Intelligent combustion and external pressure equalization control system for blast furnace hot blast stove

Abstract

The application relates to the technical field of automatic control of blast furnace ironmaking, and discloses an intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove, which comprises a detection unit, an execution module and a core cooperative control module. The core cooperative control module is used for: in the combustion process, dynamically compensating the air-fuel ratio by solving a virtual calorific value representing the change of internal energy through an energy balance equation; at the end of the combustion cycle, calculating and latching a thermal barycenter offset degree reflecting the energy storage distribution characteristics; in the pressure equalization process, implementing adaptive damping control on the opening process of the cold blast pressure equalization valve based on the thermal barycenter offset degree to soften the switching impact. The system can also input the virtual calorific value and the thermal barycenter offset degree and the like in the air supply process. The application realizes intelligent cooperation and feedforward prediction by constructing physical characteristic parameters throughout the processes, and significantly improves the combustion efficiency, switching safety and stability of the air supply temperature.

Description

Technical Field

[0001] This invention relates to the field of automated control technology for blast furnace ironmaking, specifically to an intelligent combustion and external pressure equalization control system for blast furnace hot blast stoves. Background Technology

[0002] The blast furnace hot blast stove is a key thermal equipment that provides high-temperature hot blast to the blast furnace. Its operation is cyclical, mainly consisting of two stages: combustion and blast supply. During the combustion stage, the hot blast stove transfers heat to and stores it in its internal heat storage medium by burning coal gas. In the subsequent blast supply stage, cold air is heated by the hot heat storage medium before being sent into the blast furnace. The economic efficiency and safety of the hot blast stove's operation directly depend on its level of automation control.

[0003] In existing hot blast stove control practices, combustion process regulation typically uses furnace top temperature as the primary feedback target. However, the calorific value of the fuel gas often fluctuates due to changes in blast furnace operating conditions, and the furnace top temperature, due to its significant thermal inertia, lags considerably behind changes in the calorific value of the gas. This lag prevents traditional feedback control systems from adjusting the ratio of combustion air to gas in a timely manner, frequently leading to incomplete combustion or an excessively high excess air coefficient, which reduces fuel utilization efficiency and causes flue gas heat loss.

[0004] Furthermore, during the switching between the combustion and blast stages, an external pressure equalization operation is required. This involves opening the cold air pressure equalization valve to connect the near-atmospheric pressure furnace cavity to the high-pressure cold air main pipe, allowing for rapid pressure equalization. Currently, the opening process of the pressure equalization valve generally employs a fixed, pre-set programmed control. This "one-size-fits-all" control method ignores the actual differences in the furnace's heat storage state after each combustion cycle. When excessive heat accumulates in the furnace top area due to previous combustion, the fixed rapid pressure equalization will cause a large influx of cold air, resulting in severe thermal shock and pressure oscillations on the furnace top refractory material. Over time, this can damage the furnace structure and pose safety hazards. Existing control systems lack effective means of sensing the furnace's energy storage state and fail to correlate combustion process information with subsequent pressure equalization control strategies. This leads to fragmented control across different process stages, hindering the overall improvement of the hot blast stove's operational level. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent combustion and external pressure equalization control system for blast furnace hot blast stoves, which solves the problems of lag in adjustment and insufficient adaptability in the combustion, reversing pressure equalization and air supply stages of existing blast furnace hot blast stove control systems.

[0006] This invention provides an intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove, comprising: at least one execution module, the execution module including an air-fuel ratio regulating valve, a cold air pressure equalization valve, and a cold air mixing valve; multiple detection units for collecting instantaneous volumetric flow rates of blast gas, combustion air, dome temperature, flue gas emission temperature, cold air main pipe pressure, and instantaneous pressure inside the hot blast stove cavity during the operation of the hot blast stove; and a core collaborative control module connected to the execution module and detection units, configured to perform the following operations: In the combustion process, based on the measured temperature of the dome and the flue gas emission temperature, a fictitious calorific value representing the change in energy in the heat storage chamber is calculated by solving the energy balance equation, and a compensation control command for the air-fuel ratio regulating valve is generated based on the deviation between the fictitious calorific value and the standard calorific value of the gas. At the end of the combustion cycle, the thermal centroid offset reflecting the energy storage distribution characteristics is calculated and locked by fusing the jet momentum integral value characterizing the thermal energy penetration depth with the longitudinal temperature steady-state gradient characterization term characterizing the thermal gradient distribution. In the pressure equalization process, based on the latched thermal center of gravity offset, a pressure equalization rate correction factor is generated to adjust the opening rate of the cold air pressure equalization valve, and a target full-stroke opening trajectory model of the cold air pressure equalization valve is constructed in combination with the real-time pressure difference, thereby realizing adaptive damping control.

[0007] In one embodiment of the present invention, the core collaborative control module is further configured to: after receiving the original electrical signal from the detection unit, set differentiated damping time constants for different physical variables such as pressure, temperature and flow rate, and perform first-order hysteresis smoothing processing to obtain reference data with consistent physical meaning.

[0008] In one embodiment of the present invention, the calculation process of the fictitious calorific value includes adding the heat power absorbed by the heat storage body to the heat loss power carried away by the flue gas, and dividing by the product of the combustion thermal efficiency and the instantaneous volumetric flow rate of the gas. The calculation of the heat power absorbed by the heat storage body incorporates the equivalent mass of the arch refractory material dynamically corrected according to the arch temperature range, thereby enabling the energy budget model to more accurately reflect the actual heat capacity changes of the heat storage body at different temperatures.

[0009] In one embodiment of the present invention, the core collaborative control module incorporates a low-sensitivity element when generating compensation control commands for the air-fuel ratio regulating valve. This element sets a preset dead-zone threshold, and compensation calculation is only initiated when the fictitious calorific value deviates from the standard calorific value of the gas by more than this threshold, thus avoiding unnecessary adjustments caused by minute instantaneous fluctuations in the signal.

[0010] In one embodiment of the present invention, the calculation process of the thermal center of gravity offset includes: weighting and fusing the ratio of the jet momentum integral value to the total chemical energy cumulative value input to the system, and the longitudinal temperature steady-state gradient characterization term, using a center of gravity weighting coefficient. The center of gravity weighting coefficient is adaptively calibrated according to the process operating load, so that the influence of momentum effect and thermal conduction effect on heat distribution can be reasonably allocated under different operating conditions.

[0011] In one embodiment of the present invention, the equalization rate correction factor is calculated as follows: based on an exponential mapping relationship including model sensitivity coefficients, the thermal centroid offset is mapped to a multiplicative decay limiting factor, which serves as the equalization rate correction factor. When the thermal centroid is too high, the correction factor decreases accordingly to suppress the subsequent valve opening rate.

[0012] In one embodiment of the present invention, the target full-stroke opening trajectory model is an exponential boosting model with damping attenuation characteristics. This model uses the equalization rate correction factor as an intervention parameter affecting the response speed. By adjusting the attenuation rate of the exponential term, the residence time in the low opening region is extended in the early stage of equalization, and the throttling effect is used to mitigate the pressure and temperature shock of the cold air entering the furnace in the early stage.

[0013] In one embodiment of the present invention, the core collaborative control module is further configured with safety defense logic in the pressure equalization process. This logic includes: when the furnace pressure rise rate is detected to exceed a preset threshold, triggering a self-protection constraint locking function to stop accumulating valve core opening increments; and when the pressure equalization process exceeds the global maximum tolerance time, forcibly switching to a preset conservative safety rate to advance the opening of the cold air pressure equalization valve to ensure the completion of the pressure equalization process.

[0014] In one embodiment of the present invention, the core collaborative control module is further configured to perform the following functions in the air supply process: construct a multidimensional state feature tensor containing the fictitious calorific value, thermal centroid offset, and real-time operating condition data; input the feature tensor into a pre-trained recurrent neural network model to predict the estimated value of the wind temperature attenuation slope within a future set window; and finally, based on the estimated value of the wind temperature attenuation slope, combined with the steady-state basic opening benchmark and proportional error, generate a control command with feedforward compensation for the cold air mixing valve to achieve predictive adjustment of wind temperature attenuation.

[0015] This invention also provides a method for intelligent combustion and external pressure equalization control of a blast furnace hot blast stove, the method comprising the following steps: In the combustion process, based on the collected dome temperature and flue gas emission temperature, a fictitious calorific value representing the change in energy in the heat storage chamber is calculated through the energy budget equation. Based on the deviation between the fictitious calorific value and the standard calorific value of the gas, compensation control for the air-fuel ratio is generated and executed. At the end of the combustion cycle, the thermal centroid offset reflecting the energy storage distribution characteristics is calculated and locked by fusing the jet momentum integral value characterizing the thermal energy penetration depth with the longitudinal temperature steady-state gradient characterization term characterizing the thermal gradient distribution. In the pressure equalization process, based on the latched thermal center of gravity offset, a pressure equalization rate correction factor is generated to adjust the opening rate of the cold air pressure equalization valve, and a target full-stroke opening trajectory is constructed in combination with the real-time pressure difference to achieve adaptive damping control of the cold air pressure equalization valve.

[0016] Furthermore, before calculating the fictitious calorific value, the method further includes setting differentiated damping time constants for different physical variables such as collected pressure, temperature, and flow rate, and performing first-order hysteresis smoothing to obtain an input benchmark with physical consistency.

[0017] Furthermore, the step of calculating the fictitious calorific value includes: adding the heat power absorbed by the heat storage body to the heat loss power carried away by the flue gas, and dividing by the product of the combustion thermal efficiency and the instantaneous volumetric flow rate of the gas; wherein, the calculation of the heat power absorbed by the heat storage body introduces the equivalent mass of the arch refractory material dynamically corrected according to the arch temperature range.

[0018] Furthermore, in the step of generating compensation control for the air-fuel ratio, a low-sensitivity element is built in, which only initiates compensation calculation when the fictitious calorific value deviates from the standard calorific value of the gas by more than a preset dead zone threshold.

[0019] Furthermore, the step of calculating the thermal centroid offset includes: weighting and fusing the ratio of the jet momentum integral value to the total chemical energy integral value input by the system, and the longitudinal temperature steady-state gradient characterization term, through a centroid weighting coefficient adaptively calibrated according to the process operating load.

[0020] Furthermore, the step of generating the equalization rate correction factor is as follows: based on an exponential mapping relationship that includes model sensitivity coefficients, the thermal centroid offset is mapped to a multiplicative decay limiting factor, which is used as the equalization rate correction factor.

[0021] Furthermore, the target full-stroke opening trajectory is constructed through an exponentially increasing model with damping attenuation characteristics. This model uses the equalization-pressure rate correction factor as an intervention parameter affecting the response speed, so as to prolong the residence time of the low opening region in the early stage of equalization.

[0022] Furthermore, a safety defense logic is executed during the pressure equalization process. This logic includes: when the rate of pressure rise in the furnace exceeds a preset threshold, triggering a self-protection constraint lockout to stop the cumulative valve core opening increment; and when the pressure equalization process exceeds the global maximum tolerance time, forcibly switching to a preset conservative safety rate to advance the opening of the cold air pressure equalization valve.

[0023] Furthermore, this method also includes the following steps: in the air supply process, constructing a multidimensional state feature tensor containing the fictitious calorific value, thermal center of gravity offset, and real-time operating condition data; inputting the feature tensor into a pre-trained recurrent neural network model to predict the estimated value of the wind temperature decay slope within a future set window; and based on the estimated value of the wind temperature decay slope, combined with the steady-state basic opening benchmark and proportional error, generating and executing control of the cold air mixing valve with feedforward compensation.

[0024] This invention provides an intelligent combustion and external pressure equalization control system for blast furnace hot blast stoves. It has the following beneficial effects: 1. This invention constructs a fictitious calorific value observation unit, calculates the fictitious calorific value representing the current thermal state in real time based on the principle of energy balance, and uses this value to adjust the combustion air in advance. This method effectively overcomes the lag of traditional temperature control feedback, can respond to fluctuations in the calorific value of coal gas in real time, and maintains dynamic matching of the air-fuel ratio throughout the combustion process, thereby significantly improving combustion efficiency and fuel utilization.

[0025] 2. This invention proposes the parameter of thermal center of gravity offset to accurately quantify the non-uniform distribution of heat storage in the furnace along the longitudinal direction. During the pressure equalization stage, the system applies dynamic, non-linear damping control to the opening process of the cold air pressure equalization valve based on this parameter, so that the pressure equalization rate can adapt to the actual heat storage level at the furnace top, effectively mitigating drastic changes in pressure and temperature, thereby reducing the thermal shock and mechanical stress on the furnace top refractory material, extending the service life of the hot blast stove, and improving the safety of reversing operations.

[0026] 3. This invention uses key historical data, such as the steady-state fictitious calorific value and the latched thermal center of gravity offset, obtained during the combustion process as input features to a neural network model for predicting blast temperature. Through forward inference, it predicts the future blast temperature attenuation slope and accordingly performs feedforward compensation on the cold blast mixing valve. This method profoundly utilizes the inherent data correlation between each process, effectively suppressing blast temperature overshoot and fluctuations caused by significant thermal inertia, ensuring a highly stable hot blast temperature supplied to the blast furnace, and providing a solid guarantee for the stable and smooth operation of the blast furnace. Attached Figure Description

[0027] Figure 1 This is a control system architecture diagram of an embodiment of the present invention; Figure 2 This is a flowchart of a hot blast stove control method according to an embodiment of the present invention; Figure 3 This is a digital signal processing logic distribution diagram according to an embodiment of the present invention; Figure 4 This is a calculation logic architecture diagram of the calorific value observation unit according to an embodiment of the present invention; Figure 5 This is a logical mapping diagram of the thermal centroid offset algorithm according to an embodiment of the present invention; Figure 6 A logic block diagram for generating equalization process adjustment instructions in an embodiment of the present invention; Figure 7 This is a topology diagram of the supply air mixing temperature prediction model according to an embodiment of the present invention; Figure 8 This is a comparison diagram of the combustion cycle dome temperature control effect of embodiments of the present invention; Figure 9 This is a comparison diagram of pressure changes inside the furnace during the pressure equalization process according to an embodiment of the present invention; Figure 10 This is a comparison chart of the hot air temperature stability during the air supply cycle in an embodiment of the present invention. Figure 11 A bar chart comparing the optimization effects of key performance indicators in embodiments of the present invention.

[0028] The system comprises: 10. Data acquisition module; 11. Flow detection unit; 12. Temperature detection unit; 13. Pressure detection unit; 14. Supply air flow detection unit; 15. Hot air temperature detection unit; 16. Operating condition interlock signal acquisition unit; 20. Execution module; 21. Air-fuel ratio regulating valve; 22. Cold air equalizing valve; 23. Cold air mixing valve; 30. Core collaborative control module; 31. Calorific value observation unit; 32. Feature extraction unit; 33. Feedforward coupling unit; and 34. Supply air temperature prediction unit. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] See attached document Figure 1 , Figure 1 This is a control system architecture diagram according to an embodiment of the present invention. The present invention provides an intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove, which may include: a data acquisition module 10, an execution module 20, and a core collaborative control module 30.

[0031] The data acquisition module 10 is deployed within the hot blast stove body and auxiliary gas pipeline system to acquire discrete physical quantities. The data acquisition module 10 is equipped with a flow detection unit 11, a temperature detection unit 12, and a pressure detection unit 13. The flow detection unit 11 reads the instantaneous volumetric flow rate of the gas and the instantaneous volumetric flow rate of the combustion air. The temperature detection unit 12 acquires the measured temperature of the hot blast stove dome and the flue gas emission temperature. The pressure detection unit 13 measures the measured pressure of the blast furnace cold blast main pipe and the instantaneous pressure inside the hot blast stove cavity. As a preferred embodiment, the data acquisition module 10 also includes a forced draft flow detection unit 14, a hot blast temperature detection unit 15, and a working condition interlock signal acquisition unit 16; wherein, the forced draft flow detection unit 14 is used to acquire the current flow rate of the main cold blast pipeline. Or the total air volume, the hot air temperature detection unit 15 is used to obtain the measured air temperature of the main air supply pipe. The operating condition interlock signal acquisition unit 16 is used to acquire combustion feedback position signals, combustion air switching valve lock-up signals, gas switching valve lock-up signals, pressure equalization permission signals, and air supply permission signals. The aforementioned acquisition components transmit standard analog electrical signals or switch status signals to the control network.

[0032] The execution module 20 includes an air-fuel ratio regulating valve 21, a cold air equalization valve 22, and a cold air mixing valve 23. The execution module 20 receives an analog input value and drives the servo mechanism to change the cross-sectional size of the pipeline. The air-fuel ratio regulating valve 21 is used to regulate the flow rate of combustion air, the cold air equalization valve 22 is used to regulate the cold air charging rate during the external pressure equalization process, and the cold air mixing valve 23 is used to regulate the amount of cold air mixed during the air supply stage.

[0033] The core collaborative control module 30 includes: a calorific value observation unit 31, a feature extraction unit 32, a feedforward coupling unit 33, and an air supply temperature prediction unit 34; wherein, the calorific value observation unit 31 is used to perform virtual calorific value back-calculation and output virtual calorific value parameters; the feature extraction unit 32 is used to extract the thermal center of gravity offset during the combustion cycle; the feedforward coupling unit 33 is used to generate feedforward correction amounts for the air-fuel ratio regulating valve 21, the cold air equalization valve 22, and the cold air mixing valve 23 based on the virtual calorific value, the thermal center of gravity offset, and real-time pressure, temperature, and other parameters; the air supply temperature prediction unit 34 is used to predict the air supply temperature decay trend based on the state characteristics of the preceding combustion and air supply stages, and to output feedforward correction commands to the cold air mixing valve 23.

[0034] See attached document Figure 2 , Figure 2 This is a flowchart of a hot blast stove control method according to an embodiment of the present invention. The present invention provides an intelligent combustion and external pressure equalization control method for a blast furnace hot blast stove, operating based on the hardware module architecture mentioned above, and includes the following steps: S10: Obtain multi-source discrete physical parameters during the operation of the hot blast stove and perform a first-order low-pass filter to eliminate signal glitches. S20, combined with the instantaneous flow rate and boundary temperature after cleaning, calculates the energy balance, reversely calculates the real-time fictitious calorific value and maps it to the compensation and adjustment parameters, and outputs it to the air-fuel ratio regulating valve 21 to perform the adjustment action; S30, after the combustion process is reversed, a spatial dimensionality reduction operator is established based on the downward accumulated momentum of the mixed airflow and the total heat exchange income, and the characteristic value of the thermal centroid offset is output. S40, after entering the forced pressure equalization state, calculate the dynamic pressure difference limit, input the thermal center of gravity offset characteristic value to calculate the dynamic damping coefficient, and output the superimposed and corrected fluid control command to the cold air pressure equalization valve 22. S50 After the hot blast stove completes pressure equalization and switches to the air supply mode, a state tensor containing the characteristics of the previous mode is constructed, input into the air supply temperature prediction model, the air supply temperature decay trend is calculated, and a feedforward compensation control command for the cold air mixing valve 23 is generated in combination with the real-time air supply parameters.

[0035] See attached document Figure 3 , Figure 3 This is a digital signal processing logic distribution diagram according to an embodiment of the present invention. In specific implementation, the data processing flow begins with the standardization conversion of the sensor's native electrical signal. For step S10, which involves acquiring the multi-source discrete physical parameters during the hot blast stove's operation and performing a first-order low-pass filter to eliminate signal glitches, the detailed execution path is as follows: In the specific signal acquisition phase, the control system first establishes a unified clock cycle. The electrical signals acquired by the flow detection unit 11, temperature detection unit 12, and pressure detection unit 13 included in the data acquisition module 10 are aggregated into the image register of the core collaborative control module 30 via the fieldbus. In this embodiment, the scanning sampling period is controlled. The time interval is set between 50ms and 200ms, preferably 100ms, to match the mechanical response frequency of heavy-duty valve actions. Continuous physical simulation quantities are discretized according to this period to generate physical variables. This parameter array comprehensively covers the instantaneous volumetric flow rate of coal gas. Instantaneous volumetric flow rate of combustion air Temperature measurement at the top of the hot blast stove Flue gas emission temperature Pressure measurement of the blast furnace cold blast main pipe and the instantaneous pressure inside the hot blast furnace .

[0036] For the basic circuit logic such as constant voltage drive and cold junction compensation around the sensor, those skilled in the art can select a common transmission processing scheme according to the field environment. The physical connection method is a well-known technology in the field and will not be described in detail here.

[0037] In the specific logical operation process, the core collaborative control module 30 performs first-order hysteresis smoothing on the above variables through a built-in recursive operator. Considering the stability at the initial stage of program startup, the system forcibly sets the filter value at the 0th sampling time. Equal to the original sample value This eliminates the unexpected step caused by recursive operations during the initialization phase. In each subsequent execution step... In this process, the filter eigenvalues ​​at the current time are obtained by calculating the following difference equation. : ; In the above formula, These are the discrete values ​​obtained from the current sampling. The filtered result stored in the RAM buffer from the previous control cycle. and These respectively constitute the weights of historical and current values. Key weighting factors in the formula. The configuration is adaptively determined based on the dynamic characteristics of the physical quantity, with its value strictly limited to between 0.05 and 0.3. The design principle behind this value range is that when... If the pressure is too small, it will cause severe phase lag in the system feedback, making it impossible to capture pressure changes during the equalization process in real time; while when... If the voltage is too high, it will be unable to effectively filter out the high-frequency noise interference caused by the surge of the booster compressor in the gas pipeline network.

[0038] Based on the weighted allocation requirements for control response agility, the core collaborative control module 30 implements differentiated damping time constant settings for physical variables with different attributes. (Calculation coefficients) With hardware response constant Relationship satisfaction This ensures The value is set between 0 and 1, and the filter strength is matched with the sampling period and the response speed of the measured physical quantity. This is used to process the feedback from the pressure detection unit 13. and At that time, set the time constant. Between 0.2s and 0.5s, this measurement range allows the control loop to capture and significantly reflect the actual dynamic pressure change caused by the action of the cold air equalization valve 22 within approximately three sampling cycles, thus avoiding excessive phase lag in pressure feedback relative to the action of the equalization valve. In contrast, the pressure feedback collected by the temperature detection unit 12... and Variables, setting time constants The time constant is between 2 and 5 seconds. Due to the significant thermal inertia of the thermocouple protective sheath, the large time constant can effectively filter out pseudo-random pulses caused by combustion flame disturbances, providing a smooth data envelopment line for the subsequent accurate extraction of the thermal centroid offset. Through this strategy, the filtered parameter sequence output by the system provides a physically consistent input benchmark for subsequent feedforward coupling calculations across operating conditions.

[0039] See attached document Figure 4 , Figure 4 This is a computational logic architecture diagram of a calorific value observation unit according to an embodiment of the present invention. The detailed implementation path of the fictitious calorific value estimation and compensation action involved in step S20 is described below: Upon entering the combustion process, the calorific value observation unit 31 first performs a boundary judgment on the effectiveness of the operating environment. Considering the nonlinear error of the flow transmitter at small openings and the mathematical calculation dead zone with a denominator of zero, the system presets a start-up flow limit. This value is typically taken as 15% of the rated gas flow rate of the hot blast stove. When the instantaneous volumetric gas flow rate is fed back by the flow detection unit 11... When the commutation system is in the combustion feedback position closed state, the energy budget equation solution process is triggered. If the above conditions are not met, the observer maintains the output state and displays the previous time step. As the current value, to ensure the continuity of control commands, the preset standard calorific value of the gas will be used when the system is first started or when there is no healthy historical value. As The initial value is set, and dynamic observation and calculation are initiated only after the effective combustion determination conditions are met.

[0040] The core collaborative control module 30 reads the measured temperature of the dome after smoothing in step S10. With flue gas emission temperature This maps the heat power absorbed by the heat storage body to the heat loss power carried away by the flue gas. Virtual calorific value The essential meaning is the equivalent calorific value of the gas required to cause a change in the energy within the heat storage chamber per unit time. This is an equivalent thermal state quantity constructed within the core collaborative control module 30, and does not represent the actual analytical calorific value of the gas. Its dimensions are consistent with the standard calorific value (kJ / m³). 3 The specific dynamic observation equations are as follows: ; In the above equations, the rules for determining the relevant variables and parameters are as follows: The equivalent mass (kg) of the arch refractory material participating in instantaneous heat exchange needs to be dynamically corrected based on the thermal conductivity depth of the refractory bricks.

[0041] As one possible implementation method, the baseline equivalent mass can be determined first during the commissioning calibration phase through step combustion tests or historical operating data. Then, based on the temperature range of the dome, the segmented calibration table is called to determine the current... When using a continuous correction method, the following formula can be used for calculation: ; in, and Corresponding to 8.0×10 4 kg to 1.2 × 10 5 The boundary values ​​of kg, For on-site calibration coefficients, and These are the lower and upper limits of the calibration temperature, respectively; The specific heat capacity of the refractory brick is taken as 1.1 kJ / (kg·℃). This represents the rate of temperature rise of the dome within a single sampling period; To determine the combustion thermal efficiency of the hot blast stove, considering both incomplete combustion losses and furnace shell heat dissipation, a calibration constant between 0.88 and 0.92 is selected. The weighted temperature of the inlet fluid can be taken as 30℃ as the reference constant; The average specific heat of the flue gas is taken as 1.42 kJ / (m³). 3 ·℃); This is a flue gas excess air coefficient correction factor used to correct deviations in flue gas volume under non-standard air-fuel ratio conditions. As one possible implementation method, it is first based on the standard air-fuel ratio coefficient... Calculate the real-time excess air ratio: ; Then Mapped to via amplitude limiting or calibration table : ; in, and Determined by on-site combustion calibration data.

[0042] Obtaining fictitious heat value Subsequently, the core collaborative control module 30 compares it with the preset standard calorific value of the gas. A comparison is performed to generate a feedforward gain for compensating the air-fuel ratio. As a preferred approach, to prevent malfunctions of the observer during instantaneous signal fluctuations, a low-sensitivity element with a dead zone width of ±1% is introduced into the system. Only when... Deviation Only when the amplitude exceeds the threshold is the corrected target value for the combustion air flow calculated using the following function. : ; In the formula, This is the standard air-fuel ratio coefficient of the hot blast stove at its rated calorific value. To compensate for the flow rate gain, it is typically set between 0.8 and 1.1 to ensure the damping characteristics of the regulation process. This mapping mechanism transforms the temperature change adjustment, which was originally a post-event feedback, into a pre-event predictive adjustment based on the flow rate setpoint.

[0043] Calculated The current is converted to the standard execution range. The air-fuel ratio regulating valve 21 within the execution module 20 receives a current command from the core collaborative control module 30. In this embodiment, this command signal is converted into torque to drive the servo mechanism via a linear power drive circuit. The air-fuel ratio regulating valve 21 then changes the cross-sectional area of ​​the medium flow channel, adjusting the instantaneous volumetric flow rate of the combustion air. This causes the temperature rise curve of the dome to tend towards the preset process trajectory.

[0044] For the valve position feedback verification and overcurrent protection mechanism of the air-fuel ratio regulating valve 21, those skilled in the art can use conventional servo control boards to implement it. The implementation of these hardware interfaces and drive protocols are well-known technologies in the field and will not be described in detail here.

[0045] See attached document Figure 5 , Figure 5 This is a logical mapping diagram of a thermal centroid offset algorithm according to an embodiment of the present invention. In this embodiment, the thermal centroid offset ( The extraction of energy is performed by the feature extraction unit 32 in the core collaborative control module 30. The feature extraction unit 32 provides discrete constraints for damping control during the subsequent pressure equalization period by quantifying the spatial coupling relationship between the fluid momentum integral and the internal energy gradient within the combustion cycle. Energy dynamic integration is performed using the cleaned parameter data to establish a spatial mapping model reflecting the longitudinal heat distribution of the hot blast stove. The specific implementation path of this process is as follows: During the combustion cycle, the feature extraction unit 32 performs spatiotemporal accumulation of the fluid jet kinetic energy at fixed intervals. Since the heating of the hot blast stove regenerator is a downward-penetrating heat transfer process, the greater the mechanical momentum of the fluid ejected from the burner, the deeper the high-temperature zone axis tends to shift downwards. Specifically, the system utilizes the instantaneous volumetric flow rate of the pretreated gas. Instantaneous volumetric flow rate of combustion air Calculate the total momentum flow of the mixed gas entering the regenerator. This is done by solving for the jet momentum integral. The estimated penetration depth of thermal energy along the longitudinal axis of the heat storage chamber is calculated using the following formula: ; The relevant variables in the formula are defined as follows: This represents the total number of sampling execution steps for a single combustion cycle. Real-time dynamic density of the gas mixture (kg / m³) 3 As a preferred method, this value is not a fixed constant, but is obtained based on the conservation of fluid components at each scan step, i.e. ,in and These are the calibration densities of coal gas and air under cold standard operating conditions, respectively. Effective cross-sectional area of ​​the combustion duct of the hot blast stove (m²) 2 The value is determined by the physical dimensions of the equipment as shown in the original drawings. This value is fixed at the end of combustion, and its range of variation directly reflects the intensity of momentum sinking.

[0046] As momentum accumulates, the core collaborative control module 30 synchronously extracts the thermodynamic boundary features at both ends of the heat storage body to reconstruct the heat gradient distribution model. The system tracks the measured temperature of the dome cycle by cycle. With flue gas emission temperature The system extracts the characteristic temperature difference between the two components at the end of the combustion cycle. Considering that the cold purging at the beginning of ignition and the initial fluctuations in the pipeline network often lead to pseudo-phase heat conduction, the system shields the data from the earlier stages and only extracts the last 60% of the total combustion cycle time as the effective decision integration window. Within this effective time window, the arithmetic mean temperature of the dome and the flue is calculated separately to construct a steady-state gradient characterization term for the longitudinal temperature.

[0047] Calculate the thermal centroid offset in conjunction with the above physical processes. In this embodiment, Defined as a dimensionless characteristic vector that quantitatively describes the deviation of the thermodynamic core from the design geometric center of the regenerator. The control unit incorporates the following comprehensive solution mapping equation: ; In this mapping formula: The total chemical energy (kJ) input into the system during a specific combustion cycle is accumulated. and These are the dome and the average flue gas temperature within the integration window, respectively. As a weighting coefficient for multi-source fusion, it effectively allocates the proportion of momentum effect and heat conduction effect. The value of this coefficient is adaptively calibrated based on the process operating load, typically configured in the range of 0.4 to 0.75. When the hot blast stove operates under enhanced high blast temperature conditions and air-fuel ratio... When the flow rate is too high, causing a sharp increase in flow velocity, The value is shifted towards the 0.75 side to enhance the weighted extraction of tropospheric momentum. This is to avoid issues during short-term operation of unconventional furnaces. If the value is too small, it will cause a division overflow exception in the microprocessor. The algorithm forces the denominator to be assigned a safe boundary constant. When the actual accumulated value is lower than At this time, the calculation of this formula is suspended, and the health history value of the previous complete cycle is used. Under normal calculation, The standard range will smoothly converge within the scale interval [-1, 1].

[0048] When the combustion operation enters the equalization preparation dead zone before the shutdown or transfer of air, the core collaborative control module 30 obtains the current cycle... Parameters are latched. This reduced-dimensional feature, as a state label reflecting the current transient energy storage distribution characteristics of the equipment, is transferred to step S40 to adjust the opening and closing envelope of the cooling air equalization valve 22. For example, it is observed that... Approaching the upper limit of the extreme value indicates that heat accumulation is severely retained in the surface area of ​​the dome. The normal pressurization rate may induce airflow swirl at the top and risk of refractory spalling. As a result, the system generates an intervention command to increase damping in advance.

[0049] For the local thermal stress variation of the lining material in the heat storage chamber under different operating temperature zones and its disturbance law on the gas film resistance coefficient, engineers can obtain a typical reference system based on the Metallurgical Thermal Engineering Handbook. The underlying physical laws are well-known technical frameworks and will not be elaborated here. Subsequently, the equipment condition characteristics are formally incorporated into the aerodynamic intervention and control.

[0050] See attached document Figure 6 , Figure 6 This is a logic block diagram for generating equalization process adjustment commands according to an embodiment of the present invention. In this embodiment, the feedforward coupling unit 33 receives the thermal centroid offset output by the feature extraction unit 32. (and combined with the real-time differential pressure collected by the pressure detection unit 13) A feedforward correction opening command is generated for the cold air equalizing valve 22. Combining the thermal center of gravity offset characteristics with the real-time pressure difference, a nonlinear opening command trajectory for the cold air equalizing valve 22 is generated to suppress the risks of physical oscillation and thermal shock during the switching process. The specific implementation path of this process is as follows: The equalization condition is initiated and a pressure reference is established. When the core collaborative control module 30 receives a combustion end signal, and the operating condition interlock signal acquisition unit 16 confirms that both the existing combustion air switching valve and the existing gas switching valve are in a locked feedback state, the equalization subroutine logic is triggered. The aforementioned combustion air switching valve and gas switching valve can be existing valves in the original reversing system of the hot blast stove. The control system of this invention reads their status feedback signals through hardwiring or fieldbus, without being limited to their specific mechanical structure. The pressure detection unit 13 synchronously senses and measures the pressure in the cold air main pipe. Instantaneous pressure inside the hot blast furnace The system obtains the pressure difference characteristic by subtracting the two instantaneously, i.e. This serves as the original dynamic boundary parameter for driving the cold air medium filling the furnace.

[0051] Based on thermal centroid offset Calculate the equalization rate correction factor The longitudinal non-uniformity of heat distribution within the hot blast furnace directly affects the local gas expansion and density gradient. When the thermal center of gravity is located in the upper part (manifesting as...), When the calculated value is positive and relatively large, if the equalizing valve is opened at a constant rate as usual, the rigid cold airflow will inevitably impact the high-temperature refractory material of the arch instantly, causing not only a rapid pressure pulse but also potential spalling of the refractory lattice surface. To avoid this effect, the system incorporates a correction factor. The calculation mapping relationship is as follows: ; In the above mapping equation, and The model sensitivity coefficients are designed to take into account the equipment configuration. Among them... Based on experience, the value is planned to be between 0.05 and 0.15. Typically calibrated between 0.8 and 1.5. This is due to the upward evolution of the thermal accumulation center. When the value increases, the algorithm forcibly lowers the correction factor. This establishes a reverse-action suppression source to slow down the valve's acceleration in the next stage.

[0052] A target full-stroke opening trajectory model of the cold air equalizing valve 22 is constructed. After obtaining the correction factor as the intervention parameter, the core collaborative control module 30 generates the ideal opening command for each control sampling cycle by discretizing the time coordinates. To ensure a smooth progression of the furnace pressure curve, the exponentially increasing mathematical model with damping attenuation characteristics built into this embodiment is as follows: ; The specific physical meanings of the variables involved in the formula include: Set as the final target stroke of the valve core when the pressure equalization logic terminates (the normal field configuration is 100%). The action time constant used for calibration is given based on the mechanical no-load full stroke time of the actuator matched with this series of pressure equalizing valves. In this embodiment, the range of 5s to 12s is selected. The sampling operation reference period of the microcontroller is specified, which is consistent with the control scan sampling period in step S10, and is between 50ms and 200ms, preferably 100ms. This represents the number of discrete cycle steps accumulated since the initial issuance of the equalization command. The calculated correction factor... Multiplicative decay limits are applied to the response speed of the exponential term. In terms of physical control effects, the system enhances the throttling hydraulic resistance effect at the valve orifice by lengthening the residence transition time in the low-opening region of the equalizing valve, thereby softening the airflow peaks that cause cold air backflow into the upper heat storage area.

[0053] The execution command output and operating condition closed-loop verification are performed. After the calculation is completed, the core collaborative control module 30 will output the data from each node. The signal is equivalently transformed into a 4-20mA standard industrial analog signal and sent to the pressure equalization valve servo speed control device installed in execution module 20. As a preferred method, to prevent pressure oscillations and backflow overshoot caused by sudden changes in the duct network during the later stages of pressure equalization, the system concurrently activates the pressure change rate limit. Perform hard intervention. When the real-time rate of increase in the internal pressure extracted from the feedback exceeds... When the threshold (a representative value of 0.05 MPa / s is given in this embodiment) is reached, the control side immediately activates the self-protection constraint locking function, stops accumulating the valve core opening increment, and automatically resumes operation after the gas pressure inside the furnace has been buffered and tends to stabilize. Considering the need to prevent the furnace changeover process from being interrupted due to occasional aerodynamic disturbances within this protection mechanism, an additional safety defense is incorporated into the system: a maximum tolerance time for global pressure equalization is set. (If set to 45s), if this timeout occurs, the original optimization logic permission will be revoked, and the preset conservative safety rate will be switched to continue advancing the opening degree of the cold air equalizing valve 22, while retaining the overpressure interlock, pressure change rate interlock and valve position abnormality interlock; when any interlock condition is triggered, the valve will stop opening or enter the fault holding state, instead of unconditionally opening the valve.

[0054] For the on-site sealing verification and over-torque current warning of the selected cold air equalizing valve 22, and other peripheral supporting assembly technologies, those skilled in the art can obtain solutions according to the corresponding control panel and general valve body construction specifications. This falls within the known technical scope of electromechanical control and fluid cut-off devices, and will not be elaborated here.

[0055] The completion of this step indicates a smooth transition in the pressurization process based on the predicted space energy storage conditions. Finally, when the detection... and The static pressure deviation shrinks into the equilibrium judgment tolerance zone (e.g.) After that, the main command sequence drives the cold air equalization valve 22 to lock the full stroke position, and then continuously opens the main cold air channel, so that the hot air furnace can seamlessly transition into the working air supply sequence.

[0056] See attached document Figure 7 , Figure 7This is a topological diagram of the air mixing temperature prediction model according to an embodiment of the present invention. After the hot blast stove completes pressure equalization and successfully enters the air supply process, ensuring that the high blast temperature continuously supplied to the blast furnace remains stable becomes an indicator for measuring the overall thermal efficiency of the hot blast stove. In this embodiment, the control logic shifts from reversing action constraints to steady-state air supply temperature optimization. Combining the previously extracted physical heating characteristics, the air supply temperature prediction unit 34 in the core collaborative control module 30 predicts the air supply temperature decay trend based on the embedded prediction model, and generates a feedforward compensation control command acting on the cold air mixing valve 23 through the feedforward coupling unit 33. The specific implementation path of this process is as follows: Construction of multidimensional state feature tensors during the initial stage of air supply. After the external pressure equalization process is completed and the hot blast stove officially enters the air supply mode, the core collaborative control module 30 synchronously starts the air supply optimization calculation process. At this time, the system collects real-time instantaneous operating data and assembles it with the feature parameters extracted from the previous process. In order to establish an input-side data representation that conforms to the specifications of a neural network, the system performs truncation and normalization preprocessing on the process variables.

[0057] Extract the constructed input state tensor The state variables include five dimensions: the steady-state fictitious heat value calculated in the preceding step S20. Step S30 latching thermal centroid offset Current mainstream air conditioning pipeline flow rate Initial temperature of the vault and the target blast furnace blow temperature For the aforementioned tensor, the system uses the device's historical operation database to extract the upper and lower edge values ​​of the physical limits of the corresponding parameters as a fixed scaling scale, and linearly maps the real-time data to the [0,1] interval. This avoids the neuron activation saturation phenomenon caused by abnormal peak values ​​exceeding the limit, and at the same time eliminates gradient calculation deviations caused by different dimensions.

[0058] This paper proposes a deep sensing network-based method for predicting the rate of heat loss due to heat loss. In the nonlinear heat release process of a heat storage chamber, conventional single-loop PID feedback regulation often induces heat loss oscillations due to significant hysteresis in the thermodynamic process. To improve this control performance, this embodiment constructs a recurrent neural network (LSTM) model with a long short-term memory mechanism to extract the dynamic decay law of heat release. This model is deployed within the core collaborative control module 30, and its internal hierarchy and data flow are clearly defined as follows: preprocessed data is input to an input layer containing 5 nodes; the input layer data is sequentially transmitted to the first hidden layer composed of 64 LSTM neurons to capture the heat loss time series characteristics over a time span; the state vector output by this layer is further unidirectionally fully connected to the second hidden layer containing 32 neurons to perform nonlinear spatial feature mapping; finally, the data is aggregated to the output layer containing a single node. The output result represents the estimated value of the wind loss slope within a future prediction window (e.g., 5 minutes). (℃ / min).

[0059] For this predictive model, its network parameters need to undergo offline optimization training before the equipment is put into operation. The sample set for the model training phase comes from archived historical data of the stable air supply cycle curves of the isomorphic hot blast stove over the past year. The sample label is defined as the actual rate of temperature decrease within the next 5 minutes for the corresponding data point. The mean squared error is used as the loss function to measure the bias during the training process. ; In the formula, This represents the total number of samples used in a single gradient calculation. and These refer to the inferred decay rate and the true labeled decay rate of the network at the same time slice, respectively. The network optimization process uses the Adam algorithm to update the learning rate parameter, with an initial learning efficiency pre-configured to 0.001. After multiple batches of forward mapping and error backpropagation operations, when the validation set samples... When the value stably falls back to within the set allowable deviation bandwidth (0.1 in the example), the network weight matrix is ​​extracted and encapsulated in an engineering manner. The encapsulated weight matrix, normalized upper limit value, normalized lower limit value, and network hierarchy parameters are written into the non-volatile storage area of ​​the core collaborative control module 30. During online operation, the supply air temperature prediction unit 34 only performs forward inference calculations and does not perform online backpropagation updates; when the input tensor is missing, the normalized data exceeds the limit, or the model output triggers the physical hard upper limit, the core collaborative control module 30 switches to the traditional proportional-integral-derivative (PID) control mode.

[0060] Generates a wind-mixing compensation command based on the attenuation prediction. Obtains the estimated wind temperature attenuation slope calculated by the model. Subsequently, the system performs feedforward calculations to reconstruct it. Under normal settings, the cold air mixing valve 23 is controlled by the reference table and deviation adjustment. In this embodiment, it is extended to an intervention model with forward prediction weights: ; The physical definitions and configuration rules for the relevant parameters include: That is, the final issued control opening value of the cold air mixing valve; As a steady-state baseline for wind speed opening, this value is obtained from locally stored data based on the target wind temperature using a two-dimensional interpolation algorithm. The total air volume is retrieved from the calibration mapping matrix indexed by the total air volume, which can be directly measured by the air flow detection unit 14. When only the main cold air duct flow detection is configured on site, the current main cold air duct flow rate will be used. It is used as an equivalent input to the total air supply volume in two-dimensional interpolation; Measured air temperature for the main pipeline; This is the standard proportional error suppression coefficient; As a feedforward weighting compensation coefficient, in order to prevent excessive valve actuation, its effective range is usually specified to be between 0.2 and 0.6; The preset prediction compensation advance step size constant is generally within the range of 60s to 120s. During actual operation, if the model extracts... Abnormally high temperatures indicate a situation where there is significant heat storage in the upper part but a low temperature gradient in the middle and lower parts. This allows for the early prediction of a sharp drop in air temperature during the later stages of air supply, thus enabling [further action]. The device actively applies negative compensation, prompting the control end to shrink the cross-section of the mixing valve earlier.

[0061] Algorithm constraint verification and dead-zone prevention logic execution. Considering that occasional instrument failures or sudden pipeline fluctuations may distort the air supply data, causing the model to output an overloaded attenuation prediction, this could lead to the unexpected complete blockage of the mixing channel. As a preferred approach, a threshold trap defense is pre-defined in the underlying logic sequence. When the predicted value... When the calculation results trigger the hard upper limit of the physical constants configured in the system (e.g., a set value of ≥15℃ / min), the control side triggers a safety interruption mechanism. This mechanism temporarily revokes the output authority of the feedforward compensation intervention branch and smoothly transitions the control to the traditional self-sustaining proportional-integral-derivative adjustment module for conservative maintenance. This state continues until the tensor features have been used for three consecutive model inference cycles or until the preset confirmation time is reached. The system returned to its normal operating range, and then the bypass takeover was revoked. The timeout period can be set from 5 to 30 seconds based on the inertia of the supply air temperature to avoid premature exit from the safety bypass mode due to instantaneous noise recovery. This design mechanism eliminates model prediction drift caused by the misalignment of a single sensor and ensures the continuity of the execution module's functions.

[0062] Regarding the high-voltage isolation and filtering of the cold air mixing valve 23 actuator and the bus communication error correction in the control loop, engineering personnel can directly call the industrial standard library components according to the relevant specifications. The related drive logic construction is a well-known technology in the field of conventional industrial automation and will not be elaborated here. With the smooth execution of valve position control, the entire hot air furnace equipment enters a highly efficient and steady-state air supply state that is highly matched with the target requirements.

[0063] This solution was applied to a 4500m high-rise building in a steel plant. 3 The blast furnace is equipped with a Kalugin-top top-fired hot blast stove, which consists of four hot blast stoves that operate in rotation.

[0064] Operating conditions and system configuration: Key parameters of the object: Maximum design vault temperature: 1350℃; Rated supply air temperature: 1250℃; Rated gas flow rate: 80000 m³ / h 3 / h; Rated combustion air flow rate: 72000m³ / h 3 / h; Base gas calorific value ( 3500kJ / m 3 Standard air-fuel ratio (SFR) ):0.9; Equalization time: <45s.

[0065] System hardware configuration: Core collaborative control module 30: Siemens S7-1500 series PLC is selected as the core collaborative control module, and the algorithm described in this solution is embedded.

[0066] Data Acquisition Module 10: All sensors use standard 4-20mA output industrial-grade transmitters, connected to the control network via a remote I / O station. The system controls the scanning sampling cycle. Set to 100ms.

[0067] Execution module 20: The air-fuel ratio regulating valve 21, the cold air equalizing valve 22, and the cold air mixing valve 23 are all equipped with electric regulating valves with intelligent positioners to ensure that the response time is less than 3 seconds.

[0068] Step S10: The data acquisition and filtering control system acquires various physical quantities at 100ms intervals. For example, at time k, fluctuations in the pipeline network cause pressure changes in the blast furnace cold blast main pipe. The original signal exhibited glitches, with an instantaneous value of 350.5 kPa, while the previous cycle's filtering result... The pressure is 350.1 kPa. This is based on the set pressure signal weighting factor. Apply the first-order low-pass filter formula: ; The filter value for the current period is: ; The smoothed value will be used in subsequent calculations, effectively avoiding interference from process noise.

[0069] Step S20: During the 10-minute fictitious calorific value observation and air-fuel ratio compensation combustion, the actual calorific value decreases due to fluctuations in gas composition. The calorific value observation unit 31 then uses real-time data (…) , Temperature rise rate of the dome , ), and calculate the fictitious heat value for the current period. 3380 kJ / m 3 This value deviates from the reference calorific value. The amplitude has exceeded the ±1% dead zone. The system immediately initiates compensation, based on the compensation adjustment formula: ; Calculate the target combustion air flow rate (assuming compensation gain). =1.0): ; The control module then adjusts the command of the air-fuel ratio regulating valve 21 to increase the combustion air flow. from Downgraded to The air-fuel ratio is dynamically optimized to ensure that the dome temperature rises steadily along a preset trajectory.

[0070] Step S30: Extraction of thermal centroid offset features: After 60 minutes of combustion, the reversal condition is met. Feature extraction unit 32 processes the data from the last 36 minutes (the last 60%) of the combustion cycle to obtain: Integral value of jet flow 4.5×10 8 kg·m / s; Total input chemical energy 1.6×10 11 kJ; Effective temperature uniformity of the vaulted ceiling inside the window 1345℃; Effective uniform temperature of smoke inside the window 95℃ based on weighting coefficient The mapping equation is solved by applying the thermal centroid offset: ; Calculate the thermal center of gravity offset: ; Calculated This positive value indicates that the heat accumulation during this combustion is biased towards the upper part of the front hot chamber. This characteristic value is latched to guide the subsequent pressure equalization process.

[0071] Step S40: Adaptive pressure equalization control: The furnace is switched to pressure equalization mode. The system reads the latched data. Based on the model sensitivity coefficient The mapping relationship is calculated using the equalization rate correction factor: ; The calculation yielded: ; The correction factor A trajectory model of the opening degree directly acting on the cold air equalizing valve 22. Compared to a fixed-rate opening ( The valve opening speed in the initial stage is significantly slowed down, reducing the initial impact of cold air entering the furnace and achieving flexible pressurization of high-temperature gas in the furnace, thereby avoiding direct thermal shock to the dome refractory material.

[0072] Step S50: Supply air temperature prediction and feedforward control: Pressure equalization is complete, and the furnace switches to air supply. The air supply temperature prediction unit 35 immediately constructs the normalized input tensor: ; The deployed LSTM model predicts the wind-temperature decay slope over the next 5 minutes. -9.5 ℃ / min. This is a relatively fast decay rate, indicating a risk of rapid temperature drop in the middle and later stages of the air supply due to the upward distribution of heat. The system immediately generates a feedforward compensation command. The feedforward compensation amount is calculated according to the following formula: ; Substitute the value (setting) , ): ; Final mixing valve opening command The original PID calculation results were superimposed with a 5.7% feedforward. This means that the system reduced the opening of the cold air mixing valve 23 in advance, actively reducing the amount of cold air mixed, so as to accurately offset the upcoming cooling trend and maintain a high degree of stability in the supply air temperature.

[0073] Experimental verification and effect comparison: To verify the system's effectiveness, a month-long comparative test was conducted on the same hot blast furnace. The factory's existing PID control system (control group) was used for the first two weeks, and the intelligent control system described in this solution was switched to for the last two weeks (experimental group).

[0074] Evaluation indicators: Temperature control accuracy of the dome: the average absolute deviation between the measured temperature of the dome during the combustion period and the target curve.

[0075] Pressure shock during the equalization process: the peak rate of pressure rise in the furnace during the equalization period.

[0076] Supply air temperature stability: The range of fluctuation between the hot air temperature during the supply air period and the set value (1250℃).

[0077] Gas consumption per unit: The standard volume of gas consumed per unit of heat output (GJ).

[0078] Comparison of experimental results: ; Comparative tests with traditional PID control show that this control system reduces total gas consumption by 2.6%. This energy-saving achievement is attributed to the comprehensive optimization of the system's three core stages: combustion, reversing, and air supply. Specifically, the control accuracy of the dome temperature is improved by 66.7%; pressure surges during the reversing and equalization process are reduced by 43.5%; and the stability of the air supply temperature is improved by 60%.

[0079] Appendix Figure 8 The temperature control effect during the combustion cycle was demonstrated. The hypothetical calorific value feedforward compensation introduced in this scheme enables the system to predict and offset real-time fluctuations in the calorific value of the gas. Therefore, the temperature curve of the experimental group (solid line) closely tracks the target setpoint (dashed line), while the control group with PID control exhibits significant overshoot and oscillation due to response lag.

[0080] The control differences during the commutation phase are reflected in the appendix Figure 9 The system achieves a smooth pressure equalization process. By adaptively adjusting the equalization rate based on the thermal center of gravity shift, the system achieves a gradual pressure rise, fundamentally eliminating the pressure surge upon cold air intake. In contrast, the rigid control logic of the control group produces a clear pressure spike in the graph.

[0081] The stability of the supply air temperature after entering the air supply phase is shown in the appendix. Figure 10 The LSTM model endowed the system with predictive capabilities, enabling it to intervene in advance and offset the nonlinear decay of heat release from the thermal storage body. As a result, the experimental group maintained a highly stable wind temperature throughout the entire cycle. In contrast, the lag adjustment in the control group led to the temperature tailing phenomenon commonly seen at the end of the air supply period.

[0082] Finally, attached Figure 11 The bar chart summarizes the quantitative comparison of the above key performance indicators. Whether it is control accuracy, shock suppression, or temperature stability, this solution demonstrates clear advantages, and these cumulative improvements have collectively contributed to the final energy-saving goal.

[0083] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart combustion and external pressure equalization control system for a blast furnace hot blast stove, characterized in that, include: Multiple detection units are used to collect operating parameters of the hot blast stove; At least one execution module, the execution module including an air-fuel ratio regulating valve and a cold air equalizing valve; The core collaborative control module, connected to the detection unit and the execution module, is used for: In the combustion process, based on the parameters collected by the detection unit, a fictitious calorific value representing the change in energy in the heat storage chamber is calculated by solving the energy balance equation, and a compensation control command for the air-fuel ratio regulating valve is generated based on the deviation between the fictitious calorific value and the standard calorific value of the gas. At the end of the combustion cycle, based on the parameters collected by the detection unit, the thermal center of gravity offset reflecting the energy storage distribution characteristics is calculated and locked. In the subsequent pressure equalization process, the opening process of the cold air pressure equalization valve is adaptively damped based on the thermal center of gravity offset.

2. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, When calculating the fictitious calorific value, the core collaborative control module adds the heat power absorbed by the heat storage body to the heat loss power carried away by the flue gas based on the collected dome measurement temperature and flue gas emission temperature, and divides it by the product of the combustion thermal efficiency and the instantaneous volumetric flow rate of the gas.

3. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, When generating compensation control commands for the air-fuel ratio regulating valve, the core collaborative control module initiates compensation calculations only when the fictitious calorific value deviates from the standard calorific value of the gas by more than a preset dead zone threshold.

4. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, When calculating the thermal centroid offset, the core collaborative control module uses a centroid weighting coefficient to weight and fuse the jet momentum integral value, which characterizes the thermal energy penetration depth, with the longitudinal temperature steady-state gradient characterization term, which characterizes the thermal gradient distribution.

5. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, When implementing adaptive damping control on the cold air equalizing valve, the core collaborative control module generates an equalizing rate correction factor based on the thermal center of gravity offset through an exponential mapping relationship, and substitutes this correction factor into an exponential boosting model with damping attenuation characteristics to construct the target full-stroke opening trajectory of the cold air equalizing valve.

6. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, When implementing adaptive damping control on the cold air equalization valve, the core collaborative control module also includes a safety defense logic. This logic is used to stop the valve core opening increment when the furnace pressure rise rate is detected to exceed a preset threshold; or to forcibly switch to a preset conservative safety rate when the equalization process exceeds the global maximum tolerance time.

7. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, The execution module also includes a cold air mixing valve; and the core collaborative control module is also used to predict the estimated value of the air temperature attenuation slope through a pre-trained recurrent neural network model in the air supply process, and generate a feedforward compensation control command for the cold air mixing valve based on the estimated value.

8. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 7, characterized in that, In the air supply process, the core collaborative control module is used to construct a multi-dimensional state feature tensor that includes the fictitious calorific value, thermal centroid offset, current cold air mainstream pipeline flow rate, initial dome temperature, and target air supply temperature, and uses it as the input to the recurrent neural network model.

9. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 8, characterized in that, When generating the feedforward compensation control command for the cold air mixing valve, the core collaborative control module multiplies the estimated value of the air temperature attenuation slope by a feedforward weight compensation coefficient, and calculates the final valve control opening value by combining the steady-state basic opening benchmark and the proportional error.

10. The intelligent combustion and external pressure equalization control system for a blast furnace hot blast stove according to claim 1, characterized in that, The core collaborative control module is also used to set differentiated damping time constants for different physical variables such as pressure, temperature and flow rate and perform first-order hysteresis smoothing processing before processing the parameters collected by the detection unit.

Images