Cooperative combustion control method and system for tunnel kiln

Abstract

The invention provides a collaborative combustion control method and system for a tunnel kiln, and relates to the technical field of tunnel kilns. The method comprises the following steps: dividing the tunnel kiln into a plurality of independent control areas; key process parameters of all the control areas are collected in real time, and all the key process parameters are preprocessed; according to the preprocessed key process parameters of each control area, a fuzzy PID controller, a feedforward prediction model and an AI air-fuel ratio dynamic optimization model, obtaining a fine adjustment quantity, a pre-adjustment quantity and an air-fuel ratio optimization coefficient of each control area; obtaining a final execution instruction; state data during combustion of the tunnel kiln are obtained, and the state data are fed back in real time; inputting the preprocessed key process parameters of each control area into a digital twinborn model for off-line simulation and on-line verification; according to the method, an integrated system of partition regulation and control, fuzzy self-adaption, feedforward pre-judgment and digital simulation optimization is formed, and efficient and accurate control over the combustion process is achieved.

Description

Technical Field

[0001] This application relates to the field of tunnel kiln technology, specifically to a method and system for coordinated combustion control in tunnel kilns. Background Technology

[0002] A tunnel kiln is a modern, continuous firing thermal equipment constructed from refractory, insulation, and building materials. It contains kiln cars and other transport vehicles and is mainly used for the firing of refractory materials, ceramic products, and carbon products. It is also used in metallurgical industries such as abrasives.

[0003] Combustion control in tunnel kilns is a core element in ensuring product quality and reducing energy consumption. Existing combustion control technologies have the following key defects: (1) Poor adaptability of control algorithms: Traditional PID control uses fixed parameters, which is difficult to match the large inertia, nonlinearity, and time-varying characteristics of tunnel kilns. When the type of raw materials or production load changes, it is easy to have large overshoot and response lag, resulting in insufficient temperature control accuracy; (2) Lack of prediction and coordination mechanisms: Existing controls are mostly feedback controls, which only adjust based on the current temperature deviation and do not consider the influence of factors such as the kiln car's advancing speed and the temperature at the car's entry end. The prediction of heat load changes is insufficient, and the response timeliness is poor; (3) Insufficient coordination of zone control: The heat load of different process sections such as the heating section, calcination section, and heat preservation section of the tunnel kiln is not fully coordinated. The differences in industrial needs are significant. Existing technologies mostly adopt a unified control strategy, which does not achieve precise control of zones, resulting in poor temperature uniformity in each section and affecting product consistency; (4) Energy consumption optimization and combustion efficiency are disconnected: the air-fuel ratio control is mostly a fixed set value, without dynamic optimization combined with real-time temperature and exhaust parameters, which easily leads to incomplete combustion of fuel or excessive air supply, resulting in energy waste and excessive emissions; (5) Lack of digital simulation support: the adjustment of control strategy depends on on-site testing, which is time-consuming, costly, and cannot predict the adjustment effect in advance, making it difficult to achieve the global optimal control parameters.

[0004] In existing technologies, some improvement solutions only optimize PID parameters or add simple feedforward logic, which cannot fundamentally solve the problems of accuracy, coordination and economy in tunnel kiln combustion control. Summary of the Invention

[0005] To address one of the aforementioned technical deficiencies, this application provides a method and system for coordinated combustion control in tunnel kilns.

[0006] According to the first aspect of this application, a method for co-combustion control in a tunnel kiln is provided, comprising:

[0007] Based on the physical structure, process layout and thermal characteristics of the tunnel kiln, the tunnel kiln is divided into multiple independent control zones.

[0008] Key process parameters of each control area are collected in real time, and the collected key process parameters of each control area are preprocessed; each key process parameter includes: real-time temperature value of the area, set temperature value, kiln car advance speed, gas flow rate, combustion air flow rate, flue gas temperature, temperature of key points in the kiln and oxygen content of flue gas.

[0009] Based on the key process parameters of each control region after preprocessing, the fuzzy PID controller, the feedforward prediction model, and the AI ​​air-fuel ratio dynamic optimization model, the fine adjustment amount, the pre-adjustment amount, and the air-fuel ratio optimization coefficient of each control region are obtained; the fine adjustment amount includes the gas adjustment amount and the combustion air adjustment amount, and the pre-adjustment amount includes the gas pre-adjustment amount and the combustion air pre-adjustment amount.

[0010] Based on the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control area, the final execution command is obtained;

[0011] Based on the final execution command, the state data during the combustion of the tunnel kiln is obtained, and the state data is fed back in real time; the state data includes actual temperature, actual pressure, and emission indicators.

[0012] The key process parameters of each control region after preprocessing are input into the digital twin model for offline simulation and online verification, and the simulation results and verification results are output.

[0013] Preferably, the step of obtaining the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient for each control region based on the key process parameters of each control region after preprocessing, the fuzzy PID controller, the feedforward prediction model, and the AI ​​air-fuel ratio dynamic optimization model specifically includes:

[0014] Based on the preprocessed real-time temperature value and set temperature value of each control area, the temperature deviation and temperature deviation change rate of each control area are calculated respectively.

[0015] The calculated temperature deviation and temperature deviation change rate of each control area are input into the fuzzy PID controller to obtain the gas regulation amount and combustion air regulation amount of the corresponding control area.

[0016] The acquired kiln car end temperature, historical heat load data of each control area, and preprocessed kiln car advance speed of each control area are input into the feedforward prediction model to predict the heat load change of each control area, and the gas pre-regulation amount and combustion air pre-regulation amount of the corresponding control area are calculated based on the prediction results.

[0017] The pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature in the kiln, and flue gas oxygen content of each control area are input into the AI ​​air-fuel ratio dynamic optimization model to obtain the air-fuel ratio optimization coefficient of each control area.

[0018] Preferably, after inputting the calculated temperature deviation and temperature deviation change rate of each control area into the fuzzy PID controller, the gas regulation amount and combustion air regulation amount of the corresponding control area are obtained, specifically including:

[0019] The calculated temperature deviation and temperature deviation rate of change of each control region are input into the fuzzy PID controller for fuzzification processing to obtain the temperature deviation membership degree and temperature deviation rate of change membership degree of each control region.

[0020] Based on a pre-set fuzzy rule base, fuzzy reasoning is performed on the membership degree of temperature deviation and the membership degree of temperature deviation change rate of each control area to obtain the output fuzzy set;

[0021] The obtained output fuzzy set is defuzzified to obtain parameter adjustment amounts; the parameter adjustment amounts include proportional gain (Kp) adjustment amount, integral gain (Ki) adjustment amount, and derivative gain (Kd) adjustment amount;

[0022] The updated real-time parameters are obtained based on the parameter adjustment amount and the initial parameters (or parameters from the previous cycle) of the fuzzy PID controller; the updated real-time parameters include Kp. * Ki * and Kd * ;

[0023] Based on the updated real-time parameters and the temperature deviation and temperature deviation rate of each control area, standard PID control is used to calculate the preliminary control output of each control area.

[0024] Based on the preset gas-to-heat conversion coefficient and the reference air-fuel ratio, the preliminary control output of each control area is converted into the gas regulation amount and combustion air regulation amount of the corresponding control area.

[0025] Preferably, the step of inputting the acquired kiln car end temperature, historical heat load data of each control area, and preprocessed kiln car advance speed of each control area into the feedforward prediction model to predict the heat load change of each control area, and calculating the gas pre-regulation amount and combustion air pre-regulation amount of the corresponding control area based on the prediction results, specifically includes:

[0026] Acquire the kiln car end temperature and historical heat load data for each control area;

[0027] The preprocessed kiln car advance speed and kiln car end temperature of each control area are input into the linear regression layer of the feedforward prediction model to calculate the baseline prediction value of each control area.

[0028] The historical heat load data and baseline prediction values ​​of each control area are input into the neural network layer of the feedforward prediction model to obtain the nonlinear correction amount for each control area.

[0029] The heat load change of each control area is calculated based on the baseline predicted value and nonlinear correction amount of each control area.

[0030] Based on the calculated heat load changes in each control area, the pre-regulation amounts of gas and combustion air for the corresponding control areas are calculated.

[0031] Preferably, the step of calculating the gas pre-regulation amount and combustion air pre-regulation amount for the corresponding control area based on the calculated heat load change amount for each control area specifically includes:

[0032] According to the formula The pre-regulation amount of gas ΔQr2 is calculated; where ΔQ represents the change in heat load.

[0033] According to the formula The pre-adjustment amount of combustion air ΔQa2 was calculated.

[0034] Preferably, the step of inputting the pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature in the kiln, and flue gas oxygen content of each control zone into the air-fuel ratio dynamic optimization model to obtain the air-fuel ratio optimization coefficient of each control zone specifically includes:

[0035] The pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature in the kiln, and flue gas oxygen content of each control area are input into the performance function J(λ) in the AI ​​air-fuel ratio dynamic optimization model; the performance function J(λ) takes the current comprehensive energy efficiency index as the target, where λ represents the actual air-fuel ratio;

[0036] The performance function J(λ) is optimized using reinforcement learning or extreme value search algorithms to obtain the air-fuel ratio optimization coefficient k.

[0037] Preferably, obtaining the final execution command based on the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control region includes:

[0038] Based on the fine adjustment amount and the pre-adjustment amount, the basic command value of gas and the basic command value of combustion air are calculated.

[0039] Based on the air-fuel ratio optimization coefficient and the basic combustion air command value, the corrected combustion air command value is obtained;

[0040] The final execution command is obtained based on the basic gas command value and the corrected combustion air command value.

[0041] Preferably, the step of performing fuzzy inference on the membership degrees of temperature deviation and the membership degrees of temperature deviation change rate of each control area based on a preset fuzzy rule base to obtain an output fuzzy set specifically includes:

[0042] Based on the membership degree of temperature deviation and the membership degree of temperature deviation change rate of each control area, corresponding rules are matched from the preset fuzzy rule base, and the output fuzzy set of each control area is obtained according to the matched rules; each output fuzzy set includes Kp membership degree, Ki membership degree and Kd membership degree.

[0043] Extract the minimum value of the membership degree of temperature deviation and the membership degree of temperature deviation change rate in each control region, and use it as the activation intensity of the corresponding rule;

[0044] The membership function of the output fuzzy set of the corresponding control region is pruned according to the activation intensity of the corresponding rules;

[0045] The output fuzzy sets after clipping each control region are aggregated to obtain the output fuzzy set.

[0046] Preferably, the synergistic combustion control method for tunnel kilns further includes: constructing a 1:1 three-dimensional computational fluid dynamics (CFD) simulation model, i.e., a digital twin model, based on the accurate CAD drawings of the tunnel kiln, material property parameters, combustion reaction mechanism and heat transfer principle;

[0047] The process involves inputting the preprocessed key process parameters of each control region into a digital twin model for offline simulation and online verification, and then outputting the simulation and verification results. Specifically, this includes:

[0048] The key process parameters of each control region after preprocessing are input into the digital twin model for real-time simulation calculation, and the simulation results are output in real time. The simulation results include: three-dimensional temperature field cloud map, key virtual measurement point data, and system-level energy consumption estimate.

[0049] The simulation results output in real time are compared with the key process parameters of each control area acquired in real time in time and in multiple dimensions to obtain the comparison results.

[0050] Based on preset difference threshold rules, it is determined whether the comparison results have a deviation; the difference threshold rules include temperature deviation threshold, pressure deviation threshold and system-level energy consumption estimation deviation threshold.

[0051] If deviations occur, the model and / or rule base are iteratively optimized; the model is a feedforward prediction model, and the rule base is a preset fuzzy rule base.

[0052] Validate the optimized model and / or rule base, and output the validation results.

[0053] According to a second aspect of this application, a collaborative combustion control system for a tunnel kiln is provided, characterized in that it comprises a sensing layer, a control layer, an intelligent computing layer and a digital twin layer connected in sequence.

[0054] The sensing layer includes sensors and actuators for acquiring key process parameters and executing control actions in each control area; the sensors include resistance temperature detectors / thermocouples, pressure transmitters, flow sensors, and CO concentration analyzers; the actuators include burners, gas proportional valves, and combustion air regulating valves; the flow sensors include gas flow sensors and combustion air flow sensors.

[0055] The control layer includes a DCS control system, a PLC control system, and a MES system, which are used to receive control commands, drive actuators, and collect execution feedback data.

[0056] The intelligent computing layer includes a real-time / historical database, an AI computing engine, an application server, a fuzzy PID control module, a feedforward prediction module, and an AI air-fuel ratio dynamic optimization module, which are responsible for data processing, algorithm calculation, and control command generation.

[0057] The digital twin layer includes a temperature field simulation module, a combustion process simulation module, and a control strategy optimization module, which are used to construct a 1:1 three-dimensional computational fluid dynamics (CFD) simulation model and support simulation verification and iterative optimization of control parameters.

[0058] The beneficial effects of this application are as follows:

[0059] The collaborative combustion control method for tunnel kilns provided in this application takes zoned precise control as its core, combines fuzzy PID adaptive adjustment and feedforward prediction mechanism, and integrates digital twin simulation optimization to form an integrated system of "zoned control - fuzzy adaptive adjustment - feedforward prediction - digital simulation optimization," achieving efficient and precise control of the combustion process. Furthermore, this method also has advantages such as fast response speed, excellent energy economy, high optimization efficiency, and wide adaptability. Among them, fuzzy PID control dynamically adjusts parameters (Kp, Ki, and Kd) to adapt to the nonlinear and large inertia characteristics of tunnel kilns. The feedforward prediction model compensates for changes in heat load in advance, reducing temperature deviation. Precise control of each zone ensures the uniformity of cross-sectional temperature, and the deviation between actual temperature and set curve is significantly reduced. The synergistic effect of feedforward prediction and feedback regulation avoids the lag problem of traditional feedback control, and responds more promptly to factors such as kiln car advancement and load changes, significantly reducing overshoot. The AI ​​air-fuel ratio dynamic optimization model locks the optimal operating point, ensuring complete fuel combustion while reducing energy consumption, and significantly reducing natural gas consumption per unit product and overall power consumption. The zone collaborative control logic adapts to the thermal requirements of different process sections. Combined with equipment interlocking and parameter verification mechanisms, it ensures stable combustion process and reduces production fluctuations. The digital twin model supports offline simulation and control strategy pre-verification, avoiding the blindness of on-site testing, shortening the optimization cycle, and reducing testing costs. Through AI self-learning and dynamic parameter adjustment, it can adapt to the combustion control requirements under different raw materials and different production loads, and is suitable for various large-size and high-precision tunnel kiln equipment.

[0060] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of what is pointed out in the written description and the accompanying drawings. Attached Figure Description

[0061] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0062] Figure 1 A schematic flowchart of the synergistic combustion control method for tunnel kilns provided in this application;

[0063] Figure 2 Architecture diagram of the collaborative combustion control system for tunnel kilns provided in this application;

[0064] Figure 3 The flowchart for optimizing digital twin simulation provided for this application. Detailed Implementation

[0065] To make the technical solutions and advantages of the embodiments of this application clearer, the exemplary embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0066] Addressing some problems existing in the current technology:

[0067] In a first aspect, embodiments of this application provide a method for co-combustion control of a tunnel kiln. This method can be executed by a co-combustion control device for a tunnel kiln, or by components such as chips or chip systems configured within the co-combustion control device for a tunnel kiln, or by logic modules or software having some or all of the functions of a co-combustion control device for a tunnel kiln. This application does not limit this aspect.

[0068] For example, such as Figure 1 As shown, the co-combustion control method for tunnel kilns includes:

[0069] Based on the physical structure, process layout and thermal characteristics of the tunnel kiln, the tunnel kiln is divided into multiple independent control zones along its length (such as the heating section, calcination section, and heat preservation section).

[0070] Real-time acquisition of key process parameters in each control area provides data support for subsequent intelligent control. Each key process parameter includes: real-time temperature value T, set temperature value T0, kiln car advance speed v, gas flow rate Qr, combustion air flow rate Qa, flue gas temperature Tp, key point temperature inside the kiln T′, and flue gas oxygen content (or CO concentration). In specific implementation, the kiln car end temperature T is also acquired. in The historical heat load data of each control area, and the car-end temperature of the kiln car, which is the car-end temperature when the kiln car just enters the tunnel kiln, are collected only for the control area at the kiln-entry end.

[0071] The key process parameters of each control area collected are preprocessed; in specific implementation, the key process parameters of each control area after preprocessing are also verified; the verification includes range checking, rate of change filtering and equipment status correlation verification, which can eliminate abnormal data caused by signal jumps and instrument failures, and ensure that the data input to the subsequent control model is true and reliable, thus ensuring the quality of control commands from the source.

[0072] Based on the key process parameters of each control region after preprocessing, the fuzzy PID controller, the feedforward prediction model, and the AI ​​air-fuel ratio dynamic optimization model, the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control region are obtained; the fine adjustment amount includes the gas adjustment amount and the combustion air adjustment amount, and the pre-adjustment amount includes the gas pre-adjustment amount and the combustion air pre-adjustment amount; based on the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control region, the final execution command is obtained;

[0073] According to the final execution command, the tunnel kiln is controlled to carry out combustion, and the state data of the tunnel kiln during combustion is obtained. The state data is fed back in real time and returned to the data acquisition step. The state data includes actual temperature, actual pressure, and emission indicators. The emission indicators include CO concentration and oxygen content in the flue gas.

[0074] The key process parameters of each control region after preprocessing are input into the digital twin model for offline simulation and online verification, and the simulation results and verification results are output.

[0075] Based on the above scheme, the collaborative combustion control method for tunnel kilns provided in this application takes zoned precise control as its core, combines fuzzy PID adaptive adjustment and feedforward prediction mechanism, and integrates digital twin simulation optimization to form an integrated system of "zoned control - fuzzy adaptive adjustment - feedforward prediction - digital simulation optimization," achieving efficient and precise control of the combustion process. Furthermore, this method also has advantages such as fast response speed, excellent energy economy, high optimization efficiency, and wide adaptability.

[0076] In practical applications, when dividing a tunnel kiln into multiple independent control zones along its length, a static baseline division is adopted. This involves engineers manually configuring and mapping the kiln design drawings, the spatial arrangement of the burners (i.e., burners), and the continuous thermal process stages to establish the initial physical control zones. This division ensures that the control architecture strictly corresponds to the production process. During the acquisition of key process parameters, the control parameters (e.g., temperature setpoint, air-fuel ratio target, pressure setpoint) of each zone (i.e., control area) are dynamically associated with key process variables (e.g., kiln car position, product type). This method supports calling preset or self-learning parameter sets according to different production formulas (e.g., raw material type, product specifications) or optimization goals to achieve "soft" reconfiguration of the zones, thereby adapting to different production conditions without changing the physical structure.

[0077] Based on the above scheme, the collaborative combustion control method for tunnel kilns proposed in this application, through a "dynamic-static combination" zoning strategy, achieves a high degree of matching between the controlled object and the dynamic process state, providing an optimal control structure framework for achieving cross-sectional temperature uniformity control and laying the foundation for precise control. Comprehensive and high-quality parameter acquisition provides complete, real-time, and reliable input for feedforward prediction, fuzzy adaptive PID, AI air-fuel ratio optimization, and digital twin simulation, forming an intelligent control closed loop of "perception-prediction-adjustment-verification" and constructing a closed-loop data flow. Through parameterized configuration and dynamic association, fixed physical zones can flexibly adapt to changing production conditions (such as changing raw materials or adjusting production capacity), improving the system's generalization ability and enhancing its flexibility and adaptability. Clear data flow directions and algorithm interfaces transform raw data into specific control commands and optimization suggestions, transforming the entire combustion control process from experience-driven to data and model-driven, thereby achieving data-driven decision-making.

[0078] In some possible implementations of the first aspect, obtaining the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient for each control region based on the preprocessed key process parameters of each control region, the fuzzy PID controller, the feedforward prediction model, and the AI ​​air-fuel ratio dynamic optimization model specifically includes:

[0079] Based on the preprocessed real-time temperature value and set temperature value of each control area, the temperature deviation and temperature deviation change rate of each control area are calculated respectively; among them, these two variables (i.e., temperature deviation and temperature deviation change rate) are the core inputs of the fuzzy PID controller, which are used to dynamically adjust the proportional, integral and derivative parameters.

[0080] The calculated temperature deviation and temperature deviation change rate of each control area are input into the fuzzy PID controller to obtain the gas regulation amount and combustion air regulation amount of the corresponding control area.

[0081] The acquired kiln car end temperature, historical heat load data of each control area, and preprocessed kiln car advance speed of each control area are input into the feedforward prediction model to predict the heat load change of each control area, and the gas pre-regulation amount and combustion air pre-regulation amount of the corresponding control area are calculated based on the prediction results.

[0082] The pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature in the kiln, and flue gas oxygen content (or CO concentration) of each control area are input into the AI ​​air-fuel ratio dynamic optimization model to obtain the air-fuel ratio optimization coefficient of each control area, and then the optimal air-fuel ratio operating point is found and locked.

[0083] Based on the above scheme, the fuzzy PID controller can realize dynamic adjustment of fuzzy PID parameters and is the core adaptive control link of the system. It aims to solve the inherent defects of traditional fixed-parameter PID controllers, which are difficult to adapt to the large inertia, nonlinearity, and time-varying characteristics of tunnel kilns. By introducing fuzzy logic, the operator's experience knowledge is transformed into executable rules, realizing online dynamic tuning of PID parameters (i.e., proportional gain Kp, integral gain Ki, and derivative gain Kd). This allows the controller to intelligently adjust its control behavior according to the real-time control deviation and its changing trend (i.e., rate of change). The feedforward prediction model established and applied in this application predicts the heat load changes that will enter each control zone in advance through intelligent analysis of the tunnel kiln inlet conditions and historical operating data. Based on this, it calculates the required pre-adjustment of gas and combustion air, thereby achieving advance compensation for the large inertia and nonlinear combustion system and reducing temperature fluctuations caused by factors such as kiln car advancement and raw material changes.

[0084] Optionally, the step of calculating the temperature deviation and temperature deviation change rate of each control area based on the preprocessed real-time temperature value and set temperature value of each control area specifically includes:

[0085] According to the formula The temperature deviation of each control zone is calculated separately; where, This indicates the temperature deviation at the current moment. This indicates the current set temperature value. This indicates the real-time temperature value of the area at the current moment;

[0086] According to the formula The temperature deviation change rate of each control area is calculated; where ec(k) represents the temperature deviation change rate at the current time, e(k−1) represents the temperature deviation at the previous sampling time, and Ts represents the fixed sampling time interval of the system.

[0087] In some possible implementations of the first aspect, the step of inputting the calculated temperature deviation and temperature deviation change rate of each control area into the fuzzy PID controller to obtain the gas regulation amount and combustion air regulation amount of the corresponding control area specifically includes:

[0088] The calculated temperature deviation and temperature deviation rate of change of each control region are input into the fuzzy PID controller for fuzzification processing to obtain the temperature deviation membership degree and temperature deviation rate of change membership degree of each control region.

[0089] Based on a pre-set fuzzy rule base, fuzzy inference is performed on the membership degree of temperature deviation and the membership degree of temperature deviation change rate of each control area to obtain the output fuzzy set of parameter adjustment amount (ΔKp, ΔKi, ΔKd);

[0090] The obtained output fuzzy set is defuzzified to obtain parameter adjustment values. The parameter adjustment values ​​include Kp adjustment value, Ki adjustment value, and Kd adjustment value. In specific implementation, the widely used centroid method is used for defuzzification. The center point of the area under the membership function curve of the entire output fuzzy set is calculated. The abscissa value of this point is the required precise parameter adjustment values ​​ΔKp, ΔKi, and ΔKd. These adjustment values ​​are the result of fuzzy inference that integrates temperature deviation and temperature deviation change rate. Then, the above-aggregated output fuzzy set is converted into precise values ​​that can be used for control.

[0091] The updated real-time parameters are obtained based on the parameter adjustment amount and the initial parameters (or parameters from the previous cycle) of the fuzzy PID controller; the updated real-time parameters include Kp. * Ki * and Kd * ;

[0092] Based on the updated real-time parameters and the temperature deviation and temperature deviation rate of each control area, standard PID control is used to calculate the preliminary control output of each control area.

[0093] Based on the preset gas-to-heat conversion coefficient and the reference air-fuel ratio, the preliminary control output of each control area is converted into the gas regulation amount and combustion air regulation amount of the corresponding control area.

[0094] Specifically, based on the updated real-time parameters and the temperature deviation and temperature deviation rate of change of each control area, standard PID control is used to calculate the preliminary control output of each control area, which includes:

[0095] According to standard PID control The preliminary control output u(t) for each control region is calculated; where e(t) represents the temperature deviation of the corresponding control region at the current sampling time. Kp represents the rate of change (ec) of the temperature deviation in the corresponding control area at the current sampling time. * Ki represents the updated proportional gain. * Kd represents the updated integral gain. * This represents the updated differential gain.

[0096] Specifically, the step of converting the preliminary control output of each control zone into the corresponding gas regulation amount and combustion air regulation amount based on the preset gas-to-heat conversion coefficient and the reference air-fuel ratio includes:

[0097] According to the formula The gas regulation amount ΔQr1 for the corresponding control area is calculated; where, This represents the preset gas-to-heat conversion coefficient, which reflects the proportional relationship between the initial control output u(t) and the actual gas flow rate change;

[0098] According to the formula The gas regulation amount ΔQa1 for the corresponding control area is calculated; where, This represents the baseline air-fuel ratio, the initial setting value to ensure complete combustion.

[0099] Based on the above scheme, this application constructs a parameter self-tuning intelligent fuzzy PID controller. This controller breaks the limitations of fixed parameters, enabling it to think like an experienced operator: when the deviation is large, it enhances the control action to quickly eliminate the deviation (increasing Kp); when the deviation decreases rapidly, it reduces the control action in advance to prevent overshoot (adjusting Kd); when the system approaches steady state, it finely adjusts to eliminate steady-state error (adjusting Ki), thus maintaining good control quality under various operating conditions. In addition, this process also has the following advantages: 1. Significantly improves control accuracy and stability: dynamic parameters significantly reduce the overshoot of the system to large inertia elements and enhance its adaptability to nonlinear characteristics, thereby controlling the temperature of each zone more stably near the set value and improving product consistency; 2. Enhances system robustness: when the calorific value of raw materials changes or the production load fluctuates... When time-varying factors such as equipment characteristic drift occur, fuzzy PID control can autonomously adjust parameters for compensation, reducing reliance on precise mathematical models and improving the system's anti-interference capability and long-term operational stability; 3. Realizing experience-based knowledge and automation: The fuzzy experience of "watching the fire and adjusting the wind" in manual operation is quantified, solidified, and automatically executed through a fuzzy rule base, reducing reliance on highly skilled operators and achieving standardization and inheritability of control strategies; 4. Seamless collaboration with upstream and downstream links: This step receives real-time temperature deviation e and temperature deviation change rate ec from "zone parameter acquisition," and its output optimized control quantity (i.e., fine adjustment quantity) is superimposed with the feedforward control quantity (i.e., pre-adjustment quantity) to form the final execution instruction, forming a highly efficient collaborative control pattern of "feedforward compensation for large disturbances and fuzzy PID fine adjustment."

[0100] In practical applications, the calculated temperature deviations and temperature deviation change rates of each control region are input into the fuzzy PID controller for fuzzification processing. This process converts the precise input quantities, temperature deviation e and temperature deviation change rate ec, into corresponding membership degrees. Specifically:

[0101] The universes of discourse for e and ec are defined respectively (e.g., e: [-50℃, +50℃], ec: [-10℃ / min, +10℃ / min]) and several fuzzy linguistic variables (e.g., {negative large (NB), negative medium (NM), zero (ZO), positive medium (PM), positive large (PB)}).

[0102] Based on the membership function, the membership degrees of e and ec to the corresponding fuzzy linguistic variables are calculated respectively, and a value (i.e., e or ec) may simultaneously belong to "zero" and "center" with different membership degrees. More specifically, the membership function can be a triangular membership function or a Gaussian membership function.

[0103] Optionally, the step of performing fuzzy inference on the membership degrees of temperature deviation and the membership degrees of temperature deviation change rate of each control area based on a preset fuzzy rule base to obtain an output fuzzy set specifically includes:

[0104] Based on the membership degree of temperature deviation and the membership degree of temperature deviation change rate of each control area, corresponding rules are matched from the preset fuzzy rule base, and the output fuzzy set of each control area is obtained according to the matched rules; each output fuzzy set includes Kp membership degree, Ki membership degree and Kd membership degree.

[0105] Extract the minimum value of the membership degree of temperature deviation and the membership degree of temperature deviation change rate in each control region, and use it as the activation intensity of the corresponding rule;

[0106] The membership function of the output fuzzy set of the corresponding control region is pruned according to the activation intensity of the corresponding rules; in practice, the minimum value method or the product method is usually used for pruning.

[0107] The output fuzzy sets after clipping each control region are aggregated to obtain the output fuzzy set. In practice, the maximum value method is usually used to aggregate the output fuzzy sets after clipping of all activated rules to form three (Kp, Ki, Kd) comprehensive output fuzzy sets with complex shapes.

[0108] In practical applications, the pre-defined fuzzy rule base is the core of the control strategy. It takes the form of an "IF-THEN" statement and embodies expert experience. For example: Rule 1: IF e is PB (positive large) AND ec is NB (negative large), THEN Kp is PB (positive large), Ki is ZO (zero), Kd is PS (positive small); Rule 2: IF e is ZO (zero) AND ec is NM (negative medium), THEN Kp is PM (positive medium), Ki is NS (negative small), Kd is ZO (zero). The conclusion part (after THEN) of each rule corresponds to an output fuzzy set regarding Kp, Ki, and Kd.

[0109] Optionally, the updated real-time parameters are obtained based on the parameter adjustment amount and the initial parameters (or parameters from the previous cycle) of the fuzzy PID controller, specifically including:

[0110] According to the formula The updated proportional gain is calculated; where Kp * Kp represents the updated proportional gain, Kp represents the initial proportional gain of the fuzzy PID controller (or the proportional gain of the previous cycle), and ΔKp represents the proportional gain adjustment.

[0111] According to the formula The updated integral gain is calculated; where Ki * Ki represents the updated integral gain, which represents the initial integral gain of the fuzzy PID controller (or the integral gain of the previous cycle), and ΔKi represents the integral gain adjustment.

[0112] According to the formula The updated differential gain is calculated; where Kd * The updated derivative gain is represented by Kd, which represents the initial derivative gain of the fuzzy PID controller (or the derivative gain of the previous cycle), and ΔKd represents the derivative gain adjustment.

[0113] Based on the above scheme, the parameter adjustment amount is added to the initial parameter (or the parameter of the previous cycle) of the PID controller to obtain the updated real-time parameter. This process is repeated in each control cycle (e.g., at the second level) to achieve continuous and smooth adaptive parameter control.

[0114] In some possible implementations of the first aspect, the process of inputting the acquired kiln car end temperature, historical heat load data of each control zone, and preprocessed kiln car advance speed of each control zone into a feedforward prediction model to predict the heat load change of each control zone, and calculating the gas pre-regulation amount and combustion air pre-regulation amount of the corresponding control zone based on the prediction results, specifically includes:

[0115] Acquire the kiln car end temperature and historical heat load data for each control area;

[0116] The kiln car end temperature and the kiln car advance speed of each control area after preprocessing are input into the linear regression layer of the feedforward prediction model to calculate the baseline prediction value of each control area. Among them, multiple linear regression is performed on the strongly linearly correlated features such as the kiln car advance speed and the kiln car end temperature of each zone to quickly calculate a baseline prediction value that reflects the trend of macro load change. This process has high computational efficiency and can quickly respond to clear speed or temperature step changes.

[0117] The historical heat load data and baseline prediction values ​​of each control area are input into the neural network layer of the feedforward prediction model to obtain the nonlinear correction amount of each control area. In specific implementation, the baseline prediction value output by linear regression, historical heat load data and other nonlinear features are input into a multilayer perceptron (MLP) neural network. This network learns the complex nonlinear dynamics, hysteresis effects and high-order coupling relationships between features in the thermal process through the hidden layer, and outputs a nonlinear correction amount.

[0118] The heat load change of each control area is calculated based on the baseline predicted value and nonlinear correction amount of each control area.

[0119] Based on the calculated heat load changes in each control area, the pre-regulation amounts of gas and combustion air for the corresponding control areas are calculated.

[0120] Based on the above scheme, the collaborative combustion control method for tunnel kilns provided in this application realizes a shift from a "passive feedback" to an "active feedforward prediction" control mode. By using a model to anticipate future heat load changes caused by "disturbance sources" such as kiln car advance speed and kiln inlet temperature (i.e., kiln car end temperature), and issuing adjustment commands (i.e., pre-adjustment amounts) in advance, it effectively compensates for the huge thermal inertia and transmission delay of the tunnel kiln combustion system, suppressing disturbances at the initial state. Furthermore, this process has the following advantages: 1. Significantly improved response speed and control accuracy: Compared to traditional PID control which only acts after a temperature deviation occurs, feedforward control can intervene several minutes to tens of minutes in advance, significantly reducing temperature overshoot and fluctuations, and making the actual temperature curve more closely follow the set curve; 2. Enhanced... System anti-interference capability: A quantitative prediction and compensation mechanism has been established to address major production disturbances such as intermittent kiln car entry and raw material batch changes, improving the stability of the production process under varying operating conditions; 3. Complementary advantages with the feedback system: Feedforward processing handles predictable major disturbances, while feedback (i.e., fuzzy PID control) eliminates unmodeled disturbances and accumulated errors. Together, they form a fast, accurate, and robust composite control system; 4. Provides a forward-looking foundation for intelligent optimization: The output of feedforward prediction provides predictive information on future operating conditions for subsequent combustion optimization (such as dynamic setting of air-fuel ratio), making optimization decisions more forward-looking.

[0121] In practical applications, the core input of the feedforward prediction model consists of multi-dimensional time-series features that characterize future heat load trends. These features mainly include: the current and recent historical (e.g., the previous 20 minutes) kiln car advance speed sequence, which directly affects the amount of material entering the kiln and the heat storage per unit time; the real-time value and trend of the kiln car end temperature, which reflects the initial thermal state of the material entering the kiln; historical heat load data Q(t-τ) for each zone (which can be calculated based on gas flow rate and fuel calorific value), taking data within a time window τ (e.g., the previous 1-2 hours), which reflects the recent thermal inertia and combustion state of the kiln; and auxiliary features such as the current ambient temperature and product type code (different raw materials have different heat capacities). All of the above data are input into the feedforward prediction model, which uses a fusion algorithm of linear regression and neural networks to predict the future heat load changes in each control zone. Based on this, the pre-adjustment amounts of gas and combustion air for each control zone are calculated in advance to compensate for the lag caused by the large inertia system. The linear regression layer is used for trend baseline extraction, and the neural network layer is used for nonlinear correction and refined prediction. Furthermore, the model is trained using historical data with the goal of minimizing the error between the predicted heat load change and the actual subsequent observed load change.

[0122] Optionally, the step of calculating the heat load change of each control area based on the baseline predicted value and the nonlinear correction amount of each control area specifically includes:

[0123] According to the formula The heat load change in each control area is calculated separately; where ΔQ represents the heat load change; ΔQ L Indicates the baseline forecast value; ΔQ n This represents the nonlinear correction amount; This represents the weight, which can be adaptively adjusted according to the operating conditions.

[0124] Optionally, the step of calculating the gas pre-regulation amount and combustion air pre-regulation amount for the corresponding control area based on the calculated heat load change amount of each control area specifically includes:

[0125] According to the formula The pre-regulation amount of gas ΔQr2 is calculated; where ΔQ represents the change in heat load; this process is directly related to heat demand and fuel supply.

[0126] According to the formula The pre-adjustment amount of combustion air ΔQa2 is calculated. The target air-fuel ratio is determined by the optimal value provided in real time by the AI ​​air-fuel ratio optimization model, and the air volume correction coefficient is finely adjusted according to the current exhaust oxygen content or CO concentration to ensure that the dynamic process is still close to complete combustion. This process ensures the coordination of air and gas regulation at the prediction level.

[0127] In some possible implementations of the first aspect, the input of the pre-processed gas flow rate, combustion air flow rate, flue gas temperature, kiln key point temperature, and flue gas oxygen content (or CO concentration) of each control zone into the AI ​​air-fuel ratio dynamic optimization model to obtain the air-fuel ratio optimization coefficient of each control zone specifically includes:

[0128] The pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature inside the kiln, and flue gas oxygen content (or CO concentration) of each control zone are input into the performance function J(λ) in the AI ​​air-fuel ratio dynamic optimization model; the performance function Where λ represents the actual air-fuel ratio, The thermal efficiency is estimated based on real-time parameters. For combined gas and electricity energy consumption, and These are configurable weighting coefficients used to balance thermal efficiency and energy consumption targets;

[0129] The performance function J(λ) is optimized using reinforcement learning or extreme value search algorithms to obtain the air-fuel ratio optimization coefficient (usually between 0.95 and 1.05).

[0130] In practical implementation, taking Extremum Search Control (ESC) as an example: the system takes the current comprehensive energy efficiency index (such as "thermal efficiency" or "reciprocal of gas-electric comprehensive energy consumption") as the performance function J(λ) to be optimized; the ESC algorithm will superimpose a small high-frequency sinusoidal disturbance a sin(ωt) on the current air-fuel ratio setpoint λ*, and then detect the component in the performance function output J that is in the same frequency as the disturbance; after that, by demodulating the signal, the ESC algorithm can determine the gradient direction of J as λ changes, and drive λ* to automatically move in the direction that increases J (i.e., reduces energy consumption), and finally stabilize near the extreme point (optimal point) of the performance function.

[0131] In practical applications, key parameters reflecting the combustion state (such as gas flow rate, combustion air flow rate, flue gas temperature, key point temperature inside the kiln, and oxygen content or CO concentration in the flue gas) are input into the AI ​​air-fuel ratio dynamic optimization model. This model uses reinforcement learning or extreme value search algorithms to achieve the goal of "highest thermal efficiency" or "lowest overall energy consumption". Under the premise of ensuring complete combustion, it finely adjusts the combustion air flow rate in real time and in small increments to find and lock the optimal air-fuel ratio operating point.

[0132] Based on the above scheme, the AI ​​air-fuel ratio dynamic optimization model provided in this application plays a dual role as a "command coordinator" and an "economic optimizer". This model is not only a node that physically superimposes feedforward and feedback signals, but also a key link that actively and continuously optimizes secondary economic objectives (energy consumption) while meeting the primary process objective (temperature) through intelligent algorithms. In addition, this process has the following advantages: 1. Achieving a balance between control performance and economic benefits: Through collaborative calculation, it balances the speed, stability, and accuracy of the system; through dynamic optimization of the air-fuel ratio, it actively reduces the excess air coefficient while ensuring complete combustion and meeting process temperatures, thereby reducing exhaust heat loss and fan power consumption, and directly reducing operating costs; 2. Enhancing the system's adaptability and intelligence: The AI ​​air-fuel ratio optimization model can adapt to slow time-varying factors such as fuel calorific value fluctuations, air density changes, and burner coking, keeping the system in a high-efficiency state and achieving a leap from "stable control" to "economic optimization"; 3. Forming a complete intelligent control closed loop: The model receives information from all upstream modules (partition parameters, feedforward prediction, fuzzy PID control) and generates the final execution instructions, completing a complete closed loop from "perception-decision-execution". Meanwhile, the effect of the AI ​​air-fuel ratio optimization model will be fed back into the system knowledge base for iterative updates to the feedforward prediction model or fuzzy rule base, forming a continuously evolving intelligent system; 4. Enhance operational safety and reliability: The command amplitude and rate limiting mechanism, as well as the monitoring of exhaust components (such as oxygen content O2% and CO concentration) by the AI ​​air-fuel ratio optimization model, together constitute an important guarantee for preventing safety risks such as oxygen-deficient combustion and accumulation of combustibles.

[0133] In some possible implementations of the first aspect, obtaining the final execution command based on the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control region specifically includes:

[0134] Based on the fine adjustment amount and the pre-adjustment amount, the basic command value of gas and the basic command value of combustion air are calculated.

[0135] Based on the air-fuel ratio optimization coefficient and the basic combustion air command value, the corrected combustion air command value is obtained;

[0136] Based on the basic gas command value and the corrected combustion air command value, the final execution command is obtained. In practice, the basic gas command value and the corrected combustion air command value are used together as the final execution command and issued to the actuators in the corresponding control areas.

[0137] Based on the above scheme, the process realizes zoned collaborative combustion control. The core of the final synthesis and optimization of control commands (i.e. execution commands) lies in the collaborative integration of the pre-adjustment amount of feedforward prediction and the fine adjustment amount of fuzzy PID feedback control to generate accurate actuator commands. Furthermore, an AI air-fuel ratio dynamic optimization model is introduced to perform closed-loop optimization of combustion economy, thereby minimizing energy consumption while ensuring accurate and stable process temperature.

[0138] Optionally, the calculation of the basic gas supply command value and the basic combustion air command value based on the fine adjustment amount and the pre-adjustment amount specifically includes:

[0139] According to the formula Qr * =Qr0+ΔQr1+ΔQr2, the basic gas command value Qr is calculated. * Where Qr0 represents the reference gas flow rate under the current operating conditions, ΔQr1 represents the gas regulation amount, and ΔQr2 represents the gas pre-regulation amount.

[0140] According to the formula Qa * =Qa0+ΔQa1+ΔQa2, calculate the basic command value of combustion wind Qa * Where Qa0 represents the reference flow rate of combustion air under the current operating conditions, ΔQa1 represents the adjustment amount of combustion air, and ΔQa2 represents the pre-adjustment amount of combustion air.

[0141] In practical applications, the synthesis of basic commands ensures that feedforward actions quickly compensate for predictable disturbances, and feedback actions eliminate residual deviations in real time. Furthermore, process safety upper and lower limit constraints (i.e., command limits) are applied to the calculated basic gas command values ​​and combustion air command values. For example, the gas command must not exceed the burner's maximum safe flow rate and must not be lower than the minimum flow rate required to maintain a stable flame; and their rate of change is limited (i.e., rate limit). The Qr obtained in this step... * with Qa * Together, these constitute the combustion control commands with the primary objective of meeting process temperature requirements. For example, the rate of change of the gas proportional valve opening is limited to within 5% per second to prevent impact on the combustion process and ensure smooth operation.

[0142] Optionally, obtaining the corrected combustion air command value based on the air-fuel ratio optimization coefficient and the basic combustion air command value specifically includes:

[0143] According to the formula The combustion air command value after air-fuel ratio optimization correction was calculated. In the formula, This represents the basic command value for combustion air, and k represents the air-fuel ratio optimization coefficient output by the AI ​​dynamic air-fuel ratio optimization model (typically between 0.95 and 1.05). In practice, this coefficient fine-tuning based on energy efficiency targets is continuous and automatic, ensuring combustion always occurs in the 'high-efficiency zone' while maintaining stable process temperatures. In summary, the final execution command generation logic is as follows: the basic gas command value Qr* is directly used as the gas execution command; the basic command value for combustion air... The AI-powered dynamic air-fuel ratio optimization model is input for energy efficiency fine-tuning to obtain the final combustion air execution command. .

[0144] In some possible implementations of the first aspect, the synergistic combustion control method for tunnel kilns further includes: constructing a high-fidelity 1:1 three-dimensional computational fluid dynamics (CFD) simulation model, i.e., a digital twin model, based on precise CAD drawings of the tunnel kiln, material properties, combustion reaction mechanism, and heat transfer principles. This model can simulate the complex airflow organization, temperature field distribution, fuel combustion, and heat transfer processes within the kiln.

[0145] Based on the above scheme, this application constructs a virtual digital model that is synchronously mapped and paralleled with the physical tunnel kiln. Through a closed-loop process of "offline simulation - online verification - parameter correction", iterative optimization of the feedforward prediction model and fuzzy rule base is realized, thereby continuously improving the accuracy, adaptability and economy of the entire control system.

[0146] In some possible implementations of the first aspect, the step of inputting the preprocessed key process parameters of each control region into a digital twin model for offline simulation and online verification, and outputting the simulation results and verification results, specifically includes:

[0147] The key process parameters of each control region after preprocessing are used as boundary conditions and initial states, and are synchronously input into the digital twin model to drive it to perform real-time simulation calculations synchronized with the physical world, and output simulation results in real time. The simulation results include: three-dimensional temperature field cloud map, key virtual measurement point data (such as simulated temperature and pressure at any location), and system-level energy consumption estimates (such as theoretical gas consumption and flue gas heat loss). In specific implementation, all parameters are synchronized to the digital twin model to drive three-dimensional temperature field simulation, support offline verification and iterative optimization of control strategies, and provide a panoramic internal state view that is not limited by the number and location of sensors.

[0148] The simulation results output in real time are time-aligned and compared with the key process parameters of each control area acquired in real time to obtain the comparison results. The comparison is not limited to the values ​​of key measuring points, but also includes the uniformity of temperature field distribution, the shape of heating / cooling curves of specific process sections, etc.

[0149] The comparison results are evaluated based on preset difference threshold rules, i.e., it is determined whether the comparison results produce significant deviations. The difference threshold rules include temperature deviation threshold, pressure deviation threshold, and system-level energy consumption estimation deviation threshold (e.g., the temperature deviation threshold is set to ±5℃, the pressure deviation threshold is set to ±10Pa, and the system-level energy consumption estimation deviation threshold is set to ±3%; when the comparison results exceed any of the relevant thresholds, it is determined that a significant deviation has occurred). In specific implementation, when the difference value in the comparison results (e.g., temperature difference, pressure difference, energy consumption estimation difference) exceeds the corresponding threshold, the system determines and identifies a significant deviation, which refers to the inconsistency between the simulation prediction value of the digital twin model and the actual operating value of the physical tunnel kiln.

[0150] If a significant deviation occurs, the model and / or rule base are iteratively optimized; the model is a feedforward prediction model, and the rule base is a preset fuzzy rule base.

[0151] Validate the optimized model and / or rule base, and output the validation results.

[0152] Specifically, the strategy iterative optimization is as follows:

[0153] Ⅰ. Feedforward prediction model parameter correction: If the deviation is strongly correlated with kiln car advance, load change, etc., it is determined that the accuracy of the feedforward prediction model is insufficient. At this time, an optimization algorithm (such as gradient descent or Bayesian optimization) is adopted to "minimize the root mean square error (RMSE) between the simulated predicted heat load and the actual calculated heat load". In the digital twin simulation environment, the coefficients of linear regression or the weights of neural network in the feedforward prediction model are automatically and repeatedly adjusted until the simulation results are highly consistent with the actual data.

[0154] II. Fuzzy Rule Base Optimization: If the deviation exhibits dynamic characteristics such as overshoot, oscillation, or steady-state error, the fuzzy rule base needs improvement. In this case, reinforcement learning (e.g., Q-learning) or genetic algorithms can be used in a digital twin simulation environment. The reward function can be "reaching the set temperature in the shortest time with minimal overshoot" or "minimizing the standard deviation of temperature fluctuation," allowing the agent to automatically explore and generate new, better fuzzy rules (e.g., "when e is PM (positive center) and ec is NS (negative small), Kp should be set to PS (positive small)"). Alternatively, the output membership function of existing rules can be adjusted to update the fuzzy rule base.

[0155] Based on the above scheme, this process acts as a "strategy laboratory" and "experience learning center" for the entire intelligent control system. It breaks through the bottleneck of traditional industrial optimization relying on time-consuming and labor-intensive field testing, enabling rapid trial and error, quantitative evaluation, and automatic evolution of control strategies in a virtual space, thus giving the control system continuous self-optimization capabilities. Furthermore, this process has the following advantages: 1. Significantly reduced optimization costs and risks: All adjustments to control parameters, verification of new algorithms, and testing under extreme conditions are conducted in the digital world, achieving "zero physical risk and zero material consumption" optimization. This avoids production fluctuations, equipment damage, or safety accidents that may occur during field testing, greatly shortening the optimization cycle; 2. Achieving global and forward-looking optimization: The digital twin model can reveal internal states that physical sensors cannot directly measure (such as temperature at any point, dead zones in the flow field), supporting global, mechanistic-level evaluation and optimization of control strategies; simultaneously, it can provide insights for future planned production. 1. **Virtual roasting of new product processes:** This allows for advance prediction and optimization of control curves, achieving a leap from "post-event correction" to "pre-event simulation." 2. **Deep system self-adaptation:** By continuously comparing simulation and actual conditions, the system can automatically detect model mismatches caused by slow-changing factors such as kiln aging, burner carbon buildup, and gradual changes in raw material characteristics. It proactively corrects its control model, enabling the system to maintain optimal performance over the long term and significantly improving its self-adaptive capabilities throughout its lifecycle. 3. **Forming a complete "perception-decision-execution-optimization" closed loop:** This step, together with all the aforementioned online control steps, constitutes a higher-level intelligent closed loop. Online control ensures stable real-time production, while the digital twin continuously learns and upgrades in the background. Together, they drive the entire system towards higher precision and lower energy consumption, providing crucial support for ultimately achieving "intelligent manufacturing."

[0156] More specifically, such as Figure 3 As shown, the process of digital twin simulation optimization is as follows:

[0157] S10. Real-time data synchronization and driving: Specifically, receive all real-time data (i.e., key process parameters) from the perception layer / control layer, preprocess the data and align the timestamps, and drive the digital twin model to update.

[0158] S20, high-fidelity mechanism simulation calculation; specifically, CFD / heat transfer simulation is performed, outputting three-dimensional temperature field / flow distribution, and extracting virtual measurement point data and theoretical energy consumption;

[0159] S30. Comparison and analysis of simulation and measured data; Specifically, compare key parameters such as temperature, pressure, and energy consumption, and calculate the deviation index between simulation results and measured data. The deviation index may include the mean absolute error (MAE), root mean square error (RMSE), and the maximum deviation value of key process points (such as the center temperature of the firing section) to quantify the accuracy of the model simulation.

[0160] S40, Intelligent Diagnosis and Root Cause Analysis; Specifically, analyze the spatiotemporal distribution characteristics of the deviation, match and analyze the expert rule base; The expert rule base contains diagnostic rules based on process knowledge (e.g., if the simulated value of the temperature field of the burning section is generally higher than the measured value, and the simulated value of the flue gas temperature is also higher, it may indicate that the feedforward prediction model underestimates the heat load); Output diagnostic conclusions: (1) The feedforward prediction model is inaccurate; (2) The fuzzy rules are not suitable; (3) The mechanism model parameters drift;

[0161] S50, Strategy Iteration Optimization; Specifically, perform strategy iteration optimization on the model and / or rule base;

[0162] S60, Multi-condition simulation verification; Specifically, historical data is used to verify the optimization effect and test the robustness under different production conditions;

[0163] S70. Secure deployment and updates; specifically, generate an optimized parameter package, have it confirmed / automatically reviewed by the administrator, and then hot-update it online to the fuzzy PID control module and feedforward prediction module in the intelligent computing layer.

[0164] Based on the above scheme, digital twin simulation optimization is the core driving force for the continuous self-evolution and optimization of control strategies. All model parameters and rules modified in the virtual environment need to be tested again by the digital twin model under various typical working conditions to verify their effectiveness and robustness. After verification, they are deployed online to the real-time control system of the "intelligent computing layer" through a security mechanism (such as manual confirmation or small-scale gray update) to complete a complete strategy iteration.

[0165] Secondly, this application provides a system, which is a collaborative combustion control system for a tunnel kiln. The system includes a sensing layer, a control layer, an intelligent computing layer, and a digital twin layer that are electrically connected in sequence. Figure 2 As shown, the overall four-layer architecture of the collaborative combustion control system for tunnel kilns is presented, consisting of the perception layer, control layer, intelligent computing layer, and digital twin layer from bottom to top. The main functional modules of each layer and the data communication paths between layers are also marked.

[0166] The sensing layer includes sensors and actuators for acquiring key process parameters and executing control actions in each control area. The sensors include resistance temperature detectors (RTDs) / thermocouples (installed on the kiln top, car bottom, exhaust, and side walls), pressure transmitters (installed inside the kiln and on the car bottom), flow sensors, and CO concentration analyzers. The actuators include burners, gas proportional valves, and combustion air regulating valves. The flow sensors include gas flow sensors and combustion air flow sensors.

[0167] The control layer includes a DCS control system, a PLC control system, and a MES system. It communicates with the perception layer and the intelligent computing layer via the OPC UA / Modbus TCP protocol to receive control commands, drive actuators, and collect execution feedback data.

[0168] The intelligent computing layer includes a real-time / historical database, an AI computing engine, an application server, a fuzzy PID control module, a feedforward prediction module, and an AI air-fuel ratio dynamic optimization module (i.e., the air-fuel ratio dynamic optimization module in the diagram), which is responsible for data processing, algorithm calculation, and control command generation. Specifically, the fuzzy PID control module provides fuzzy rule configuration, dynamic adjustment of PID parameters, and deployment of zoned control logic; the feedforward prediction module supports feedforward prediction model training, input parameter configuration, and pre-adjustment calculation; and the AI ​​air-fuel ratio optimization module achieves real-time air-fuel ratio optimization based on reinforcement learning or extreme value search algorithms.

[0169] The digital twin layer includes a temperature field simulation module, a combustion process simulation module, and a control strategy optimization module, which are used to construct a 1:1 three-dimensional computational fluid dynamics (CFD) simulation model and support simulation verification and iterative optimization of control parameters.

[0170] Specifically, the coordinated combustion control system for tunnel kilns further includes:

[0171] Data preprocessing module: used to perform parameter range checks, rate of change filtering, and equipment status correlation verification to ensure the reliability of the collected data;

[0172] Digital twin simulation module: used to provide 3D visualization, temperature field rendering, and control strategy simulation verification functions;

[0173] Alarm and Log Module: Used to record information such as abnormal parameters and equipment failures during the control process, and supports historical tracing.

[0174] Thirdly, this application provides a device that can be any device capable of implementing a collaborative combustion control method for tunnel kilns. This device can be various terminal devices, such as desktop computers, laptops, tablets, handheld devices, etc., and can be implemented through software and / or hardware.

[0175] For example, the device includes: a memory; a processor; and a computer program;

[0176] The computer program is stored in the memory and configured to be executed by the processor to implement the co-combustion control method for tunnel kilns as described above.

[0177] Fourthly, this application provides a computer-readable storage medium, which may be a ROM, RAM, disk, or optical disk, etc.

[0178] For example, the computer-readable storage medium stores a computer program; the computer program is executed by a processor to implement the co-combustion control method for tunnel kilns as described above.

[0179] In summary, the synergistic combustion control method for tunnel kilns provided in this application has produced many significant effects in the combustion control process of tunnel kilns, and has obvious advantages compared with traditional combustion control technologies:

[0180] 1. High control precision: Fuzzy PID control dynamically adjusts parameters (Kp, Ki and Kd) to adapt to the nonlinear and large inertia characteristics of tunnel kilns. The feedforward prediction model compensates for heat load changes in advance, reducing temperature deviation. Precise control of each zone ensures the uniformity of cross-sectional temperature, and the deviation between the actual temperature and the set curve is significantly reduced.

[0181] 2. Fast response speed: The synergistic effect of feedforward prediction and feedback adjustment avoids the lag problem of traditional feedback control, and responds more promptly to factors such as kiln car advancement and load changes, with a significant reduction in overshoot.

[0182] 3. Excellent energy efficiency: The AI ​​air-fuel ratio dynamic optimization model locks in the optimal operating point, ensuring complete combustion of fuel while reducing energy consumption, resulting in a significant decrease in natural gas consumption and overall electricity consumption per unit product;

[0183] 4. Strong operational stability: The zoned collaborative control logic adapts to the thermal requirements of different process sections, and combined with equipment interlocking and parameter verification mechanisms, it ensures stable combustion process and reduces production fluctuations;

[0184] 5. High optimization efficiency: Digital twin models support offline simulation and control strategy pre-verification, avoiding the blindness of field tests, shortening the optimization cycle, and reducing test costs;

[0185] 6. Wide adaptability: Through AI self-learning and dynamic parameter adjustment, it can adapt to the combustion control requirements under different raw materials and production loads, and is suitable for various large-size and high-precision tunnel kiln equipment.

[0186] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as C, VHDL, Verilog, the object-oriented programming language Java, and the interpreted scripting language JavaScript.

[0187] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] In the description of this application, it should be understood that the terms "length", "top", "bottom", "inner", "outer", etc., indicating orientation or positional relationship are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0191] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0192] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0193] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for co-combustion control in tunnel kilns, characterized in that: include: Based on the physical structure, process layout and thermal characteristics of the tunnel kiln, the tunnel kiln is divided into multiple independent control zones. Key process parameters of each control area are collected in real time, and the collected key process parameters of each control area are preprocessed; each key process parameter includes: real-time temperature value of the area, set temperature value, kiln car advance speed, gas flow rate, combustion air flow rate, flue gas temperature, temperature of key points in the kiln and oxygen content of flue gas. Based on the key process parameters of each control region after preprocessing, the fuzzy PID controller, the feedforward prediction model, and the AI ​​air-fuel ratio dynamic optimization model, the fine adjustment amount, the pre-adjustment amount, and the air-fuel ratio optimization coefficient of each control region are obtained; the fine adjustment amount includes the gas adjustment amount and the combustion air adjustment amount, and the pre-adjustment amount includes the gas pre-adjustment amount and the combustion air pre-adjustment amount. Based on the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control area, the final execution command is obtained; Based on the final execution command, the state data during the combustion of the tunnel kiln is obtained, and the state data is fed back in real time; the state data includes actual temperature, actual pressure, and emission indicators. The key process parameters of each control region after preprocessing are input into the digital twin model for offline simulation and online verification, and the simulation results and verification results are output.

2. The method for co-combustion control in a tunnel kiln according to claim 1, characterized in that: The process involves obtaining the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient for each control region based on the key process parameters of each control region after preprocessing, the fuzzy PID controller, the feedforward prediction model, and the AI ​​air-fuel ratio dynamic optimization model. Specifically, this includes: Based on the preprocessed real-time temperature value and set temperature value of each control area, the temperature deviation and temperature deviation change rate of each control area are calculated respectively. The calculated temperature deviation and temperature deviation change rate of each control area are input into the fuzzy PID controller to obtain the gas regulation amount and combustion air regulation amount of the corresponding control area. The acquired kiln car end temperature, historical heat load data of each control area, and preprocessed kiln car advance speed of each control area are input into the feedforward prediction model to predict the heat load change of each control area, and the gas pre-regulation amount and combustion air pre-regulation amount of the corresponding control area are calculated based on the prediction results. The pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature in the kiln, and flue gas oxygen content of each control area are input into the AI ​​air-fuel ratio dynamic optimization model to obtain the air-fuel ratio optimization coefficient of each control area.

3. The method for co-combustion control in a tunnel kiln according to claim 2, characterized in that: After inputting the calculated temperature deviation and temperature deviation change rate of each control area into the fuzzy PID controller, the gas regulation amount and combustion air regulation amount of the corresponding control area are obtained, specifically including: The calculated temperature deviation and temperature deviation rate of change of each control region are input into the fuzzy PID controller for fuzzification processing to obtain the temperature deviation membership degree and temperature deviation rate of change membership degree of each control region. Based on a pre-set fuzzy rule base, fuzzy reasoning is performed on the membership degree of temperature deviation and the membership degree of temperature deviation change rate of each control area to obtain the output fuzzy set; The obtained output fuzzy set is defuzzified to obtain parameter adjustment values; the parameter adjustment values ​​include Kp adjustment value, Ki adjustment value and Kd adjustment value; The updated real-time parameters are obtained based on the parameter adjustment amount and the initial parameters (or parameters from the previous cycle) of the fuzzy PID controller; the updated real-time parameters include Kp. * Ki * and Kd * ; Based on the updated real-time parameters and the temperature deviation and temperature deviation rate of each control area, standard PID control is used to calculate the preliminary control output of each control area. Based on the preset gas-to-heat conversion coefficient and the reference air-fuel ratio, the preliminary control output of each control area is converted into the gas regulation amount and combustion air regulation amount of the corresponding control area.

4. The method for co-combustion control in a tunnel kiln according to claim 2, characterized in that: The process involves using the acquired kiln car end temperature, historical heat load data for each control area, preprocessed kiln car advance speed for each control area, and input feedforward prediction model to predict the heat load change in each control area. Based on the prediction results, the pre-regulation amounts of the gas and combustion air for the corresponding control areas are calculated. Specifically, this includes: Acquire the kiln car end temperature and historical heat load data for each control area; The kiln car end temperature and the kiln car advance speed of each control area after preprocessing are input into the linear regression layer of the feedforward prediction model to calculate the baseline prediction value of each control area. The historical heat load data and baseline prediction values ​​of each control area are input into the neural network layer of the feedforward prediction model to obtain the nonlinear correction amount for each control area. The heat load change of each control area is calculated based on the baseline predicted value and nonlinear correction amount of each control area. Based on the calculated heat load changes in each control area, the pre-regulation amounts of gas and combustion air for the corresponding control areas are calculated.

5. The method for co-combustion control in a tunnel kiln according to claim 4, characterized in that: The calculation of the gas pre-regulation and combustion air pre-regulation for each control area based on the calculated heat load change in each control area specifically includes: According to the formula The pre-regulation amount of gas ΔQr2 is calculated; where ΔQ represents the change in heat load. According to the formula The pre-adjustment amount of combustion air ΔQa2 was calculated.

6. The method for co-combustion control in a tunnel kiln according to claim 2, characterized in that: The process involves inputting the pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key kiln temperature, and flue gas oxygen content of each control zone into the AI ​​air-fuel ratio dynamic optimization model to obtain the air-fuel ratio optimization coefficient for each control zone. Specifically, this includes: The pre-processed gas flow rate, combustion air flow rate, flue gas temperature, key point temperature in the kiln, and flue gas oxygen content of each control area are input into the performance function J(λ) in the AI ​​air-fuel ratio dynamic optimization model; the performance function J(λ) takes the current comprehensive energy efficiency index as the target, where λ represents the actual air-fuel ratio; The performance function J(λ) is optimized using reinforcement learning or extreme value search algorithms to obtain the air-fuel ratio optimization coefficient.

7. The method for co-combustion control in a tunnel kiln according to claim 1, characterized in that: The final execution command is obtained based on the fine adjustment amount, pre-adjustment amount, and air-fuel ratio optimization coefficient of each control region, specifically including: Based on the fine adjustment amount and the pre-adjustment amount, the basic command value of gas and the basic command value of combustion air are calculated. Based on the air-fuel ratio optimization coefficient and the basic combustion air command value, the corrected combustion air command value is obtained; The final execution command is obtained based on the basic gas command value and the corrected combustion air command value.

8. The method for co-combustion control in a tunnel kiln according to claim 3, characterized in that: The method, based on a preset fuzzy rule base, performs fuzzy inference on the membership degrees of temperature deviation and the membership degrees of temperature deviation change rate of each control area to obtain an output fuzzy set, specifically including: Based on the membership degree of temperature deviation and the membership degree of temperature deviation change rate of each control area, corresponding rules are matched from the preset fuzzy rule base, and the output fuzzy set of each control area is obtained according to the matched rules; each output fuzzy set includes Kp membership degree, Ki membership degree and Kd membership degree. Extract the minimum value of the membership degree of temperature deviation and the membership degree of temperature deviation change rate in each control region, and use it as the activation intensity of the corresponding rule; The membership function of the output fuzzy set of the corresponding control region is pruned according to the activation intensity of the corresponding rules; The output fuzzy sets after clipping each control region are aggregated to obtain the output fuzzy set.

9. The method for co-combustion control in a tunnel kiln according to claim 3, characterized in that: Also includes: Based on the precise CAD drawings of the tunnel kiln, material properties, combustion reaction mechanism and heat transfer principle, a 1:1 three-dimensional computational fluid dynamics (CFD) simulation model, namely a digital twin model, is constructed. The process involves inputting the preprocessed key process parameters of each control region into a digital twin model for offline simulation and online verification, and then outputting the simulation and verification results. Specifically, this includes: The key process parameters of each control region after preprocessing are input into the digital twin model for real-time simulation calculation, and the simulation results are output in real time. The simulation results include: three-dimensional temperature field cloud map, key virtual measurement point data, and system-level energy consumption estimate. The simulation results output in real time are compared with the key process parameters of each control area acquired in real time in time and in multiple dimensions to obtain the comparison results. Based on preset difference threshold rules, it is determined whether the comparison results have a deviation; the difference threshold rules include temperature deviation threshold, pressure deviation threshold and system-level energy consumption estimation deviation threshold. If deviations occur, the model and / or rule base are iteratively optimized; the model is a feedforward prediction model, and the rule base is a preset fuzzy rule base. Validate the optimized model and / or rule base, and output the validation results.

10. A co-combustion control system for a tunnel kiln, characterized in that: It includes a perception layer, a control layer, an intelligent computing layer, and a digital twin layer that are electrically connected in sequence; The sensing layer includes sensors and actuators for acquiring key process parameters and executing control actions in each control area; the sensors include resistance temperature detectors / thermocouples, pressure transmitters, flow sensors, and CO concentration analyzers; the actuators include burners, gas proportional valves, and combustion air regulating valves; the flow sensors include gas flow sensors and combustion air flow sensors. The control layer includes a DCS control system, a PLC control system, and a MES system, which are used to receive control commands, drive actuators, and collect execution feedback data. The intelligent computing layer includes a real-time / historical database, an AI computing engine, an application server, a fuzzy PID control module, a feedforward prediction module, and an AI air-fuel ratio dynamic optimization module, which are responsible for data processing, algorithm calculation, and control command generation. The digital twin layer includes a temperature field simulation module, a combustion process simulation module, and a control strategy optimization module, which are used to construct a 1:1 three-dimensional computational fluid dynamics (CFD) simulation model and support simulation verification and iterative optimization of control parameters.

Images