Hearth auxiliary adjustment control system

Abstract

The invention relates to the technical field of control and regulation systems, in particular to a hearth auxiliary regulation control system which comprises a data processing module, an analysis module, a control module, a regulation and control module, an optimization module and a safety early warning module. The data processing module collects temperature data of a hearth combustor area, a water cooling wall and a tail flue in real time through a distributed temperature sensor network, and a dynamic combustion field three-dimensional model is generated through Kalman filtering denoising, redundancy cross validation and a temperature compensation algorithm in combination with combustor flame position signals and an air door opening state. And the optimization module updates the strategy weight based on the combustion efficiency improvement rate, and reinforces learning to extend the adaptive boundary of the strategy library. And the safety early warning module calculates the matching coefficient of the pipe wall temperature and the allowable stress in real time, triggers a grading intervention strategy to reduce the overtemperature risk, forms full-process closed-loop control, and improves the thermal efficiency and the operation safety of the boiler.

Description

Technical Field

[0001] The present invention relates to the technical field of control and regulation systems, and particularly to a furnace auxiliary adjustment control system. Background Art

[0002] In existing thermal power plants, different combustion conditions of boilers directly affect the thermal efficiency of the boilers. During the operation of thermal power generating units, the real-time monitoring and dynamic adjustment of the combustion state of boilers directly affect the thermal efficiency and operation safety of the units. Existing control methods mainly rely on the empirical judgment of operators on scattered monitoring data, suffering from problems such as low efficiency of manual analysis, lag in working condition response, and inconsistent operation standards. Since the combustion process inside the furnace involves the coupling effect of multi-dimensional parameters, the existing manual monitoring mode is difficult to comprehensively capture the distribution characteristics of the combustion field and the dynamic heat transfer law. Especially when the load fluctuates or the fuel characteristics change, operators cannot quickly establish a spatial correlation model of the combustion state, resulting in a lack of data support for key operations such as combustion air distribution optimization and thermal radiation balance, affecting the economic operation level of the unit and increasing safety risks. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the present invention provides a furnace auxiliary adjustment control system, which is used to solve the problems of low thermal efficiency of the unit and operation safety risks caused by response lag, inconsistent operation standards due to relying on manual experience, and the lack of a spatial correlation model of the combustion state in the existing boiler combustion control.

[0004] To solve the above technical problems, the specific technical solutions of the present invention are as follows: A furnace auxiliary adjustment control system provided by the present invention includes: A data processing module, which is used to collect system data. The system data includes temperature data of the furnace burner area, water wall and tail flue, burner flame position signals and damper opening states, and reconstructs the continuous temperature field of the furnace through a preset three-dimensional space coordinate mapping relationship to generate a three-dimensional model of the dynamic combustion field; An analysis module, which is used to extract the geometric characteristics of the combustion core area based on the three-dimensional model of the dynamic combustion field, combine the combustion air distribution volume, fuel calorific value and load change data, calculate the dynamic relationship between combustion efficiency and oxygen content and primary air rate through a preset multiple regression model, identify abnormal distribution patterns of the combustion field, and generate a combustion stability evaluation index; A control module, which is used to perform pattern matching on the historical optimization operation records and the combustion stability evaluation index output by the analysis module by using a preset machine learning algorithm to generate an initial control strategy, and verify the combustion efficiency and emission index of the candidate strategy through a digital twin model simulation, and screen out an optimized control strategy that meets the safety constraints of the preset temperature resistance limit parameters of the heating surface material; A regulation module, which is used to decompose the optimized control strategy selected by the control module into a burner inclination adjustment instruction and a damper opening adjustment sequence, correct the action deviation of the actuator through a preset feedforward and feedback composite control mechanism, and monitor the dynamic response of the temperature field based on the dynamic three-dimensional model of the combustion field updated in real time by the data processing module; An optimization module, which is used to calculate the combustion efficiency improvement rate, the flue gas temperature drop amplitude, and the heating surface temperature uniformity index after the execution of the optimized control strategy according to the dynamic response data of the temperature field fed back by the regulation module; A safety warning module, which is used to superimpose the temperature resistance limit parameters of the heating surface material defined in the safety constraint filtering unit of the control module in the dynamic three-dimensional model of the combustion field, calculate the matching coefficient between the temperature distribution and the allowable stress of the heating surface material in real time, and trigger a hierarchical intervention control strategy and push a warning message to the man-machine interface when it is detected that the matching coefficient exceeds the preset threshold.

[0005] Furthermore, for the furnace auxiliary adjustment control system of the present invention, the data processing module includes: A sensor network hierarchical deployment unit, which configures an infrared temperature measurement array and anti-interference thermocouples to form a three-level temperature measurement layer in the burner area, the water wall, and the flue, and is used to collect the temperature data of the furnace burner area, the water wall, and the tail flue; A data fusion unit, which is used to perform Kalman filtering denoising and redundant cross-verification on the original temperature data output by the sensor network hierarchical deployment unit, and eliminate the measurement deviation caused by the furnace pressure fluctuation through a temperature compensation algorithm; A three-dimensional field reconstruction unit, which establishes a furnace space grid topology based on the Delaunay triangulation algorithm, maps the discrete temperature points processed by the data fusion unit to the three-dimensional grid nodes by using the radial basis function interpolation method, dynamically corrects the interpolation weight coefficient in combination with the burner flame position signal, and outputs a three-dimensional cloud map of the temperature gradient distribution to generate the dynamic three-dimensional model of the combustion field; Among them, the output of the sensor network hierarchical deployment unit is connected to the data fusion unit, the output of the data fusion unit is connected to the three-dimensional field reconstruction unit, and the output of the three-dimensional field reconstruction unit is transmitted to the analysis module.

[0006] Furthermore, for the furnace auxiliary adjustment control system of the present invention, the analysis module includes: A combustion core area identification unit, which is used to extract the continuous area with a temperature ≥ 1000 °C from the dynamic three-dimensional model of the combustion field, determine the geometric center coordinates and the volume change rate through an edge detection algorithm, and calculate the combustion intensity index in combination with the data of the external flue gas composition analyzer; The multi-parameter coupling modeling unit constructs a multiple regression model of combustion efficiency with oxygen content, primary air rate, and fuel volatile matter, and uses a sliding time window to statistically analyze the parameter change trends of the combustion air distribution, fuel calorific value, and load change data during load fluctuations to generate a combustion stability feature vector; The abnormal pattern matching unit extracts the temperature field distribution template under typical working conditions based on historical operation data, uses the dynamic time warping algorithm to match the similarity between the current combustion field and the template in real time, and outputs an abnormal deviation index to generate the combustion stability evaluation index; Among them, the output of the combustion core area identification unit is connected to the multi-parameter coupling modeling unit, the output of the multi-parameter coupling modeling unit is connected to the abnormal pattern matching unit, and the output of the abnormal pattern matching unit is transmitted to the control module.

[0007] Furthermore, for the furnace auxiliary adjustment control system of the present invention, the control module includes: The operation mode matching unit uses a convolutional neural network to extract features from historical optimization operation records, takes the temperature distribution map generated by the dynamic combustion field three-dimensional model as input, associates the corresponding burner tilt angle and damper opening label data, and generates an initial recommended strategy set; The digital twin verification unit imports the current combustion field boundary conditions into the CFD simulation engine, calculates the flue gas recirculation rate and NOx generation amount under different damper combinations, and screens the feasible strategies that meet the thermal efficiency threshold and emission constraints; The safety constraint filtering unit takes the heat stress threshold of the heating surface tube wall as a hard constraint condition, uses the Lagrange multiplier method to correct the candidate strategies output by the digital twin verification unit, and excludes the adjustment schemes that cause over-temperature risks to generate the optimized control strategy; Among them, the output of the operation mode matching unit is connected to the digital twin verification unit, the output of the digital twin verification unit is connected to the safety constraint filtering unit, and the output of the safety constraint filtering unit is transmitted to the regulation module.

[0008] Furthermore, for the furnace auxiliary adjustment control system of the present invention, the regulation module includes: The instruction decomposition unit analyzes the optimized control strategy generated by the safety constraint filtering unit into the angle increment of the burner servo motor, the target value of the secondary damper opening, and the fuel distribution valve adjustment coefficient, and generates a timing coordination instruction queue according to the actuator response delay test data; The dynamic compensation unit sets a feedforward control compensation coefficient based on the actuator response delay test data, and combines PID feedback control to eliminate the steady-state error between the actual opening and the target value; Among them, the output of the instruction decomposition unit is connected to the dynamic compensation unit, the output of the dynamic compensation unit is connected to the burner servo mechanism and the damper actuator, and the actuator action signal is fed back to the optimization module.

[0009] Further, for the furnace auxiliary adjustment control system of the present invention, the optimization module includes: An effect evaluation unit, which is used to collect the average flue gas temperature, the entropy increase rate of the steam-water system, and the standard deviation of the temperature difference of the heating surface in the dynamic response data of the temperature field fed back by the regulation module, calculate the contribution degree of the combustion efficiency improvement, and update the weight coefficient of the control strategy generation rule of the control module; A knowledge base expansion unit, when detecting a load mutation or a fuel switching event in the combustion air distribution, fuel calorific value, and load change data output by the analysis module, triggers a reinforcement learning mechanism to extract effective operation characteristics in the adjustment sequence data of the regulation module, and expands the adaptation boundary of the strategy library of the control module through a transfer learning algorithm.

[0010] Further, for the furnace auxiliary adjustment control system of the present invention, the safety warning module includes: A life prediction unit, which is used to calculate the equivalent operation time of the tube wall temperature and the design life in real time based on the Larson-Miller parameter in the temperature resistance limit parameters of the heating surface material defined in the safety constraint filtering unit of the control module, and triggers an alarm when the cumulative damage coefficient ≥ 0.8; A hierarchical intervention unit, which is used to push adjustment suggestions to the human-machine interface for a first-level warning according to the warning signal of the life prediction unit, and trigger the combustion core area offset control strategy generated by the control module for a second-level warning to forcibly reduce the regional combustion intensity.

[0011] Further, for the furnace auxiliary adjustment control system of the present invention, a communication redundancy check module is provided between the sensor network hierarchical deployment unit and the data fusion unit, which is used to detect the transmission packet loss rate of the temperature data output by the sensor network hierarchical deployment unit, and trigger a data retransmission mechanism to maintain the integrity of the input data of the three-dimensional field reconstruction unit.

[0012] Further, for the furnace auxiliary adjustment control system of the present invention, a strategy conflict detection module is provided between the digital twin verification unit and the safety constraint filtering unit, which is used to compare the combustion efficiency improvement value of the candidate strategy output by the digital twin verification unit with the safety threshold of the temperature resistance limit parameters of the heating surface material defined in the safety constraint filtering unit, identify the conflict points, and preferentially execute the safety constraint filtering operation of the safety constraint filtering unit.

[0013] Furthermore, in the furnace auxiliary adjustment control system of the present invention, an action timing calibration module is provided between the dynamic compensation unit and the actuator, which is used to decompose the mechanical response delay characteristics of the damper actuator in the actuator response delay test data stored in the instruction decomposition unit of the regulation module according to the instructions of the regulation module, and insert a time buffer interval into the timing coordination instruction queue generated by the instruction decomposition unit.

[0014] Advantages of the present invention; The advantages of the present invention are as follows: real-time collection of multi-dimensional temperature data of the furnace through a distributed temperature sensor network, combined with Kalman filtering and redundant cross-validation to eliminate measurement noise, generating a high-precision three-dimensional model of the dynamic combustion field, and realizing holographic perception of the combustion state; based on a multiple regression model and a dynamic time warping algorithm, associating the combustion efficiency with the dynamic relationship of multiple parameters, identifying abnormal patterns and generating stability evaluation indicators, providing a quantitative basis for control strategies; extracting historical operation characteristics through a convolutional neural network and combining with digital twin simulation verification, screening an optimization strategy set that meets the requirements of improving thermal efficiency and safety constraints, and standardizing control rules to reduce the differences in manual intervention; a feedforward and feedback composite control mechanism and dynamic timing calibration eliminate the action deviation of the actuator, improving the response speed and synchronization accuracy; the closed-loop optimization module dynamically updates the strategy weights based on the combustion efficiency improvement rate and the temperature difference index of the heating surface, and reinforcement learning expands the strategy library to adapt to complex working conditions; the safety warning module real-time evaluates the wall cumulative damage coefficient through the Larson-Miller parameter, triggers a hierarchical intervention strategy to reduce the over-temperature risk, forming a full-process closed-loop control from data perception, intelligent decision-making to safety protection, effectively improving the boiler thermal efficiency and reducing the operation safety risk. Description of the drawings

[0015] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, other drawings can be obtained according to the drawings without creative efforts.

[0016] Figure 1 It is a system architecture diagram of a furnace auxiliary adjustment control system provided by an embodiment of the present invention. Detailed implementation manners

[0017] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention. The following will describe in detail the technical solutions provided by each embodiment of the present invention with reference to the drawings. To better understand the objectives of the present invention, the present invention will be further described in detail below.

[0018] Please refer to Figure 1 , a furnace auxiliary adjustment control system provided by the present invention includes: A data processing module, configured to collect system data. The system data includes temperature data of the furnace burner area, water wall, and tail flue, burner flame position signals, and damper opening states. The system data is used to reconstruct the continuous temperature field of the furnace through a preset three-dimensional space coordinate mapping relationship, and generate a three-dimensional model of the dynamic combustion field; An analysis module, configured to extract geometric features of the combustion core area based on the three-dimensional model of the dynamic combustion field, combine combustion air distribution, fuel calorific value, and load change data, calculate the dynamic relationship between combustion efficiency and oxygen content and primary air rate through a preset multiple regression model, identify abnormal patterns of the combustion field distribution, and generate a combustion stability evaluation index; A control module, configured to perform pattern matching on historical optimization operation records and the combustion stability evaluation index output by the analysis module using a preset machine learning algorithm, generate an initial control strategy, and verify the combustion efficiency and emission indexes of the candidate strategy through a digital twin model simulation, and screen out an optimized control strategy that meets the safety constraints of the preset temperature resistance limit parameters of the heating surface material; A regulation module, configured to decompose the optimized control strategy selected by the control module into burner tilt adjustment instructions and damper opening adjustment sequences, correct the action deviation of the actuator through a preset feedforward and feedback composite control mechanism, and monitor the dynamic response of the temperature field based on the three-dimensional model of the dynamic combustion field updated in real time by the data processing module; An optimization module, configured to calculate the combustion efficiency improvement rate, flue gas temperature drop, and heating surface temperature uniformity index after the execution of the optimized control strategy according to the temperature field dynamic response data fed back by the regulation module; A safety warning module, configured to superimpose the temperature resistance limit parameters of the heating surface material defined in the safety constraint filtering unit of the control module on the three-dimensional model of the dynamic combustion field, calculate the matching coefficient between the temperature distribution and the allowable stress of the heating surface material in real time, and when it is detected that the matching coefficient exceeds the preset threshold, trigger a hierarchical intervention control strategy and push a warning message to the human-machine interface; Among them, the output of the data processing module is connected to the analysis module, the output of the analysis module is connected to the control module, the output of the control module is connected to the regulation module, the output of the regulation module is fed back to the optimization module, the output of the optimization module is connected to the control module, and the safety warning module receives the combustion stability evaluation index output by the analysis module and the adjustment sequence data of the regulation module in real time and outputs a warning signal.

[0019] A furnace auxiliary adjustment control system provided by the present invention realizes real-time monitoring and dynamic adjustment of the furnace combustion state through hierarchical modular design. The system first collects temperature data of the furnace burner area, water wall and tail flue through a distributed temperature sensor network, and synchronously obtains the burner flame position signal and the damper opening state. The data processing module uses Kalman filtering to denoise the original temperature data, eliminates outliers by combining redundant cross-validation, and corrects the measurement deviation caused by furnace pressure fluctuation through a temperature compensation algorithm. Based on the Delaunay triangulation algorithm, a furnace space grid topology is established, and the radial basis function interpolation method is used to map discrete temperature points to three-dimensional grid nodes. Combining the flame position signal, the interpolation weight is dynamically adjusted to reconstruct a three-dimensional dynamic combustion field model with temperature gradient distribution characteristics, providing a spatio-temporal fusion data basis for subsequent analysis.

[0020] The analysis module extracts the continuous area with a temperature ≥ 1000 °C from the three-dimensional dynamic combustion field model as the combustion core area, determines the geometric center coordinates and volume change rate of the core area through an edge detection algorithm, and calculates the combustion intensity index in combination with the data of the flue gas composition analyzer. A multiple regression model of combustion efficiency with oxygen content, primary air rate and fuel volatile matter is constructed, and the parameter trends of combustion air distribution, fuel calorific value and load change during load fluctuation are statistically analyzed using a sliding time window to generate a characteristic vector representing combustion stability. Based on historical operation data, a temperature field distribution template under typical working conditions is extracted, and the dynamic time warping algorithm is used to match the similarity between the current combustion field and the template in real time, and an abnormal deviation index is output to complete the identification of the abnormal pattern of the combustion field distribution.

[0021] The control module uses a convolutional neural network to extract features from historical optimization operation records, associates the three-dimensional temperature distribution map of the dynamic combustion field with the burner tilt angle and damper opening label data, and generates an initial recommended strategy set. The current combustion field boundary conditions are imported into the CFD simulation engine through a digital twin model, and the flue gas recirculation rate and NOx generation amount under different damper combinations are calculated, and feasible strategies that meet the thermal efficiency threshold and comply with emission constraints are screened. Taking the heat stress threshold of the heating surface tube wall as a hard constraint condition, the Lagrange multiplier method is used to correct the candidate strategy to exclude adjustment schemes that cause local overheating risks, and a safe and optimized control strategy set is formed.

[0022] The regulation module analyzes the optimized control strategy into the angle increment of the burner servo motor, the target value of the secondary air damper opening, and the adjustment coefficient of the fuel distribution valve, and generates a timing coordination instruction queue based on the actuator response delay test data. Based on the feedforward controller, the lag effect of the temperature field response is predicted, and combined with the PID feedback control to eliminate the action deviation of the actuator. The temperature change of the dynamic combustion field three-dimensional model is monitored in real time to form a closed-loop control link. The actuator action signal is fed back to the optimization module to continuously adjust the control accuracy of the drive system.

[0023] The optimization module collects the average flue gas temperature, the entropy increase rate of the steam-water system, and the standard deviation of the temperature difference of the heating surface after the implementation of the strategy, quantifies the contribution degree of the combustion efficiency improvement, and updates the weight coefficient of the control strategy generation rule. When a load mutation or fuel switching event is detected, the reinforcement learning mechanism is triggered to extract effective operation features, and the adaptation boundary of the strategy library is extended through the transfer learning algorithm to enhance the system's ability to cope with complex working conditions.

[0024] The safety warning module superimposes the temperature resistance limit parameters of the heating surface material in the dynamic combustion field three-dimensional model, and calculates the matching coefficient of the tube wall temperature and the allowable stress in real time. The Larson-Miller parameter is used to evaluate the cumulative damage degree of the tube wall. When the damage coefficient ≥ 0.8, a warning is triggered. Hierarchical intervention is performed according to the warning level: the first-level warning pushes adjustment suggestions to the operation interface; the second-level warning automatically triggers the combustion core area offset control strategy to forcibly reduce the combustion intensity in the over-temperature area and achieve long-term safe operation of the equipment. Each module transmits key parameters through a standardized data interface to form a closed-loop control link from state perception, strategy generation to execution and regulation.

[0025] Specifically, for the furnace auxiliary adjustment control system described in the present invention, the data processing module includes: The sensor network hierarchical deployment unit is configured with an infrared temperature measurement array and anti-interference thermocouples to form a three-level temperature measurement layer in the burner area, the water wall and the flue, and is used to collect the temperature data of the furnace burner area, the water wall and the tail flue; The data fusion unit is used to perform Kalman filtering denoising and redundant cross-validation on the original temperature data output by the sensor network hierarchical deployment unit, and eliminate the measurement deviation caused by the furnace pressure fluctuation through the temperature compensation algorithm; The three-dimensional field reconstruction unit establishes the furnace space grid topology based on the Delaunay triangulation algorithm, uses the radial basis function interpolation method to map the discrete temperature points processed by the data fusion unit to the three-dimensional grid nodes, dynamically corrects the interpolation weight coefficient in combination with the burner flame position signal, and outputs a three-dimensional cloud map of the temperature gradient distribution to generate the dynamic combustion field three-dimensional model; Among them, the output of the sensor network hierarchical deployment unit is connected to the data fusion unit, the output of the data fusion unit is connected to the three-dimensional field reconstruction unit, and the output of the three-dimensional field reconstruction unit is transmitted to the analysis module.

[0026] In the furnace auxiliary adjustment control system of the present invention, the data processing module collects multi-dimensional temperature data in real time through a hierarchically deployed sensor network. The sensor network hierarchical deployment unit configures a high-frequency infrared temperature measurement array in the burner area, and uses its fast response characteristics to capture the transient fluctuations of the flame; anti-interference armored thermocouples are arranged in the water wall area, and physical isolation design is used to reduce high-temperature radiation interference; infrared thermal imagers are used in the flue gas area to realize continuous monitoring of the large-range temperature field. The three-level temperature measurement layer covers the key heat transfer areas of the furnace, forming a space-time synchronous data acquisition system, providing basic data support for subsequent modeling.

[0027] The data fusion unit receives the original temperature data output by the sensor network hierarchical deployment unit, uses the Kalman filter algorithm to eliminate random noise interference, combines the redundant data of adjacent sensor nodes for cross-verification, and identifies and eliminates abnormal measurement values. Aiming at the sensor drift problem caused by the internal pressure fluctuation of the furnace, a temperature compensation algorithm is introduced to establish a pressure-temperature correction model to dynamically correct the measurement deviation and improve the data reliability. The temperature data after fusion processing retains the effective operating conditions characteristics and provides a standardized input for three-dimensional field reconstruction.

[0028] The three-dimensional field reconstruction unit constructs the furnace space grid topology based on the Delaunay triangulation algorithm, and adapts to the complex furnace geometry through unstructured grid division. Map the discrete temperature measurement points output by the data fusion unit to the grid nodes, and use the radial basis function interpolation algorithm to calculate the temperature distribution in the unmeasured area. Dynamically adjust the interpolation weight coefficient in combination with the burner flame position signal, increase the interpolation density in the flame active area, and improve the temperature field resolution of the combustion core area. The reconstruction process generates a three-dimensional cloud map with temperature gradient distribution characteristics, intuitively reflecting the distribution law of the internal heat radiation intensity of the furnace, and providing a visual data model for combustion state analysis. The reconstruction result is transmitted to the analysis module through a standardized interface, forming a complete link of data acquisition - processing - modeling.

[0029] Specifically, in the furnace auxiliary adjustment control system of the present invention, the analysis module includes: A combustion core area identification unit, which is used to extract the continuous area with a temperature ≥ 1000 °C from the three-dimensional model of the dynamic combustion field, determine the geometric center coordinates and volume change rate through an edge detection algorithm, and calculate the combustion intensity index in combination with the data of the external flue gas analyzer; The multi-parameter coupling modeling unit constructs a multiple regression model of combustion efficiency with oxygen content, primary air rate, and fuel volatile matter. Using a sliding time window, it statistically analyzes the parameter change trends of combustion air distribution, fuel calorific value, and load change data during load fluctuations to generate a combustion stability feature vector. The abnormal pattern matching unit extracts the temperature field distribution template under typical operating conditions based on historical operation data, and uses the dynamic time warping algorithm to match the similarity between the current combustion field and the template in real time, and outputs an abnormal deviation index to generate the combustion stability evaluation index. Among them, the output of the combustion core area identification unit is connected to the multi-parameter coupling modeling unit, the output of the multi-parameter coupling modeling unit is connected to the abnormal pattern matching unit, and the output of the abnormal pattern matching unit is transmitted to the control module.

[0030] In the furnace auxiliary adjustment control system of the present invention, the analysis module quantitatively evaluates the combustion state based on the dynamic combustion field three-dimensional model. The combustion core area identification unit screens out continuous areas with a temperature ≥ 1000 °C from the three-dimensional model, extracts the boundary contour of the combustion core area through the Canny edge detection algorithm, and determines the geometric center coordinates and volume change rate by combining morphological operations. Synchronously access the data of the external flue gas analyzer, calculate the ratio of oxygen concentration to carbon monoxide content, and construct a combustion intensity index in combination with the volume parameters of the combustion core area to characterize the activity of the current combustion reaction. The dynamic parameters of the combustion core area provide a spatial correlation basis for multi-parameter coupling modeling.

[0031] The multi-parameter coupling modeling unit establishes a multiple regression model of combustion efficiency with oxygen sensor data, primary air rate adjustment signal, and fuel volatile matter parameters, and fits the parameter weight coefficients under historical operating conditions by the least squares method. Using a sliding time window to intercept the combustion air distribution, fuel calorific value, and load change data during the load fluctuation period, it statistically analyzes the standard deviation and covariance matrix of each parameter to generate a feature vector reflecting combustion stability. The feature vector includes the fluctuation amplitude of combustion efficiency, the coupling strength between parameters, and the response delay characteristics, providing multi-dimensional input for abnormal pattern matching.

[0032] The abnormal pattern matching unit extracts the temperature field distribution template under typical operating conditions (such as full load steady state, low load peak shaving) from the historical operation database, uses the dynamic time warping algorithm to align the time series data of the current combustion field and the template, and calculates the cumulative distance of the temperature gradient distribution difference. It judges the deviation degree of the current combustion field by setting a similarity threshold and outputs an abnormal deviation index. When the deviation degree exceeds the preset range, it triggers a downgrade signal of the combustion stability evaluation index, driving the control module to generate an intervention strategy. The combustion core area parameters, multi-parameter feature vector, and abnormal deviation index are transmitted to the control module through a standardized interface, forming a data closed-loop from combustion state analysis to control decision-making.

[0033] Specifically, for the furnace auxiliary adjustment control system of the present invention, the control module includes: An operation mode matching unit that uses a convolutional neural network to extract features from historical optimization operation records, takes the temperature distribution map generated by the dynamic combustion field three-dimensional model as input, correlates the corresponding burner tilt angle and damper opening label data, and generates an initial recommended strategy set; A digital twin verification unit that imports the current combustion field boundary conditions into the CFD simulation engine, calculates the flue gas recirculation rate and NOx generation amount under different damper combinations, and filters out feasible strategies that meet the thermal efficiency threshold and emission constraints; A safety constraint filtering unit that takes the heat stress threshold of the heating surface tube wall as a hard constraint condition, uses the Lagrange multiplier method to correct the candidate strategies output by the digital twin verification unit, and excludes adjustment schemes that cause over-temperature risks to generate the optimized control strategy; Among them, the output of the operation mode matching unit is connected to the digital twin verification unit, the output of the digital twin verification unit is connected to the safety constraint filtering unit, and the output of the safety constraint filtering unit is transmitted to the regulation module.

[0034] In the furnace auxiliary adjustment control system of the present invention, the control module generates a safe and reliable control strategy through a multi-level verification mechanism. The operation mode matching unit uses a convolutional neural network to extract the feature patterns in the historical optimization operation records, inputs the temperature distribution map generated by the dynamic combustion field three-dimensional model into the network, captures the spatial distribution characteristics of the temperature gradient through the convolutional layer, extracts the key area features through the pooling layer, and associates the burner tilt angle and damper opening label data through the fully connected layer to output an initial recommended strategy set that matches the current combustion state. This strategy set covers multi-dimensional operation schemes for burner angle adjustment, damper opening combination, and fuel distribution ratio, providing a candidate basis for subsequent verification.

[0035] After receiving the initial recommended strategy set, the digital twin verification unit imports the real-time boundary conditions of the current combustion field into the computational fluid dynamics (CFD) simulation engine to construct a multi-physics field coupling model including flue gas flow, heat transfer, and chemical reactions. By discretely solving the flow field distribution under different damper combinations, calculates the influence of the flue gas recirculation rate on the combustion efficiency, and synchronously predicts the NOx generation amount. Combining the preset thermal efficiency threshold and emission constraint conditions, filters out feasible strategies that meet the requirements for improving combustion efficiency and meet environmental protection standards. The simulation results generate a strategy evaluation report including operation parameters and expected performance indicators, providing input data for safety constraint filtering.

[0036] The safety constraint filtering unit receives the candidate strategies output by the digital twin verification unit, and uses the thermal stress threshold of the heating surface tube wall material as a hard constraint condition. The Lagrange multiplier method is used to establish an association model between the optimization objective function and the constraint conditions, and the burner tilt angle and damper opening parameters in the candidate strategies are corrected for feasibility. Through iterative calculations, the parameter combinations that cause local over-temperature risks are eliminated, and the adjustment plans within the thermal stress safety range are retained to form the final optimized control strategy. The optimized strategy is transmitted to the regulation module through a standardized interface to drive the actuator to act in coordination. The strategies generation, simulation verification, and safety filtering process closed-loop among the units are realized through data interfaces to ensure that the control strategy takes into account both efficiency improvement and equipment safety.

[0037] Specifically, for the furnace auxiliary adjustment control system described in the present invention, the regulation module includes: An instruction decomposition unit that analyzes the optimized control strategy generated by the safety constraint filtering unit into the angle increment of the burner servo motor, the target value of the secondary damper opening, and the adjustment coefficient of the fuel distribution valve, and generates a time-sequential coordination instruction queue according to the actuator response delay test data; A dynamic compensation unit that sets a feedforward control compensation coefficient based on the actuator response delay test data, and combines PID feedback control to eliminate the steady-state error between the actual opening and the target value; Among them, the output of the instruction decomposition unit is connected to the dynamic compensation unit, the output of the dynamic compensation unit is connected to the burner servo mechanism and the damper actuator, and the actuator action signal is fed back to the optimization module.

[0038] In the furnace auxiliary adjustment control system described in the present invention, the regulation module converts the optimized strategy into precise action instructions for the actuator. The instruction decomposition unit receives the optimized control strategy generated by the safety constraint filtering unit, analyzes the operation parameters such as the angle increment of the burner servo motor, the target value of the secondary damper opening, and the adjustment coefficient of the fuel distribution valve, and generates an instruction queue with time tags according to the mechanical response time, action accuracy, and coordination time sequence requirements in the actuator response delay test data. The time-sequential coordination arrangement avoids instruction conflicts during parallel operation of multiple actuators, reduces the risk of mechanical overload, and maintains the continuity of combustion adjustment.

[0039] Based on the step response characteristics in the actuator response delay test data, the dynamic compensation unit sets the feedforward control compensation coefficient to predict the lag effect during the adjustment process of the damper opening. In combination with the PID feedback controller, it real-time collects the actuator position signal, calculates the deviation between the actual opening and the target value, and comprehensively adjusts the output through the proportional, integral, and differential terms to eliminate the steady-state error. The feedforward and feedback composite control mechanism achieves a balance between fast response and stable tracking, improving the synchronous control accuracy of the burner inclination and the damper opening. The actuator action signal is real-time fed back to the optimization module, providing dynamic response data for the evaluation of the strategy effect, and forming a closed-loop control link from command generation to execution verification.

[0040] Specifically, in the furnace auxiliary adjustment control system of the present invention, the optimization module includes: An effect evaluation unit, which is used to collect the average flue gas temperature, the entropy increase rate of the steam-water system, and the standard deviation of the temperature difference of the heating surface in the dynamic response data of the temperature field fed back by the regulation module, calculate the contribution degree of the combustion efficiency improvement, and update the weight coefficient of the control strategy generation rule of the control module; A knowledge base expansion unit, when detecting the load mutation or fuel switching event, triggers the reinforcement learning mechanism to extract the effective operation characteristics in the adjustment sequence data of the regulation module, and expands the adaptation boundary of the strategy library through the transfer learning algorithm; Among them, the output of the effect evaluation unit is connected to the knowledge base expansion unit, and the output of the knowledge base expansion unit is fed back to the control module.

[0041] In the furnace auxiliary adjustment control system of the present invention, the optimization module realizes the continuous iteration of the control strategy through a closed-loop feedback mechanism. The effect evaluation unit real-time collects the dynamic response data of the temperature field fed back by the regulation module, and extracts three core indicators: the average flue gas temperature, the entropy increase rate of the steam-water system, and the standard deviation of the temperature difference of the heating surface. Based on the combustion efficiency benchmark model, it calculates the contribution degree weight of each indicator to the efficiency improvement, and uses the weighted fusion algorithm to generate a comprehensive evaluation value, dynamically adjusting the priority coefficient of the strategy generation rule in the control module. The weight coefficient update process introduces a moving average filtering process to avoid frequent switching of the strategy caused by short-term fluctuations in the working conditions, and enhances the system stability.

[0042] The knowledge base expansion unit monitors the load command signal and fuel characteristic parameters. When it detects that the load mutation exceeds the set threshold or the fuel calorific value difference exceeds the allowable range, it triggers the reinforcement learning mechanism. It extracts operation feature vectors such as the burner tilt adjustment rate, the air damper opening change trajectory, and the fuel distribution ratio from the historical adjustment sequence data of the regulation module, and calculates the optimal action value function under different working conditions through the Q-learning algorithm. It uses transfer learning technology to map the operation modes in the existing policy library to the new working condition feature space, constructs a cross-working condition generalization policy model, and expands the adaptability of the policy library to unvalidated working conditions. The updated policy library is synchronized to the control module through a standardized interface, forming a closed-loop optimization link from effect evaluation to knowledge iteration. The output data of the effect evaluation unit is used as the input constraint for knowledge base expansion, restricting the reinforcement learning search space and preventing the policy from deviating from the safety boundary.

[0043] Specifically, in the furnace auxiliary adjustment control system of the present invention, the safety warning module includes: The life prediction unit is used to calculate the equivalent operation time of the tube wall temperature and the design life in real time based on the Larson-Miller parameter in the temperature resistance limit parameters of the heating surface material, and triggers a warning when the cumulative damage coefficient ≥ 0.8; The hierarchical intervention unit is used to push adjustment suggestions to the man-machine interface for a first-level warning according to the warning signal of the life prediction unit, and trigger the combustion core area offset control strategy for a second-level warning to forcibly reduce the regional combustion intensity; Among them, the output of the life prediction unit is connected to the hierarchical intervention unit, and the output of the hierarchical intervention unit is connected to the regulation module and the man-machine interaction interface.

[0044] In the furnace auxiliary adjustment control system of the present invention, the safety warning module ensures the safe operation of the heating surface through a dynamic monitoring and hierarchical response mechanism. The life prediction unit receives the tube wall temperature distribution data in the dynamic combustion field three-dimensional model in real time, combines the Larson-Miller parameter in the temperature resistance limit parameters of the heating surface material, and calculates the ratio of the equivalent operation time to the design life. Through the cumulative damage model, time integration is performed on historical over-temperature events. When the cumulative damage coefficient ≥ 0.8, a warning signal is triggered, indicating that the tube wall material enters the accelerated creep stage. This unit synchronously outputs the coordinates of the over-temperature area and the temperature change trend map, providing a positioning basis for hierarchical intervention.

[0045] The hierarchical intervention unit executes differential control strategies according to the early warning signal levels. The first-level early warning corresponds to the situation where the local over-temperature ≤ 10°C and the duration < 5 minutes. It pushes the fine-tuning suggestions for the burner inclination angle, the optimization plan for the secondary air damper opening, and the operation instructions for the heating surface soot blowing to the human-machine interface, prompting the operator to intervene and adjust. The second-level early warning targets the high-risk state of over-temperature > 10°C or continuous over-temperature ≥ 5 minutes, automatically triggering the combustion core area offset control strategy. Through the control module, it forcibly adjusts the inclination angle of the corresponding area burner, reduces the fuel distribution ratio, and increases the secondary air volume, migrating the high-temperature core area to the low-temperature side to reduce the tube wall heat load. During the execution process, it continuously monitors the dynamic response of the adjusted temperature field until the matching coefficient returns to the safe threshold range. After the early warning signal is lifted, the system automatically resumes to the optimized control mode to maintain the dynamic balance between combustion efficiency and safe operation. The damage assessment data of the life prediction unit and the control instructions of the hierarchical intervention unit are interacted with the control module through a standardized interface, forming a closed-loop protection link from risk identification to active intervention.

[0046] Specifically, in the furnace auxiliary adjustment control system of the present invention, a communication redundancy check module is provided between the sensor network hierarchical deployment unit and the data fusion unit, which is used to detect the transmission packet loss rate of the temperature data output by the sensor network hierarchical deployment unit and trigger a data retransmission mechanism to maintain the integrity of the input data of the three-dimensional field reconstruction unit.

[0047] In the furnace auxiliary adjustment control system of the present invention, the communication redundancy check module constructs a data transmission reliability guarantee link between the sensor network hierarchical deployment unit and the data fusion unit. This module receives the temperature data stream output by the sensor network hierarchical deployment unit, attaches a sequence number and a cyclic redundancy check code (CRC) to each data packet, and verifies the integrity of the data packet through the CRC matching. It uses a sliding window statistical mechanism to calculate the data transmission packet loss rate in real time. When it detects continuous data packet loss or the error rate of CRC verification exceeds the set threshold, it triggers the automatic repeat request (ARQ) mechanism, sends a NACK signal to the sensor network to request the retransmission of the specified sequence number data packet until the receiving end returns an ACK confirmation signal.

[0048] The data fusion unit receives the temperature data stream processed by the redundancy check module and restores the complete data set through timestamp alignment and data packet recombination. For the problem of data timing disorder caused by communication delay, a buffer queue is used to temporarily store out-of-order data packets, which are re-sorted by sequence number and then output to the subsequent processing unit. Packets that fail the check and cannot be repaired by retransmission are marked as invalid data and replaced by redundant data interpolation of adjacent nodes to avoid the influence of outliers on the accuracy of three-dimensional field reconstruction. The processed standardized data is output to the three-dimensional field reconstruction unit to maintain the continuity of temperature data at the spatial grid nodes. The communication status monitoring log records the packet loss rate, retransmission times, and check error types in real time, providing a basis for evaluating the health status of the transmission link for the human-machine interface, and forming a closed-loop reliability control for the entire process from data acquisition, transmission check to exception handling.

[0049] Specifically, in the furnace auxiliary adjustment control system of the present invention, a policy conflict detection module is provided between the digital twin verification unit and the safety constraint filtering unit, which is used to compare the combustion efficiency improvement value of the candidate policy output by the digital twin verification unit with the safety threshold of the temperature resistance limit parameter of the heating surface material, identify the conflict point and give priority to performing the safety constraint filtering operation.

[0050] In the furnace auxiliary adjustment control system of the present invention, the policy conflict detection module establishes a multi-objective optimization decision-making mechanism between the digital twin verification unit and the safety constraint filtering unit. This module receives the combustion efficiency improvement value of the candidate policy output by the digital twin verification unit and the corresponding damper opening combination parameters, and simultaneously obtains the safety threshold data in the temperature resistance limit parameters of the heating surface material. By establishing an efficiency-safety double-objective optimization model and using the Pareto front analysis method to compare the combustion efficiency gain of the candidate policy with the safety margin of the tube wall temperature, a set of conflict strategies with the combustion efficiency improvement value exceeding the allowable range of the safety threshold is identified. For the conflicting strategies, mark the key operation parameter combinations that cause the over-temperature risk and generate a conflict feature vector.

[0051] The conflict feature vector includes conflict parameters such as the adjustment range of the burner tilt angle, the change rate of the secondary damper opening, and the fuel distribution ratio, and is input to the Lagrange multiplier method optimization model of the safety constraint filtering unit. First, the parameters exceeding the safety threshold in the conflicting strategies are corrected for boundary constraints, and the tube wall temperature is limited below the critical value of material creep damage through iterative calculation. The corrected strategy retains the maximum feasible combustion efficiency improvement space, and at the same time adjusts the thermal stress distribution of the heating surface to the safe area. The conflict detection result and the corrected strategy are transmitted to the control module through a standardized interface to drive the actuator to execute the safety-first operation instruction. The conflict detection log records the policy conflict type, correction amplitude, and final optimization effect, providing decision transparency support for the human-machine interface, and forming a closed-loop optimization link from policy verification, conflict identification to safety correction.

[0052] Specifically, in the furnace auxiliary adjustment control system of the present invention, an action timing calibration module is provided between the dynamic compensation unit and the actuator, which is used to insert a time buffer interval into the timing coordination instruction queue generated by the instruction decomposition unit according to the mechanical response delay characteristics of the damper actuator in the actuator response delay test data, so as to avoid the superposition conflict of control instructions.

[0053] In the furnace auxiliary adjustment control system of the present invention, the action timing calibration module establishes a timing coordination control mechanism between the dynamic compensation unit and the actuator. This module receives the mechanical delay parameters of the damper actuator in the actuator response delay test data, including the start delay time, the full stroke action time, and the return hysteresis characteristic. The dynamic response curve under different load conditions is obtained through the step response test, and the maximum delay time is extracted as the reference value. Combining with the interval requirements of adjacent instructions in the timing coordination instruction queue generated by the instruction decomposition unit, the time buffer interval to be inserted is dynamically calculated. The setting of the buffer interval follows the principle of "minimum necessary delay", minimizing the control lag to the greatest extent on the premise of avoiding instruction superposition.

[0054] The action timing calibration module dynamically adjusts the timing coordination instruction queue. It preferentially inserts a buffer interval for the long stroke action instruction of the damper actuator, and the short stroke fine-tuning instruction adopts a parallel execution mode. For the multi-actuator coordination scenario, an instruction priority queue is established, and the fast response instruction of the burner servo motor and the slow action instruction of the damper actuator are arranged with staggered peaks. The time window segmentation technology is used to avoid the torque conflict caused by the mechanical action overlap. During the action process of the actuator, the actual position signal is real-time fed back through the servo motor encoder and the damper opening sensor. The calibration module compares the timing deviation between the target position and the actual position, and dynamically corrects the buffer interval parameters of the subsequent instruction queue.

[0055] The calibrated timing coordination instruction queue is transmitted to the actuator to drive the burner servo motor and the damper actuator to perform actions according to the optimized timing. The actual opening data, action delay records, and conflict event logs generated during the execution process are real-time fed back to the instruction decomposition unit for updating the actuator response delay database. After being analyzed by the strategy evaluation unit of the optimization module, the feedback data generates a weight adjustment suggestion for the timing calibration parameters, forming a closed-loop optimization link from instruction generation, timing calibration to execution feedback, improving the stability and reliability of multi-actuator coordination control.

[0056] The technical features involved in the technical solution of the present invention and their explanations are as follows: Distributed Temperature Sensor Network: A three - level temperature measurement layer composed of an infrared temperature measurement array, anti - interference thermocouples, and an infrared thermal imager group, which are respectively deployed in the furnace burner area, water - cooled wall, and flue area. The infrared temperature measurement array captures the transient temperature fluctuations of the flame at a sampling frequency of ≥10Hz; the armored thermocouple reduces the high - temperature radiation interference through physical shielding design and realizes the measurement with an accuracy of ±1.5°C; the infrared thermal imager group covers a large area of the flue, forming a space - time synchronous data acquisition system, providing the original temperature data for 3D modeling.

[0057] Kalman Filter and Redundant Cross - Validation: The core algorithms used in the data processing module to eliminate measurement noise. The Kalman filter filters out the random noise in the temperature signal through state prediction and measurement update iteration; the redundant cross - validation identifies and eliminates abnormal measurement values through the spatial correlation analysis of adjacent sensor data, improving data reliability.

[0058] Dynamic Combustion Field 3D Model: Based on the Delaunay triangulation algorithm, an unstructured space grid of the furnace is constructed. The radial basis function interpolation method is used to map discrete temperature points to grid nodes, and the interpolation weight coefficient is dynamically adjusted in combination with the flame position signal to generate a temperature gradient distribution cloud map with a resolution of 0.1m³. This model intuitively reflects the internal heat radiation intensity and the spatial distribution characteristics of the combustion core area of the furnace, providing high - precision input for combustion state analysis.

[0059] Combustion Core Area Identification and Combustion Intensity Index: Extract the continuous area with a temperature ≥1000°C from the 3D model, determine the geometric boundary through the Canny edge detection algorithm, and calculate the volume change rate and geometric center coordinates. Combining the ratio of oxygen concentration to carbon monoxide content obtained by the flue gas composition analyzer, a combustion intensity index is constructed to quantify the activity degree of the combustion reaction.

[0060] Multiple Regression Model and Dynamic Time Warping Algorithm: The core technologies used in the analysis module for combustion stability evaluation. The multiple regression model establishes the mathematical relationship between combustion efficiency, oxygen content, primary air rate, and fuel volatile matter, and statistically analyzes the parameter fluctuation trend through a 30s sliding time window; the dynamic time warping algorithm aligns the time - series data of the historical template and the real - time combustion field, calculates the cumulative distance of the temperature gradient distribution difference, and identifies abnormal patterns with a deviation exceeding 15%.

[0061] Convolutional Neural Network and Digital Twin Verification: The technologies for strategy generation in the control module. The convolutional neural network extracts the spatial features in historical optimization operations (such as the correlation between the temperature distribution map and operation labels) to generate an initial strategy set; the digital twin model simulates the flow field and chemical reactions under different damper combinations through the CFD simulation engine, predicts the improvement of thermal efficiency and NOx emission indicators, and screens feasible strategies that meet the requirements of thermal efficiency ≥2% and NOx ≤200mg / Nm³.

[0062] Feedforward and feedback composite control and timing collaborative instructions: The core mechanism for actuators to act in the regulation module. Feedforward control anticipates the lag effect based on the step response curve of the actuator and sets the dynamic compensation coefficient; PID feedback control corrects the deviation between the actual opening and the target value in real time. The timing collaborative instruction queue inserts buffer intervals according to the mechanical response delay characteristics (such as the 8s full stroke time of the damper actuator) to avoid conflicts in the actions of multiple actuators.

[0063] Larson-Miller parameter and cumulative damage model: The technical basis for life prediction in the safety warning module. Through the equivalent operating time calculation of the pipe wall temperature and the material temperature resistance limit parameters, combined with the time integral of historical over-temperature events, when the cumulative damage coefficient ≥ 0.8, a warning is triggered, indicating that the pipe wall has entered the accelerated creep stage, providing a quantitative basis for hierarchical intervention.

[0064] Reinforcement learning and transfer learning: The core methods for expanding the policy library in the optimization module. Reinforcement learning extracts effective operation features in load mutation or fuel switching events based on the Q-learning algorithm; transfer learning maps historical policies to the new operating condition feature space, expands the adaptation boundary of the policy library, and improves the generalization ability of the system to complex operating conditions.

[0065] The above technical features form a closed-loop control link of "data acquisition - modeling - analysis - decision - execution - optimization - warning" through modular design, replacing traditional manual experience judgment, and achieving the dual goals of improving combustion efficiency and reducing operation safety risks. Key parameters (such as temperature field model, policy weight, warning signal) are transmitted between modules through standardized data interfaces to achieve the implementability and verifiability of the technical solution.

[0066] In the boiler control scenario of a thermal power plant, the present invention realizes the real-time monitoring and intelligent adjustment of the furnace combustion state through the following technical solutions. The specific implementation methods are as follows: Three temperature measurement layers are arranged in the furnace burner area, water wall and tail flue. In the burner area, a high-frequency infrared temperature measurement array is used to capture the transient temperature fluctuations of the flame, and the sampling frequency is ≥10Hz; anti-interference armored thermocouples are deployed in the water wall area, and the high-temperature radiation interference is eliminated through physical shielding design, and the temperature measurement accuracy is controlled within ±1.5°C; an infrared thermal imager group is configured in the flue area to achieve continuous monitoring of the large-scale temperature field. The original temperature data collected is processed by the Kalman filter algorithm to remove random noise, and the redundant data of adjacent sensor nodes is cross-validated to eliminate abnormal measurement values. The measurement deviation caused by the furnace pressure fluctuation is corrected through the temperature compensation algorithm to form a standardized temperature data set. Based on the Delaunay triangulation algorithm, the furnace space grid topology is constructed, and the radial basis function interpolation method is used to map the discrete temperature points to the three-dimensional grid nodes. The interpolation weight coefficient is dynamically adjusted in combination with the burner flame position signal to generate a three-dimensional dynamic combustion field model with temperature gradient distribution characteristics, and the spatial resolution reaches 0.1m³, providing high-precision space-time fusion data for combustion state analysis.

[0067] In the three-dimensional dynamic combustion field model, the continuous area with a temperature ≥1000°C is extracted as the combustion core area. The geometric center coordinates and volume change rate of the core area are determined through the Canny edge detection algorithm. The combustion intensity index is calculated by combining the ratio of oxygen concentration to carbon monoxide content obtained by the flue gas composition analyzer. A multiple regression model of combustion efficiency with oxygen content, primary air rate and fuel volatile matter is constructed. The standard deviation and covariance matrix of the combustion air volume, fuel calorific value and load change data during load fluctuations are statistically analyzed using a 30s sliding time window to generate a characteristic vector representing combustion stability. Based on the historical operation database, the temperature field distribution templates under full-load steady state and low-load peak-shaving conditions are extracted. The dynamic time warping algorithm is used to align the time series data of the current combustion field with the template, and the cumulative distance of the temperature gradient distribution difference is calculated. When the deviation exceeds 15%, a combustion anomaly warning is triggered, and a combustion stability evaluation index is generated and transmitted to the control module.

[0068] The control module uses a convolutional neural network to extract features from the historical optimization operation records, inputs the temperature distribution map into the network to extract spatial features, correlates the burner tilt angle and damper opening label data, and generates an initial strategy set including the angle adjustment amount, damper opening combination, and fuel distribution ratio. The current combustion field boundary conditions are imported into the CFD simulation engine through the digital twin model to simulate the flue gas flow and heat transfer processes under different damper combinations, calculate the flue gas recirculation rate and NOx generation amount, and screen out the feasible strategies that can increase the thermal efficiency by ≥2% and reduce the NOx emission to ≤200 mg / Nm³. Combining with the thermal stress threshold of the heating surface tube wall material, the Lagrange multiplier method is used to correct the feasibility of the candidate strategies, exclude the parameter combinations that cause local overheating risks, and output the safe optimization control strategy. The regulation module analyzes the optimization scheme into the angle increment with an accuracy of 0.1° for the burner servo motor, the opening target value with a step size of 5% for the secondary damper, and the fuel valve adjustment coefficient, generates a time-sequence collaborative instruction queue based on the actuator response delay test data, and inserts an 8s buffer interval into the long-stroke action instructions to avoid mechanical conflicts. The feedforward controller anticipates the lag effect based on the step response curve, combines with the PID feedback control to eliminate the steady-state error, and monitors the dynamic combustion field temperature change in real time and feeds it back to the optimization module. The optimization module calculates indicators such as the average decrease in the exhaust gas temperature and the standard deviation of the temperature difference of the heating surface within 30 minutes after the execution of the optimization control strategy, updates the strategy weight coefficient, and triggers the reinforcement learning mechanism to extract effective operation features when detecting a load mutation exceeding 20% or a fuel calorific value difference exceeding 10%, and expands the strategy library through transfer learning to adapt to the boundary. The safety warning module calculates the design life matching coefficient of the tube wall temperature and the Larson-Miller parameter in real time. When the cumulative damage coefficient ≥0.8, it triggers a secondary warning and forcibly adjusts the burner tilt angle in the corresponding area, reducing the combustion intensity to the safe threshold, forming a full-process closed-loop control from data perception, strategy optimization to safety protection.

[0069] The present invention constructs a full-process closed-loop control system to solve the problems of response lag and inconsistent operation standards caused by the dependence on manual experience in traditional boiler combustion control. The system collects the temperature data of the furnace burner area, water wall, and tail flue in real time based on a distributed temperature sensor network, combines the burner flame position signal and the damper opening state, and reconstructs the three-dimensional model of the dynamic combustion field through the three-dimensional space coordinate mapping relationship. This model integrates the Kalman filter denoising, redundant cross-validation, and temperature compensation algorithms to eliminate the interference of measurement noise and pressure fluctuations, generates a high-precision temperature gradient distribution cloud map, provides a spatio-temporal fusion data basis for combustion state analysis, replaces the traditional manual experience judgment, and realizes the holographic perception of the combustion state.

[0070] For the problem of missing combustion state space correlation model, the system adopts an intelligent decision-making mechanism that combines machine learning and digital twin. The analysis module correlates the dynamic relationship between combustion efficiency, oxygen content, and primary air ratio through a preset multiple regression model, extracts typical operating condition templates based on historical operation data, and identifies abnormal combustion field distribution patterns. The control module uses a convolutional neural network to extract historical optimization operation features, generates an initial control strategy set, verifies the combustion efficiency and emission indicators of candidate strategies through a CFD simulation engine, and performs safety constraint filtering in combination with the temperature resistance limit parameters of the heating surface material to screen out optimized control strategies that balance efficiency and safety. This process standardizes the control strategy generation rules and eliminates the subjective differences in manual operations.

[0071] The system improves the response speed and safety through a preset feedforward and feedback composite control mechanism and a closed-loop optimization link. The regulation module decomposes the optimization strategy into time-series collaborative instructions, dynamically compensates for action deviations in combination with the response delay characteristics of the actuator, and monitors the dynamic response of the temperature field in real time. The optimization module evaluates the strategy effect based on the flue gas temperature, entropy increase rate, and temperature difference standard deviation, updates the strategy weights, and expands the strategy library to adapt to the boundary. The safety warning module calculates the matching coefficient between the heating surface temperature and the allowable stress in real time, triggers a hierarchical intervention control strategy, and forcibly adjusts the combustion intensity in the over-temperature area. Each module forms a closed-loop link of perception - decision - execution - feedback through a data interface to achieve a dynamic balance between combustion efficiency and operation safety.

Claims

1. A furnace auxiliary adjustment control system, characterized in that: include: The data processing module is used to collect system data, including the temperature data of the furnace burner area, water-cooled wall and tail flue, the burner flame position signal and the air door opening state, and reconstruct the furnace continuous temperature field through the preset three-dimensional space coordinate mapping relationship of the system data to generate a dynamic combustion field three-dimensional model; An analysis module is used to extract the geometric features of the combustion core area based on the three-dimensional model of the dynamic combustion field, and to calculate the dynamic relationship between the combustion efficiency and the oxygen content and the primary air rate through a preset multivariate regression model in combination with the combustion air distribution, the fuel calorific value and the load change data, so as to identify the abnormal distribution mode of the combustion field and generate a combustion stability evaluation index; A control module, which is used to use a preset machine learning algorithm to perform pattern matching on the historical optimization operation records and the combustion stability evaluation index output by the analysis module, generate an initial control strategy, and verify the combustion efficiency and emission index of the candidate strategy through digital twin model simulation, and select the optimization control strategy that meets the safety constraints of the preset heating surface material temperature resistance limit parameters; A control module, used for decomposing the optimized control strategy selected by the control module into burner inclination adjustment instructions and air door opening adjustment sequences, correcting the actuator action deviation through a preset feedforward and feedback composite control mechanism, and monitoring the dynamic response of the temperature field based on the dynamic combustion field three-dimensional model updated in real time by the data processing module; An optimization module, used to calculate the combustion efficiency improvement rate, smoke temperature reduction range and heating surface temperature uniformity index after the optimization control strategy is executed according to the temperature field dynamic response data fed back by the control module; The safety warning module is used to superimpose the temperature resistance limit parameters of the heating surface material defined in the safety constraint filtering unit of the control module in the three-dimensional model of the dynamic combustion field, calculate the matching coefficient between the temperature distribution and the allowable stress of the heating surface material in real time, and trigger the hierarchical intervention control strategy and push the warning information to the human-machine interface when it is detected that the matching coefficient exceeds the preset threshold.

2. The furnace auxiliary adjustment control system according to claim 1, characterized in that: The data processing module comprises: The sensor network hierarchical deployment unit is configured with an infrared temperature measurement array and an anti-interference thermocouple to form a three-level temperature measurement layer in the burner area, water-cooled wall and flue to collect temperature data of the furnace burner area, water-cooled wall and tail flue; A data fusion unit, used to perform Kalman filtering denoising and redundant cross-validation on the original temperature data output by the sensor network hierarchical deployment unit, and eliminate the measurement deviation caused by furnace pressure fluctuation through a temperature compensation algorithm; A three-dimensional field reconstruction unit is used to establish a furnace space grid topology based on a Delaunay triangulation algorithm, and to map the discrete temperature points processed by the data fusion unit to three-dimensional grid nodes using a radial basis function interpolation method, and to dynamically correct the interpolation weight coefficient in combination with the burner flame position signal, and to output a three-dimensional cloud map of the temperature gradient distribution to generate the three-dimensional model of the dynamic combustion field; The output of the sensor network hierarchical deployment unit is connected to the data fusion unit, the output of the data fusion unit is connected to the three-dimensional field reconstruction unit, and the output of the three-dimensional field reconstruction unit is transmitted to the analysis module.

3. The furnace auxiliary adjustment control system according to claim 2, characterized in that: The analysis module comprises: A combustion core area identification unit is used to extract a continuous area with a temperature ≥ 1000°C from the three-dimensional model of the dynamic combustion field, determine the geometric center coordinates and volume change rate through an edge detection algorithm, and calculate the combustion intensity index in combination with data from an external flue gas composition analyzer; The multi-parameter coupling modeling unit constructs a multivariate regression model of combustion efficiency and oxygen content, primary air rate and fuel volatile matter, uses a sliding time window to statistically analyze the parameter change trends of the combustion air distribution volume, fuel calorific value and load change data when the load fluctuates, and generates a combustion stability characteristic vector; An abnormal pattern matching unit extracts a temperature field distribution template under typical working conditions based on historical operating data, uses a dynamic time warping algorithm to match the similarity between the current combustion field and the template in real time, and outputs an abnormal deviation index to generate the combustion stability evaluation index; Among them, the output of the combustion core area identification unit is connected to the multi-parameter coupling modeling unit, the output of the multi-parameter coupling modeling unit is connected to the abnormal pattern matching unit, and the output of the abnormal pattern matching unit is transmitted to the control module.

4. The furnace auxiliary adjustment control system according to claim 3, characterized in that: The control module comprises: The operation mode matching unit uses a convolutional neural network to extract features from the historical optimization operation records, takes the temperature distribution map generated by the dynamic combustion field three-dimensional model as input, associates the corresponding burner inclination angle and air door opening label data, and generates an initial recommendation strategy set; The digital twin verification unit imports the current combustion field boundary conditions into the CFD simulation engine, calculates the flue gas recirculation rate and NOx generation under different damper combinations, and selects feasible strategies that meet the thermal efficiency threshold and emission constraints; A safety constraint filtering unit, which takes the thermal stress threshold of the heated surface tube wall as a hard constraint condition, uses the Lagrange multiplier method to correct the candidate strategy output by the digital twin verification unit, and excludes the adjustment scheme that causes the risk of overtemperature to generate the optimization control strategy; Among them, the output of the operation mode matching unit is connected to the digital twin verification unit, the output of the digital twin verification unit is connected to the safety constraint filtering unit, and the output of the safety constraint filtering unit is transmitted to the control module.

5. The furnace auxiliary adjustment control system according to claim 4, characterized in that: The control module includes: An instruction decomposition unit, which parses the optimization control strategy generated by the safety constraint filtering unit into a burner servo motor angle increment, a secondary air door opening target value and a fuel distribution valve adjustment coefficient, and generates a timing coordination instruction queue according to the actuator response delay test data; A dynamic compensation unit, which sets a feedforward control compensation coefficient based on the actuator response delay test data and combines PID feedback control to eliminate a steady-state error between the actual opening and the target value; The output of the instruction decomposition unit is connected to the dynamic compensation unit, the output of the dynamic compensation unit is connected to the burner servo mechanism and the damper actuator, and the actuator action signal is fed back to the optimization module.

6. The furnace auxiliary adjustment control system according to claim 5, characterized in that: The optimization module includes: An effect evaluation unit is used to collect the average exhaust temperature, the entropy increase rate of the steam-water system and the standard deviation of the temperature difference of the heating surface from the dynamic response data of the temperature field fed back by the control module, calculate the contribution to the improvement of the combustion efficiency and update the weight coefficient of the control strategy generation rule of the control module; The knowledge base expansion unit, when detecting a sudden load change or a fuel switching event in the combustion air volume, fuel calorific value and load change data output by the analysis module, triggers a reinforcement learning mechanism to extract effective operating features in the adjustment sequence data of the control module, and expands the adaptation boundary of the strategy library of the control module through a transfer learning algorithm.

7. The furnace auxiliary adjustment control system according to claim 6, characterized in that: The safety warning module includes: A life prediction unit, for calculating the equivalent operating time of the pipe wall temperature and the design life in real time based on the Larson-Miller parameter in the temperature resistance limit parameter of the heating surface material defined in the safety constraint filtering unit of the control module, and triggering an early warning when the cumulative damage coefficient is ≥0.8; The hierarchical intervention unit is used to push adjustment suggestions to the human-machine interface for the first-level warning according to the warning signal of the life prediction unit, and trigger the combustion core area offset control strategy generated by the control module for the second-level warning to force the reduction of regional combustion intensity.

8. The furnace auxiliary adjustment control system according to claim 7, characterized in that: A communication redundancy check module is provided between the sensor network hierarchical deployment unit and the data fusion unit, which is used to detect the transmission packet loss rate of the temperature data output by the sensor network hierarchical deployment unit and trigger a data retransmission mechanism to maintain the input data integrity of the three-dimensional field reconstruction unit.

9. The furnace auxiliary adjustment control system according to claim 8, characterized in that: A strategy conflict detection module is provided between the digital twin verification unit and the safety constraint filtering unit, which is used to compare the combustion efficiency improvement value of the candidate strategy output by the digital twin verification unit with the safety threshold of the temperature resistance limit parameter of the heating surface material defined in the safety constraint filtering unit, identify the conflict point and give priority to the safety constraint filtering operation of the safety constraint filtering unit.

10. The furnace auxiliary adjustment control system according to claim 9, characterized in that: An action timing calibration module is provided between the dynamic compensation unit and the actuator, which is used to insert a time buffer interval in the timing coordination instruction queue generated by the instruction decomposition unit according to the mechanical response delay characteristics of the damper actuator in the actuator response delay test data stored in the instruction decomposition unit of the control module.

Images