A furnace temperature field prediction method and device integrating actual measurement point data and mechanism model

Abstract

The present invention belongs to the technical field related to coal-fired boiler detection. It discloses a furnace temperature field prediction method and device that integrates a mechanism model of actual measurement point data. The method and device include the following steps: completing the installation of a radiation temperature measurement device based on determined furnace radiation temperature measurement points, and obtaining the furnace radiation temperature data of each measurement point online. Then, the current unit operating parameters and furnace radiation temperature data are input into a furnace temperature field machine learning model, and the furnace temperature field machine learning model outputs the furnace temperature field; wherein, a furnace temperature field machine learning model based on a multi-layer perceptron neural network is established, and the furnace temperature field machine learning model is trained using a training data set to obtain the furnace temperature field machine learning model. The present invention solves the problem of low furnace temperature field prediction accuracy of existing machine learning models.

Description

Technical Field

[0001] The present invention belongs to the technical field related to coal-fired boiler detection, and more specifically, relates to a furnace temperature field prediction method and device that integrates actual measurement point data and a mechanism model. Background Art

[0002] While developing new energy technologies aims to reduce reliance on fossil fuels, the instability of renewable energy sources results in insufficient load-shaving capabilities. Consequently, coal-fired generators are required to undertake more peak-shaving tasks to ensure a balanced supply and demand of electricity. Under low load, the proper distribution of the furnace temperature field in coal-fired generator boilers is crucial to their safety, economy, and pollutant emissions. Therefore, online monitoring of the furnace temperature field is essential when coal-fired generators participate in deep peak-shaving operations.

[0003] Combustion numerical simulation techniques based on combustion process mechanism models and numerical solution methods can accurately obtain the furnace temperature field. However, obtaining simulation results often requires several hours of iterative calculations, making it difficult to reflect the distribution of the furnace temperature field in real time. Reference [Lou Chun. Engineering Combustion Diagnostics [M]. Beijing: China Electric Power Press, 2016] describes the practical methods for online monitoring of boiler furnace temperature fields, primarily acoustics, absorption spectroscopy, and thermal radiation imaging. These methods require opening observation holes in the boiler furnace water-cooled wall and installing acoustic signal transceivers, laser signal transceivers, and thermal radiation imaging detectors at the observation holes to acquire the furnace's acoustic, laser, and thermal radiation signals, respectively. However, it should be noted that when the furnace combustion conditions or test environment become adverse, the results obtained using these online temperature field monitoring methods may be distorted. For example, combustion noise can interfere with the transmission of acoustic signals, and ash and slag produced by combustion can contaminate the detectors, affecting signal reception. In addition, it should be noted that due to the limitation of the measurement path, the spatial resolution of the furnace temperature field obtained by these three online monitoring methods is limited. Compared with the combustion numerical simulation technology, they cannot accurately give the fine changes in the temperature distribution of the primary air and secondary air nozzle areas of the burner. These areas are exactly the areas that need to be focused on for stable combustion when the boiler is operating at low load.

[0004] In recent years, with the development of artificial intelligence technology, machine learning has been applied to the prediction of combustion parameters in coal-fired boilers due to its strong learning and generalization capabilities. Reference [Jia Yonghui, Du Jianqiao, Wang Chaoyang, et al. Prediction of temperature distribution of coal-fired boilers based on BP neural network [J]. Thermal Energy and Power Engineering, 2020, 35(07): 130-138.] Using the furnace temperature fields under multiple working conditions calculated by Fluent numerical simulation as training data, a BP neural network based on operating parameters, coordinate position and temperature field data was established and trained to achieve online prediction of the furnace temperature field. The published patent document [Boiler temperature field prediction method combining computational fluid dynamics and deep learning, application number 2022107263905] proposes a boiler temperature field prediction method combining computational fluid dynamics (CFD) and deep learning. This method obtains the temperature field of the boiler during steady-state combustion based on CFD numerical simulation calculations, establishes a temperature field reconstruction neural network, takes the characteristic values ​​of the temperature near the burner and the process parameters reflecting the boiler operating conditions as network inputs, and uses the temperature data of the plane where the burner is located as the network output to complete the neural network training. Then, through the calculation of the neural network, the furnace temperature field at the burner is obtained. The accuracy of this type of online furnace temperature field prediction method based on machine learning models is heavily dependent on the training data set, namely the CFD numerical simulation results. However, the operating parameters of coal-fired boilers include hundreds of parameters such as load, coal feed rate, and air volume. The CFD numerical simulation conditions used for the model training data set are several discrete steady-state conditions and cannot cover all actual operating conditions. When the coal-fired boiler operates under variable conditions, some unlearned operating conditions will inevitably appear. At this time, the machine learning model's prediction results for the furnace temperature field will be seriously deviated. Summary of the Invention

[0005] In response to the above-mentioned defects or improvement needs of the prior art, the present invention provides a furnace temperature field prediction method and device that integrates the actual measurement point data mechanism model, which aims to solve the problem of low prediction accuracy of the furnace temperature field by the existing machine learning model.

[0006] To achieve the above objectives, according to one aspect of the present invention, a furnace temperature field prediction method integrating actual measurement point data with a mechanism model is provided. The prediction method comprises the following steps:

[0007] The radiation temperature measuring device is installed based on the determined furnace radiation temperature measuring points, and the furnace radiation temperature data of each measuring point is obtained online. Then, the current unit operating parameters and the furnace radiation temperature data are input into the furnace temperature field machine learning model, and the furnace temperature field machine learning model outputs the furnace temperature field. The construction of the furnace temperature field machine learning model includes the following steps:

[0008] (1) Determine the number and location of furnace radiation temperature measurement points based on the furnace temperature field distribution characteristics;

[0009] (2) establishing a thermal radiation imaging mechanism model based on the furnace geometric parameters and the positions of the radiation temperature measurement points, obtaining the furnace radiation temperature data of each measurement point based on the thermal radiation imaging mechanism model, and establishing a training data set including the furnace radiation temperature data, unit operating parameters and furnace temperature field data;

[0010] (3) Establish a furnace temperature field machine learning model based on a multi-layer perceptron neural network, and use the training data set to train the furnace temperature field machine learning model.

[0011] Furthermore, each radiation temperature measurement point data T D A mathematical relationship is established between (m) and the furnace temperature field T = {T(n), n = 1, 2, ..., N}, where m = 1, 2, ..., M, and the corresponding formula is:

[0012]

[0013] Where A = {A(m,n)} is the coefficient matrix, which is calculated by the DRESOR method of solving the radiation transfer equation after the furnace geometric parameters and the positions of the radiation temperature measurement points are known.

[0014] Furthermore, the quantitative relationship formula among the unit operating parameters, furnace radiation temperature and furnace temperature field is:

[0015] T(n)=f[T D (1),...,T D (M),R1,R2,...,R P ] (4)

[0016] Where R1, R2,…, R P is the unit operating parameter; the quantitative relationship formula is also the mathematical expression of the furnace temperature field machine learning model.

[0017] Furthermore, the combustion numerical simulation working conditions are set, and the furnace temperature field of each working condition is obtained by CFD numerical simulation. The number and position of the furnace radiation temperature measurement points are determined according to the distribution characteristics of the furnace temperature field.

[0018] Furthermore, a thermal radiation imaging mechanism model is established based on the furnace geometric parameters and the positions of the radiation temperature measurement points. The furnace temperature field of each operating condition obtained by CFD and the number and position of the furnace radiation temperature measurement points are input into the thermal radiation imaging mechanism model to obtain the furnace radiation temperature data of each measurement point. A training data set and a test data set including the furnace radiation temperature data, unit operating parameters and furnace temperature field data are established.

[0019] Furthermore, a furnace temperature field machine learning model based on a multi-layer perceptron neural network was established, and the furnace temperature field machine learning model was trained using the obtained training data set to obtain the quantitative relationship formula between the unit operating parameters, furnace radiation temperature and furnace temperature field, and the furnace temperature field machine learning model was tested using a test data set.

[0020] Furthermore, the installation positions and number of the radiation temperature measuring devices installed on the furnace wall are respectively consistent with the arrangement positions and number of the furnace radiation temperature measuring points determined in step (1).

[0021] Furthermore, the obtained training data set is used to train the MLP neural network model, and the preprocessed parameters are input into the input layer of the MLP neural network model. After nonlinear calculation of the multi-layer hidden layers of the MLP neural network model, the calculation results are output to the output layer of the MLP neural network model. Then, the output vector is reconstructed to obtain the furnace temperature field, and the weights of all neurons in the hidden layer are iteratively updated through the back propagation process to complete the training of the MLP neural network model.

[0022] The present invention also provides a furnace temperature field prediction system that integrates actual measurement point data and a mechanism model, characterized in that: the system includes a memory and a processor, the memory stores a computer program, and the processor executes the furnace temperature field prediction method that integrates the actual measurement point data and a mechanism model when executing the computer program.

[0023] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the furnace temperature field prediction method that integrates the actual measurement point data mechanism model as described above.

[0024] In general, the above technical solutions conceived by the present invention, compared with the prior art, provide a multi-stage layout method and device that considers the number of templates, which has the following beneficial effects:

[0025] 1. In addition to using the key operating parameters of the unit as input data for the machine learning model, radiation temperature measurement point data is also added as input data, which can provide more furnace temperature information for the MLP neural network, thereby improving the prediction ability of the MLP neural network; moreover, based on the thermal radiation imaging mechanism model, the mechanistic relationship between radiation temperature data and furnace temperature field is considered, and the established furnace temperature field machine learning model is more accurate, thereby improving the prediction accuracy of the furnace temperature field.

[0026] 2. The data of the additional furnace radiation temperature measurement points can be obtained using radiation temperature measuring devices such as pyrometers, which are simple to install and easy to implement.

[0027] 3. Establish a furnace temperature field machine learning model based on a multi-layer perceptron neural network, and use the obtained training data set to train the furnace temperature field machine learning model to obtain the quantitative relationship between the unit operating parameters, furnace radiation temperature and furnace temperature field. Use the test data set to test the generalization ability of the furnace temperature field machine learning model and the prediction accuracy of the furnace temperature field to ensure the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a furnace temperature field prediction method that integrates actual measurement point data and a mechanism model provided by the present invention;

[0029] Figure 2 It is a structural diagram of the MLP neural network model;

[0030] Figure 3 Schematic diagram of the arrangement of furnace radiation temperature measurement points in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0032] The furnace of a coal-fired power generation unit boiler is equipped with multiple temperature measurement points. The temperature data obtained from these measurement points can reflect the changes in the combustion temperature field within the furnace. If a mechanistic model linking the measurement point data and the furnace temperature field can be established and integrated into a machine learning model for furnace temperature field prediction, the generalization capability of the machine learning model will be improved, thereby ensuring the accuracy of online furnace temperature field prediction. Therefore, this paper proposes a method for online furnace temperature field prediction by integrating the mechanistic model of actual measurement point temperature data with the machine learning model of the furnace temperature field.

[0033] This method determines the number and location of radiation temperature measurement points on the furnace based on the furnace temperature field results obtained by CFD simulation calculations; secondly, based on the thermal radiation imaging mechanism model, the radiation temperature data of each measurement point is calculated according to the simulated furnace temperature field, and a data set containing radiation temperature data, unit operating parameters and furnace temperature field data is established; then, the furnace temperature field machine learning model is trained and tested using the data set containing radiation temperature measurement point data; finally, the radiation temperature of each measurement point is obtained by installing a radiation temperature measuring device on the furnace, the unit operating parameters and radiation temperature measurement point data are input into the furnace temperature field machine learning model, and the furnace temperature field prediction results are output online.

[0034] in, Figure 1 The implementation process of the present invention includes four steps: determining the furnace radiation temperature measurement point, establishing a data set including the furnace radiation temperature, establishing and training the furnace temperature field machine learning model, and online prediction of the furnace temperature field; Figure 2 This is a structural diagram of the MLP neural network model, including the input layer, hidden layer, output layer and reconstruction layer.

[0035] The furnace temperature field prediction method provided by the present invention, which integrates the actual measurement point data mechanism model, mainly includes the following steps: completing the installation of the radiation temperature measurement device based on the determined furnace radiation temperature measurement points, and obtaining the furnace radiation temperature data of each measurement point online, and then inputting the current unit operating parameters and furnace radiation temperature data into the furnace temperature field machine learning model, and the furnace temperature field machine learning model outputs the furnace temperature field;

[0036] The construction of the furnace temperature field machine learning model includes the following steps:

[0037] (1) Determine the number and location of furnace radiation temperature measurement points based on the distribution characteristics of the furnace temperature field.

[0038] Specifically, the combustion numerical simulation working conditions are set, and the furnace temperature field of each working condition is obtained by CFD numerical simulation. The number and position of the furnace radiation temperature measurement points are determined according to the distribution characteristics of the furnace temperature field.

[0039] In one embodiment, step (1) includes the following sub-steps:

[0040] S11: When using CFD to conduct numerical simulation in the furnace to obtain the furnace temperature field, it is necessary to set the combustion conditions first. The total number of conditions is the combination of various parameters. Combined with the unit design parameters and operating parameters, consider l parameters including load, main combustion air door opening, burnout air door opening, burnout air swing angle, coal mill operation mode, etc. Each parameter considers a situation, so the total number of conditions is a l indivual.

[0041] S12: Divide the furnace into I, J, and K grids along the three-dimensional directions of length, width, and height, with a total of N = I × J × K grids. After the CFD numerical simulation of each working condition is completed, the furnace temperature field T = {T(i,j,k), i = 1, 2, ..., I,j = 1, 2, ..., J,k = 1, 2, ..., K} of each working condition can be obtained. The location of the high-temperature zone of the combustion flame in the furnace is an area of ​​key concern. The present invention proposes to use a method of calculating the center of gravity of the three-dimensional temperature field of the furnace to determine the position X (length direction), Y (width direction), and Z (height direction) of the high-temperature zone of the combustion flame in the furnace in the three-dimensional space of the furnace, as shown in formula (1):

[0042]

[0043] Where, T(x,y,z) is the temperature of each point in the furnace temperature field; x, y, z are the three-dimensional coordinates of the furnace.

[0044] Discretizing formula (1), we can get:

[0045]

[0046] When the combustion conditions set in step S11 change, the position of the high-temperature zone of the combustion flame in the furnace also changes in the length, width, and height directions of the furnace; therefore, the radiation temperature measuring points arranged on the furnace wall must consider the changing characteristics of the high-temperature zone position X, Y, and Z; among them, at least two measuring points are arranged in the horizontal direction formed by the length and width to determine the position changes in the horizontal direction X and Y; the radiation temperature measuring points arranged in the height direction can cover the change range of Z.

[0047] (2) A thermal radiation imaging mechanism model is established according to the furnace geometric parameters and the positions of the radiation temperature measurement points. The furnace radiation temperature data of each measurement point is obtained based on the thermal radiation imaging mechanism model, and a training data set containing the furnace radiation temperature data, unit operating parameters and furnace temperature field data is established.

[0048] Specifically, a thermal radiation imaging mechanism model is established based on the furnace geometric parameters and the positions of the radiation temperature measuring points. The parameters such as the furnace temperature field of each working condition, the number and position of the furnace radiation temperature measuring points obtained by CFD in step (1) are input into the thermal radiation imaging mechanism model to obtain the furnace radiation temperature data of each measuring point, and a training data set and a test data set including the furnace radiation temperature data, the unit operating parameters and the furnace temperature field data are established.

[0049] In one embodiment, step (2) includes the following sub-steps:

[0050] S21: According to the method of step (1), a total of M radiation temperature measurement points are arranged on the furnace wall. Based on the furnace thermal radiation imaging mechanism model, after the furnace geometric parameters and the radiation temperature measurement point positions are determined, each radiation temperature measurement point data T D A mathematical relationship can be established between (m) and the furnace temperature field T = {T(n), n = 1, 2, ..., N}, where m = 1, 2, ..., M, as shown in formula (3):

[0051]

[0052] Where A = {A(m,n)} is the coefficient matrix, which can be calculated by the DRESOR method of solving the radiation transfer equation when the furnace geometric parameters and the positions of the radiation temperature measurement points are known.

[0053] S22: Input the furnace temperature field T of each working condition obtained by CFD in step (1) into formula (3) established based on the thermal radiation imaging mechanism model to obtain the furnace radiation temperature data T of each measuring point. D (m). In terms of constructing a data set for the furnace temperature field machine learning model, the present invention proposes to establish a training data set and a test data set including furnace radiation temperature data, unit operating parameters and temperature field data; wherein, the key operating parameters of the unit (R1, R2, ..., R P ) includes P parameters such as load, water supply, total fuel quantity, total primary air volume, total secondary air volume, main steam flow, etc.

[0054] (3) Establish a furnace temperature field machine learning model based on a multi-layer perceptron neural network, and use the training data set to train the furnace temperature field machine learning model.

[0055] Specifically, a furnace temperature field machine learning model based on a multi-layer perceptron (MLP) neural network is established, and the furnace temperature field machine learning model is trained using the training data set obtained in step (2) to obtain the quantitative relationship between the unit operating parameters, furnace radiation temperature and furnace temperature field. The generalization ability of the furnace temperature field machine learning model and the prediction accuracy of the furnace temperature field are tested using a test data set.

[0056] In one embodiment, step (3) includes the following sub-steps:

[0057] S31: Build an MLP neural network model (furnace temperature field machine learning model) with unit operating parameters and radiation temperature data as input and furnace temperature field data as output. The input of each set of sample data in the data set contains L=M+P parameters, and the output contains N parameters. Since the numerical ranges of the input parameters vary greatly, in order to improve the convergence speed and prediction accuracy of the MLP neural network model, the input parameters are preprocessed. The MLP neural network model is trained using the training data set obtained in step (2). The preprocessed parameters are input into the input layer. After the nonlinear calculation of multiple hidden layers, the calculation results are output to the output layer. The output vector is then reconstructed to obtain the furnace temperature field. The weights of all neurons in the hidden layer are iteratively updated through the back-propagation process to complete the training of the MLP neural network model.

[0058] S32: After training the MLP neural network model, the quantitative relationship between the unit operating parameters, furnace radiation temperature and furnace temperature field can be obtained:

[0059] T(n)=f[T D (1),...,T D (M),R1,R2,...,R P ] (4)

[0060] Then, the input data of the test data set is put into the MLP neural network model to obtain the predicted value of the furnace temperature field, and it is compared with the target value of the furnace temperature field under the corresponding working conditions to test the prediction accuracy and generalization ability of the MLP neural network model.

[0061] Online prediction consists of the following sub-steps:

[0062] S41: When installing the radiation temperature measuring device on the furnace wall, its installation position and number must be consistent with the arrangement position and number of the furnace radiation temperature measuring points determined in step (1). After the radiation temperature measuring device is installed, the radiation temperature data of each measuring point is obtained online.

[0063] S42: Obtain key operating parameters such as load, water supply, and total fuel quantity from the power plant DCS system. Call the MLP neural network model established in step (3) and input the M radiation temperature data and P key operating parameters into the MLP neural network model to output the furnace temperature field distribution, completing the online prediction of the furnace temperature field.

[0064] The present invention also provides a furnace temperature field prediction system that integrates actual measurement point data and a mechanism model. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the furnace temperature field prediction method that integrates actual measurement point data and a mechanism model as described above.

[0065] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the furnace temperature field prediction method that integrates the actual measurement point data mechanism model as described above.

[0066] The present invention is further described in detail below with reference to specific embodiments.

[0067] Example 1

[0068] (1) Determine the furnace radiation temperature measurement point

[0069] A 350MW supercritical boiler combustion numerical simulation was conducted using Fluent. Combining unit design and operating parameters, a probability density analysis was performed on five parameters: load, main combustion damper opening, burnout damper opening, burnout air swing angle, and pulverizer operation mode. Three scenarios were considered for each parameter, resulting in 243 combustion conditions. The furnace model was constructed with a length, width, and height of 14m, 14m, and 34m, respectively. The furnace model was divided into 40, 40, and 40 grids along the length, width, and height directions, for a total of N = 64,000 grids. Numerical calculations for each condition yielded the furnace temperature field distribution T(i,j,k), where i = 1, 2, …, 40; j = 1, 2, …, 40; and k = 1, 2, …, 40. Then, the center of gravity of the three-dimensional temperature field of the furnace is calculated by combining formula (2) to determine the X, Y, and Z positions of the high-temperature zone of the combustion flame in the furnace in the three-dimensional space of the furnace under each combustion condition. According to the position of the high-temperature zone of the combustion flame under each combustion condition, the radiation temperature measurement points are arranged at the heights of 8m and 17m to cover the range of change of the high-temperature zone of the combustion flame in the Z direction. At the same time, in order to determine the position change in the horizontal directions X and Y, four radiation temperature measurement points are arranged on the furnace wall at these two heights, and the horizontal coordinates of the four measurement points are (4.7, 0), (9.3, 14), (14, 4.7), and (0, 9.3). The furnace radiation temperature measurement points are arranged as follows: Figure 3 shown.

[0070] (2) Establish a data set including furnace radiation temperature

[0071] According to the method of step (1), a total of 8 radiation temperature measurement points are arranged on the furnace wall. After the furnace geometric parameters and the positions of the radiation temperature measurement points are determined, the furnace temperature field T of each working condition obtained by Fluent in step (1) is input into the formula (3) established based on the thermal radiation imaging mechanism model to calculate the furnace radiation temperature data T of each measurement point under each working condition. D(m), m=1,2,…,8. Then, a data set including furnace radiation temperature data, unit operating parameters and furnace temperature field data can be constructed, among which the key operating parameters of the unit (R1, R2,…, R P ) includes 6 parameters: load, total fuel quantity, feed water quantity, main steam flow, total primary air volume and total secondary air volume.

[0072] (3) Establishment and training of furnace temperature field machine learning model

[0073] The constructed dataset contains 243 data sets, each with L = 14 inputs and N = 64,000 outputs. 195 of these sets were selected as training data sets, and the remaining 48 sets were used as test data sets. Each set of training data was standardized. An MLP neural network model was then established, using the Adam algorithm with an adaptive learning rate as the optimization algorithm. The Leaky Relu function, optimized from the Relu function, was used as the activation function. The mean squared error (MSE) was used as the loss function, and L2 regularization was added to prevent overfitting. Preprocessed key operating parameters and radiation temperature data were input into the MLP neural network model. After nonlinear calculations in six hidden layers, the temperature field data was output as a one-dimensional vector. After reconstruction, the furnace temperature field distribution was obtained. After 1,000 training iterations, the training loss function was reduced to 0.06122.

[0074] After training the MLP neural network model using the training dataset, a quantitative relationship between the unit's key operating parameters, furnace radiation temperature, and the furnace temperature field was obtained. The generalization ability and prediction accuracy of the MLP neural network model were then verified using a test dataset. The input data for each sample set in the test dataset was standardized and then fed into the MLP neural network model. The predicted values ​​for the furnace temperature field were then output. The predicted values ​​were compared with the target values, and the average relative error of the furnace temperature field predicted by the model was calculated to be 3.924%.

[0075] (4) Online prediction of furnace temperature field

[0076] When installing the radiation temperature measuring device on the furnace wall, its installation location and number are consistent with the location and number of the furnace radiation temperature measurement points determined in step (1). After the radiation temperature measuring device is installed, the radiation temperature data of the eight measurement points can be obtained in real time. Then, key operating parameters such as load, water supply, and total fuel quantity are exported from the power plant DCS system. The key operating parameters and the eight radiation temperature data obtained by the radiation temperature measuring device are input into the established MLP neural network model, and the furnace temperature field can be output online.

[0077] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A furnace temperature field prediction method integrating actual measurement point data and mechanism model, characterized in that: The prediction method includes the following steps: The radiation temperature measuring device is installed based on the determined furnace radiation temperature measuring points, and the furnace radiation temperature data of each measuring point is obtained online. Then, the current unit operating parameters and the furnace radiation temperature data are input into the furnace temperature field machine learning model, and the furnace temperature field machine learning model outputs the furnace temperature field. The construction of the furnace temperature field machine learning model includes the following steps: (1) Determine the number and location of furnace radiation temperature measurement points based on the furnace temperature field distribution characteristics; (2) establishing a thermal radiation imaging mechanism model based on the furnace geometric parameters and the positions of the radiation temperature measurement points, obtaining the furnace radiation temperature data of each measurement point based on the thermal radiation imaging mechanism model, and establishing a training data set including the furnace radiation temperature data, unit operating parameters and furnace temperature field data; (3) Establish a furnace temperature field machine learning model based on a multi-layer perceptron neural network, and use the training data set to train the furnace temperature field machine learning model.

2. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 1 is characterized in that: Each radiation temperature measurement point data T D A mathematical relationship is established between (m) and the furnace temperature field T = {T(n), n = 1, 2, ..., N}, where m = 1, 2, ..., M, and the corresponding formula is: Where A = {A(m,n)} is the coefficient matrix, which is calculated by the DRESOR method of solving the radiation transfer equation after the furnace geometric parameters and the positions of the radiation temperature measurement points are known.

3. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 2 is characterized in that: The quantitative relationship formula among the unit operating parameters, furnace radiation temperature and furnace temperature field is: T(n)=f[T D (1),...,T D (M),R1,R2,...,R P ] (4) Where R1, R2,…, R P is the unit operating parameter; the quantitative relationship formula is also the mathematical expression of the furnace temperature field machine learning model.

4. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 1 is characterized in that: The combustion numerical simulation working conditions are set, and the furnace temperature field of each working condition is obtained by CFD numerical simulation. The number and location of the furnace radiation temperature measuring points are determined according to the distribution characteristics of the furnace temperature field.

5. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 4 is characterized in that: A thermal radiation imaging mechanism model was established based on the furnace geometric parameters and the positions of the radiation temperature measurement points. The furnace temperature field under each operating condition obtained by CFD and the number and positions of the furnace radiation temperature measurement points were input into the thermal radiation imaging mechanism model to obtain the furnace radiation temperature data of each measurement point. A training data set and a test data set including the furnace radiation temperature data, unit operating parameters and furnace temperature field data were established.

6. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 5 is characterized in that: A furnace temperature field machine learning model based on a multi-layer perceptron neural network was established, and the furnace temperature field machine learning model was trained using the obtained training data set. The quantitative relationship formula between the unit operating parameters, furnace radiation temperature and furnace temperature field was obtained, and the furnace temperature field machine learning model was tested using a test data set.

7. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 6 is characterized in that: The installation positions and number of the radiation temperature measuring devices installed on the furnace wall are respectively consistent with the arrangement positions and number of the furnace radiation temperature measuring points determined in step (1).

8. The furnace temperature field prediction method integrating actual measurement point data and a mechanism model according to claim 6 is characterized in that: The MLP neural network model is trained using the obtained training data set. The preprocessed parameters are input into the input layer of the MLP neural network model. After nonlinear calculation of the multi-layer hidden layers of the MLP neural network model, the calculation results are output to the output layer of the MLP neural network model. The output vector is then reconstructed to obtain the furnace temperature field. The weights of all neurons in the hidden layer are iteratively updated through the back propagation process to complete the training of the MLP neural network model.

9. A furnace temperature field prediction system integrating actual measurement point data and mechanism model, characterized by: The system includes a memory and a processor, the memory stores a computer program, and the processor executes the furnace temperature field prediction method integrating the actual measurement point data mechanism model according to any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions prompt the processor to implement the furnace temperature field prediction method that integrates the actual measurement point data mechanism model according to any one of claims 1 to 8.

Images