Method and system for evaluating combustion instability during low-load operation

Abstract

According to the combustion instability evaluation method and system during low-load operation provided by the invention, the thermal power generation combustion state abnormity analysis model is trained through the comparison result between the second type division result example and the fusion identification example; the trained thermal power generation combustion state abnormity analysis model is used for determining an object type analysis model, and a thermal power generation combustion state instability evaluation result is determined through the object type analysis model. In the process of model updating training, the initial thermal power generation combustion data example does not need to be obtained, only the newly added thermal power generation combustion data example is used for training the initial object type analysis model, the model performance can be improved when new data is learned, and therefore the accuracy and precision of thermal power generation combustion state evaluation can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of instability assessment, and in particular to a method and system for assessing combustion instability during low-load operation. Background Art

[0002] Combustion instability in thermal power generation refers to instability in the combustion system of a thermal power plant's boiler during operation, leading to decreased combustion efficiency, equipment damage, and even accidents. Combustion instability manifests itself primarily as combustion fluctuations and cyclical changes in pressure and temperature, which can, in severe cases, cause boiler fires to extinguish. When combustion instability occurs, it also reduces power generation. Therefore, proactive assessment of combustion instability is necessary to ensure normal power generation. However, how to assess combustion instability is currently a difficult technical challenge. Summary of the Invention

[0003] In order to improve the technical problems existing in the related technologies, the present application provides a method and system for evaluating combustion instability during low-load operation.

[0004] In a first aspect, a method for evaluating combustion instability during low-load operation is provided, comprising:

[0005] Acquire a newly added thermal power generation combustion data example of the object example; the newly added thermal power generation combustion data example records a newly added object type;

[0006] Based on the initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting an initial thermal power generation combustion data example of the object example into a target model for object type parsing training; the initial thermal power generation combustion data example records an initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier;

[0007] Inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state classification model for object type analysis processing to obtain a first type classification result example;

[0008] Performing fusion processing on the first type classification result example and the newly added object type identifier to obtain a fusion identifier example;

[0009] Inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a second type classification result example;

[0010] The thermal power generation combustion state abnormality parsing model is trained by comparing the results of the second type division result example with the fusion identification example. The trained thermal power generation combustion state abnormality parsing model is used to determine the object type parsing model. The thermal power generation combustion state instability assessment result is determined by the object type parsing model. The object type parsing model is used to parse the newly added object type identification in the thermal power generation combustion state data.

[0011] In the present application, after acquiring the newly added thermal power generation combustion data example of the object example, the method further includes:

[0012] Performing data filtering processing on the newly added thermal power generation combustion data examples to obtain a data filtering result, wherein the filtering processing is used to characterize the arrangement of the newly added thermal power generation combustion data examples;

[0013] inputting the data filtering results into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain first loss information;

[0014] The training of the thermal power generation combustion state abnormality analysis model based on the comparison result between the second type classification result example and the fusion identification example includes:

[0015] Determining second loss information through a comparison result between the second-type division result example and the fusion identification example;

[0016] determining target loss information based on the first loss information and the second loss information;

[0017] The model parameters of the thermal power generation combustion state abnormality analysis model are corrected using the target loss information.

[0018] In the present application, the data filtering results are input into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain the first loss information, including:

[0019] Inputting the data filtering result into the thermal power generation combustion state classification model for object type analysis processing to obtain a third type classification result example;

[0020] Inputting the data filtering result into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a fourth type classification result example;

[0021] The first loss information is determined by comparing the fourth type division result example with the third type division result example.

[0022] In this application, the method further comprises:

[0023] After the number of training times reaches the first target number of times, the thermal power generation combustion state classification model is updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model. The number of training times refers to the number of times the thermal power generation combustion state abnormality analysis model is trained using the newly added thermal power generation combustion data examples.

[0024] In the present application, after the number of training times reaches the first target number of times, the thermal power generation combustion state classification model is updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model, including:

[0025] updating the thermal power generation combustion state classification model based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain an updated thermal power generation combustion state classification model;

[0026] During the training process after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model.

[0027] In this application, the number of training times

[0028] During the training process after the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated based on the model parameters of the thermal power generation combustion state abnormality parsing model to obtain the object type parsing model, including:

[0029] After the number of training times reaches the first target number of times, each time a second target number of training times passes, the updated thermal power generation combustion state classification model is updated using the model parameters of the thermal power generation combustion state abnormality analysis model;

[0030] After the training is terminated, the updated thermal power generation combustion state classification model is determined as the object type parsing model.

[0031] In the present application, the updating of the thermal power generation combustion state classification model is performed using the model parameters of the thermal power generation combustion state abnormality analysis model after each second target number of training times, including:

[0032] After each second target number of training times, real-time model parameters of the thermal power generation combustion state abnormality analysis model are obtained;

[0033] Processing the real-time model parameters based on a parameter update model to obtain target model parameters;

[0034] The updated thermal power generation combustion state classification model is updated using the target model parameters.

[0035] In this application, the real-time model parameters are processed based on the parameter update model to obtain the target model parameters, including:

[0036] Determining a first similarity of a thermal power generation combustion state classification model and a second similarity of a thermal power generation combustion state abnormality analysis model based on the number of training times corresponding to the real-time model parameters; wherein the first similarity decreases as the number of training times increases, and the second similarity increases as the number of training times increases;

[0037] The real-time model parameters are processed using the first similarity, the second similarity, and the parameter update model to obtain the target model parameters.

[0038] In this application, the training step of the initial object type parsing model includes:

[0039] Acquiring the initial thermal power generation combustion data example of the object example;

[0040] Inputting the initial thermal power generation combustion data example into the target model for object type parsing processing to obtain an initial type classification result example;

[0041] Determining example loss information by comparing the initial type classification result example with the initial object example type identifier;

[0042] The model parameters of the target model are modified using the example loss information until the training termination requirements are met, and the target model at the time of training termination is determined as the initial object type parsing model.

[0043] In a second aspect, a combustion instability assessment system during low-load operation is provided, comprising a processor and a memory communicating with each other, wherein the processor is configured to read and execute a computer program from the memory to implement the above-mentioned method.

[0044] The embodiment of the present application provides a method and system for evaluating combustion instability during low-load operation, which obtains a newly added thermal power generation combustion data example of an object example; the newly added thermal power generation combustion data example records a newly added object type identifier; based on an initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting the initial thermal power generation combustion data example of the object example into a target model for object type parsing training; the initial thermal power generation combustion data example records the initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state classification model for object type parsing The first type of classification result example is obtained by performing analysis processing; the first type of classification result example and the newly added object type identifier are fused to obtain a fusion identifier example; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state abnormality analysis model for object type analysis processing to obtain a second type of classification result example; the thermal power generation combustion state abnormality analysis model is trained through the comparison result between the second type of classification result example and the fusion identifier example, and the trained thermal power generation combustion state abnormality analysis model is used to determine the object type analysis model, and the thermal power generation combustion state instability assessment result is determined through the object type analysis model, and the object type analysis model is used to parse the newly added object type identifier in the thermal power generation combustion state data. In the process of model update training, the present application does not need to obtain the initial thermal power generation combustion data example, and only uses the newly added thermal power generation combustion data example to train the initial object type analysis model, which can improve the model performance when learning new data, thereby improving the accuracy and precision of thermal power generation combustion state assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 This is a flow chart of a method for evaluating combustion instability during low-load operation provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to better understand the above technical solution, the technical solution of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0048] See also Figure 1 , shows a method for evaluating combustion instability during low-load operation, which may include the technical solutions described in the following steps S201-S211.

[0049] S201: Acquire a newly added thermal power generation combustion data example of an object example; the newly added thermal power generation combustion data example records a newly added object type.

[0050] Illustratively, examples of thermal power generation combustion data include: 1. Uneven mixing of fuel and oxidant: Uneven mixing of fuel and oxidant can lead to incomplete combustion, resulting in local high and low temperature areas, which in turn cause combustion fluctuations.

[0051] 2. Heat release fluctuations: Fluctuations in the heat release rate during combustion can cause changes in pressure and temperature, leading to unstable combustion.

[0052] 3. Disturbance in flow and mixing process: The flow and mixing process in the combustion chamber is affected by many factors, such as intake velocity, fuel injection method, etc. These disturbances will affect combustion stability.

[0053] 4. Chemical reaction kinetics: The nonlinear characteristics of the combustion reaction make the response to small disturbances nonlinear, resulting in uneven heat release in the combustion system, which in turn causes combustion instability.

[0054] 5. Combustion chamber structural characteristics: The shape, size and internal structure of the combustion chamber (such as flame holder, baffle, etc.) will affect the propagation of sound waves and the stability of the flame.

[0055] 6. External excitation: Changes in external conditions such as vibration and pressure fluctuations may also lead to unstable combustion.

[0056] S203: Based on the initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting the initial thermal power generation combustion data example of the object example into the target model for object type parsing training; the initial thermal power generation combustion data example records the initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier.

[0057] In one possible implementation embodiment, the training step of the initial object type parsing model includes the following content.

[0058] S301: Acquire the initial thermal power generation combustion data example of the object example;

[0059] S303: Inputting the initial thermal power generation combustion data example into the target model for object type analysis to obtain an initial type classification result example;

[0060] S305: Determine example loss information by comparing the initial type classification result example with the initial object example type identifier;

[0061] S307: Modify the model parameters of the target model using the example loss information until the training termination requirements are met, and determine the target model at the time of training termination as the initial object type parsing model.

[0062] In the present application, the initial thermal power generation combustion data example and the newly added thermal power generation combustion data example are thermal power generation combustion state data corresponding to the same type of object example, and are of different types; the target model can be a classification model, and the initial thermal power generation combustion data example can be input into the target model to extract the thermal power generation combustion state data features to obtain the example initial thermal power generation combustion state data features, and then the example initial thermal power generation combustion state data features are subjected to object type parsing processing to obtain the initial type classification result example; then, the example loss information is calculated by comparing the initial type classification result example with the initial object example type identifier; and the model parameters of the target model are corrected according to the example loss information until the training termination requirements are met. The training termination requirements may include but are not limited to the example loss information being less than the target loss target value, the example loss information being less than the target loss target value, and the number of model training iterations reaching the target number. Finally, the target model at the time of training termination is determined to be the initial object type parsing model.

[0063] In this application, the target model can be trained based on the initial thermal power generation combustion data example to quickly and accurately obtain the initial object type parsing model, so that when there are new thermal power generation combustion data examples, the model can be updated based on the initial object type parsing model to improve the model update efficiency.

[0064] S205: Inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state classification model to perform object type analysis processing to obtain a first type classification result example.

[0065] In the present application, the newly added thermal power generation combustion data example can be input into the thermal power generation combustion state classification model, and the thermal power generation combustion state data features of the thermal power generation combustion data example can be extracted based on the thermal power generation combustion state classification model, and object type parsing processing can be performed based on the extracted thermal power generation combustion state data features to obtain a first type classification result example; the first type classification result example and the newly added object type are identified as the same type of data.

[0066] In the present application, after constructing a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality analysis model based on the initial object type analysis model, a newly added thermal power generation combustion data example can be input into the thermal power generation combustion state classification model for object type analysis processing to obtain a first type classification result example; thereby, the thermal power generation combustion state abnormality analysis model is trained according to the output result of the thermal power generation combustion state classification model.

[0067] S207: Fusing the first type classification result example and the newly added object type identifier to obtain a fusion identifier example.

[0068] For example, fusion can be understood as integration, etc. Identification can be understood as labeling.

[0069] S209: Inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a second type classification result example.

[0070] For example, the abnormal analysis model of the thermal power generation combustion state can be understood as a neural convolution model, which can analyze the abnormality of the thermal power generation combustion state and obtain the corresponding abnormal points.

[0071] In the present application, a newly added thermal power generation combustion data example can be input into the thermal power generation combustion state abnormality analysis model, and the thermal power generation combustion state data features of the newly added thermal power generation combustion data example can be extracted based on the thermal power generation combustion state abnormality analysis model, and object type analysis processing can be performed based on the extracted thermal power generation combustion state data features to obtain a second type classification result example.

[0072] S2011: The thermal power generation combustion state abnormality parsing model is trained by comparing the second type division result example with the fusion identification example to obtain an object type parsing model.

[0073] In the present application, target loss information can be constructed based on the comparison results between the second type division result example and the fusion identification example, and the thermal power generation combustion state abnormality analysis model can be trained through the target loss information, so that the trained thermal power generation combustion state abnormality analysis model can be used as the object type analysis model, and the object type analysis model can be further obtained based on the trained model.

[0074] In relation to a possible implementation embodiment, after obtaining the newly added thermal power generation combustion data example of the object example, the method further includes: performing data filtering processing on the newly added thermal power generation combustion data example to obtain a data filtering result, wherein the filtering processing is used to characterize the arrangement of the newly added thermal power generation combustion data example; and inputting the data filtering result into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain first loss information.

[0075] S211: The thermal power generation combustion state abnormality analysis model is trained through the comparison result between the second type division result example and the fusion identification example. The trained thermal power generation combustion state abnormality analysis model is used to determine the object type analysis model. The thermal power generation combustion state instability assessment result is determined through the object type analysis model. The object type analysis model is used to parse the newly added object type identification in the thermal power generation combustion state data.

[0076] It should be understood that a newly added thermal power generation combustion data example of an object example is obtained; the newly added thermal power generation combustion data example records a newly added object type identifier; based on the initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting the initial thermal power generation combustion data example of the object example into the target model for object type parsing training; the initial thermal power generation combustion data example records the initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state classification model for object type parsing processing to obtain a first type classification Result example; the first type division result example and the newly added object type identifier are fused to obtain a fusion identifier example; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state abnormality analysis model for object type analysis processing to obtain a second type division result example; the thermal power generation combustion state abnormality analysis model is trained through the comparison result between the second type division result example and the fusion identifier example, and the trained thermal power generation combustion state abnormality analysis model is used to determine the object type analysis model, and the thermal power generation combustion state instability assessment result is determined through the object type analysis model, and the object type analysis model is used to parse the newly added object type identifier in the thermal power generation combustion state data. In the process of model update training, the present application does not need to obtain the initial thermal power generation combustion data example, and only uses the newly added thermal power generation combustion data example to train the initial object type analysis model, which can improve the model performance when learning new data, thereby improving the accuracy and precision of thermal power generation combustion state assessment.

[0077] In the present application, the data filtering results are input into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain the first loss information, including the following contents.

[0078] S401: Inputting the data filtering result into the thermal power generation combustion state classification model for object type analysis processing to obtain a third type classification result example.

[0079] S403: Inputting the data filtering result into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a fourth type classification result example.

[0080] S405: Determine the first loss information by comparing the fourth type division result example with the third type division result example.

[0081] In the present application, the data filtering results can be input into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain the third type division result example and the fourth type division result example; and unsupervised training is performed based on the third type division result example and the fourth type division result example, and the first loss information is determined by comparing the results between the fourth type division result example and the third type division result example.

[0082] Exemplarily, the training of the thermal power generation combustion state abnormality analysis model by comparing the second type division result example with the fusion identification example includes the following content.

[0083] S20111: Determine second loss information through a comparison result between the second type division result example and the fusion identification example.

[0084] In the present application, detection training can be performed based on the second type segmentation result example and the fusion identification example, and the second loss information is obtained by calculating the comparison result between the second type segmentation result example and the fusion identification example.

[0085] S20113: Determine the target loss information through the first loss information and the second loss information; in the present application, the sum of the first loss information and the second loss information can be calculated to obtain the target loss information; the similarity corresponding to the first loss information and the second loss information can also be set respectively, and the weighted sum of the first loss information and the second loss information can be calculated to obtain the target loss information.

[0086] S20115: Correcting the model parameters of the thermal power generation combustion state abnormality analysis model based on the target loss information.

[0087] In some embodiments, the model parameters of the thermal power generation combustion state abnormality analysis model can be corrected through the target loss information until the training termination requirements are met, and the thermal power generation combustion state abnormality analysis model at the time of training termination is used as the object type analysis model; the thermal power generation combustion state classification model can also be mined through the model parameters of the thermal power generation combustion state abnormality analysis model at the time of training termination, and the final thermal power generation combustion state classification model can be used as the object type analysis model.

[0088] In the present application, the method further includes the following steps.

[0089] S601: Correcting the model parameters of the thermal power generation combustion state abnormality analysis model using the target loss information until the training termination requirements are met, and using the thermal power generation combustion state abnormality analysis model at the time of training termination as the updated thermal power generation combustion state abnormality analysis model.

[0090] S603: The thermal power generation combustion state classification model is updated by updating the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model.

[0091] In the present application, the model parameters of the thermal power generation combustion state abnormality analysis model can be corrected according to the target loss information until the training termination requirements are met. The training termination requirements may include that the target loss information is less than the target target value or the target loss information is less than the target target value and the number of iterations reaches the target number; then the thermal power generation combustion state abnormality analysis model at the time of training termination is used as the updated thermal power generation combustion state abnormality analysis model.

[0092] In the present application, the thermal power generation combustion state classification model is updated by updating the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model, including: updating the thermal power generation combustion state classification model by updating the model parameters of the thermal power generation combustion state abnormality analysis model to obtain an updated thermal power generation combustion state classification model; iteratively updating the updated thermal power generation combustion state abnormality analysis model and the updated thermal power generation combustion state classification model to obtain the object type analysis model.

[0093] In the present application, the updating of the thermal power generation combustion state classification model by updating the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model includes the following steps.

[0094] S6031: Acquire real-time model parameters of the updated thermal power generation combustion state abnormality analysis model.

[0095] S6033: Process the real-time model parameters based on the parameter update model to obtain target model parameters.

[0096] S6035: Update the thermal power generation combustion state classification model using the target model parameters to obtain the object type parsing model.

[0097] In this application, it means that the real-time model parameters of the thermal power generation combustion state abnormality analysis model can be processed through the parameter update model to obtain the target model parameters; and the target model parameters are used to replace the model parameters in the thermal power generation combustion state classification model to achieve the update of the thermal power generation combustion state classification model and obtain the object type analysis model.

[0098] In some embodiments, the method also includes: after the number of training times reaches a first target number, updating the thermal power generation combustion state classification model based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model, and the number of training times refers to the number of times the thermal power generation combustion state abnormality analysis model is trained using the newly added thermal power generation combustion data examples.

[0099] Exemplarily, after the number of training times reaches the first target number of times, the thermal power generation combustion state classification model is updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model, including: updating the thermal power generation combustion state classification model based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain an updated thermal power generation combustion state classification model; during the training process after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model.

[0100] In the present application, after the number of training times reaches the first target number of times, the thermal power generation combustion state classification model can be updated according to the model parameters of the thermal power generation combustion state abnormality analysis model to obtain an updated thermal power generation combustion state classification model; and during the training process after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated according to the model parameters of the thermal power generation combustion state abnormality analysis model, so that the object type analysis model can be quickly trained.

[0101] In some embodiments, during the training process after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model, including: after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is updated by the model parameters of the thermal power generation combustion state abnormality analysis model every time the second target number of training times passes; after the training is terminated, the updated thermal power generation combustion state classification model is determined as the object type analysis model.

[0102] In the present application, the updated thermal power generation combustion state classification model is updated by the model parameters of the thermal power generation combustion state abnormality analysis model every time the second target number of training times passes, including: obtaining the real-time model parameters of the thermal power generation combustion state abnormality analysis model every time the second target number of training times passes; processing the real-time model parameters based on the parameter update model to obtain target model parameters; and updating the updated thermal power generation combustion state classification model by the target model parameters.

[0103] In the present application, after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is updated by the model parameters of the thermal power generation combustion state abnormality analysis model every time the second target number of training times passes; thereby, a high-accuracy object type analysis model can be quickly trained.

[0104] In some embodiments, the real-time model parameters are processed based on the parameter update model to obtain target model parameters, including: determining a first similarity of a thermal power generation combustion state classification model and a second similarity of a thermal power generation combustion state abnormality analysis model through the number of training times corresponding to the real-time model parameters; the first similarity decreases with an increase in the number of training times, and the second similarity increases with an increase in the number of training times; the real-time model parameters are processed through the first similarity, the second similarity and the parameter update model to obtain the target model parameters.

[0105] Exemplarily, the target loss information is used to correct the model parameters of the thermal power generation combustion state abnormality analysis model until the training termination requirements are met, and the thermal power generation combustion state abnormality analysis model at the time of training termination is used as the updated thermal power generation combustion state abnormality analysis model, including the following steps.

[0106] S6011: Input the newly added thermal power generation combustion data example into the thermal power generation combustion state abnormality analysis model for one training to obtain target loss information, which is determined as initial loss information.

[0107] S6013: Correct the model parameters of the thermal power generation combustion state abnormality analysis model using the initial loss information until the number of training times reaches the first target number, and use the thermal power generation combustion state abnormality analysis model at the end of the training as the updated thermal power generation combustion state abnormality analysis model.

[0108] In the present application, the model parameters of the thermal power generation combustion state abnormality analysis model can be corrected according to the initial loss information until the training times reach the first target times, and the thermal power generation combustion state abnormality analysis model at the end of the training is used as the updated thermal power generation combustion state abnormality analysis model.

[0109] Exemplarily, the updated thermal power generation combustion state abnormality analysis model can be trained using the newly added thermal power generation combustion data examples, and when the number of training times in the training process reaches the second target number, the model parameters of the real-time thermal power generation combustion state abnormality analysis model are obtained; and then the updated thermal power generation combustion state classification model is updated using the model parameters of the real-time thermal power generation combustion state abnormality analysis model to obtain the object type analysis model.

[0110] In the present application, the iterative updating of the updated thermal power generation combustion state abnormality parsing model and the updated thermal power generation combustion state classification model to obtain the object type parsing model includes the following steps.

[0111] S901: Using the updated thermal power generation combustion state abnormality analysis model as the real-time thermal power generation combustion state abnormality analysis model, and using the updated thermal power generation combustion state classification model as the real-time thermal power generation combustion state classification model.

[0112] S903: The real-time thermal power generation combustion state abnormality analysis model is trained using the newly added thermal power generation combustion data example until the number of training times reaches a second target number of times.

[0113] S905: re-using the real-time thermal power generation combustion state abnormality analysis model at the time of training termination as the real-time thermal power generation combustion state abnormality analysis model, and obtaining model parameters of the real-time thermal power generation combustion state abnormality analysis model.

[0114] S907: updating the real-time thermal power generation combustion state classification model using the model parameters of the real-time thermal power generation combustion state abnormality analysis model, and re-using the updated real-time thermal power generation combustion state classification model as the real-time thermal power generation combustion state classification model.

[0115] S909: Repeat the steps of training the real-time thermal power generation combustion state abnormality analysis model through the newly added thermal power generation combustion data example until the number of training times reaches the second target number, updating the real-time thermal power generation combustion state classification model through the model parameters of the real-time thermal power generation combustion state abnormality analysis model, and re-using the updated real-time thermal power generation combustion state classification model as the real-time thermal power generation combustion state classification model until the update termination requirements are met.

[0116] S9011: Determine the real-time thermal power generation combustion state classification model that meets the update termination requirements as the object type parsing model.

[0117] In the present application, the updated thermal power generation combustion state abnormality analysis model is used as the real-time thermal power generation combustion state abnormality analysis model, and the updated thermal power generation combustion state classification model is used as the real-time thermal power generation combustion state classification model; the real-time thermal power generation combustion state abnormality analysis model is trained by the newly added thermal power generation combustion data example until the training times reach the second target times; the real-time thermal power generation combustion state abnormality analysis model at the time of training termination is used again as the real-time thermal power generation combustion state abnormality analysis model, and the model parameters of the real-time thermal power generation combustion state abnormality analysis model are obtained; the real-time thermal power generation combustion state classification model is updated by the model parameters of the real-time thermal power generation combustion state abnormality analysis model, and the updated real-time thermal power generation combustion state abnormality analysis model is used as the real-time thermal power generation combustion state abnormality analysis model. The electric combustion state classification model is re-used as the real-time thermal power generation combustion state classification model; then the step of training the real-time thermal power generation combustion state abnormality analysis model through the newly added thermal power generation combustion data example until the training times reach the second target times and repeating the subsequent steps, when the training times in the training process reach the second target times each time, the model parameters of the real-time thermal power generation combustion state abnormality analysis model are obtained; the updated thermal power generation combustion state classification model is updated through the model parameters of the real-time thermal power generation combustion state abnormality analysis model to obtain the object type analysis model, thereby continuously and alternately updating the model parameters of the learning model and the thermal power generation combustion state classification model during the training process, thereby improving the accuracy of the object type analysis model.

[0118] The prior art simultaneously inputs historical data into a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality analysis model, determines the first loss data based on the regression analysis results output by the two models, obtains an updated thermal power generation combustion state abnormality analysis model based on the first loss data, and then uses only the new data to update the updated learning model, that is, inputs the new data into the updated thermal power generation combustion state abnormality analysis model, determines the second loss data based on the third regression analysis result output by the updated thermal power generation combustion state abnormality analysis model and the object type identifier corresponding to the new data, and obtains the application model based on the second loss data; whereas, in this embodiment, the new data (newly added thermal power generation combustion data examples) are simultaneously input into the thermal power generation combustion state classification model and the learning model, trains the learning model through the output results of the thermal power generation combustion state classification model, and then performs knowledge precipitation on the thermal power generation combustion state classification model through the trained thermal power generation combustion state abnormality analysis model, and finally uses the thermal power generation combustion state classification model as the application model. The specific training process is as follows: the initial object examples are input into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain the first historical type result and the second historical type result, and the example loss information is calculated; when there are new thermal power generation combustion data examples, the new thermal power generation combustion data examples are input into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model obtained in the last training respectively, and the first loss information is determined according to the output results of the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model; according to the first type output of the thermal power generation combustion state classification model The division result example and the newly added object type identification corresponding to the newly added thermal power generation combustion data example are used to obtain a fusion identification example, and then the second loss information is obtained according to the second type division result example and the fusion identification example output by the thermal power generation combustion state abnormality analysis model; finally, the target loss information is determined according to the first loss information and the second loss information; then, the parameters of the thermal power generation combustion state abnormality analysis model are corrected according to the target loss information, and then the knowledge of the thermal power generation combustion state classification model is precipitated through the trained thermal power generation combustion state abnormality analysis model, and finally the thermal power generation combustion state classification model is used as the application model (object type analysis model).

[0119] It can be seen from the technical solutions provided by the above embodiments of this specification that the embodiments of this specification disclose a method for evaluating combustion instability during low-load operation, including: obtaining a newly added thermal power generation combustion data example of an object example; the newly added thermal power generation combustion data example records a newly added object type identifier; based on the initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting the initial thermal power generation combustion data example of the object example into the target model for object type parsing training; the initial thermal power generation combustion data example records the initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality parsing model. The class model performs object type parsing processing to obtain a first type division result example; the first type division result example and the newly added object type identifier are fused to obtain a fusion identifier example; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state abnormality parsing model for object type parsing processing to obtain a second type division result example; the thermal power generation combustion state abnormality parsing model is trained through the comparison result between the second type division result example and the fusion identifier example, and the trained thermal power generation combustion state abnormality parsing model is used to determine the object type parsing model, and the thermal power generation combustion state instability assessment result is determined through the object type parsing model, and the object type parsing model is used to parse the newly added object type identifier in the thermal power generation combustion state data. In the process of model update training, the present application does not need to obtain the initial thermal power generation combustion data example, and only uses the newly added thermal power generation combustion data example to train the initial object type parsing model, which can improve the model performance when learning new data, thereby improving the accuracy and precision of the thermal power generation combustion state assessment.

[0120] The embodiment of this specification also provides an object type parsing method, which includes the following steps.

[0121] S1201: Obtain the thermal power generation combustion state data to be analyzed of the object to be analyzed; in the present application, the object to be analyzed and the object example are objects in the same area, and the thermal power generation combustion state data to be analyzed may be the thermal power generation combustion state data obtained by analyzing the object to be analyzed, or the thermal power generation combustion state data extracted from the data to be analyzed corresponding to the object to be analyzed, and one object to be analyzed may correspond to one or more thermal power generation combustion state data to be analyzed.

[0122] S1203: Inputting the thermal power generation combustion state data to be analyzed into an object type analysis model for object type analysis processing to obtain a target object type of the object to be analyzed; wherein the object type analysis model is trained based on the above-mentioned training method.

[0123] In this application, when the object to be analyzed corresponds to several thermal power generation combustion state data to be analyzed, the target object type of the object to be analyzed can be determined by regressing the analysis results based on the types of the several thermal power generation combustion state data to be analyzed.

[0124] Based on the above, a combustion instability assessment device during low-load operation is provided, comprising:

[0125] A data acquisition module, configured to acquire a newly added thermal power generation combustion data example of an object example; the newly added thermal power generation combustion data example records a newly added object type;

[0126] A model building module is configured to build a thermal power generation combustion state classification model and a thermal power generation combustion state anomaly analysis model based on an initial object type analysis model; the initial object type analysis model is obtained by inputting an initial thermal power generation combustion data example of the object example into a target model for object type analysis training; the initial thermal power generation combustion data example records an initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier;

[0127] a first result obtaining module, configured to input the newly added thermal power generation combustion data example into the thermal power generation combustion state classification model for object type analysis processing to obtain a first type classification result example;

[0128] an identification fusion module, configured to fuse the first type classification result example and the newly added object type identification to obtain a fused identification example;

[0129] A second result obtaining module is configured to input the newly added thermal power generation combustion data example into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a second type classification result example;

[0130] A result evaluation module is used to train the thermal power generation combustion state abnormality analysis model by comparing the results of the second type division result example with the fusion identification example. The trained thermal power generation combustion state abnormality analysis model is used to determine the object type analysis model. The thermal power generation combustion state instability assessment result is determined by the object type analysis model. The object type analysis model is used to parse the newly added object type identification in the thermal power generation combustion state data.

[0131] Based on the above, a combustion instability assessment system during low-load operation is shown, which includes a processor and a memory that communicate with each other. The processor is used to read and execute a computer program from the memory to implement the above method.

[0132] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0133] In summary, based on the above scheme, a newly added thermal power generation combustion data example of the object example is obtained; the newly added thermal power generation combustion data example records the newly added object type identifier; based on the initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting the initial thermal power generation combustion data example of the object example into the target model for object type parsing training; the initial thermal power generation combustion data example records the initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier; the newly added thermal power generation combustion data example is input into the thermal power generation combustion state classification model for object type parsing processing, and a first type An example of a division result; fusing the first type of division result example and the newly added object type identifier to obtain an example of a fusion identifier; inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state abnormality analysis model for object type analysis processing to obtain an example of a second type of division result; training the thermal power generation combustion state abnormality analysis model through the comparison result between the second type of division result example and the fusion identifier example, the trained thermal power generation combustion state abnormality analysis model is used to determine the object type analysis model, and the thermal power generation combustion state instability assessment result is determined through the object type analysis model, and the object type analysis model is used to parse the newly added object type identifier in the thermal power generation combustion state data. In the process of model update training, the present application does not need to obtain the initial thermal power generation combustion data example, and only uses the newly added thermal power generation combustion data example to train the initial object type analysis model, which can improve the model performance when learning new data, thereby improving the accuracy and precision of the thermal power generation combustion state assessment.

[0134] It should be understood that the system and its modules shown above can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer-executable instructions and / or contained in processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the system and its modules of the present application. Not only can hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc. be implemented, they can also be implemented using software executed by various types of processors, and can also be implemented by the above hardware circuits and software (for example, firmware).

[0135] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other possible beneficial effects.

Claims

1. A method for evaluating combustion instability during low-load operation, characterized in that: The method comprises: Acquire a newly added thermal power generation combustion data example of the object example; the newly added thermal power generation combustion data example records a newly added object type; Based on the initial object type parsing model, a thermal power generation combustion state classification model and a thermal power generation combustion state abnormality parsing model are constructed; the initial object type parsing model is obtained by inputting an initial thermal power generation combustion data example of the object example into a target model for object type parsing training; the initial thermal power generation combustion data example records an initial object example type identifier; the initial object example type identifier is different from the newly added object type identifier; Inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state classification model for object type analysis processing to obtain a first type classification result example; Performing fusion processing on the first type classification result example and the newly added object type identifier to obtain a fusion identifier example; Inputting the newly added thermal power generation combustion data example into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a second type classification result example; The thermal power generation combustion state abnormality parsing model is trained by comparing the results of the second type division result example with the fusion identification example. The trained thermal power generation combustion state abnormality parsing model is used to determine the object type parsing model. The thermal power generation combustion state instability assessment result is determined by the object type parsing model. The object type parsing model is used to parse the newly added object type identification in the thermal power generation combustion state data.

2. The method according to claim 1, wherein After acquiring the newly added thermal power generation combustion data example of the object example, the method further includes: Performing data filtering processing on the newly added thermal power generation combustion data examples to obtain a data filtering result, wherein the filtering processing is used to characterize the arrangement of the newly added thermal power generation combustion data examples; inputting the data filtering results into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model respectively to obtain first loss information; The training of the thermal power generation combustion state abnormality analysis model based on the comparison result between the second type classification result example and the fusion identification example includes: Determining second loss information through a comparison result between the second type division result example and the fusion identification example; determining target loss information based on the first loss information and the second loss information; The model parameters of the thermal power generation combustion state abnormality analysis model are corrected using the target loss information.

3. The method according to claim 2, wherein The step of inputting the data filtering results into the thermal power generation combustion state classification model and the thermal power generation combustion state abnormality analysis model to obtain first loss information includes: Inputting the data filtering result into the thermal power generation combustion state classification model for object type analysis processing to obtain a third type classification result example; Inputting the data filtering result into the thermal power generation combustion state abnormality analysis model to perform object type analysis processing to obtain a fourth type classification result example; The first loss information is determined by comparing the fourth type division result example with the third type division result example.

4. The method according to claim 2, wherein The method further comprises: After the number of training times reaches the first target number of times, the thermal power generation combustion state classification model is updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model. The number of training times refers to the number of times the thermal power generation combustion state abnormality analysis model is trained using the newly added thermal power generation combustion data examples.

5. The method according to claim 4, wherein After the number of training times reaches the first target number of times, the thermal power generation combustion state classification model is updated based on the model parameters of the thermal power generation combustion state abnormality parsing model to obtain the object type parsing model, including: updating the thermal power generation combustion state classification model based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain an updated thermal power generation combustion state classification model; During the training process after the number of training times reaches the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated based on the model parameters of the thermal power generation combustion state abnormality analysis model to obtain the object type analysis model.

6. The method according to claim 5, wherein The number of training times During the training process after the first target number of times, the updated thermal power generation combustion state classification model is iteratively updated based on the model parameters of the thermal power generation combustion state abnormality parsing model to obtain the object type parsing model, including: After the number of training times reaches the first target number of times, each time a second target number of training times passes, the updated thermal power generation combustion state classification model is updated using the model parameters of the thermal power generation combustion state abnormality analysis model; After the training is terminated, the updated thermal power generation combustion state classification model is determined as the object type parsing model.

7. The method according to claim 6, wherein The updating of the thermal power generation combustion state classification model is performed by using the model parameters of the thermal power generation combustion state abnormality analysis model after each second target number of training times, including: After each second target number of training times, real-time model parameters of the thermal power generation combustion state abnormality analysis model are obtained; Processing the real-time model parameters based on a parameter update model to obtain target model parameters; The updated thermal power generation combustion state classification model is updated using the target model parameters.

8. The method according to claim 7, wherein The processing of the real-time model parameters based on the parameter update model to obtain target model parameters includes: Determining a first similarity of a thermal power generation combustion state classification model and a second similarity of a thermal power generation combustion state abnormality analysis model based on the number of training times corresponding to the real-time model parameters; wherein the first similarity decreases as the number of training times increases, and the second similarity increases as the number of training times increases; The real-time model parameters are processed using the first similarity, the second similarity, and the parameter update model to obtain the target model parameters.

9. The method according to claim 1, wherein The step of training the initial object type parsing model includes: Acquiring the initial thermal power generation combustion data example of the object example; Inputting the initial thermal power generation combustion data example into the target model for object type parsing processing to obtain an initial type classification result example; Determining example loss information by comparing the initial type classification result example with the initial object example type identifier; The model parameters of the target model are modified using the example loss information until the training termination requirements are met, and the target model at the time of training termination is determined as the initial object type parsing model.

10. A combustion instability assessment system during low-load operation, characterized in that: The method comprises a processor and a memory communicating with each other, wherein the processor is used to read a computer program from the memory and execute the computer program to implement the method according to any one of claims 1 to 9.

Images