Coke oven whole furnace heat flow assignment method based on computational fluid dynamics method

Abstract

The application provides a coke oven whole-oven heat flow assignment method based on a computational fluid dynamics method, and relates to the technical field of coke oven data processing. The application comprises the following steps: collecting physical structure parameters and production working condition parameters of a coke oven; constructing a target feature unit model based on a target combustion chamber unit and partial regions of carbonization chambers on both sides of the target combustion chamber unit; simulating the working condition of the target feature unit model based on the computational fluid dynamics method to obtain heat flow density distribution data of a heating wall between the combustion chamber and the carbonization chamber at different moments; and establishing a heat flow data time sequence mapping rule according to the running time sequence of the coke oven, which is used to map the heat flow density distribution data to the outer surfaces of the heating walls of all combustion chambers based on a phase difference to form whole-oven dynamic heat flow boundary parameters.

Description

Technical Field

[0001] This invention relates to the field of coke oven data processing technology, specifically a method for assigning coke oven heat flux based on computational fluid dynamics. Background Technology

[0002] Coke ovens are core thermal equipment in the iron and steel metallurgy and coal chemical industries. They convert blended coal into coke, coal gas, and chemical products through high-temperature dry distillation. The quality of the coke directly determines the efficiency of blast furnace ironmaking and the quality of the finished product. Modern large-scale regenerating coke ovens typically consist of dozens of alternating carbonization chambers and combustion chambers. However, research on coke ovens often focuses on local areas such as a single vertical flue, a single pair of double vertical flues, or a single carbonization chamber, completely ignoring the interrelationship of the operation of each combustion chamber. This leads to significant deviations in the prediction of temperature and pyrolysis process across the entire oven, making it difficult to accurately guide production control. If a three-dimensional model of the entire coke oven is used for transient simulation, there are problems such as huge computational load, excessively long calculation cycle, and complex boundary condition settings for the entire oven, which lacks engineering practicality. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to propose a method for assigning the heat flux of the entire coke oven based on computational fluid dynamics, which can solve the technical problems of inaccurate prediction of the heat flux of the entire coke oven and excessive simulation computing power in existing technologies.

[0004] The coke oven used in this invention includes multiple carbonization chambers and multiple combustion chambers, which are arranged alternately. Each combustion chamber is composed of multiple combustion chamber units, and each combustion chamber unit includes a double-linked vertical flue and heating walls on both sides of the double-linked vertical flue.

[0005] Methods for assigning coke oven heat flux based on computational fluid dynamics include:

[0006] Collect physical structural parameters and production condition parameters of the coke oven;

[0007] A target feature unit model is constructed based on the target combustion chamber unit and a portion of the carbonization chamber on both sides of the target combustion chamber unit;

[0008] Based on computational fluid dynamics, the target feature unit model is simulated under working conditions to obtain heat flux density distribution data of the heating wall between the combustion chamber and the carbonization chamber at different times. The heat flux density distribution data includes heat flux data and temperature data.

[0009] Based on the operating sequence of the coke oven, a heat flux data time sequence mapping rule is established. The operating sequence represents the phase difference of the operating cycle between adjacent combustion chambers. The heat flux data time sequence mapping rule is used to map the heat flux density distribution data to the outer surface of the heating wall of all combustion chambers based on the phase difference, forming dynamic heat flux boundary parameters for the entire furnace.

[0010] Furthermore, the simulation of the coke oven's operation during the specified time sequence, including periodically reading the heat flux density distribution data between the left coke oven sub-model, the right coke oven model, and the heating flue sub-model, includes:

[0011] Based on a first fixed time interval, heat flux density distribution data is transferred between the left coke oven sub-model and the heating flue sub-model, and between the right coke oven model and the heating flue model.

[0012] Based on a second fixed time interval, a heating wall reversal heating operation is performed in the heating flue sub-model to ensure that the furnace wall temperature is uniform in the left coke oven sub-model and the right coke oven sub-model that are in contact with the heating wall.

[0013] The second fixed time interval is greater than the first fixed time interval.

[0014] Furthermore, the second fixed time interval is an integer multiple of the first fixed time interval.

[0015] Furthermore, the left coke oven sub-model and the right coke oven sub-model respectively store the heat flux density distribution curve at the interface of the heating wall on the opposite side, and the heating flue sub-model stores two temperature distribution curves at the interface of the heating walls on both sides.

[0016] The transfer of heat flux density distribution data between the left coke oven sub-model and the heating flue sub-model, and between the right coke oven model and the heating flue model, based on a first fixed time interval, includes:

[0017] Based on the first fixed time interval, a two-step model synchronization process is initiated. The heat flux density distribution curves of the left coke oven sub-model and the right coke oven sub-model are read through the heating flue sub-model. The heat flux direction is reversed, and the temperature distribution curves of the opposite side are read through the left coke oven model and the right coke oven model.

[0018] Furthermore, it also includes:

[0019] The initial boundary conditions of the left coke oven model and the right coke oven model are switched multiple times to simulate the operation process of the coke oven in the runtime sequence.

[0020] Furthermore, it also includes:

[0021] Based on different coal parameters, the working conditions of the coal cake pyrolysis process in the carbonization chamber are simulated to obtain the coal cake temperature distribution, volatile matter release rate and coke maturity.

[0022] A database of the entire furnace carbonization chamber is constructed based on the coal cake temperature distribution, the volatile matter emission rate, and the coke maturity data.

[0023] A proxy model is constructed using machine learning algorithms. The input layer of the proxy model includes the coal parameters, the dynamic heat flux boundary parameters of the entire furnace, and the coking time. The output layer of the proxy model includes the coal cake temperature distribution, volatile matter emission rate, and coke maturity. The machine learning algorithm includes one of the following: feedforward neural network, support vector machine, convolutional neural network, recurrent neural network, and random forest.

[0024] The surrogate model is trained using the database of the entire furnace carbonization chamber to obtain an accurate surrogate model;

[0025] The precise proxy model is used as the core prediction module. It is integrated with the three-dimensional geometric model of the entire coke oven carbonization chamber, the time-series mapping module, and the data interaction module to construct a digital twin model of the entire coke oven carbonization chamber. The time-series mapping module is used to align the working cycle of the combustion chamber, and the data interaction module is used to attach the collected heat flow data and temperature data to the three-dimensional geometric model of the entire coke oven carbonization chamber.

[0026] Furthermore, the step of training the surrogate model using the entire furnace carbonization chamber database to obtain an accurate surrogate model includes:

[0027] The entire furnace carbonization chamber database is preprocessed and feature extracted to extract the coal parameters, the dynamic heat flux boundary parameters of the entire furnace, and the coking time, forming a standardized training dataset.

[0028] Based on a preset ratio, the standardized training dataset is divided into a training set, a validation set, and a test set;

[0029] The training set is input into the surrogate model, and iterative training is performed using gradient descent. The parameters of the surrogate model are then updated using the backpropagation algorithm.

[0030] The generalization ability of the proxy model is monitored using the validation set to obtain the trained proxy model;

[0031] The test set is input into the trained proxy model to obtain the prediction results;

[0032] Based on the prediction results, the accuracy of the trained agent model is predicted.

[0033] If the accuracy prediction result meets the standard, the trained proxy model is used as the accurate proxy model;

[0034] If the accuracy prediction result fails to meet the standard, the type of machine learning algorithm is changed, and the proxy model is reconstructed.

[0035] The beneficial effects of adopting the above technical solution are as follows: By constructing characteristic units of the combustion chamber's unit structure, determining the refined heat flux density distribution data of these characteristic units, and establishing a time-series mapping rule based on the temporal pattern of coke oven operation, the refined heat flux density distribution data of the characteristic units is mapped to all combustion units, and subsequently to all combustion chambers, thereby forming dynamic heat flux boundary parameters for the entire furnace from the heating walls of the combustion chambers. Therefore, this invention, through the refined data of the unit model and the time-series mapping method, projects the data to all combustion chambers, ensuring data accuracy while reducing the computational load of the model simulation. Thus, it can solve the technical problems of inaccurate prediction of heat flux assignment for the entire coke oven and excessively high simulation computational costs in related technologies. Attached Figure Description

[0036] Figure 1 This is a top view of a partial coke oven group in an embodiment of the present invention;

[0037] Figure 2 This is a schematic flowchart of a method for assigning heat flux to the entire coke oven based on computational fluid dynamics in an embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram of the coupling of a three-dimensional model in an embodiment of the present invention. Detailed Implementation

[0039] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0040] First, the structure of the coke oven involved in this application will be described. Figure 1 This is a top view of a partial coke oven group in an embodiment of the present invention, such as... Figure 1 As shown, this coke oven is a modern large-scale reheating coke oven, comprising multiple carbonization chambers and multiple combustion chambers. These carbonization chambers and combustion chambers are arranged alternately, separated by heating walls in the combustion chambers, which are typically constructed of silica bricks. Each combustion chamber includes multiple double-connected vertical flues, each flue consisting of two vertical flues connected to a bottom circulation hole via a top cross-hole, forming a structurally consistent and functionally independent combustion and heat exchange unit. Figure 1The heating flue assembly typically includes a double-linked vertical flue, a heat storage chamber, a connecting passage, and a supply system. Since the double-linked vertical flue is the core of the heating flue assembly, and other parts are not specifically involved in this application, the heating flue assembly is used here to refer to the double-linked vertical flue.

[0041] To address the problems existing in the prior art, this invention provides a method for assigning total heat flux in a coke oven based on computational fluid dynamics. Figure 2 This is a schematic flowchart of a method for assigning total heat flux in a coke oven based on computational fluid dynamics, as described in an embodiment of the present invention. Figure 2 As shown, the method may include the following steps:

[0042] Step 1: Collect the physical structure parameters and production condition parameters of the coke oven.

[0043] For example, a coke oven includes a carbonization chamber, a combustion chamber, a double vertical flue for the combustion chamber, a heating wall for the combustion chamber, and overall structural parameters of the coke oven.

[0044] The geometric dimensions of the carbonization chamber include its length, width, height, and taper along the height direction; the geometric dimensions of the combustion chamber include its length, width, and height; the geometric dimensions of the twin vertical flues include the number of flues, flue spacing, flue cross-sectional dimensions, the position and size of the crossing holes, the position and size of the circulation holes, and the size and installation position of the burners at the bottom of each twin vertical flue; the thickness of the heating wall, and the thermal conductivity, specific heat capacity, density, emissivity, and other thermophysical properties of the heating wall material; and the overall structural parameters of the coke oven include the total number of carbonization chambers, the total number of combustion chambers, and the overall length and height of the coke oven.

[0045] Production operating parameters include heating regime parameters, specifically the flow rate of heating gas in the twin-channel flue, heating gas composition, excess air coefficient, and gas / air preheating temperature; coking time parameters, namely the standard coking cycle and the fluctuation range of coking time in actual production; coal charging parameters in the carbonization chamber, including single-hole coal charging amount, coal cake bulk density, initial coal cake temperature, and industrial analysis composition of the coal; operating sequence parameters, including coke pushing plan, phase difference between adjacent combustion chamber cycles, and reversing cycle; and environmental condition parameters, including the external ambient temperature of the coke oven and the temperature of the oven roof space.

[0046] Step 2: Construct a target feature unit model based on the target combustion chamber unit and a portion of the carbonization chamber on both sides of the target combustion chamber unit.

[0047] For example, the above feature unit model is as follows: Figure 2 As shown in the dashed box, the target combustion chamber unit includes a portion of the carbonization chamber on both sides of the target combustion chamber unit. The target combustion chamber unit includes a double-linked vertical flue and heating walls on both sides of the double-linked vertical flue.

[0048] Step 3: Based on computational fluid dynamics, perform operating condition simulation on the target feature unit model to obtain heat flux density distribution data of the heating wall between the combustion chamber and the carbonization chamber at different times. The heat flux density distribution data includes heat flux data and temperature data.

[0049] It should be noted that CFD (Computational Fluid Dynamics) is an important tool for revealing the combustion and heat transfer mechanisms of coke ovens. It is an interdisciplinary field that uses computers and numerical methods to simulate and analyze fluid flow, heat transfer, and related physical phenomena in a virtual environment. It predicts fluid behavior by solving the fundamental physical laws describing fluid motion and is a mature computational method.

[0050] For example, based on the CFD method, a multi-physics model including turbulence, combustion, radiation and heat transfer is used. By changing the boundary conditions of the model, the characteristic unit model is simulated under multiple operating conditions, covering different heating regimes, coking cycles and operating parameters. The heat flow data and temperature data of the heating wall between the coke oven combustion chamber and the carbonization chamber at different times are obtained, thereby obtaining the heat flux density distribution data.

[0051] In some examples, the large amount of heat flux density distribution data can be stored to form a feature cell database for easy management.

[0052] It should be noted that steps 2 and 3 involve constructing a unit model and using mature computational methods to simulate the working conditions of the unit model, thereby obtaining accurate data for the unit.

[0053] In some examples, step 3 above may include:

[0054] Step 3.1: Based on the actual size of the coke oven, construct a three-dimensional geometric model of the target feature unit model. The three-dimensional geometric model includes a left coke oven sub-model, a right coke oven model, and a heating flue sub-model.

[0055] For example, the coke oven model on the left and the coke oven model on the right correspond to... Figure 2 The heating flue sub-model corresponds to a portion of the two carbonization chambers located at opposite top and bottom. Figure 2 The heating flue assembly in the middle.

[0056] It should be noted that the terms "left" and "right" mentioned above are merely for distinction and do not limit the actual left and right.

[0057] Step 3.2, couple the three-dimensional geometric model, set the initial boundary conditions and initial operating conditions. The initial boundary conditions of the left coke oven sub-model are the initial state, the initial boundary conditions of the right coke oven sub-model are the transient solutions of the running sequence, and the initial boundary conditions of the heating flue sub-model are the transient solutions that match the left coke oven sub-model and the right coke oven sub-model.

[0058] It should be noted that coupling the above three-dimensional geometric model refers to coupling the left coke oven model, the right coke oven model, and the heating flue sub-model through multiphysics.

[0059] For example, Figure 3 This is a schematic diagram of the coupling of a three-dimensional model in an embodiment of the present invention, such as... Figure 3 As shown in Figure (a), the initial boundary condition of the coke oven sub-model on the left is the initial state, i.e. the working condition of ambient temperature wet coal. The initial boundary condition of the coke oven sub-model on the right is the transient solution of half a coking cycle, which is a constant boundary condition. In addition, in this embodiment, the running sequence of the coke oven is half a coking cycle. The initial boundary condition of the intermediate heating flue sub-model is the transient solution matching the left and right coke oven models, which is the transient solution at the end of the coking process, and is a constant boundary condition.

[0060] Step 3.3: Simulate the operation process of the coke oven in the running sequence, and periodically read the heat flux density distribution data between the left coke oven sub-model, the right coke oven model and the heating flue sub-model.

[0061] For example, Figure 3 Figures (a), (b), and (c) sequentially illustrate the simulated operation of a coke oven during half a coking cycle. Temperature and heat flow are periodically transferred between the coke oven sub-model on the left and the heating flue sub-model on the right, allowing the heat flux density distribution data at the coupling interface to be read.

[0062] In some examples, step 3.3 above may include:

[0063] Step 3.31: Based on a first fixed time interval, heat flux density distribution data is transferred between the left coke oven sub-model and the heating flue sub-model, and between the right coke oven model and the heating flue sub-model.

[0064] For example, heat flow and temperature data are transferred between the coke oven sub-model and the heating flue sub-model at fixed time intervals to facilitate data recording and processing.

[0065] In some examples, the heat flux density distribution curves at the interface of the heating walls on the left and right sides are stored in the coke oven model and the heating flue model, respectively, and two temperature distribution curves at the interface of the two heating walls are stored in the heating flue model. Step 3.31 above may include:

[0066] Based on the first fixed time interval, a two-step model synchronization process is initiated. The heat flux density distribution curves of the left coke oven sub-model and the right coke oven sub-model are read through the heating flue sub-model. The heat flux direction is reversed, and the temperature distribution curves of the opposite side are read through the left coke oven model and the right coke oven model.

[0067] It should be noted that the two-step model synchronization process is a mechanism for achieving high-precision time synchronization. This mechanism is used in the embodiments of this application to more accurately simulate the operation of the coke ovens on both sides, thereby improving the overall integrity of the feature unit model.

[0068] For example, such as Figure 3 As shown in Figure (b), the two-step model synchronization process is initiated every 3 minutes. The two heat flux density distribution curves output by the left and right coke ovens are read through the heating flue sub-model and the heat flux direction is reversed. At the same time, the temperature distribution curve of the corresponding side heating wall is read through the coke oven sub-model.

[0069] Step 3.32: Based on the second fixed time interval, perform a heating wall reversal heating operation in the heating flue sub-model to ensure that the furnace wall temperature is uniform in the left coke oven sub-model and the right coke oven sub-model that are in contact with the heating wall; wherein, the second fixed time interval is greater than the first fixed time interval.

[0070] For example, in order to ensure uniform temperature of the coke oven walls on both sides in contact with the heating flue, a reversing heating operation is performed on the heating walls at fixed time intervals. During the temperature field switching of the heating walls, the boundary conditions of the corresponding left and right coke oven models are also replaced synchronously.

[0071] It should be noted that, in order to ensure that the conditions do not change during a single data sampling, the heating reversal time should be longer than the sampling time.

[0072] In some examples, the second fixed time interval is an integer multiple of the first fixed time interval.

[0073] For example, the first fixed time interval is 3 minutes and the second fixed time interval is 21 minutes. This ensures that there is no sampling data across the heating direction and improves the quality of the collected data.

[0074] According to some embodiments, the above-mentioned method for assigning coke oven heat flux based on computational fluid dynamics further includes:

[0075] Step 3.4: Switch the initial boundary conditions of the left coke oven model and the right coke oven model multiple times to simulate the operation process of the coke oven in the runtime sequence.

[0076] For example, such as Figure 3 As shown in Figure (c), at the end of half a coking cycle, the initial boundary conditions of the left and right coke oven models are swapped, i.e., the right coke oven model is simulated. This process is repeated multiple times to obtain a stable periodic solution.

[0077] Step 4: Based on the operating sequence of the coke oven, establish a heat flow data time sequence mapping rule. The operating sequence represents the phase difference of the operating cycle between adjacent combustion chambers. The heat flow data time sequence mapping rule is used to map the heat flow density distribution data to the outer surface of the heating wall of all combustion chambers based on the phase difference, forming dynamic heat flow boundary parameters for the whole furnace.

[0078] For example, the above-described runtime sequence represents the phase difference of the operating cycles between adjacent combustion chambers, which is determined by the coke pushing cycle and the number of combustion chambers. The heat flux density distribution data obtained from individual feature units is assigned to the outer surface of the heating wall of the combustion chamber of the entire coke oven according to the corresponding time phase, thereby forming the dynamic heat flux boundary parameters of the entire oven.

[0079] This application proposes a method for assigning overall heat flux to a coke oven based on computational fluid dynamics. By constructing characteristic units of the combustion chamber's unit structure, determining the refined heat flux density distribution data of these characteristic units, and establishing a time-series mapping rule based on the temporal patterns of coke oven operation, the refined heat flux density distribution data of the characteristic units is mapped to all combustion units, and subsequently to all combustion chambers. This allows the formation of dynamic heat flux boundary parameters for the entire oven from the heating walls of the combustion chambers. Therefore, this invention, by projecting refined data from the unit model to all combustion chambers using time-series mapping, ensures data accuracy while reducing the computational burden of model simulation. Thus, it solves the technical problems of inaccurate prediction of overall coke oven heat flux and excessive computational cost in related technologies.

[0080] Digital twin technology is gradually being applied in fields such as aviation, manufacturing, and metallurgy, becoming one of the key technologies to promote the intelligent transformation of industry. In view of this, this application proposes a method for constructing a digital twin model in the context of coke ovens, combining the aforementioned method. This method can accurately predict the temperature distribution of coal cakes and the pyrolysis process, and provide data support for the subsequent intelligent production control of coke ovens.

[0081] According to some embodiments, the above-mentioned method for assigning coke oven heat flux based on computational fluid dynamics further includes:

[0082] Step 5: Based on different coal parameters, simulate the working conditions of the coal cake pyrolysis process in the carbonization chamber to obtain the coal cake temperature distribution, volatile matter release rate and coke maturity.

[0083] For example, CFD methods are used to simulate the pyrolysis process of coal cake in a carbonization chamber under multiple operating conditions. Specifically, the pyrolysis kinetic model can be coupled to the mass source term and energy source term of the coal cake.

[0084] The above pyrolysis kinetic model can be represented as follows:

[0085] ;

[0086] in, Pre-exponential factor, , As the apparent activation energy, , Let be the ideal gas constant. Absolute temperature , Indicates a chemical species. For a moment Product of the Times mass fraction, As time approaches infinity The value of .

[0087] The component decomposition rate and endothermic / exothermic reaction process of the coal cake were calculated using the above pyrolysis kinetic model. The temperature distribution, volatile matter release rate and coke maturity of the coal cake were obtained by calculation using an appropriate physical model (such as a pyrolysis kinetic model or a distributed activation energy model).

[0088] Step 6: Construct a database of the entire furnace carbonization chamber based on the coal cake temperature distribution, the volatile matter emission rate, and the coke maturity data.

[0089] Step 7: Construct a proxy model using a machine learning algorithm. The input layer of the proxy model includes the coal parameters, the dynamic heat flux boundary parameters of the entire furnace, and the coking time. The output layer of the proxy model includes the coal cake temperature distribution, the volatile matter emission rate, and the coke maturity. The machine learning algorithm includes one of the following: feedforward neural network, support vector machine, convolutional neural network, recurrent neural network, and random forest.

[0090] For example, based on the strong nonlinearity and spatiotemporal coupling characteristics of the coke oven coal cake pyrolysis process, a surrogate model is constructed by selecting one of the following: feedforward neural network, support vector machine, convolutional neural network, recurrent neural network, and random forest. The surrogate model uses coal parameters, the above-mentioned dynamic heat flow boundary parameters of the whole furnace, and coking time as the model input layer, and coal cake temperature distribution, volatile matter release rate, and coke maturity as the model output layer.

[0091] Step 8: Train the surrogate model using the whole furnace carbonization chamber database to obtain an accurate surrogate model.

[0092] It should be noted that the whole furnace carbonization chamber database is established based on the reliable computing method of CFD. Using this whole furnace carbonization chamber database as a training database can improve the accuracy of the constructed neural network model.

[0093] In some examples, step 8 may include:

[0094] Step 8.1: Perform preprocessing and feature extraction on the whole furnace carbonization chamber database to extract the coal parameters, the whole furnace dynamic heat flow boundary parameters and the coking time to form a standardized training dataset.

[0095] For example, the above-mentioned data from the entire coking chamber is preprocessed to remove outliers and missing values. The input features and output labels are normalized and mapped to the [0,1] interval to eliminate dimensional differences. Feature engineering is then performed in conjunction with the physical logic of coke oven production to extract key features such as coal parameters, dynamic heat flow boundary parameters, and coking time. Core input variables are then selected to form a standardized training dataset.

[0096] Step 8.2: Based on a preset ratio, the standardized training dataset is divided into a training set, a validation set, and a test set.

[0097] For example, a standardized training dataset can be divided into a training set, a validation set, and a test set in a ratio of 7:2:1.

[0098] Step 8.3: Input the training set into the surrogate model, perform iterative training using gradient descent, and update the parameters of the surrogate model using the backpropagation algorithm.

[0099] Step 8.4: Use the validation set to monitor the generalization ability of the proxy model to obtain the trained proxy model.

[0100] It should be noted that overfitting of the model should be prevented.

[0101] Step 8.5: Input the test set into the trained proxy model to obtain the prediction results;

[0102] Step 8.6 Based on the prediction results, perform accuracy prediction on the trained agent model;

[0103] Step 8.6.1: If the accuracy prediction result meets the standard, the trained proxy model is used as the accurate proxy model.

[0104] Step 8.6.2: If the accuracy prediction result does not meet the standard, change the type of the machine learning algorithm and rebuild the surrogate model.

[0105] For example, the coefficient of determination (R²) can be used as an evaluation metric to verify the model's prediction accuracy. If the model's prediction accuracy meets the standard, the surrogate model is used as the accurate model for the next step. If the model's prediction accuracy does not meet the standard, for example, R² ≥ 0.95, then return to step 7, change the algorithm type, and rebuild the surrogate model.

[0106] In some examples, multiple evaluation metrics can be set to validate the model, such as mean absolute error (MAE) and root mean square error (RMSE).

[0107] Step 9: Using the surrogate model as the core prediction module, integrate it with the three-dimensional geometric model of the entire coke oven carbonization chamber, the time-series mapping module, and the data interaction module to construct a digital twin model of the entire coke oven carbonization chamber.

[0108] For example, the timing mapping module is used to align the working cycle of the combustion chamber, and the data interaction module is used to attach the collected heat flow data and temperature data to the three-dimensional geometric model of the entire coke oven carbonization chamber.

[0109] In some examples, computer graphics technology can be further used to build a digital twin visualization system for the entire coke oven, enabling real-time dynamic display of coal cake temperature distribution and pyrolysis process.

[0110] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for assigning total heat flux in a coke oven based on computational fluid dynamics, characterized in that, The coke oven includes multiple carbonization chambers and multiple combustion chambers, which are arranged alternately. Each combustion chamber is composed of multiple combustion chamber units, and each combustion chamber unit includes a double-linked vertical flue and heating walls on both sides of the double-linked vertical flue. The method for assigning the total heat flux of the coke oven includes: Collect physical structural parameters and production condition parameters of the coke oven; A target feature unit model is constructed based on the target combustion chamber unit and a portion of the carbonization chamber on both sides of the target combustion chamber unit; Based on computational fluid dynamics, the target feature unit model is simulated under working conditions to obtain heat flux density distribution data of the heating wall between the combustion chamber and the carbonization chamber at different times. The heat flux density distribution data includes heat flux data and temperature data. Based on the operating sequence of the coke oven, a heat flux data time sequence mapping rule is established. The operating sequence represents the phase difference of the operating cycle between adjacent combustion chambers. The heat flux data time sequence mapping rule is used to map the heat flux density distribution data to the outer surface of the heating wall of all combustion chambers based on the phase difference, forming dynamic heat flux boundary parameters for the entire furnace.

2. The method for assigning coke oven heat flux based on computational fluid dynamics as described in claim 1, characterized in that, The operation simulation of the target feature unit model based on computational fluid dynamics includes: Based on the actual dimensions of the coke oven, a three-dimensional geometric model of the target feature unit model is constructed. The three-dimensional geometric model includes a left coke oven sub-model, a right coke oven model, and a heating flue sub-model. The three-dimensional geometric model is coupled, and initial boundary conditions and initial operating conditions are set. The initial boundary conditions of the left coke oven sub-model are the initial state, the initial boundary conditions of the right coke oven model are the transient solutions of the running sequence, and the initial boundary conditions of the heating flue sub-model are the transient solutions that match the left coke oven model and the right coke oven model. The operation of the coke oven in the specified operating sequence is simulated by periodically reading the heat flux density distribution data between the left coke oven sub-model, the right coke oven sub-model, and the heating flue sub-model.

3. The method for assigning coke oven heat flux based on computational fluid dynamics as described in claim 2, characterized in that, The simulation of the coke oven's operation during the specified time sequence includes periodically reading the heat flux density distribution data between the left coke oven sub-model, the right coke oven model, and the heating flue sub-model, including: Based on a first fixed time interval, heat flux density distribution data is transferred between the left coke oven sub-model and the heating flue sub-model, and between the right coke oven model and the heating flue model. Based on a second fixed time interval, a heating wall reversal heating operation is performed in the heating flue sub-model to ensure that the furnace wall temperature is uniform in the left coke oven sub-model and the right coke oven sub-model that are in contact with the heating wall. The second fixed time interval is greater than the first fixed time interval.

4. The method for assigning total heat flux of a coke oven based on computational fluid dynamics as described in claim 3, characterized in that, The second fixed time interval is an integer multiple of the first fixed time interval.

5. The method for assigning coke oven heat flux based on computational fluid dynamics as described in claim 3, characterized in that, The left coke oven model and the right coke oven model respectively store the heat flux density distribution curve at the interface of the heating wall on the opposite side, and the heating flue sub-model stores two temperature distribution curves at the interface of the heating walls on both sides. The transfer of heat flux density distribution data between the left coke oven sub-model and the heating flue sub-model, and between the right coke oven model and the heating flue model, based on a first fixed time interval, includes: Based on the first fixed time interval, a two-step model synchronization process is initiated. The heat flux density distribution curves of the left coke oven sub-model and the right coke oven sub-model are read through the heating flue sub-model. The heat flux direction is reversed, and the temperature distribution curves of the opposite side are read through the left coke oven model and the right coke oven model.

6. The method for assigning total heat flux of a coke oven based on computational fluid dynamics as described in claim 3, characterized in that, Also includes: The initial boundary conditions of the left coke oven model and the right coke oven model are switched multiple times to simulate the operation process of the coke oven in the runtime sequence.

7. The method for assigning coke oven heat flux based on computational fluid dynamics according to any one of claims 1 to 6, characterized in that, Also includes: Based on different coal parameters, the working conditions of the coal cake pyrolysis process in the carbonization chamber are simulated to obtain the coal cake temperature distribution, volatile matter release rate and coke maturity. A database of the entire furnace carbonization chamber is constructed based on the coal cake temperature distribution, the volatile matter emission rate, and the coke maturity data. A proxy model is constructed using machine learning algorithms. The input layer of the proxy model includes the coal parameters, the dynamic heat flux boundary parameters of the entire furnace, and the coking time. The output layer of the proxy model includes the coal cake temperature distribution, volatile matter emission rate, and coke maturity. The machine learning algorithm includes one of the following: feedforward neural network, support vector machine, convolutional neural network, recurrent neural network, and random forest. The surrogate model is trained using the database of the entire furnace carbonization chamber to obtain an accurate surrogate model; The precise proxy model is used as the core prediction module. It is integrated with the three-dimensional geometric model of the entire coke oven carbonization chamber, the time-series mapping module, and the data interaction module to construct a digital twin model of the entire coke oven carbonization chamber. The time-series mapping module is used to align the working cycle of the combustion chamber, and the data interaction module is used to attach the collected heat flow data and temperature data to the three-dimensional geometric model of the entire coke oven carbonization chamber.

8. The method for assigning coke oven heat flux based on computational fluid dynamics as described in claim 7, characterized in that, The step of training the surrogate model using the whole furnace carbonization chamber database to obtain an accurate surrogate model includes: The entire furnace carbonization chamber database is preprocessed and feature extracted to extract the coal parameters, the dynamic heat flux boundary parameters of the entire furnace, and the coking time, forming a standardized training dataset. Based on a preset ratio, the standardized training dataset is divided into a training set, a validation set, and a test set; The training set is input into the surrogate model, and iterative training is performed using gradient descent. The parameters of the surrogate model are then updated using the backpropagation algorithm. The generalization ability of the proxy model is monitored using the validation set to obtain the trained proxy model; The test set is input into the trained proxy model to obtain the prediction results; Based on the prediction results, the accuracy of the trained agent model is predicted. If the accuracy prediction result meets the standard, the trained proxy model is used as the accurate proxy model; If the accuracy prediction result fails to meet the standard, the type of machine learning algorithm is changed, and the proxy model is reconstructed.

Images