Gas turbine waste heat boiler heating surface oxidation monitoring method and system

Abstract

The invention discloses a gas turbine waste heat boiler heating surface oxidation monitoring method and system, and belongs to the technical field of gas turbine waste heat boilers. According to the method, internal flow heat transfer and oxidation details and performance evaluation parameters of the gas turbine waste heat boiler are obtained through a mechanism model and numerical simulation, the calculation cost is reduced, and the monitoring precision is improved; and the oxidation condition of the heating surface of the gas turbine waste heat boiler under any working condition is predicted through the multi-task self-attention convolution model, and the safety performance monitoring requirement of the gas turbine waste heat boiler can be met.

Description

Technical Field

[0001] The invention belongs to the technical field of waste heat boilers for gas turbines, and in particular relates to a method and system for monitoring oxidation of a heating surface of a waste heat boiler for a gas turbine. Background Art

[0002] Industrial energy consumption accounts for about 70% of total energy consumption, of which about 60% is recyclable waste heat resources. Gas turbine waste heat boilers are key equipment for waste heat recovery and utilization. They reuse waste heat from the gas turbine outlet to generate hot water or steam for use in other sections. The heat exchanger system composed of multi-layer finned tube heat exchangers is the core component of gas turbine waste heat boilers. The system absorbs waste heat through a large number of finned tube heat exchangers to generate high-temperature steam. Steam parameters are constantly improving. For example, in large gas turbine waste heat boilers, the main steam temperature can reach 600°C and the total length of the finned tubes can reach 300,000 meters.

[0003] Due to the large size of the gas turbine waste heat boiler and the dense fins on the heating surface, the flue gas forms an uneven flow after passing through the inlet flue diffusion section. The heating surface tube wall is prone to peeling and erosion after oxidation under the action of high-temperature flue gas and high-temperature steam. It is difficult to directly monitor the full-scale flow heat transfer and oxidation process of the gas turbine waste heat boiler; Chinese patent CN103672846A uses a neural network to predict the metal wall temperature of the boiler heating surface, and calculates the oxide scale thickness of the heating surface through a heat transfer and oxidation mathematical model. However, this monitoring method has low accuracy and it is difficult to obtain the oxide scale thickness at a specific position of the heating surface; based on this, it is necessary to propose a heating surface oxidation monitoring method for a gas turbine waste heat boiler to meet the safety performance monitoring of the gas turbine waste heat boiler. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the prior art and to provide a method and system for monitoring the oxidation of the heating surface of a waste heat boiler of a gas turbine.

[0005] In order to achieve the above-mentioned invention object, the present invention specifically adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a method for monitoring oxidation of a heating surface of a waste heat boiler of a gas turbine, comprising the following steps:

[0007] S1. Acquire the real-time and historical operation data of the target gas turbine waste heat boiler and perform filtering processing; wherein the real-time and historical operation data of the target gas turbine waste heat boiler include the inlet and outlet temperature, pressure, flow rate of the flue gas of the gas turbine waste heat boiler, the inlet and outlet flue gas temperature and steam temperature data of each heating surface, the historical data sequence of the wall temperature of the heating surface containing time information, the structural parameters of the gas turbine waste heat boiler and the physical property parameters of the working fluid;

[0008] S2. Input the flue gas inlet and outlet temperature, pressure, flow rate, inlet and outlet flue gas temperature and steam temperature data of each heating surface, structural parameters of the waste heat boiler and working fluid physical parameters in the filtered real-time and historical operation data as boundary conditions into the heating surface CFD oxidation simulation model according to the working conditions, and calculate the oxide scale thickness of the heating surface by the heating surface CFD oxidation simulation model to obtain evaluation parameters; wherein the evaluation parameters include the inlet flue non-uniformity coefficient of the waste heat boiler, the heat exchange efficiency and the oxide scale thickness monitoring data at different positions at different times;

[0009] S3. Input the physical parameters and boundary conditions of the gas turbine waste heat boiler to be tested into the trained multi-task self-attention convolution model to obtain the prediction results of the evaluation parameters, and complete the oxidation monitoring of the heating surface of the gas turbine waste heat boiler; wherein, the multi-task self-attention convolution model is trained on a training set, and the training set is constructed by the physical parameters, boundary conditions and evaluation parameters of the gas turbine waste heat boiler under different operating conditions; the multi-task self-attention convolution model is composed of a convolutional neural network and a multi-task attention network, and the multi-task attention network includes a one-dimensional convolution block and a fully connected layer.

[0010] Based on the above scheme, each step can be implemented in the following preferred specific manner.

[0011] As a preferred embodiment of the first aspect, in step S1, the real-time and historical operation data of the target gas turbine waste heat boiler are read from the unit DCS server and the boiler sensor.

[0012] As a preferred embodiment of the first aspect, the specific process of step S2 is as follows:

[0013] S21. Simplify the finned tubes in the heating surface of the waste heat boiler of the gas turbine, establish a simplified geometric model according to the structural parameters of the finned tubes, and perform CFD simulation on the simplified geometric model to obtain the pressure drop and heat transfer correlation of the finned area of ​​the finned tubes, equate the finned area of ​​the finned tubes to porous media, calculate the equivalent porous medium parameters according to the pressure drop and heat transfer correlation of the finned area of ​​the finned tubes, build an equivalent porous medium model according to the equivalent porous medium parameters, and verify the accuracy of the equivalent porous medium model through CFD simulation results: if the CFD simulation results are consistent with the calculation results of the equivalent porous medium model, proceed to the next step; otherwise, recalculate the equivalent porous medium parameters and generate a new equivalent porous medium model until the CFD simulation results are consistent with the calculation results of the equivalent porous medium model;

[0014] S22. Use three-dimensional modeling software to establish an overall geometric model of the waste heat boiler of the gas turbine, and extract the volume of the overall geometric model of the waste heat boiler of the gas turbine to obtain a calculation domain;

[0015] S23, meshing different regions in the calculation domain obtained in S22 respectively, and meshing different calculation domains using a common node division method, so that the number and quality of meshes meet the calculation accuracy requirements;

[0016] S24, importing the grid model obtained after grid division into the CFD fluid calculation platform, inputting physical parameters and boundary conditions to obtain a simulation calculation model, performing steady-state simulation calculation by the simulation calculation model, obtaining the flow field and temperature field inside the gas turbine waste heat boiler when there is no oxidation calculation, and correcting the simulation calculation model by the real-time and historical operation data of the target gas turbine waste heat boiler obtained in S1;

[0017] S25. Calculate the temperature distribution of the steam side wall in the porous medium domain according to the temperature field obtained in S24, calculate the oxide scale thickness at different positions of the tube wall according to the oxidation kinetic equation, update the grid according to the oxide scale thickness, update the calculated wall temperature under the new grid condition, couple the wall temperature and oxide scale thickness to iteratively calculate until the set time is reached, and obtain the evaluation parameters.

[0018] Furthermore, in step S21, the structural parameters of the fin tube include the base tube diameter of the fin tube, the fin spacing and the fin height; the equivalent porous medium parameters include the porosity, thermal conductivity, inertial resistance coefficient and viscous resistance coefficient in the X, Y and Z directions.

[0019] Furthermore, the equivalent porous medium parameters are obtained by the following formula:

[0020]

[0021] Where Δp is the pressure drop per unit length; a 1 and a 2 is the fitting coefficient; v represents the speed; C g is the inertial resistance coefficient in the inertial term; ρ is the fluid density; Δn is the thickness of the porous medium; a n is the viscous drag coefficient; μ is the dynamic viscosity of the fluid.

[0022] Furthermore, in step S22, the calculation domain includes a flue gas domain, a steam domain and a porous medium domain; the flue gas domain is obtained by extracting the volume of the flue gas flow area; the steam domain is obtained by extracting the volume of the steam flow area in the fin tube; and the porous medium domain is established by equivalently establishing the fin area of ​​the fin tube.

[0023] Furthermore, the physical property parameters include flue gas and steam physical property parameters; the boundary conditions include flue gas inlet temperature, flue gas inlet flow rate, steam inlet temperature, steam inlet total flow rate and equivalent porous medium parameters of the waste heat boiler of the gas turbine.

[0024] Furthermore, the calculation method of the non-uniformity coefficient of the inlet flue of the gas turbine waste heat boiler is:

[0025]

[0026] Among them, u i is the velocity of the i-th point in the inlet flue, is the average inlet flue gas velocity.

[0027] As a preferred embodiment of the first aspect, in step S3, the specific processing process in the multi-task self-attention convolution model is as follows:

[0028] S31. Inputting the physical parameters and boundary conditions of the waste heat boiler to be detected into the convolutional neural network, and outputting a reduced-dimensional feature vector by the convolutional neural network;

[0029] S32. Inputting the reduced-dimensional feature vector into the multi-task attention network, in which the reduced-dimensional feature vector is first multiplied by the learnable self-attention weight to obtain a weighted feature vector;

[0030] S33. Input the weighted feature vector into the first one-dimensional convolution block for feature fusion to obtain a first fused feature vector;

[0031] S34. Input the weighted feature vector and the first fused feature vector together into the second one-dimensional convolution block for feature fusion to obtain a second fused feature vector;

[0032] S35. After passing the first fused feature vector through the first fully connected layer, the prediction result of the unevenness coefficient of the flue inlet of the gas turbine waste heat boiler is output; after passing the second fused feature vector through the second fully connected layer, the prediction result of the heat exchange efficiency is output; after passing the second fused feature vector through the third fully connected layer, the prediction result of the oxide scale thickness monitoring data at different positions at different times is output; the prediction result of the evaluation parameter is composed of the prediction result of the unevenness coefficient of the flue inlet of the gas turbine waste heat boiler, the prediction result of the heat exchange efficiency and the prediction result of the oxide scale thickness monitoring data at different positions at different times.

[0033] In a second aspect, the present invention provides a heating surface oxidation monitoring system for a waste heat boiler of a gas turbine, comprising:

[0034] The data acquisition and preprocessing module is used to obtain the real-time and historical operating data of the target gas turbine waste heat boiler and perform filtering processing; wherein, the real-time and historical operating data of the target gas turbine waste heat boiler include the inlet and outlet temperature, pressure, flow rate of the flue gas of the gas turbine waste heat boiler, the inlet and outlet flue gas temperature and steam temperature data of each heating surface, the historical data sequence of the wall temperature of the heating surface containing time information, the structural parameters of the gas turbine waste heat boiler, and the physical property parameters of the working fluid;

[0035] A data storage module is used to store all the relevant data of the gas turbine waste heat boiler, including the real-time and historical operation data after filtering and processing by the data acquisition and preprocessing module;

[0036] A calculation module is used to obtain the real-time and historical operation data after filtering from the data storage module, and input the flue gas inlet and outlet temperature, pressure, flow rate, inlet and outlet flue gas temperature and steam temperature data of each heating surface, structural parameters of the waste heat boiler and physical properties of the working fluid in the real-time and historical operation data after filtering as boundary conditions into the heating surface CFD oxidation simulation model according to the working conditions, and the heating surface CFD oxidation simulation model calculates the oxide scale thickness of the heating surface to obtain evaluation parameters; wherein the evaluation parameters include the inlet flue non-uniformity coefficient of the waste heat boiler, the heat exchange efficiency and the oxide scale thickness monitoring data at different positions at different times;

[0037] A prediction module is used to input the physical parameters and boundary conditions of the gas turbine waste heat boiler to be tested into a trained multi-task self-attention convolution model to obtain the prediction results of the evaluation parameters and complete the oxidation monitoring of the heating surface of the gas turbine waste heat boiler; wherein, the multi-task self-attention convolution model is trained on a training set, and the training set is constructed by the physical parameters, boundary conditions and evaluation parameters of the gas turbine waste heat boiler under different operating conditions; the multi-task self-attention convolution model is composed of a convolutional neural network and a multi-task attention network, and the multi-task attention network includes a one-dimensional convolution block and a fully connected layer.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention performs porous medium equivalence on the fin part of the heating surface with complex structure, retains the flow details on the flue gas side of the heating surface in the gas turbine waste heat boiler, and provides a basis for optimizing the flow field distribution;

[0040] 2. The present invention greatly reduces the calculation cost by performing porous medium equivalence on the actual physical model. The proposed method is applicable to simulation calculations of waste heat boilers of different types of gas turbines;

[0041] 3. The present invention monitors the oxide scale thickness on the steam side according to the inner wall temperature of the heat exchange tube on the heated surface, predicts where oxidation peeling is most likely to occur, and guides subsequent optimization.

[0042] 4. The multi-task self-attention convolution model proposed in the present invention can distribute the weights of the input parameters, improve the model prediction accuracy, take the inlet flue uniformity as the one-dimensional convolution block input of the heat transfer efficiency and oxide scale thickness, consider the influence of the inlet flue shape on the heat exchange system, and improve the model prediction accuracy and optimization accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the steps of the present invention;

[0044] Figure 2 It is a schematic diagram of an equivalent porous medium model of a spiral fin tube heat exchanger of the present invention;

[0045] Figure 3 It is a schematic diagram of the calculation domain of the waste heat boiler of a gas turbine of the present invention;

[0046] Figure 4 A flow chart of flow heat transfer calculation and heated surface oxidation monitoring according to an embodiment of the present invention;

[0047] Figure 5 Schematic diagram of the structure of the multi-task self-attention convolution model of the present invention;

[0048] Figure 6 is a system block diagram of the present invention;

[0049] In the figure: simplified model of spiral fin tube 1, equivalent porous medium model 2, flue gas domain 3, steam domain 4, porous medium domain 5. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation mode of the present invention is described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the present invention can be combined accordingly without conflicting with each other.

[0051] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for the purpose of distinguishing descriptions, and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0052] like Figure 1 As shown, in a preferred implementation of the present invention, the above-mentioned method for monitoring oxidation of heating surface of waste heat boiler of a gas turbine comprises the following steps S1 to S3. The specific implementation process thereof will be described in detail below.

[0053] S1. Obtain the real-time and historical operation data of the target gas turbine waste heat boiler and perform filtering processing. The real-time and historical operation data of the target gas turbine waste heat boiler include the inlet and outlet temperature, pressure, flow rate of the flue gas of the gas turbine waste heat boiler, the inlet and outlet flue gas temperature and steam temperature data of each heating surface, the historical data sequence of the wall temperature of the heating surface containing time information, the structural parameters of the gas turbine waste heat boiler, and the physical property parameters of the working fluid.

[0054] It should be noted that, in step S1 of the present invention, the real-time and historical operating data of the target gas turbine waste heat boiler are read from the unit DCS server and the boiler sensor.

[0055] S2. Input the flue gas inlet and outlet temperature, pressure, flow rate of the waste heat boiler after filtering in the real-time and historical operating data after filtering, the flue gas temperature and steam temperature data of each heating surface inlet and outlet, the waste heat boiler structural parameters and the working fluid physical parameters as boundary conditions into the heating surface CFD oxidation simulation model, and calculate the oxide scale thickness of the heating surface by the heating surface CFD oxidation simulation model to obtain evaluation parameters; wherein, the evaluation parameters include the inlet flue non-uniformity coefficient of the waste heat boiler, the heat exchange efficiency and the oxide scale thickness monitoring data at different positions at different times.

[0056] It should be noted that if Figure 4 As shown, the specific process of step S2 of the present invention is as follows:

[0057] S21. Simplify the finned tubes in the heating surface of the waste heat boiler of the gas turbine, establish a simplified geometric model according to the structural parameters of the finned tubes, and perform CFD simulation on the simplified geometric model to obtain the pressure drop and heat transfer correlation of the fin area of ​​the finned tubes. The fin area of ​​the finned tubes is equivalent to a porous medium, and the equivalent porous medium parameters are calculated by the pressure drop and heat transfer correlation of the fin area of ​​the finned tubes. An equivalent porous medium model is constructed according to the equivalent porous medium parameters, and the accuracy of the equivalent porous medium model is verified by the CFD simulation results: if the CFD simulation results are consistent with the calculation results of the equivalent porous medium model, proceed to the next step; otherwise, recalculate the equivalent porous medium parameters and generate a new equivalent porous medium model until the CFD simulation results are consistent with the calculation results of the equivalent porous medium model.

[0058] Furthermore, in step S21, the structural parameters of the fin tube include the base tube diameter of the fin tube, the fin spacing and the fin height.

[0059] Furthermore, in step S21, the equivalent porous medium parameters include porosity, thermal conductivity, inertial resistance coefficient and viscous resistance coefficient in the X, Y and Z directions.

[0060] In this embodiment, the structural parameters of the finned tube and the corresponding equivalent porous medium parameters are shown in Table 1:

[0061] Table 1 Structural parameters of finned tubes and corresponding equivalent porous media parameters

[0062] Fin height (mm) 16 Fin spacing (mm) 4.5 Viscous drag coefficient (X, Y, Z directions) 2.8E5, 5.9E5, 2.8E5 Inertial drag coefficient (X, Y, Z direction) 36.0、257.4、36.0 Porosity 0.82 Thermal conductivity (W / (m·K)) 57

[0063] It should be noted that, since the fin tube has periodic symmetry, the fin tube can be simplified, and a periodic symmetric boundary condition will be set in the CFD simulation to reduce the calculation cost. In step S21 of this embodiment, the schematic diagram of the equivalent porous medium model constructed is as follows: Figure 2 The structural parameters of the finned tube include the base tube diameter, fin spacing and fin height. After selecting different inlet velocities to perform CFD simulation on the simplified geometric model, the pressure drop and heat transfer coefficient in the X, Y and Z directions are obtained. The equivalent porous medium parameters including the porosity, thermal conductivity, inertial resistance coefficient and viscous resistance coefficient in the X, Y and Z directions are obtained by the following formula:

[0064]

[0065] Where Δp is the pressure drop per unit length; a 1 and a 2 is the fitting coefficient; v represents the speed; C g is the inertial resistance coefficient in the inertial term; ρ is the fluid density; Δn is the thickness of the porous medium; a n is the viscous drag coefficient; μ is the dynamic viscosity of the fluid.

[0066] S22. Execute the overall modeling step: use three-dimensional modeling software to establish an overall geometric model of the waste heat boiler of the gas turbine, and perform volume extraction on the overall geometric model of the waste heat boiler of the gas turbine to obtain a calculation domain.

[0067] Further, in step S22, the calculation domain diagram is as follows Figure 3 As shown, the calculation domain includes a flue gas domain, a steam domain and a porous medium domain; the flue gas domain is obtained by extracting the volume of the flue gas flow area; the steam domain is obtained by extracting the volume of the steam flow area in the fin tube; and the porous medium domain is established by equivalently establishing the fin area of ​​the fin tube.

[0068] S23, executing the meshing step: meshing different areas in the calculation domain obtained in S22 respectively, and meshing between different calculation domains adopts a common node division method so that the number and quality of meshes meet the calculation accuracy requirements.

[0069] S24, execute flow heat transfer calculation steps: import the grid model obtained after grid division into the CFD fluid calculation platform, input physical parameters and boundary conditions to obtain the simulation calculation model, perform steady-state simulation calculation by the simulation calculation model, obtain the flow field and temperature field inside the gas turbine waste heat boiler when there is no oxidation calculation, and modify the simulation calculation model with the real-time and historical operation data of the target gas turbine waste heat boiler obtained in S1.

[0070] Furthermore, in step S24, the physical property parameters include flue gas and steam physical property parameters; the boundary conditions include flue gas inlet temperature, flue gas inlet flow rate, steam inlet temperature, steam inlet total flow rate and equivalent porous medium parameters of the waste heat boiler of the gas turbine.

[0071] In this embodiment, ANSYS-Fluent is selected as the CFD fluid calculation platform, and the RNG ke model is used to simulate turbulent flow. Physical parameters including flue gas and steam physical parameters are input, and the flue gas inlet temperature, flue gas inlet flow rate, steam inlet temperature, steam inlet total flow rate of the gas turbine waste heat boiler and the equivalent porous medium parameters obtained in step S21 are used as boundary conditions to perform steady-state simulation calculations.

[0072] In this embodiment, the boundary conditions are shown in Table 2. In Table 2, the inlet temperature (K) of the high temperature superheater 1 and the inlet temperatures (K) of the high temperature superheaters 2 and 3 belong to the steam inlet temperature, and the total flow rate of the high temperature superheater is used to represent the total flow rate of the steam inlet. Here, the equivalent porous medium parameters are consistent with those in Table 1.

[0073] Table 2 Boundary conditions of gas turbine waste heat boiler

[0074] parameter Numeric Flue gas inlet flow rate (kg / s) 143 Total flow rate of high temperature superheater (kg / s) 30.14 Flue gas inlet temperature (K) 865 High temperature superheater 1 inlet temperature (K) 713 High temperature superheater 2, 3 inlet temperature (K) 518 Viscosity coefficient (X, Y, Z directions) 2.8E5, 5.9E5, 2.8E5 Inertia coefficient (X, Y, Z direction) 36.0、257.4、36.0 Porosity 0.82 Effective thermal conductivity (W / (m·K)) 57

[0075] S25, execute the oxide scale thickness calculation step: calculate the steam side wall temperature distribution in the porous medium domain according to the temperature field obtained in S24, calculate the oxide scale thickness at different positions of the tube wall according to the oxidation kinetic equation, update the grid according to the oxide scale thickness, update the calculated wall temperature under the new grid condition, couple the iterative calculation of the wall temperature and the oxide scale thickness until the set time is reached to obtain the above evaluation parameters.

[0076] Furthermore, the calculation method of the non-uniformity coefficient of the inlet flue of the gas turbine waste heat boiler is:

[0077]

[0078] Among them, u i is the velocity of the i-th point in the inlet flue, is the average inlet flue gas velocity.

[0079] In step S25 of this embodiment, consistent with step S24, this step also uses ANSYS-Fluent as the computing platform, and uses the RNG ke model to simulate turbulent flow. Thus, the monitoring calculation shows that the inlet flue non-uniformity coefficient of the waste heat boiler of the gas turbine under the set working condition is 76.65%, the heat transfer coefficient is 823W / m2·K, and the average oxide scale thickness is 276μm.

[0080] S3. Input the physical parameters and boundary conditions of the gas turbine waste heat boiler to be tested into the trained multi-task self-attention convolution model to obtain the prediction results of the evaluation parameters, and complete the oxidation monitoring of the heating surface of the gas turbine waste heat boiler; the multi-task self-attention convolution model is trained on a training set, and the training set is constructed by the physical parameters, boundary conditions and evaluation parameters of the gas turbine waste heat boiler under different working conditions; the multi-task self-attention convolution model is composed of a convolutional neural network and a multi-task attention network, and the multi-task attention network includes a one-dimensional convolution block and a fully connected layer.

[0081] It should be noted that in step S3, the above steps are performed on the gas turbine waste heat boiler under different working conditions to obtain the uneven coefficient of the inlet flue of the gas turbine waste heat boiler under different working conditions, the heat transfer efficiency and the monitoring data of the oxide scale thickness at different positions at different times, and then the physical parameters and boundary conditions of the gas turbine waste heat boiler under different working conditions are used as input, and the uneven coefficient of the inlet flue of the gas turbine waste heat boiler, the heat transfer efficiency and the monitoring data of the oxide scale thickness at different positions at different times are used as output to construct a data set, and the data set is divided to obtain a training set and a test set. Then, when the training set is used to train the multi-task self-attention convolution model, a conventional training method can be used. The key is to set the corresponding loss function, and to optimize the parameters of the multi-task self-attention convolution model by minimizing the loss function through the gradient descent method, so as to obtain the optimal multi-task self-attention convolution model configuration. Then, the test set is used to evaluate the performance of the multi-task self-attention convolution model. When the error meets the requirements, the multi-task self-attention convolution model is constructed. In addition, in the multi-task self-attention convolutional model constructed above, for any sample input, multiple target parameters can be predicted through convolution and self-attention mechanisms, and its prediction accuracy needs to be improved through continuous parameter optimization of training data.

[0082] It should be noted that in step S3, the trained multi-task self-attention convolution model is used to predict the evaluation parameters of the gas turbine waste heat boiler and the oxide scale thickness monitoring data at different positions of the heating surface under any working conditions. Among them, the input layer parameters of the multi-task self-attention convolution model are flue gas and steam physical parameters, gas turbine waste heat boiler flue gas inlet temperature, flue gas inlet flow, steam inlet temperature, steam inlet total flow and equivalent porous medium parameters, and the output layer is the inlet non-uniformity coefficient, heat exchange efficiency and oxide scale thickness monitoring data at different positions at different times.

[0083] It should be noted that in step S3 of the present invention, the multi-task self-attention convolution model used includes two parts: a convolutional neural network and a multi-task attention network. The internal structure of the convolutional neural network part includes a convolutional layer, a linear rectifier layer (Relu) and a pooling layer. The input vector (physical parameters and boundary conditions of the gas turbine waste heat boiler to be detected) passes through the convolutional layer, the linear rectifier layer and the pooling layer in turn to obtain a reduced-dimensional feature vector. In the convolutional neural network, the main function of the convolutional layer is to extract local features from the input vector. The linear rectifier layer (Relu) follows the convolutional layer, activates the local features through the ReLU function, and then compresses and reduces the dimensions of the local features by the pooling layer, simplifies the computational complexity and extracts the main features of the input vector. The convolutional neural network is connected to the multi-task attention network, and the reduced-dimensional feature vector output by the convolutional neural network is then input into the multi-task attention network. In the multi-task attention network, the reduced-dimensional feature vector obtained by the convolutional neural network is multiplied by the learnable self-attention weight and input into the one-dimensional convolution block for feature fusion, and finally the output result is obtained through the fully connected layer.

[0084] In this embodiment, if Figure 5 As shown, the specific processing process in the multi-task self-attention convolution model is introduced below to facilitate understanding by technical personnel in this field.

[0085] S31. Inputting the physical parameters and boundary conditions of the waste heat boiler to be detected into the convolutional neural network, and outputting a reduced-dimensional feature vector by the convolutional neural network;

[0086] S32. Inputting the reduced-dimensional feature vector into the multi-task attention network, in which the reduced-dimensional feature vector is first multiplied by the learnable self-attention weight to obtain a weighted feature vector;

[0087] S33. Input the weighted feature vector into the first one-dimensional convolution block for feature fusion to obtain a first fused feature vector;

[0088] S34. Input the weighted feature vector and the first fused feature vector together into the second one-dimensional convolution block for feature fusion to obtain a second fused feature vector;

[0089] S35. After passing the first fused feature vector through the first fully connected layer, the prediction result of the unevenness coefficient of the flue inlet of the gas turbine waste heat boiler is output; after passing the second fused feature vector through the second fully connected layer, the prediction result of the heat exchange efficiency is output; after passing the second fused feature vector through the third fully connected layer, the prediction result of the oxide scale thickness monitoring data at different positions at different times is output; the prediction result of the evaluation parameter is composed of the prediction result of the unevenness coefficient of the flue inlet of the gas turbine waste heat boiler, the prediction result of the heat exchange efficiency and the prediction result of the oxide scale thickness monitoring data at different positions at different times.

[0090] It should be noted that in the multi-task attention network, the focus allocation of different features is realized based on the self-attention mechanism. The weights can be continuously learned and updated during training. The higher the weight, the greater the influence of the corresponding feature on the output, so that attention can be focused on the key features. Since the inlet flue non-uniformity coefficient in the output parameters will affect the heat transfer efficiency and the average oxide scale thickness, a multi-task learning strategy is adopted in the design of the multi-task self-attention convolution model, and the first one-dimensional convolution block is used to obtain the prediction results of the inlet flue non-uniformity coefficient of the gas turbine waste heat boiler, and the second one-dimensional convolution block is used to obtain the prediction results of the heat transfer efficiency and the prediction results of the oxide scale thickness monitoring data at different times and positions. In addition, the output results of the first one-dimensional convolution block are added to the input parameters of the second one-dimensional convolution block, and its weight is set to 1.

[0091] In summary, the method of the present invention can accurately monitor the oxide scale thickness on the heating surface during practical application, and its monitoring result is consistent with the actual situation. Therefore, it has a high guiding significance for subsequent research on boiler safety performance.

[0092] It should also be noted that the heating surface oxidation monitoring method of the waste heat boiler of a gas turbine in the above embodiment can essentially be executed by a computer program or module. Therefore, similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a heating surface oxidation monitoring system of a waste heat boiler of a gas turbine corresponding to the heating surface oxidation monitoring method of the waste heat boiler of a gas turbine provided in the above embodiment, such as Figure 6 As shown, it includes:

[0093] The data acquisition and preprocessing module is used to obtain the real-time and historical operating data of the target gas turbine waste heat boiler and perform filtering processing; wherein, the real-time and historical operating data of the target gas turbine waste heat boiler include the inlet and outlet temperature, pressure, flow rate of the flue gas of the gas turbine waste heat boiler, the inlet and outlet flue gas temperature and steam temperature data of each heating surface, the historical data sequence of the wall temperature of the heating surface containing time information, the structural parameters of the gas turbine waste heat boiler, and the physical property parameters of the working fluid;

[0094] A data storage module is used to store all the relevant data of the gas turbine waste heat boiler, including the real-time and historical operation data after filtering and processing by the data acquisition and preprocessing module;

[0095] A calculation module is used to obtain the real-time and historical operation data after filtering from the data storage module, and input the flue gas inlet and outlet temperature, pressure, flow rate, inlet and outlet flue gas temperature and steam temperature data of each heating surface, structural parameters of the waste heat boiler and physical properties of the working fluid in the real-time and historical operation data after filtering as boundary conditions into the heating surface CFD oxidation simulation model according to the working conditions, and the heating surface CFD oxidation simulation model calculates the oxide scale thickness of the heating surface to obtain evaluation parameters; wherein the evaluation parameters include the inlet flue non-uniformity coefficient of the waste heat boiler, the heat exchange efficiency and the oxide scale thickness monitoring data at different positions at different times;

[0096] A prediction module is used to input the physical parameters and boundary conditions of the gas turbine waste heat boiler to be tested into a trained multi-task self-attention convolution model to obtain the prediction results of the evaluation parameters and complete the oxidation monitoring of the heating surface of the gas turbine waste heat boiler; wherein, the multi-task self-attention convolution model is trained on a training set, and the training set is constructed by the physical parameters, boundary conditions and evaluation parameters of the gas turbine waste heat boiler under different operating conditions; the multi-task self-attention convolution model is composed of a convolutional neural network and a multi-task attention network, and the multi-task attention network includes a one-dimensional convolution block and a fully connected layer.

[0097] The above-described embodiment is only a preferred solution of the present invention, but it is not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for monitoring oxidation of heating surface of waste heat boiler of a gas turbine, characterized in that: The following steps are involved: S1. Acquire the real-time and historical operation data of the target gas turbine waste heat boiler and perform filtering processing; wherein the real-time and historical operation data of the target gas turbine waste heat boiler include the inlet and outlet temperature, pressure, flow rate of the flue gas of the gas turbine waste heat boiler, the inlet and outlet flue gas temperature and steam temperature data of each heating surface, the historical data sequence of the wall temperature of the heating surface containing time information, the structural parameters of the gas turbine waste heat boiler and the physical property parameters of the working fluid; S2. Input the flue gas inlet and outlet temperature, pressure, flow rate, inlet and outlet flue gas temperature and steam temperature data of each heating surface, structural parameters of the waste heat boiler and working fluid physical parameters in the filtered real-time and historical operation data as boundary conditions into the heating surface CFD oxidation simulation model according to the working conditions, and calculate the oxide scale thickness of the heating surface by the heating surface CFD oxidation simulation model to obtain evaluation parameters; wherein the evaluation parameters include the inlet flue non-uniformity coefficient of the waste heat boiler, the heat exchange efficiency and the oxide scale thickness monitoring data at different positions at different times; S3. Input the physical parameters and boundary conditions of the gas turbine waste heat boiler to be tested into the trained multi-task self-attention convolution model to obtain the prediction results of the evaluation parameters, and complete the oxidation monitoring of the heating surface of the gas turbine waste heat boiler; wherein, the multi-task self-attention convolution model is trained on a training set, and the training set is constructed by the physical parameters, boundary conditions and evaluation parameters of the gas turbine waste heat boiler under different operating conditions; the multi-task self-attention convolution model is composed of a convolutional neural network and a multi-task attention network, and the multi-task attention network includes a one-dimensional convolution block and a fully connected layer.

2. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine according to claim 1, characterized in that: In step S1, the real-time and historical operating data of the target gas turbine waste heat boiler are read from the unit DCS server and boiler sensor.

3. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 1, characterized in that: The specific process of step S2 is as follows: S21. Simplify the finned tubes in the heating surface of the waste heat boiler of the gas turbine, establish a simplified geometric model according to the structural parameters of the finned tubes, and perform CFD simulation on the simplified geometric model to obtain the pressure drop and heat transfer correlation of the finned area of ​​the finned tubes, equate the finned area of ​​the finned tubes to porous media, calculate the equivalent porous medium parameters according to the pressure drop and heat transfer correlation of the finned area of ​​the finned tubes, build an equivalent porous medium model according to the equivalent porous medium parameters, and verify the accuracy of the equivalent porous medium model through CFD simulation results: if the CFD simulation results are consistent with the calculation results of the equivalent porous medium model, proceed to the next step; otherwise, recalculate the equivalent porous medium parameters and generate a new equivalent porous medium model until the CFD simulation results are consistent with the calculation results of the equivalent porous medium model; S22. Use three-dimensional modeling software to establish an overall geometric model of the waste heat boiler of the gas turbine, and extract the volume of the overall geometric model of the waste heat boiler of the gas turbine to obtain a calculation domain; S23, meshing different regions in the calculation domain obtained in S22 respectively, and meshing different calculation domains using a common node division method, so that the number and quality of meshes meet the calculation accuracy requirements; S24, importing the grid model obtained after grid division into the CFD fluid calculation platform, inputting physical parameters and boundary conditions to obtain a simulation calculation model, performing steady-state simulation calculation by the simulation calculation model, obtaining the flow field and temperature field inside the gas turbine waste heat boiler when there is no oxidation calculation, and correcting the simulation calculation model by the real-time and historical operation data of the target gas turbine waste heat boiler obtained in S1; S25. Calculate the temperature distribution of the steam side wall in the porous medium domain according to the temperature field obtained in S24, calculate the oxide scale thickness at different positions of the tube wall according to the oxidation kinetic equation, update the grid according to the oxide scale thickness, update the calculated wall temperature under the new grid condition, couple the wall temperature and oxide scale thickness to iteratively calculate until the set time is reached, and obtain the evaluation parameters.

4. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 3, characterized in that: In step S21, the structural parameters of the fin tube include the base tube diameter, fin spacing and fin height of the fin tube; the equivalent porous medium parameters include the porosity, thermal conductivity, inertial resistance coefficient and viscous resistance coefficient in the X, Y and Z directions.

5. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 4, characterized in that: The equivalent porous medium parameters are obtained by the following formula: Where Δp is the pressure drop per unit length; a1 and a2 are fitting coefficients; v represents the velocity; C g is the inertial resistance coefficient in the inertial term; ρ is the fluid density; Δn is the thickness of the porous medium; a n is the viscous drag coefficient; μ is the dynamic viscosity of the fluid.

6. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 3, characterized in that: In step S22, the calculation domain includes a flue gas domain, a steam domain and a porous medium domain; the flue gas domain is obtained by extracting the volume of the flue gas flow area; the steam domain is obtained by extracting the volume of the steam flow area in the fin tube; and the porous medium domain is established by equivalently establishing the fin area of ​​the fin tube.

7. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 1 or 3, characterized in that: The physical property parameters include flue gas and steam physical property parameters; the boundary conditions include flue gas inlet temperature, flue gas inlet flow rate, steam inlet temperature, steam inlet total flow rate and equivalent porous medium parameters of the waste heat boiler of the gas turbine.

8. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 1 or 3, characterized in that: The calculation method of the inlet flue non-uniformity coefficient of the gas turbine waste heat boiler is: Among them, u i is the velocity of the i-th point in the inlet flue, is the average inlet flue gas velocity.

9. A method for monitoring oxidation of heating surface of a waste heat boiler of a gas turbine as claimed in claim 1, characterized in that: In step S3, the specific processing process in the multi-task self-attention convolution model is as follows: S31. Inputting the physical parameters and boundary conditions of the waste heat boiler to be detected into the convolutional neural network, and outputting a reduced-dimensional feature vector by the convolutional neural network; S32. Inputting the reduced-dimensional feature vector into the multi-task attention network, in which the reduced-dimensional feature vector is first multiplied by the learnable self-attention weight to obtain a weighted feature vector; S33. Input the weighted feature vector into the first one-dimensional convolution block for feature fusion to obtain a first fused feature vector; S34. Input the weighted feature vector and the first fused feature vector together into the second one-dimensional convolution block for feature fusion to obtain a second fused feature vector; S35. After passing the first fused feature vector through the first fully connected layer, the prediction result of the unevenness coefficient of the flue inlet of the gas turbine waste heat boiler is output; after passing the second fused feature vector through the second fully connected layer, the prediction result of the heat exchange efficiency is output; after passing the second fused feature vector through the third fully connected layer, the prediction result of the oxide scale thickness monitoring data at different positions at different times is output; the prediction result of the evaluation parameter is composed of the prediction result of the unevenness coefficient of the flue inlet of the gas turbine waste heat boiler, the prediction result of the heat exchange efficiency and the prediction result of the oxide scale thickness monitoring data at different positions at different times.

10. A heating surface oxidation monitoring system for a waste heat boiler of a gas turbine, characterized in that: include: The data acquisition and preprocessing module is used to obtain the real-time and historical operating data of the target gas turbine waste heat boiler and perform filtering processing; wherein, the real-time and historical operating data of the target gas turbine waste heat boiler include the inlet and outlet temperature, pressure, flow rate of the flue gas of the gas turbine waste heat boiler, the inlet and outlet flue gas temperature and steam temperature data of each heating surface, the historical data sequence of the wall temperature of the heating surface containing time information, the structural parameters of the gas turbine waste heat boiler, and the physical property parameters of the working fluid; A data storage module is used to store all the relevant data of the gas turbine waste heat boiler, including the real-time and historical operation data after filtering and processing by the data acquisition and preprocessing module; A calculation module is used to obtain the real-time and historical operation data after filtering from the data storage module, and input the flue gas inlet and outlet temperature, pressure, flow rate, inlet and outlet flue gas temperature and steam temperature data of each heating surface, structural parameters of the waste heat boiler and physical properties of the working fluid in the real-time and historical operation data after filtering as boundary conditions into the heating surface CFD oxidation simulation model according to the working conditions, and the heating surface CFD oxidation simulation model calculates the oxide scale thickness of the heating surface to obtain evaluation parameters; wherein the evaluation parameters include the inlet flue non-uniformity coefficient of the waste heat boiler, the heat exchange efficiency and the oxide scale thickness monitoring data at different positions at different times; A prediction module is used to input the physical parameters and boundary conditions of the gas turbine waste heat boiler to be tested into a trained multi-task self-attention convolution model to obtain the prediction results of the evaluation parameters and complete the oxidation monitoring of the heating surface of the gas turbine waste heat boiler; wherein, the multi-task self-attention convolution model is trained on a training set, and the training set is constructed by the physical parameters, boundary conditions and evaluation parameters of the gas turbine waste heat boiler under different operating conditions; the multi-task self-attention convolution model is composed of a convolutional neural network and a multi-task attention network, and the multi-task attention network includes a one-dimensional convolution block and a fully connected layer.

Images