Method and system for optimizing and regulating the hydrogen-rich gas combustion state of a blast furnace raceway

Abstract

This invention discloses a method and system for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone, relating to the field of blast furnace smelting technology. This invention constructs a basic correlation mechanism model by monitoring physical, chemical, and furnace condition field correlation data in the swirl zone, combined with hydrogen-rich combustion kinetic equations and fluid dynamic characteristics. The mechanism model is then corrected in real time using an LSTM network to generate a hybrid prediction model combining mechanism and data. A multi-objective optimization function is set, targeting combustion efficiency, temperature field uniformity, and furnace condition adaptability, and solved using the NSGA-III algorithm to obtain a dynamic prediction model for hydrogen-rich combustion. Based on a graded control strategy, and combined with a PID algorithm, the deviation between predicted and measured values ​​is compared, and the control parameters are dynamically corrected in stages to achieve precise optimization of the hydrogen-rich combustion state in the blast furnace swirl zone. This method and system effectively solve problems such as low combustion efficiency, uneven temperature distribution, and unstable furnace conditions in the original process by optimizing and controlling the injection pipe and injected gas, and has broad application value.

Description

Technical Field

[0001] This invention relates to the field of blast furnace smelting technology, specifically to a method and system for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone, which integrates hydrogen-rich gas combustion dynamics, fluid mechanics, time-series neural networks, and multi-objective optimization algorithms, with the blast furnace swirl zone as the core of regulation. Background Technology

[0002] As hydrogen-rich blast furnace smelting becomes a core direction for the green transformation of the steel industry due to its significant carbon reduction advantages, related technologies have gradually entered the industrial demonstration stage. The tuyeres swirl zone is the only oxidizing area in the blast furnace; after injecting hydrogen-rich gas, the hydrogen-rich gas... and edging Mixed here, a violent combustion reaction occurs, if Unable to When the pyrolysis temperature is reached through complete combustion, carbon is produced during the pyrolysis process. Since the gas velocity at the tuyere can reach over 200 m / s, the carbon produced during pyrolysis will leave the combustion zone with the gas flow. This not only significantly reduces the heat gain of the furnace hearth, but also deteriorates the furnace hearth operating conditions due to the large amount of activated carbon mixed with the slag. Therefore, it is necessary to study the operating conditions of the tuyere swirl zone and the impact of hydrogen-rich injection parameters on... The effects of combustion, thus providing Favorable conditions are created for full combustion in the ventilated area, and the relevant technologies have gradually entered the industrial demonstration stage.

[0003] Chinese patent (publication number CN115341057A) discloses a blast furnace hydrogen-rich smelting system and method, which produces hydrogen by electrolyzing water using renewable energy and preheating it for injection, thereby achieving basic emission reduction. However, the method has insufficient accuracy in predicting the combustion state and does not combine the temporal characteristics and multi-field coupling law of hydrogen-rich combustion, making it difficult to accurately match the dynamic correlation between hydrogen concentration, hot air parameters and combustion temperature and flame length.

[0004] Chinese patent (publication number CN115522003A) discloses a hydrogen-rich blast furnace ironmaking system based on energy-mass conversion and its production control method, which achieves hydrogen-rich injection by integrating hydrogen production through water electrolysis with the blast furnace; however, the method does not design a multi-objective optimization scheme for the core reaction area of ​​the swirl zone, and cannot take into account combustion efficiency, temperature field uniformity and furnace condition adaptability.

[0005] In summary, existing technical solutions are slow to respond to fluctuations in hydrogen concentration and changes in furnace conditions, which can easily lead to unstable combustion and low waste heat utilization efficiency. A method that can solve the core challenges of changes in furnace temperature distribution and insufficient hydrogen utilization after hydrogen-rich injection has become a key need for the industry. Summary of the Invention

[0006] Based on the aforementioned technical problems, this application discloses a method for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone, specifically including:

[0007] By monitoring the correlation data of the physical field, chemical field and furnace condition field in the blast furnace swirling zone, and based on the chemical reaction kinetic equation of hydrogen-rich combustion and the hydrodynamic characteristics of the swirling zone, a basic correlation mechanism model of combustion temperature, flame length and hydrogen concentration and hot blast parameters is established.

[0008] Based on the associated data, the mechanistic model is corrected in real time using LSTM to obtain a hybrid prediction model that combines the mechanism and the data.

[0009] A multi-objective optimization function was defined, and the objectives were solved using the non-dominated sorting genetic algorithm NSGA-Ⅲ to obtain a dynamic prediction model for hydrogen-rich combustion.

[0010] Based on a graded control strategy, the jet velocity, gas jet angle, jet point distribution and hydrogen-rich gas mixing ratio of the jet pipe are dynamically adjusted.

[0011] Collect and control data, compare the deviation with the predicted values ​​of the hybrid prediction model, and dynamically correct the control parameters through the PID algorithm.

[0012] Preferably, the basic correlation mechanism model is specifically: based on the chemical reaction kinetics equation of hydrogen-rich combustion, the hydrodynamic equation of the swirling zone, and the heat transfer equation, the model is constructed through parameter correlation, wherein the input parameters are hydrogen concentration, hot air parameters, and jet velocity of the blowpipe, and the output parameters are combustion temperature and flame length;

[0013] Hydrogen concentration is directly substituted into the combustion reaction rate equation as the core influencing factor of the reaction rate; hot air temperature is used as the initial boundary condition for combustion temperature; hot air pressure is used to calculate the axial pressure gradient in the fluid dynamics equation; hot air velocity and jet velocity from the blowpipe jointly determine the axial velocity after mixing; the core equation is solved by a sequential coupling method.

[0014] By introducing a turbulence correction factor, the effective viscosity coefficient is corrected to reflect the influence of airflow disturbance on velocity distribution; by introducing a radiation heat transfer correction factor into the heat transfer equation, the emissivity of the furnace charge surface is corrected, and the accuracy of temperature calculation is improved.

[0015] Preferably, the solution of the core equation using the sequential coupling method specifically involves substituting the input parameters—hydrogen concentration, hot air parameters, and jet velocity—into the fluid dynamics equations to obtain the axial velocity distribution in the swirling zone, as shown in the formula:

[0016]

[0017] in, The axial velocity in the vortex region, For hydrogen concentration, For axial pressure gradient, The radius of the cyclone region, Radial coordinates, The effective viscosity coefficient, The jet velocity;

[0018] The volume of the combustion reaction zone is determined based on the axial velocity distribution. Substituting the total heat release into the heat transfer equilibrium equation, the combustion temperature can be obtained by solving the equation. Substituting the hydrogen concentration and reaction zone volume into the combustion reaction rate equation, the H2 consumption rate and total heat release are calculated using the following formula:

[0019]

[0020] in, The reaction is exothermic. The rate of H2 consumption reaction. for, For the total heat release, the Stefan-Boltzmann constant is given. For the surface emissivity of the furnace charge, This refers to the radiative heat exchange area between the combustion zone and the furnace charge. The surface temperature of the furnace charge;

[0021] Based on the axial flow velocity and combustion reaction rate, the flame length is calculated by integration, using the following formula:

[0022]

[0023]

[0024] in, The length of the flame. axial The cross-sectional area of ​​the reaction zone at that location, The preset H2 conversion rate threshold, The initial hydrogen concentration, Input volumetric flow rate for hydrogen-rich gas.

[0025] Preferably, the hybrid prediction model specifically involves: preprocessing the collected associated data to obtain time-series samples, designing a multi-layer LSTM network based on the time-series characteristics of hydrogen-rich combustion; inputting the preprocessed time-series samples into the LSTM network, updating the network parameters through the backpropagation algorithm until the loss function converges; and evaluating the model performance using validation set data after training.

[0026] The real-time monitored correlation data is synchronously input into the basic correlation mechanism model and the LSTM correction model. The basic correlation mechanism model calculates the predicted combustion temperature T and the predicted flame length L at the current moment based on the real-time input data.

[0027] The LSTM correction model, based on the same real-time time series input data, predicts the temperature deviation ΔT and length deviation ΔL at the current moment.

[0028] By combining the mechanism output with the LSTM prediction bias correction logic, the predicted value of the target combustion state is calculated, forming a hybrid prediction model that combines mechanism and data. The formula is as follows:

[0029]

[0030]

[0031] At regular intervals, the output value of the hybrid prediction model is compared with the real value collected by the perception system. When the continuous deviation is too large, the latest collected time series data is added to the training set to incrementally train the LSTM correction model, update the network parameters, and achieve adaptive optimization of the model.

[0032] Preferably, the hydrogen-rich combustion dynamic prediction model specifically comprises: based on the combustion state parameters output by the hybrid prediction model, defining the optimization objectives and constraints; searching for the Pareto optimal solution set that satisfies all objectives using the NSGA-Ⅲ algorithm; integrating the model parameters corresponding to the optimal solution into the hybrid prediction model to construct a dynamic prediction model that combines high-precision prediction with multi-objective optimization characteristics.

[0033] Preferably, the specific optimization objectives and constraints are as follows: determining the combustion efficiency based on the ratio of the actual amount of hydrogen participating in combustion to the total amount of input hydrogen and setting a target threshold for combustion efficiency; determining the temperature field uniformity based on the reciprocal of the temperature difference between the core area and the edge area of ​​the swirling zone and setting a target threshold for temperature field uniformity; determining the furnace condition adaptability based on the fluctuation of the material column permeability and setting a maximum threshold for fluctuation.

[0034] The optimization objectives are determined by maximizing combustion efficiency, maximizing temperature field uniformity, and achieving optimal furnace condition adaptability.

[0035] Set constraints on hydrogen concentration, temperature, unburned hydrogen concentration, and control parameters.

[0036] Preferably, the construction of a dynamic prediction model that combines high-precision prediction and multi-objective optimization features specifically involves: normalizing each optimization objective to eliminate dimensional differences; setting weight coefficients for each objective based on industrial needs, and constructing a maximization objective function, the formula of which is:

[0037]

[0038] in, To optimize the variable vector, , , These are the normalized combustion efficiency, temperature uniformity, and furnace condition adaptability, respectively. , , The weights for combustion efficiency, temperature uniformity, and furnace condition adaptability are respectively.

[0039] The non-dominated sorting genetic algorithm NSGA-III is used to solve the constructed multi-objective optimization function. Its core advantage is that it can handle multi-objective conflict problems and obtain a uniformly distributed Pareto optimal solution set. ;

[0040] The optimal solution set The corresponding optimization variables serve as the dynamic baseline parameters of the hybrid prediction model. The prediction parameters are dynamically adjusted in conjunction with real-time monitoring data to obtain the dynamic prediction model for hydrogen-rich combustion.

[0041] Preferably, the graded control strategy specifically involves: setting a preset hydrogen concentration fluctuation threshold. Temperature field temperature difference threshold Combustion efficiency and furnace condition adaptability threshold Based on the output of the dynamic prediction model for hydrogen-rich combustion and multi-dimensional sensing data, key parameters are adjusted according to the hierarchical logic of rapid response, optimized adaptation, and long-term adaptation.

[0042] Among them, the first-level regulation rapid response is in response to hydrogen concentration fluctuations. or temperature field temperature difference It is triggered in time, and the injection flow rate of the injection pipe is adjusted in real time through the electromagnetic proportional valve, and the hot air oxygen supply ratio is adjusted simultaneously to quickly maintain combustion stability.

[0043] The results of the secondary control optimization and adaptation to the combustion efficiency and furnace condition adaptability of the hybrid prediction model output. The system is triggered in time to adjust the injection angle and injection point distribution of hydrogen-rich gas, thereby achieving temperature field homogenization.

[0044] The three-level control system is adapted to the long-term needs of the furnace by dynamically adjusting the hydrogen-rich gas mixing ratio based on furnace condition trend data, thus establishing an adaptation relationship between hydrogen concentration and furnace condition.

[0045] Preferably, the step of dynamically correcting the control parameters using the PID algorithm specifically involves setting the proportional coefficient of the PID algorithm according to the different objects of graded control. Integral coefficient Differential coefficients Every so often, the deviation of different control targets is calculated based on the predicted and measured values ​​of the hydrogen-rich combustion dynamic prediction model. The deviation signal is smoothed by a low-pass filtering algorithm and then input into the PID algorithm to obtain the correction amount. The output correction amounts of the three links are superimposed to obtain the total control correction amount.

[0046] Based on the parameter characteristics of different control objects, the total correction amount is mapped to specific parameter correction values; the correction values ​​are superimposed with the current control parameter values ​​to obtain the updated control parameter values.

[0047] The hydrogen-rich gas combustion state optimization and control system adapted to the blast furnace swirl zone includes a dynamic sensing module, a hydrogen-rich combustion dynamic prediction module, a graded control module, and an intelligent feedback correction module, as detailed below:

[0048] The dynamic sensing module is used to collect all-element correlation data of the physical field, chemical field and furnace condition field in the blast furnace swirl zone, including the physical field monitoring unit, chemical field monitoring unit, furnace condition field correlation monitoring unit and data preprocessing unit;

[0049] The hydrogen-rich combustion dynamic prediction module receives signals from the dynamic sensing module to achieve high-precision prediction of the combustion state, including a basic correlation mechanism model unit, an LSTM correction model unit, and a multi-objective optimization solution unit.

[0050] The graded control module receives the dynamic prediction results of hydrogen-rich combustion output by the dynamic prediction module of hydrogen-rich combustion and uses them to execute graded control actions.

[0051] The intelligent feedback correction module establishes bidirectional signal connections with the multi-dimensional sensing and acquisition module and the hierarchical control and execution module to achieve dynamic assurance of control accuracy.

[0052] Compared with the prior art, the technical solution of this application has the following technical effects:

[0053] This invention establishes a fundamental correlation mechanism model based on the chemical reaction kinetics, fluid mechanics, and heat transfer equations of hydrogen-rich combustion. It also uses an LSTM network to correct the deviation of the mechanism model in real time, which solves the problem that the single model in the existing technology is difficult to match the time sequence characteristics and multi-field coupling law of hydrogen-rich combustion. This significantly improves the prediction accuracy of combustion temperature and flame length, and provides accurate data support for control strategies.

[0054] This invention aims to maximize combustion efficiency, temperature field uniformity, and furnace condition adaptability. It solves the multi-objective optimization function using the NSGA-Ⅲ algorithm to obtain a uniformly distributed Pareto optimal solution set. Combined with a graded control strategy, it balances combustion stability and furnace condition adaptability, effectively improving hydrogen utilization and reducing the risk of unburned hydrogen emissions.

[0055] This invention is based on a PID algorithm. By comparing the deviation between the predicted values ​​of the mixed prediction model and the measured values ​​after regulation, it dynamically corrects the injection velocity, angle, injection point distribution, and hydrogen-rich gas mixing ratio of the injection pipe, ensuring the continuous and efficient operation of blast furnace smelting.

[0056] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0057] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0059] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0060] Figure 1 Overall flowchart of the method for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone;

[0061] Figure 2 The neural network structure diagram of the dynamic prediction model for hydrogen-rich combustion is shown.

[0062] Figure 3 Overall architecture diagram of the system for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone;

[0063] Figure 4 An experimental architecture diagram for optimizing and controlling a vanadium-titanium magnetite blast furnace using this method and system;

[0064] Figure 5 A diagram showing the outlet velocity distribution in the vortex zone of shale gas injection nozzles with different pipe diameters;

[0065] Figure 6 This graph shows the variation of the highest temperature of shale gas at the outlet of the vortex zone under different pipe diameters.

[0066] Figure 7A comparison chart of shale gas combustion rate and gas state for different pipe diameters.

[0067] Figure 8 This diagram illustrates the effect of the injection pipe inlet location on the shale gas combustion rate.

[0068] Figure 9 This is a graph showing the effect of blast oxygen enrichment rate on shale gas combustion rate.

[0069] Figure 10 This is a data graph showing the impact of different shale gas injection rates and oxygen content on the swirling zone. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0071] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0072] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0073] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0074] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0075] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0076] Example 1 mainly describes a method for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone, such as... Figure 1 As shown, it specifically includes:

[0077] By monitoring the correlation data of the physical field, chemical field and furnace condition field in the blast furnace swirling zone, and based on the chemical reaction kinetic equation of hydrogen-rich combustion and the hydrodynamic characteristics of the swirling zone, a basic correlation mechanism model of combustion temperature, flame length and hydrogen concentration and hot blast parameters is established.

[0078] Based on the associated data, the mechanistic model is corrected in real time using LSTM to obtain a hybrid prediction model that combines the mechanism and the data.

[0079] A multi-objective optimization function was defined, and the objectives were solved using the non-dominated sorting genetic algorithm NSGA-Ⅲ to obtain a dynamic prediction model for hydrogen-rich combustion.

[0080] Based on a graded control strategy, the jet velocity, gas jet angle, jet point distribution and hydrogen-rich gas mixing ratio of the jet pipe are dynamically adjusted.

[0081] Collect and control data, compare the deviation with the predicted values ​​of the hybrid prediction model, and dynamically correct the control parameters through the PID algorithm.

[0082] Furthermore, the basic correlation mechanism model is specifically as follows: based on the chemical reaction kinetics equation of hydrogen-rich combustion, the hydrodynamic equation of the swirling zone, and the heat transfer equation, the model is constructed through parameter correlation, wherein the input parameters are hydrogen concentration, hot air parameters, and jet velocity of the blowpipe, and the output parameters are combustion temperature and flame length.

[0083] Hydrogen concentration is directly substituted into the combustion reaction rate equation as the core influencing factor of the reaction rate; hot air temperature is used as the initial boundary condition for combustion temperature; hot air pressure is used to calculate the axial pressure gradient in the fluid dynamics equation; hot air velocity and jet velocity from the blowpipe jointly determine the axial velocity after mixing; the core equation is solved by a sequential coupling method.

[0084] By introducing a turbulence correction factor, the effective viscosity coefficient is corrected to reflect the influence of airflow disturbance on velocity distribution; by introducing a radiation heat transfer correction factor into the heat transfer equation, the emissivity of the furnace charge surface is corrected, and the accuracy of temperature calculation is improved.

[0085] Furthermore, the solution of the core equation using the sequential coupling method specifically involves substituting the input parameters—hydrogen concentration, hot air parameters, and jet velocity—into the fluid dynamics equations to obtain the axial velocity distribution in the swirling zone. The formula is as follows:

[0086]

[0087] in, The axial velocity in the vortex region, For hydrogen concentration, For axial pressure gradient, The radius of the cyclone region, Radial coordinates, The effective viscosity coefficient, The jet velocity;

[0088] The volume of the combustion reaction zone is determined based on the axial velocity distribution. Substituting the total heat release into the heat transfer equilibrium equation, the combustion temperature can be obtained by solving the equation. Substituting the hydrogen concentration and reaction zone volume into the combustion reaction rate equation, the H2 consumption rate and total heat release are calculated using the following formula:

[0089]

[0090] in, The reaction is exothermic. The rate of H2 consumption reaction. for, For the total heat release, the Stefan-Boltzmann constant is given. For the surface emissivity of the furnace charge, This refers to the radiative heat exchange area between the combustion zone and the furnace charge. The surface temperature of the furnace charge;

[0091] Based on the axial flow velocity and combustion reaction rate, the flame length is calculated by integration, using the following formula:

[0092]

[0093]

[0094] in, The length of the flame. axial The cross-sectional area of ​​the reaction zone at that location, The preset H2 conversion rate threshold, The initial hydrogen concentration, Input volumetric flow rate for hydrogen-rich gas.

[0095] Furthermore, the hybrid prediction model specifically involves: preprocessing the collected associated data to obtain time-series samples; designing a multi-layer LSTM network based on the time-series characteristics of hydrogen-rich combustion; inputting the preprocessed time-series samples into the LSTM network; updating the network parameters through the backpropagation algorithm until the loss function converges; and evaluating the model performance using validation set data after training.

[0096] The real-time monitored correlation data is synchronously input into the basic correlation mechanism model and the LSTM correction model. The basic correlation mechanism model calculates the predicted combustion temperature T and the predicted flame length L at the current moment based on the real-time input data.

[0097] The LSTM correction model, based on the same real-time time series input data, predicts the temperature deviation ΔT and length deviation ΔL at the current moment.

[0098] By combining the mechanism output with the LSTM prediction bias correction logic, the predicted value of the target combustion state is calculated, forming a hybrid prediction model that combines mechanism and data. The formula is as follows:

[0099]

[0100]

[0101] At regular intervals, the output value of the hybrid prediction model is compared with the real value collected by the perception system. When the continuous deviation is too large, the latest collected time series data is added to the training set to incrementally train the LSTM correction model, update the network parameters, and achieve adaptive optimization of the model.

[0102] Furthermore, considering the temporal characteristics of hydrogen-rich combustion, a multi-layer LSTM network was designed with the following structural parameters:

[0103] The number of neurons in the input layer equals the dimension of the input features, i.e., the number of input parameters is 7, including hydrogen concentration, hot air temperature, hot air pressure, hot air velocity, jet velocity from the blowpipe, material column permeability, and furnace top pressure. The hidden layer uses 3 LSTM units, with 64 neurons in each layer. Dropout layers are used with a dropout rate of 0.2 to suppress overfitting, and the activation function is tanh. The output layer has 2 neurons, corresponding to temperature deviation ΔT and length deviation ΔL, respectively, and the activation function is a linear function.

[0104] Furthermore, mean squared error (MSE) was chosen as the loss function to measure the difference between prediction bias and actual bias. The optimizer was Adam, with an initial learning rate of 0.001. A learning rate decay strategy was adopted, halving the learning rate every 100 epochs. The training batch size was set to 256, and the number of training epochs was set to 100. Simultaneously, 80% of the data was divided into a training set and 20% into a validation set. Training was stopped when the validation set loss function did not decrease for 10 consecutive epochs to avoid overfitting.

[0105] Furthermore, such as Figure 2 The diagram shows the neural network structure of the dynamic prediction model. Specifically, the dynamic prediction model for hydrogen-rich combustion is as follows: based on the combustion state parameters output by the hybrid prediction model, the optimization objectives and constraints are defined; the Pareto optimal solution set that satisfies all objectives is searched using the NSGA-Ⅲ algorithm; the model parameters corresponding to the optimal solution are integrated into the hybrid prediction model to construct a dynamic prediction model that combines high-precision prediction and multi-objective optimization characteristics.

[0106] Furthermore, the specific optimization objectives and constraints are as follows: determining the combustion efficiency based on the ratio of the actual amount of hydrogen participating in combustion to the total amount of input hydrogen and setting a target threshold for combustion efficiency; determining the temperature field uniformity based on the reciprocal of the temperature difference between the core and edge regions of the swirling zone and setting a target threshold for temperature field uniformity; determining the furnace condition adaptability based on the fluctuation of the material column permeability and setting a maximum threshold for fluctuation.

[0107] The optimization objectives are determined by maximizing combustion efficiency, maximizing temperature field uniformity, and achieving optimal furnace condition adaptability.

[0108] Set constraints on hydrogen concentration, temperature, unburned hydrogen concentration, and control parameters.

[0109] Furthermore, the construction of a dynamic prediction model that combines high-precision prediction and multi-objective optimization features specifically involves: normalizing each optimization objective to eliminate dimensional differences; setting weight coefficients for each objective based on industrial needs; and constructing a maximization objective function, the formula of which is:

[0110]

[0111] in, To optimize the variable vector, , , These are the normalized combustion efficiency, temperature uniformity, and furnace condition adaptability, respectively. , , The weights for combustion efficiency, temperature uniformity, and furnace condition adaptability are respectively.

[0112] The non-dominated sorting genetic algorithm NSGA-III is used to solve the constructed multi-objective optimization function. Its core advantage is that it can handle multi-objective conflict problems and obtain a uniformly distributed Pareto optimal solution set. ;

[0113] The optimal solution set The corresponding optimization variables serve as the dynamic baseline parameters of the hybrid prediction model. The prediction parameters are dynamically adjusted in conjunction with real-time monitoring data to obtain the dynamic prediction model for hydrogen-rich combustion.

[0114] Furthermore, the process of the NSGA-Ⅲ algorithm to solve the Pareto optimal solution is as follows: taking the optimization variable vector X as the gene, an initial population is randomly generated, the population size is set to 200, each individual corresponds to a set of optimization variable values, and satisfies the constraints; the number of iterations is set to 100, the crossover probability is set to 0.8, and the mutation probability is set to 0.05.

[0115] Calculate the objective function value F(X) for each individual in the population, and rank the population based on the non-dominance relationship—divide the population into different non-dominance levels, level 1 is the set of optimal individuals that are not dominated by other individuals, level 2 is the set of individuals that are dominated only by individuals of level 1, and so on; at the same time, calculate the crowding distance for each individual.

[0116] The roulette wheel selection method is used to select superior individuals from the current population to enter the mating pool; the single-point crossover method is used to perform crossover operations on the selected individuals to generate offspring; and the random mutation method is used to mutate the genes of some individuals to ensure population diversity.

[0117] Merge the parent and offspring populations, perform non-dominated sorting and crowding calculations again, retain individuals with high non-dominated levels and high crowding, and update the population size to 200; repeat the above steps until the set number of iterations is reached, at which point a stable Pareto optimal solution set is obtained;

[0118] The solution with the best overall performance is selected from the Pareto optimal solution set, denoted as... The combustion efficiency, temperature field uniformity, and furnace condition adaptability of the solution all meet the engineering objectives.

[0119] Furthermore, the graded control strategy specifically involves: setting a preset hydrogen concentration fluctuation threshold. Temperature field temperature difference threshold Combustion efficiency and furnace condition adaptability threshold Based on the output of the dynamic prediction model for hydrogen-rich combustion and multi-dimensional sensing data, key parameters are adjusted according to the hierarchical logic of rapid response, optimized adaptation, and long-term adaptation.

[0120] Among them, the first-level regulation rapid response is in response to hydrogen concentration fluctuations. or temperature field temperature difference It is triggered in time, and the injection flow rate of the injection pipe is adjusted in real time through the electromagnetic proportional valve, and the hot air oxygen supply ratio is adjusted simultaneously to quickly maintain combustion stability.

[0121] The results of the secondary control optimization and adaptation to the combustion efficiency and furnace condition adaptability of the hybrid prediction model output. The system is triggered in time to adjust the injection angle and injection point distribution of hydrogen-rich gas, thereby achieving temperature field homogenization.

[0122] The three-level control system is adapted to the long-term needs of the furnace by dynamically adjusting the hydrogen-rich gas mixing ratio based on furnace condition trend data, thus establishing an adaptation relationship between hydrogen concentration and furnace condition.

[0123] Furthermore, the dynamic correction of control parameters using the PID algorithm specifically involves setting the proportional coefficient of the PID algorithm according to the different objects of graded control. Integral coefficient Differential coefficients Every so often, the deviation of different control targets is calculated based on the predicted and measured values ​​of the hydrogen-rich combustion dynamic prediction model. The deviation signal is smoothed by a low-pass filtering algorithm and then input into the PID algorithm to obtain the correction amount. The output correction amounts of the three links are superimposed to obtain the total control correction amount.

[0124] Based on the parameter characteristics of different control objects, the total correction amount is mapped to specific parameter correction values; the correction values ​​are superimposed with the current control parameter values ​​to obtain the updated control parameter values.

[0125] Furthermore, the specific initial parameters for adjusting the jet velocity of the blowpipe are set as follows: The specific initial parameters for adjusting the hydrogen-rich gas injection angle are set as follows: The specific initial parameters for adjusting the hydrogen-rich gas mixing ratio are set as follows: .

[0126] Furthermore, based on the parameter characteristics of different control objects, the specific adjustments are as follows: In the proportional stage, the adjustment amount is directly output according to the magnitude of the deviation to achieve a rapid response and offset the current deviation; in the integral stage, the accumulated historical deviation is adjusted to eliminate the static deviation; in the derivative stage, the deviation trend is predicted according to the rate of change of the deviation, and the reverse adjustment amount is output in advance to suppress the rapid change of the deviation and improve the system stability. The output correction amounts of the three stages are superimposed to obtain the target PID control output.

[0127] This embodiment details a method for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone. By monitoring the physical, chemical, and furnace condition data of the swirl zone, and combining the hydrogen-rich combustion kinetics equations with fluid dynamics characteristics, a basic correlation mechanism model is constructed. An LSTM network is used to correct the mechanism model in real time, generating a hybrid prediction model combining mechanism and data. A multi-objective optimization function is set, targeting combustion efficiency, temperature field uniformity, and furnace condition adaptability. The NSGA-III algorithm is used to solve this function, resulting in a dynamic prediction model for hydrogen-rich combustion. A graded control strategy is employed, adjusting the injection velocity, angle, injection point distribution, and hydrogen-rich gas mixing ratio of the injection pipe according to threshold levels. A PID algorithm is used to compare the deviations between predicted and measured values, dynamically correcting the control parameters to achieve precise optimization of the hydrogen-rich combustion state in the blast furnace swirl zone.

[0128] Based on Example 1, this example describes in detail a system for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace vortex zone, such as... Figure 3 The system architecture diagram shown includes a dynamic sensing module, a hydrogen-rich combustion dynamic prediction module, a graded control module, and an intelligent feedback correction module, as detailed below:

[0129] The dynamic sensing module is used to collect comprehensive data on the physical field, chemical field, and furnace condition field in the blast furnace swirl zone. It includes a physical field monitoring unit, a chemical field monitoring unit, a furnace condition field correlation monitoring unit, and a data preprocessing unit. The physical field monitoring unit collects temperature, pressure, and flow rate data; the chemical field monitoring unit detects the volume fractions of H2, CO, CO2, and H2O in combustion products online; the furnace condition field correlation monitoring unit collects auxiliary parameters such as blast furnace top pressure, burden permeability, and hot blast temperature; and the data preprocessing unit identifies outliers, smooths time-series data, and fuses multi-source data using a weighted average method.

[0130] The hydrogen-rich combustion dynamic prediction module receives signals from the dynamic sensing module to achieve high-precision prediction of the combustion state. It includes a basic correlation mechanism model unit, an LSTM correction model unit, and a multi-objective optimization solution unit. The basic correlation mechanism model unit establishes a basic correlation model between combustion temperature, flame length, hydrogen concentration, and hot air parameters based on the chemical reaction kinetic equation of hydrogen-rich combustion and the hydrodynamic characteristics of the swirling zone. The LSTM correction model unit uses the historical correlation data preprocessed by the dynamic sensing module as input to train a hybrid prediction model. The multi-objective optimization solution unit constructs a hydrogen-rich combustion dynamic prediction model with the objectives of maximizing combustion efficiency, maximizing temperature field uniformity, and optimizing furnace condition adaptability.

[0131] The graded control module receives the dynamic prediction results of hydrogen-rich combustion from the dynamic prediction module and executes graded control actions, including a primary control unit, a secondary control unit, and a tertiary control unit. The primary control unit adjusts the injection velocity of the injection pipe when there are large fluctuations in hydrogen concentration or large temperature differences in the temperature field, and simultaneously links the hot air oxygen supply adjustment mechanism to fine-tune the oxygen supply ratio. The secondary control unit adjusts the hydrogen-rich gas injection angle based on the combustion efficiency and furnace condition adaptability results output by the mixing prediction model. The tertiary control unit dynamically adjusts the hydrogen-rich gas mixing ratio based on the trend of changes in the permeability of the fuel column.

[0132] The intelligent feedback correction module establishes bidirectional signal connections with the multi-dimensional sensing and acquisition module and the graded control execution module to achieve dynamic assurance of control accuracy. The module collects the measured data after control according to a preset cycle, calculates the deviation with the prediction results of the combustion state prediction subsystem, and corrects the control parameters of the graded control execution module through the built-in PID parameter self-tuning algorithm.

[0133] This embodiment details an optimized control system for the combustion state of hydrogen-rich gas in the blast furnace swirl zone, comprising four main modules: multi-dimensional dynamic sensing, dynamic prediction of hydrogen-rich combustion, graded control, and intelligent feedback correction. The sensing module includes physical field, chemical field, and furnace condition field monitoring units, collecting and preprocessing data on temperature, gas composition, and charge permeability. The prediction module constructs a dynamic prediction model with both theoretical support and data-driven characteristics using a basic correlation mechanism model, an LSTM correction model, and a multi-objective optimization solution unit, achieving accurate prediction of the combustion state. The graded control module adjusts the injection velocity, angle, and hydrogen-rich gas mixing ratio of the injection pipe according to threshold levels, rapidly responding to changes in operating conditions. The feedback correction module compares the predicted and measured values, dynamically correcting the control parameters using a PID algorithm.

[0134] Based on Example 1, this example details an experiment using this method and system for optimized control of a vanadium-titanium magnetite blast furnace. This blast furnace primarily smelts vanadium-titanium magnetite and now employs shale gas as a hydrogen-rich injection medium to upgrade the injection process by replacing all pulverized coal with shale gas. Addressing issues in the original process such as low shale gas combustion rate, uneven temperature distribution in the swirl zone, and significant furnace condition fluctuations, this invention is applied for precise control. The specific implementation process is as follows:

[0135] Real-time physical field data is collected using temperature sensors, pressure sensors, and flow rate monitoring devices, including temperature and pressure in the vortex zone, and shale gas inlet and outlet flow rates.

[0136] The chemical field data of combustion products, including the volume fractions of H2, CO, CO2, and H2O, are detected by an online gas analysis device, with a focus on monitoring the residual amount of CH4.

[0137] The data preprocessing unit identifies outliers and smooths time-series data in the collected associated data. The weighted average method is used to fuse multi-source data to obtain the inlet velocity data of shale gas injection with different pipe diameters as shown in Table 1.

[0138] Table 1. Inlet velocities of shale gas injection pipes with different diameters

[0139]

[0140] According to Table 1 and Figure 5 The figure shows the velocity distribution at the outlet of the vortex zone of shale gas injection under different pipe diameters. (a) shows the velocity change of shale gas at the outlet center of the vortex zone under different pipe diameters. As can be seen from the figure, when the pipe diameter is 16 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the outlet center velocity increases from 112.3 m / s to 134.3 m / s, an increase of 19.6%; when the injection pipe diameter is 20 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the outlet center velocity increases from 102.6 m / s to 128.1 m / s, an increase of 24.9%; when the injection pipe diameter is 24 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the outlet center velocity increases from 97.1 m / s to 118 m / s, an increase of 21.5%. This shows that as the pipe diameter increases from 16mm to 24mm, the outlet center velocity gradually decreases, with the largest increase occurring at a pipe diameter of 20mm. This phenomenon indicates that, under a certain injection volume, the gas flow velocity at the center can be adjusted by changing the pipe diameter, thereby controlling the combustion behavior and achieving optimal combustion state and temperature.

[0141] (b) shows the average exit velocity of shale gas in the vortex zone under different pipe diameters. As can be seen from the figure, when the pipe diameter is 16 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the average exit velocity increases from 49.5 m / s to 66 m / s, an increase of 33.3%; when the pipe diameter is 20 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the average exit velocity increases from 48.7 m / s to 56.4 m / s, an increase of 15.8%; when the pipe diameter is 24 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the average exit velocity increases from 49.2 m / s to 53.5 m / s, an increase of 8.7%. It can be seen that the increase in average outlet velocity gradually decreases with the increase in pipe diameter, and the increase in average outlet velocity is the largest when the pipe diameter is 16mm. The reason for this is that, under the condition of the same increase in gas volume, the smaller the pipe diameter, the greater the increase in inlet velocity, the greater the increase in combustion rate, the greater the increase in gas volume, and the greater the increase in average outlet velocity.

[0142] Based on the data on the outlet temperature and increase of shale gas in the vortex zone under different pipe diameters, the following table is obtained:

[0143] Table 2. Data on shale gas outlet temperature and its increase in the vortex zone for different pipe diameters.

[0144]

[0145] According to Table 2 and Figure 6 The figure shows the variation of the highest temperature of shale gas at the outlet of the vortex zone under different pipe diameters. Figure 6(a) shows the variation of the highest temperature of shale gas at the outlet of the vortex zone under different pipe diameters. As can be seen from the figure, when the pipe diameter is 16 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the highest temperature of shale gas increases from 1523.5℃ to 1689.1℃, an increase of 10.8%; when the pipe diameter is 20 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the highest temperature of shale gas increases from 1514.5℃ to 1628.8℃, an increase of 7.5%; when the pipe diameter is 24 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the highest temperature of shale gas increases from 1527.8℃ to 1618.6℃, an increase of only 6%. The above phenomena indicate that, under the same injection rate, the maximum temperature of shale gas in the tuyeres decreases with increasing pipe diameter. Furthermore, the increase in maximum temperature also decreases with increasing pipe diameter, with the largest increase observed when the pipe diameter is 16 mm. This demonstrates that the smaller the pipe diameter, the greater the velocity of the shale gas at the tuyeres, resulting in more complete combustion and a higher temperature at the outlet.

[0146] Figure 6(b) shows the average temperature variation of shale gas at the outlet of the vortex zone under different pipe diameters. As can be seen from the figure, when the pipe diameter is 16 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the average temperature of the shale gas increases from 1395.5℃ to 1593.8℃, an increase of 14.2%; when the pipe diameter is 20 mm and the injection rate increases from 60 kg / tHM to 120 kg / tHM, the average temperature of the shale gas increases from 1382.5℃ to 1489.1℃, an increase of [missing value]. 7.7%; When the pipe diameter is 24mm and the injection rate increases from 60kg / tHM to 120kg / tHM, the average outlet temperature increases from 1390.8℃ to 1448.5℃, an increase of 4.1%. This phenomenon indicates that under the same injection rate, the average temperature change at the outlet of shale gas in the vortex zone is basically consistent with the maximum temperature change. As the injection rate increases, the increase in average temperature is also consistent with the increase in maximum temperature. When the pipe diameter is 16mm, the increase in average outlet temperature is the largest.

[0147] 6(c) and 6(d) represent the outlet surface of the vortex zone under different conditions of 20 mm pipe diameter and shale gas injection rate. and A schematic diagram of volume fraction distribution; image data proves that as the injection volume increases, the outlet surface of the air vortex zone... The volume fraction content first increased and then decreased. The volume fraction continuously decreases. This proves that under these conditions, the combustion rate of shale gas increases with the increase of the injection rate.

[0148] Based on the data on shale gas combustion rate and gas composition variations under different pipe diameters, the following table is obtained:

[0149] Table 3. Data on shale gas combustion rate and gas composition variations under different pipe diameters.

[0150]

[0151] According to Table 3 and Figure 7 The figures show a comparison of shale gas combustion rate and gas state at different pipe diameters. Figure 7(a) shows the variation of shale gas combustion rate at the outlet of the tuyeres. As can be seen from the figure, the shale gas combustion rate first decreases and then increases with increasing injection rate. When the pipe diameter is 16mm and the injection rate is less than 90kg / tHM, the combustion rate varies between 32.5% and 52.7%; when the pipe diameter is 20mm and the injection rate is less than 150kg / tHM, the combustion rate ranges from 30.8% to 65.9%; and when the pipe diameter is 24mm and the injection rate is less than 150kg / tHM, the combustion rate is 31.3% to 39.5%. Figure 7(b) shows the variation of shale gas combustion rate at different locations in the tuyeres. As can be seen from the figure... Within 1.2m of the tuyeres, the shale gas combustion rate is very low. It increases continuously with increasing injection rate, reaching its highest point when the pipe diameter is 20mm and the injection rate is 120kg / tHM of shale gas. The diagram shows the distribution of CH4 and O2 volume fractions at the outlet of the vortex zone under different injection rates and pipe diameters (20mm). It clearly shows that as the injection rate increases, the CH4 volume fraction at the outlet of the vortex zone first increases and then decreases, while the O2 volume fraction continuously decreases. This demonstrates that under these conditions, the shale gas combustion rate increases with increasing injection rate.

[0152] Based on the chemical reaction kinetics equations of hydrogen-rich combustion, the fluid dynamics equations of the swirling zone, and the heat transfer equations, a basic correlation mechanism model between combustion temperature, flame length, hydrogen concentration, and hot air parameters is established.

[0153] 80% of the preprocessed time-series sample data was selected as the training set, and 20% as the validation set. A 3-layer LSTM network was designed, with 7 neurons in the input layer, 64 neurons in each hidden layer, a dropout rate of 0.2, and the activation function tanh. The output layer had 2 neurons, corresponding to temperature deviation ΔT and length deviation ΔL. The mean squared error (MSE) was selected as the loss function, and the Adam optimizer was used. The initial learning rate was set to 0.001, the batch size was set to 256, and the number of training epochs was set to 100. After training, the LSTM-corrected model can accurately predict temperature and length deviations under different operating conditions, and is combined with the basic correlation mechanism model to form a hybrid prediction model.

[0154] Based on the combustion state parameters output by the hybrid prediction model, optimization objectives were set to maximize combustion efficiency, maximize temperature field uniformity, and optimize furnace condition adaptability. Based on previous experimental data, constraints were set as follows: hydrogen concentration constrained to a reasonable range for the corresponding injection rate, temperature constrained to 1380-1700℃, unburned CH4 concentration constrained to <5%, and control parameter constraints: injection pipe velocity 217.4-271.8 m / s and oxygen enrichment rate 3%-9%. The multi-objective optimization function was solved using the NSGA-Ⅲ algorithm to obtain the Pareto optimal solution set, where the optimal injection pipe velocity was 245 m / s, the oxygen enrichment rate adjustment range was 3%-7%, and the injection angle was 15°.

[0155] Based on the output of the dynamic prediction model for hydrogen-rich combustion and multi-dimensional sensing data, combined with, for example Figure 8 , Figure 9 The data showing the impact of shale gas injection rate and injection pipe status on the swirl zone are used to implement control actions according to a graded control strategy. When hydrogen concentration fluctuations are detected to be greater than 5%, the injection velocity of the injection pipe is adjusted in real time via an electromagnetic proportional valve. For example, when the shale gas injection rate increases from 120 kg / tHM to 150 kg / tHM, the hydrogen concentration fluctuation reaches 6.2%, triggering the first-level control. The injection velocity of the injection pipe is adjusted from 245 m / s to 271.8 m / s, and the hot air oxygen supply ratio is simultaneously fine-tuned to quickly maintain combustion stability and keep the hydrogen concentration fluctuations within the threshold range.

[0156] Under initial operating conditions, the combustion efficiency was 58.2%, which was less than the preset combustion efficiency threshold of 60%. By adjusting the hydrogen-rich gas injection angle to 15° and optimizing the injection point distribution, the temperature difference between the core and edge areas of the swirling zone was reduced from 95°C to 72°C, the temperature field uniformity was significantly improved, and the combustion efficiency was increased to 65.9%, meeting the furnace condition adaptability requirements.

[0157] The hydrogen-rich gas mixing ratio is dynamically adjusted based on furnace condition trend data, combined with, for example... Figure 10The data show the impact of different shale gas injection rates and oxygen content on the vortex zone. Through three-level regulation, the oxygen enrichment rate was gradually increased from 3% to 7%, the combustion rate was increased from 65.9% to 78.7%, and the permeability fluctuation of the feed column was monitored to decrease from 8% to 3%, and the furnace stability was significantly enhanced.

[0158] This embodiment details an experiment using the method and system to optimize and control a vanadium-titanium magnetite blast furnace. By optimizing and controlling the combustion state of shale gas in the swirl zone of the vanadium-titanium magnetite blast furnace, the problems of low combustion efficiency, uneven temperature distribution, and unstable furnace conditions in the original process are effectively solved. This fully verifies the engineering applicability, reliability, and economy of the method and system of this invention in the hydrogen-rich smelting scenario of vanadium-titanium magnetite blast furnace.

[0159] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A method for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone, characterized in that, include: By monitoring the correlation data of the physical field, chemical field and furnace condition field in the blast furnace swirling zone, and based on the chemical reaction kinetic equation of hydrogen-rich combustion and the hydrodynamic characteristics of the swirling zone, a basic correlation mechanism model of combustion temperature, flame length and hydrogen concentration and hot blast parameters is established. Based on the associated data, the mechanistic model is corrected in real time using LSTM to obtain a hybrid prediction model that combines the mechanism and the data. A multi-objective optimization function was defined, and the objectives were solved using the non-dominated sorting genetic algorithm NSGA-Ⅲ to obtain a dynamic prediction model for hydrogen-rich combustion. Based on a graded control strategy, the jet velocity, gas jet angle, jet point distribution and hydrogen-rich gas mixing ratio of the jet pipe are dynamically adjusted. Collect and control data, compare the deviation with the predicted values ​​of the hybrid prediction model, and dynamically correct the control parameters through the PID algorithm.

2. The method according to claim 1, characterized in that, The basic correlation mechanism model is specifically: based on the chemical reaction kinetics equation of hydrogen-rich combustion, the hydrodynamic equation of the swirling zone, and the heat transfer equation, the model is constructed through parameter correlation, wherein the input parameters are hydrogen concentration, hot air parameters, and jet velocity of the blowpipe, and the output parameters are combustion temperature and flame length. Hydrogen concentration is directly substituted into the combustion reaction rate equation as the core influencing factor of the reaction rate; hot air temperature is used as the initial boundary condition for combustion temperature; hot air pressure is used to calculate the axial pressure gradient in the fluid dynamics equation; hot air velocity and jet velocity from the blowpipe jointly determine the axial velocity after mixing; the core equation is solved by a sequential coupling method. By introducing a turbulence correction factor, the effective viscosity coefficient is corrected to reflect the influence of airflow disturbance on velocity distribution; by introducing a radiation heat transfer correction factor into the heat transfer equation, the emissivity of the furnace charge surface is corrected, and the accuracy of temperature calculation is improved.

3. The method according to claim 2, characterized in that, The method of solving the core equations via sequential coupling involves substituting the input parameters—hydrogen concentration, hot air parameters, and jet velocity—into the fluid dynamics equations to obtain the axial velocity distribution in the vortex region. The formula is as follows: in, The axial velocity in the vortex region, For hydrogen concentration, For axial pressure gradient, The radius of the cyclone region, Radial coordinates, The effective viscosity coefficient, The jet velocity; The volume of the combustion reaction zone is determined based on the axial velocity distribution. Substituting the total heat release into the heat transfer equilibrium equation, the combustion temperature can be obtained by solving the equation. Substituting the hydrogen concentration and reaction zone volume into the combustion reaction rate equation, the H2 consumption rate and total heat release are calculated using the following formula: in, The reaction is exothermic. The rate of H2 consumption reaction. for, For the total heat release, the Stefan-Boltzmann constant is given. For the surface emissivity of the furnace charge, This refers to the radiative heat exchange area between the combustion zone and the furnace charge. The surface temperature of the furnace charge; Based on the axial flow velocity and combustion reaction rate, the flame length is calculated by integration, using the following formula: in, The length of the flame. axial The cross-sectional area of ​​the reaction zone at that location, The preset H2 conversion rate threshold, The initial hydrogen concentration, Input volumetric flow rate for hydrogen-rich gas.

4. The method according to claim 1, characterized in that, The hybrid prediction model is specifically as follows: after preprocessing the collected correlation data to obtain time series samples, a multi-layer LSTM network is designed in combination with the time series characteristics of hydrogen-rich combustion; the preprocessed time series samples are input into the LSTM network, and the network parameters are updated through the backpropagation algorithm until the loss function converges; after training, the model performance is evaluated using validation set data. The real-time monitored correlation data is synchronously input into the basic correlation mechanism model and the LSTM correction model. The basic correlation mechanism model calculates the predicted combustion temperature T and the predicted flame length L at the current moment based on the real-time input data. The LSTM correction model, based on the same real-time time series input data, predicts the temperature deviation ΔT and length deviation ΔL at the current moment. By combining the mechanism output with the LSTM prediction bias correction logic, the predicted value of the target combustion state is calculated, forming a hybrid prediction model that combines mechanism and data. The formula is as follows: At regular intervals, the output value of the hybrid prediction model is compared with the real value collected by the perception system. When the continuous deviation is too large, the latest collected time series data is added to the training set to incrementally train the LSTM correction model, update the network parameters, and achieve adaptive optimization of the model.

5. The method according to claim 1, characterized in that, The hydrogen-rich combustion dynamic prediction model is specifically defined as follows: based on the combustion state parameters output by the hybrid prediction model, the optimization objectives and constraints are clarified; the Pareto optimal solution set that satisfies all objectives is searched using the NSGA-Ⅲ algorithm; the model parameters corresponding to the optimal solution are integrated into the hybrid prediction model to construct a dynamic prediction model that combines high-precision prediction and multi-objective optimization characteristics.

6. The method according to claim 5, characterized in that, The specific optimization objectives and constraints are as follows: the combustion efficiency is determined based on the ratio of the actual amount of hydrogen participating in combustion to the total amount of input hydrogen, and a target threshold for combustion efficiency is set; the temperature field uniformity is determined by the reciprocal of the temperature difference between the core and edge zones of the swirling zone, and a target threshold for temperature field uniformity is set; the furnace condition adaptability is determined based on the fluctuation of the material column permeability, and a maximum threshold for fluctuation is set. The optimization objectives are determined by maximizing combustion efficiency, maximizing temperature field uniformity, and achieving optimal furnace condition adaptability. Set constraints on hydrogen concentration, temperature, unburned hydrogen concentration, and control parameters.

7. The method according to claim 5, characterized in that, The construction of a dynamic prediction model that combines high-precision prediction and multi-objective optimization features specifically involves: normalizing each optimization objective to eliminate dimensional differences; setting weight coefficients for each objective based on industrial needs; and constructing a maximization objective function, the formula of which is: in, To optimize the variable vector, , , These are the normalized combustion efficiency, temperature uniformity, and furnace condition adaptability, respectively. , , The weights for combustion efficiency, temperature uniformity, and furnace condition adaptability are respectively. The non-dominated sorting genetic algorithm NSGA-III is used to solve the constructed multi-objective optimization function. Its core advantage is that it can handle multi-objective conflict problems and obtain a uniformly distributed Pareto optimal solution set. ; The optimal solution set The corresponding optimization variables serve as the dynamic baseline parameters of the hybrid prediction model. The prediction parameters are dynamically adjusted in conjunction with real-time monitoring data to obtain the dynamic prediction model for hydrogen-rich combustion.

8. The method according to claim 1, characterized in that, The graded control strategy specifically involves: setting a preset hydrogen concentration fluctuation threshold. Temperature field temperature difference threshold Combustion efficiency and furnace condition adaptability threshold Based on the output of the dynamic prediction model for hydrogen-rich combustion and multi-dimensional sensing data, key parameters are adjusted according to the hierarchical logic of rapid response, optimized adaptation, and long-term adaptation. Among them, the first-level regulation rapid response is in response to hydrogen concentration fluctuations. or temperature field temperature difference It is triggered in time, and the injection flow rate of the injection pipe is adjusted in real time through the electromagnetic proportional valve, and the hot air oxygen supply ratio is adjusted simultaneously to quickly maintain combustion stability. The results of the secondary control optimization and adaptation to the combustion efficiency and furnace condition adaptability of the hybrid prediction model output. The system is triggered in time to adjust the injection angle and injection point distribution of hydrogen-rich gas, thereby achieving temperature field homogenization. The three-level control system is adapted to the long-term needs of the furnace by dynamically adjusting the hydrogen-rich gas mixing ratio based on furnace condition trend data, thus establishing an adaptation relationship between hydrogen concentration and furnace condition.

9. The method according to claim 1, characterized in that, The dynamic correction of control parameters through the PID algorithm specifically involves setting the proportional coefficient of the PID algorithm according to the different objects of graded control. Integral coefficient Differential coefficients Every so often, the deviation of different control targets is calculated based on the predicted and measured values ​​of the hydrogen-rich combustion dynamic prediction model. The deviation signal is smoothed by a low-pass filtering algorithm and then input into the PID algorithm to obtain the correction amount. The output correction amounts of the three links are superimposed to obtain the total control correction amount. Based on the parameter characteristics of different control objects, the total correction amount is mapped to specific parameter correction values; the correction values ​​are superimposed with the current control parameter values ​​to obtain the updated control parameter values.

10. A system for optimizing and controlling the combustion state of hydrogen-rich gas in the blast furnace swirl zone, used to implement any of the methods of claims 1-9, characterized in that, It includes a dynamic sensing module, a hydrogen-rich combustion dynamic prediction module, a graded control module, and an intelligent feedback correction module, as detailed below: The dynamic sensing module is used to collect all-element correlation data of the physical field, chemical field and furnace condition field in the blast furnace swirl zone, including the physical field monitoring unit, chemical field monitoring unit, furnace condition field correlation monitoring unit and data preprocessing unit; The hydrogen-rich combustion dynamic prediction module receives signals from the dynamic sensing module to achieve high-precision prediction of the combustion state, including a basic correlation mechanism model unit, an LSTM correction model unit, and a multi-objective optimization solution unit. The graded control module receives the dynamic prediction results of hydrogen-rich combustion output by the dynamic prediction module of hydrogen-rich combustion and uses them to execute graded control actions. The intelligent feedback correction module establishes bidirectional signal connections with the multi-dimensional sensing and acquisition module and the hierarchical control and execution module to achieve dynamic assurance of control accuracy.

Images