In this step, it is first determined that the embodiment provided in this application uses the cross-entropy loss function to determine the total training loss, and the embodiment provided in this application uses positive sample data and negative sample data respectively to determine different types of sub-training losses between the predicted value of sample performance degradation and the label information. Among them, the sub-training loss determined according to the difference between the predicted value of sample performance degradation and the label information corresponding to each positive sample data is the first training loss.

[0108] Sub-step 2: Determine the second training loss based on the difference between the predicted sample performance degradation value and the label information corresponding to each negative sample data.

[0109] In this step, the sub-training loss determined based on the difference between the predicted performance degradation value of each positive sample data and the label information is the second training loss.

[0110] Sub-step 3: Determine the total training loss based on the first training loss and the second training loss.

[0111] In this step, the purpose of determining the first training loss and the second training loss is to re-weight the coefficients of the second training loss. Specifically, the weights of the second training loss are weighted for the negative sample data that has failed, i.e., the samples whose performance degradation prediction value is greater than the performance range. This is to achieve a balance of sample data and thus enhance the prediction accuracy of the trained fault prediction model.

[0112] It's important to note that cross-entropy is a concept from information entropy theory. Originally used to estimate average code length, in deep learning training models, cross-entropy can be viewed as representing the difficulty of probability distribution p(x) through probability distribution q(x). Cross-entropy characterizes the distance between two probability distributions; that is, the smaller the cross-entropy value, i.e., the smaller the relative entropy value, the closer the two probability distributions are. The specific formula for the cross-entropy loss function is as follows:

[0113]

[0114] Where p(x) is used to characterize the target distribution; q(x) is used to characterize the predicted matching distribution.

[0115] In one embodiment, the boiler fault detection method further includes the following steps:

[0116] Sub-step 1: Determine the first weight coefficient corresponding to the first training loss.

[0117] Sub-step 2: Based on the preset coefficient weighting rule and the second weight coefficient, determine the second weighting coefficient corresponding to the second training loss, and the sum of the first weight coefficient and the second weight coefficient is 1.

[0118] For example, the embodiments provided in this application first need to determine the ratio between positive sample data and negative sample data, then determine the preset weighting coefficient based on the preset coefficient weighting rule and the ratio, and then determine the second weighting coefficient corresponding to the second training loss according to the preset weighting coefficient and the second weighting coefficient, wherein the preset weighting coefficient is greater than 1.

[0119] It should be noted that because there are relatively few negative samples with faults in the sample data, the sample data is imbalanced, which affects the prediction results of the trained fault prediction model. This causes the prediction results to favor more values, meaning that the model trains and iterates more in the direction where there are no faults. Therefore, it is necessary to increase the second weight coefficient of the minority negative sample data according to the preset weighting coefficients (determine the preset weighting coefficients). Specifically, a weighted cross-entropy loss function is used, and the formula for the weighted cross-entropy loss function is as follows:

[0120]

[0121] Where w is used to represent the preset weighting coefficient.

[0122] The embodiments provided in this application described above can be specifically applied to... This term is multiplied by a preset weighting coefficient w greater than 1, so that when y = 0, the second training loss of this term contributes more to the weighted cross-entropy loss function, thereby prompting the initial fault prediction model to pay more attention to the case where the predicted value is 0, so that the sample performance degradation prediction value can be more reasonably biased towards 0, thereby improving the prediction ability of boiler faults.

[0123] Here, the preset coefficient weighting rule in the embodiments provided in this application can be specifically, but not limited to, being determined based on empirical values; or determined based on an external positive-negative sample ratio-weighting coefficient mapping table.

[0124] Figure 2 This illustration shows a loss training curve in boiler fault detection according to an embodiment of this application. For example, as shown... Figure 2 As shown, Figure 2 The embodiment provided in this application illustrates the process of using a weighted cross-entropy loss function for loss training until the loss converges.

[0125] In one embodiment, the present application provides an embodiment that: obtaining sample data corresponding to the boiler to be tested includes the following sub-steps:

[0126] Sub-step 1: Obtain the initial sample data corresponding to the boiler to be tested.

[0127] In this step, various types of external sensor components are first used to collect initial sample data corresponding to the boiler to be tested. Here, the initial sample data in the embodiment provided in this application is specifically m=39 dimensions.

[0128] Sub-step 2: Based on the preset parameter variation range, the initial sample data is screened for fault parameters. The initial sample data related to the fault parameters in the initial sample data is determined as the first candidate sample data. It is assumed that the first candidate sample data in the embodiment provided in this application is specifically n=29 dimensions.

[0129] In this step, after determining the initial sample data, the embodiments provided in this application need to determine the sample degradation characteristic parameters of emission changes in the time series from the initial sample data according to the preset parameter change range, and determine the screened initial sample data as the first candidate sample data.

[0130] It should be noted that the preset parameter variation range in the embodiments provided in this application can be obtained directly from the experience value of the operator, or it can be preset based on an external database.

[0131] For example, taking the lubricating oil pressure and other parameters of the boiler under test as an example, it can be found that the corresponding initial sample data has no trend of change and is all 0. Therefore, the initial sample data cannot characterize the operation process of the boiler under test at all, so it cannot be used as the first candidate sample data and needs to be discarded.

[0132] Sub-step 3: Standardize the first sample data to obtain the second candidate sample data corresponding to the boiler to be tested.

[0133] In this step, in order to eliminate the influence of dimensions between the first sample data, i.e. the difference in numerical range, the real-time stream provided in this application needs to standardize the first sample data of the above 29 dimensions, and at the same time divide the training set and the test set. Here, the absolute difference of the second candidate sample data after standardization is eliminated, while the relative difference is retained.

[0134] For example, the standardization process for the first sample data in the embodiments provided in this application can be specifically as follows: the first sample data is subjected to min-max normalization, and the second candidate sample data obtained after normalization is scaled to the range [0,1]. The formula for max-min normalization is as follows:

[0135]

[0136] Where, x′ i Used to characterize the normalized second candidate sample data; x i Used to characterize the i-th first sample data; x min Used to characterize the minimum value in the first sample data; x max The table is used to represent the maximum value in the first sample data.

[0137] Sub-step 4: Expand the sample data for the second candidate sample data to obtain the sample data corresponding to the boiler to be tested.

[0138] In this step, since the amount and type of the second candidate sample data in the embodiments provided in this application are relatively small, in order to prevent the initial fault prediction model of deep learning from overfitting, it is necessary to expand the second candidate sample data.

[0139] For example, the embodiments provided in this application use a sliding window to process the second candidate sample data, creating more second candidate sample data for the initial fault prediction model. Specifically, the time series data is converted into an array format suitable for the input of the initial fault prediction model using a sliding window. For example, i is used to represent the total length of the time series; l is used to represent the length of the sliding window (the number of time steps contained in the window); k is used to represent the width of the sliding window (the number of features contained in each time step); and s is used to represent the sliding step size. Then, a total of n = (i-l+s) / s training sequences with a shape size of l×k can be constructed.

[0140] Here, the width of the sliding window is set to 29, the length to 10, and the sliding step size to 1. The number of samples slides forward by one step each time, which can construct more training sequences, i.e., sample data, for the initial fault prediction model, thereby improving the training efficiency of the initial fault prediction model.

[0141] Compared with the prior art, the boiler fault detection method provided in this application determines the fault detection result of the boiler under test by inputting the target operating parameters into the trained fault prediction model. When the fault detection result indicates that the boiler under test has a fault, the target maintenance strategy for the boiler under test is determined. Compared with the fault determination method of manual inspection, this application reduces the fault rate while improving the efficiency and reliability of fault determination, and significantly reduces labor costs. Furthermore, the embodiments provided in this application generate accurate and rich sample data by performing multiple noise processing, standardization, and sample expansion processes on the initial sample data collected by different sensing devices, thereby improving the stability, robustness, and training efficiency of the trained fault prediction model. This application solves the gradient vanishing problem in long sequence data processing of traditional RNNs by employing GRU, and merges the input gate and forget gate in LSTM, reducing model parameters. It can effectively capture long-term dependencies in data and identify and retain key information under limited resources. Moreover, the embodiments provided in this application improve the recognition accuracy and generalization ability of the trained fault prediction model by using functions including the cross-entropy function to determine the total training loss.

[0142] Figure 3 This diagram illustrates the structural block diagram of a boiler fault detection device provided in an embodiment of this application. For example,... Figure 3 As shown, the boiler fault detection device 300 includes:

[0143] The acquisition module 310 is used to acquire the target operating parameters corresponding to the boiler under test in the current operating state, wherein the target operating parameters are used to characterize the attribute feature parameters of the boiler under test.

[0144] The first determining module 320 is used to input the target operating parameters into the trained fault prediction model to determine the fault detection result of the boiler to be tested.

[0145] The second determining module 330 is used to determine the target maintenance strategy for the boiler under test if the fault detection result indicates that the boiler under test has a fault.

[0146] In one embodiment, the first determining module 320 is specifically used for:

[0147] By inputting the target operating parameters into the trained fault prediction model, the predicted performance degradation value of the boiler under test is obtained.

[0148] Based on the predicted performance degradation values, the fault detection results of the boiler under test are determined.

[0149] In one embodiment, the fault detection result includes whether a fault exists. Based on the performance degradation prediction value, the fault detection result of the boiler under test is determined, including:

[0150] Based on the performance range within which the predicted performance degradation value falls, it is determined whether the boiler under test has a fault.

[0151] In one embodiment, the acquisition module 310 determines the trained fault prediction model in the following manner:

[0152] Obtain sample data corresponding to the boiler to be tested. The sample data includes historical sample operating parameters and the label information corresponding to the historical sample operating parameters. The label information is used to characterize the true value of sample performance degradation corresponding to the historical sample operating parameters.

[0153] Based on the sample data, the initial fault prediction model is trained at least once until the training termination condition is met, thus obtaining a trained fault prediction model.

[0154] The training operations include:

[0155] For any sample data, input the historical sample operating parameters into the initial fault prediction model to obtain the sample performance degradation prediction value.

[0156] The total training loss is determined based on the difference between the predicted performance degradation value and the label information for each sample data.

[0157] The parameters of the initial prediction model are adjusted based on the total training loss, and the adjusted initial prediction model is used as the prediction model for the next training operation.

[0158] In one embodiment, the sample data includes positive sample data and negative sample data. Based on the difference between the predicted performance degradation value and the label information corresponding to each sample data point, the total training loss is determined, including:

[0159] The first training loss is determined based on the difference between the predicted performance degradation value and the label information corresponding to each positive sample data.

[0160] The second training loss is determined based on the difference between the predicted performance degradation value of each negative sample and the label information.

[0161] The total training loss is determined based on the first training loss and the second training loss.

[0162] In one embodiment, the first determining module 320 is further specifically used for:

[0163] Determine the first weight coefficient corresponding to the first training loss.

[0164] Based on the preset coefficient weighting rule and the second weight coefficient, the second weighting coefficient corresponding to the second training loss is determined, and the sum of the first weight coefficient and the second weight coefficient is 1.

[0165] In one embodiment, determining the second weighted weight coefficient corresponding to the second training loss based on a preset coefficient weighting rule and a second weight coefficient includes:

[0166] Determine the ratio between positive and negative sample data.

[0167] The preset weighting coefficients are determined based on the preset weighting rules and ratios.

[0168] The second weighting coefficient corresponding to the second training loss is determined based on the preset weighting coefficient and the second weighting coefficient, wherein the preset weighting coefficient is greater than 1.

[0169] In one embodiment, obtaining sample data corresponding to the boiler to be tested includes:

[0170] Obtain the initial sample data corresponding to the boiler to be tested.

[0171] Based on the preset parameter variation range, the initial sample data is screened for fault parameters, and the initial sample data related to the fault parameters is determined as the first candidate sample data.

[0172] The first sample data is standardized to obtain the second candidate sample data corresponding to the boiler to be tested.

[0173] The sample data of the second candidate sample is expanded to obtain the sample data corresponding to the boiler to be tested.

[0174] Compared with the prior art, the boiler fault detection device 300 in the embodiments of this application determines the fault detection result of the boiler under test by inputting the target operating parameters into the trained fault prediction model. When the fault detection result indicates that the boiler under test has a fault, the target maintenance strategy of the boiler under test is determined. Compared with the fault determination method of manual inspection, this application reduces the fault rate while improving the efficiency and reliability of fault determination, and greatly reduces labor costs. In addition, the embodiments of this application generate accurate and rich sample data by performing multiple noise processing, standardization processing and sample expansion on the initial sample data collected by different sensing devices, thereby improving the stability, robustness and training efficiency of the trained fault prediction model. This application solves the gradient vanishing problem in the long sequence data processing of traditional RNN by adopting GRU, and merges the input gate and forget gate in LSTM, reducing model parameters. It can effectively capture long-term dependencies in data and identify and retain key information under limited resources. In addition, the embodiments of this application improve the recognition accuracy and generalization ability of the trained fault prediction model by using the indicated function including the cross-entropy function to determine the total training loss.

[0175] Please see Figure 4 , Figure 4 This application provides a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.

[0176] Memory 420 stores machine-readable instructions executable by processor 410. When electronic device 400 is running, processor 410 and memory 420 communicate via bus 430. When the machine-readable instructions are executed by processor 410, they can perform the operations described above. Figures 1 to 2 The steps of the boiler fault detection method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0177] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figures 1 to 2 The steps of the boiler fault detection method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0178] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0179] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0180] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0181] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0182] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0183] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0184] This application also provides a computer program product, which includes computer software instructions that, when executed on a processing device, cause the processing device to execute a process for a boiler fault detection method.

[0185] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0186] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0187] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0188] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0189] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0190] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0191] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

[0192] Although preferred embodiments have been described in this specification, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this specification.

[0193] Obviously, those skilled in the art can make various modifications and variations to this specification without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims and their equivalents, this specification is also intended to include such modifications and variations.

Claims

1. A method for detecting boiler malfunctions, characterized in that, The boiler fault detection method includes: Obtain the target operating parameters corresponding to the boiler under test in its current operating state, wherein the target operating parameters are used to characterize the attribute feature parameters of the boiler under test; The target operating parameters are input into the trained fault prediction model to determine the fault detection result of the boiler to be tested. If the fault detection result indicates that the boiler under test has a fault, then the target maintenance strategy for the boiler under test is determined.

2. The boiler fault detection method according to claim 1, characterized in that, The step of inputting the target operating parameters into the trained fault prediction model to determine the fault detection result of the boiler under test includes: By inputting the target operating parameters into the trained fault prediction model, the predicted performance degradation value of the boiler under test is obtained. Based on the predicted performance degradation value, the fault detection result of the boiler under test is determined.

3. The boiler fault detection method according to claim 1, characterized in that, The fault detection result includes whether a fault exists. Determining the fault detection result of the boiler under test based on the performance degradation prediction value includes: Based on the performance range within which the predicted performance degradation value falls, it is determined whether the boiler under test has a fault.

4. The boiler fault detection method according to claim 1, characterized in that, The trained fault prediction model is determined in the following way: Obtain sample data corresponding to the boiler to be tested, wherein the sample data includes historical sample operating parameters and label information corresponding to the historical sample operating parameters, and the label information is used to characterize the true value of sample performance degradation corresponding to the historical sample operating parameters. Based on the sample data, the initial fault prediction model is trained at least once until the training termination condition is met, and a trained fault prediction model is obtained. The training operations include: For any of the sample data, the historical sample operating parameters are input into the initial fault prediction model to obtain the sample performance degradation prediction value; The total training loss is determined based on the difference between the predicted performance degradation value of each sample data and the label information. The parameters of the initial prediction model are adjusted based on the total training loss, and the adjusted initial prediction model is used as the prediction model for the next training operation.

5. The boiler fault detection method according to claim 4, characterized in that, The sample data includes positive sample data and negative sample data. The step of determining the total training loss based on the difference between the predicted performance degradation value of each sample and the label information includes: The first training loss is determined based on the difference between the predicted performance degradation value and the label information corresponding to each positive sample data. The second training loss is determined based on the difference between the predicted performance degradation value of each negative sample and the label information. Based on the first training loss and the second training loss, the total training loss is determined.

6. The boiler fault detection method according to claim 5, characterized in that, The method for detecting boiler malfunctions also includes: Determine the first weight coefficient corresponding to the first training loss; Based on the preset coefficient weighting rule and the second weight coefficient, the second weighting coefficient corresponding to the second training loss is determined, and the sum of the first weight coefficient and the second weight coefficient is 1.

7. The boiler fault detection method according to claim 6, characterized in that, The step of determining the second weighted weight coefficient corresponding to the second training loss based on the preset coefficient weighting rule and the second weight coefficient includes: Determine the ratio between positive and negative sample data; Based on the preset weighting rules and the ratio, the preset weighting coefficients are determined; The second weighting coefficient corresponding to the second training loss is determined based on the preset weighting coefficient and the second weighting coefficient, wherein the preset weighting coefficient is greater than 1.

8. The boiler fault detection method according to claim 4, characterized in that, The acquisition of sample data corresponding to the boiler to be tested includes: Obtain the initial sample data corresponding to the boiler to be tested; Based on the preset parameter variation range, the initial sample data is screened for fault parameters, and the initial sample data related to the fault parameters is determined as the first candidate sample data. The first sample data is standardized to obtain the second candidate sample data corresponding to the boiler to be tested. The second candidate sample data is expanded to obtain the sample data corresponding to the boiler to be tested.

9. A boiler fault detection device, characterized in that, The boiler fault detection device includes: The acquisition module is used to acquire the target operating parameters corresponding to the boiler under test in the current operating state, wherein the target operating parameters are used to characterize the attribute feature parameters of the boiler under test; The first determining module is used to input the target operating parameters into the trained fault prediction model to determine the fault detection result of the boiler to be detected; The second determining module is used to determine the target maintenance strategy for the boiler under test if the fault detection result indicates that the boiler under test has a fault.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the boiler fault detection method as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the boiler fault detection method as described in any one of claims 1-8.

Images