Torch combustion efficiency control method and system capable of achieving real-time feedback adjustment

Abstract

The invention relates to the technical field of intelligent control analysis, in particular to a torch combustion efficiency control method and system capable of achieving real-time feedback adjustment. The method comprises the following steps that waste gas combustion potential state interval analysis is conducted through torch waste gas monitoring data and torch combustion external condition data, and waste gas combustion potential state interval data are generated; performing torch combustion efficiency dominant analysis on the torch combustion monitoring data to generate torch combustion efficiency dominant data; based on the waste gas combustion potential state interval data and the torch combustion efficiency dominant data, fuel related characteristic analysis of combustion failure is carried out, and combustion failure related characteristic data is generated; intelligent torch combustion efficiency control parameters which are fed back and adjusted in real time are designed based on combustion failure related characteristic data; and intelligent torch combustion control operation is executed according to the intelligent torch combustion efficiency control parameters which are fed back and adjusted in real time. According to the invention, accurate instant feedback intelligent control of torch combustion efficiency is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent control and analysis technology, and in particular to a control method and system for flare combustion efficiency that can be adjusted in real time. Background Technology

[0002] In industrial production processes such as petrochemicals, coal chemicals, and natural gas extraction and processing, large quantities of process waste gases containing combustible components are generated. To ensure production safety and prevent environmental pollution caused by the leakage of toxic and harmful gases, flare systems, as core environmental protection and safety assurance equipment, are widely used to treat these waste gases through combustion. Flare combustion efficiency directly determines the thoroughness of waste gas treatment, affecting not only whether pollutant emissions meet standards but also the energy recovery and utilization rate and the safety of system operation. However, existing flare combustion efficiency control methods are mostly based on simple closed-loop or fixed-parameter open-loop control using a single monitoring index, failing to consider multiple sources of parameters such as waste gas operating conditions, external environment, and combustion state. This makes it impossible to comprehensively capture the dynamic changes in the combustion system and accurately determine combustion efficiency. Furthermore, the lack of system analysis and interval division of waste gas combustion potential makes it impossible to predict combustion stability, resulting in insufficient targeting of control strategies and a tendency to cause combustion instability and flameout. Moreover, it is difficult to assess combustion efficiency and distinguish between fuel-limited and non-fuel-limited failure causes. Control intervention methods are prone to energy waste and pollution, lack real-time perception and rapid response mechanisms, and are slow to respond to sudden changes in operating conditions, making it difficult to maintain stable and efficient combustion. Summary of the Invention

[0003] Based on this, the present invention provides a method and system for controlling the flare combustion efficiency that can be adjusted in real time, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for controlling flare combustion efficiency with real-time feedback adjustment includes the following steps: Step S1: Use the multi-source monitoring sensors set up at the flare channel entrance to perform multi-source correlation data monitoring and processing of flare combustion to generate multi-source correlation monitoring data of flare combustion, wherein the multi-source correlation monitoring data of flare combustion includes flare exhaust gas monitoring data, flare combustion external condition data and flare combustion monitoring data; Step S2: Perform range analysis of exhaust gas combustion potential state using flare exhaust gas monitoring data and flare combustion external condition data to generate exhaust gas combustion potential state range data; Step S3: Perform explicit analysis on the flare combustion monitoring data to generate explicit flare combustion efficiency data; Step S4: Based on the exhaust gas combustion potential state range data and the flare combustion efficiency explicit data, perform fuel-related characteristic analysis of combustion failure and generate combustion failure-related characteristic data; Step S5: Design intelligent control parameters for flare combustion efficiency that are adjusted in real time based on combustion failure related characteristic data; execute intelligent flare combustion control operations according to the intelligent control parameters for flare combustion efficiency adjusted in real time.

[0005] Furthermore, step S2 includes the following steps: Step S21: Analyze the exhaust gas condition status type of the flare exhaust gas monitoring data and generate exhaust gas condition status type data; Step S22: Analyze the impact of exhaust gas operating conditions on flare combustion stability based on exhaust gas operating condition data, and generate exhaust gas operating condition combustion stability impact data. Step S23: Identify and process the combustible components of the flare gas monitoring data to generate combustible component data of the flare gas; Step S24: Design the dilution and suppression weights for exhaust gas flammability based on the exhaust gas flammability component data, generate exhaust gas flammability dilution and suppression weight parameters, and perform exhaust gas flammability characteristic analysis on the exhaust gas flammability component data through the exhaust gas flammability dilution and suppression weight parameters to generate exhaust gas flammability characteristic data. Step S25: Based on the external conditions data of flare combustion and the data on the impact of exhaust gas combustion stability on exhaust gas operating conditions and the combustibility characteristics data of exhaust gas, perform exhaust gas combustion potential characteristic analysis to generate exhaust gas combustion potential characteristic data. Step S26: Perform exhaust gas combustion potential state interval discrimination processing based on exhaust gas combustion potential characteristic data to generate exhaust gas combustion potential state interval data.

[0006] Furthermore, the exhaust gas combustion potential state range data mentioned in step S26 includes exhaust gas combustion potential full state range data and exhaust gas combustion potential limited state range data.

[0007] Furthermore, step S3 includes the following steps: Step S31: Analyze the flare combustion morphology characteristics based on the flare combustion monitoring data to generate flare combustion morphology characteristic data; Step S32: Analyze the torch combustion morphology pattern based on the torch combustion morphology characteristic data to generate torch combustion morphology pattern data; Step S33: Based on the exhaust gas combustion potential state range data and the flare combustion morphology pattern data, perform combustion morphology correlation deviation and combustion efficiency response relationship analysis under the combustion potential state range constraint, and generate combustion morphology correlation deviation-efficiency response relationship data. Step S34: Design a mapping index for flare combustion mode and combustion efficiency based on the combustion mode correlation deviation-efficiency response relationship data, and generate a flare mode-efficiency mapping index. Step S35: Perform explicit analysis of flare combustion efficiency on the flare combustion monitoring data based on the flare shape-efficiency mapping index to generate explicit flare combustion efficiency data.

[0008] Furthermore, step S33 includes the following steps: Step S331: Analyze the correlation deviation characteristics between the flare combustion mode and the exhaust gas combustion potential state interval using the exhaust gas combustion potential state interval data, and generate combustion mode correlation deviation characteristic data of the combustion potential state. Step S332: Perform combustion efficiency sensitivity analysis on the combustion mode correlation deviation feature data of combustion potential state to generate combustion mode correlation deviation feature efficiency sensitivity data. Step S333: Based on the efficiency sensitivity data of combustion mode correlation deviation characteristics, analyze the relationship between combustion mode correlation deviation and combustion efficiency response under the constraint of combustion potential state interval, and generate combustion mode correlation deviation-efficiency response relationship data.

[0009] Furthermore, step S4 includes the following steps: Step S41: Perform trend feature analysis on the explicit data of flare combustion efficiency to generate explicit trend feature data of flare combustion efficiency; Step S42: Based on the characteristics of the apparent trend of flare combustion efficiency, conduct combustion efficiency trend anomaly assessment processing to generate combustion efficiency trend anomaly assessment data; Step S43: Analyze the characteristics of combustion failure precursors based on the abnormal combustion efficiency trend assessment data, and generate combustion failure precursor characteristic data; Step S44: Analyze combustion failure-related features of combustion failure precursor features using exhaust gas combustion potential state range data to generate combustion failure-related feature data.

[0010] Furthermore, the combustion failure-related feature data in step S44 includes combustion failure-non-fuel-limited feature data or combustion failure-fuel-limited feature data.

[0011] Furthermore, step S44 includes the following steps: When the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential full state interval data, the non-fuel-limited feature analysis of combustion failure is performed on the combustion failure precursor feature data to generate combustion failure-non-fuel-limited feature data; or, when the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential limited state interval data, the fuel-limited feature analysis of combustion failure is performed on the combustion failure precursor feature data to generate combustion failure-fuel-limited feature data.

[0012] Furthermore, step S5 includes the following steps: Step S51: When the combustion failure related characteristic data is combustion failure-non-fuel-limited characteristic data, analyze the combustion organization regulation relationship of the flare combustion monitoring data to generate combustion organization regulation relationship data; based on the combustion organization regulation relationship data and the combustion failure-non-fuel-limited characteristic data, analyze the minimum intervention control parameters for combustion organization regulation to generate the minimum intervention control parameters for combustion organization regulation. Step S52: When the combustion failure related characteristic data is combustion failure-fuel limitation characteristic data, analyze the minimum intervention control parameters for fuel supply regulation based on the exhaust gas combustion potential characteristic data and the combustion failure-fuel limitation characteristic data, and generate the minimum intervention control parameters for fuel supply regulation. Step S53: Design intelligent control parameters for flare combustion efficiency that are adjusted in real time by adjusting the minimum intervention control parameters through combustion organization and / or fuel replenishment; Step S54: Execute intelligent flare combustion control operation based on the intelligent control parameters of flare combustion efficiency adjusted according to real-time feedback.

[0013] This specification provides a control system for flare combustion efficiency that can be adjusted in real time, used to execute the control method for flare combustion efficiency that can be adjusted in real time as described above. The control system for flare combustion efficiency that can be adjusted in real time includes: The flare combustion multi-source correlation monitoring module is used to monitor and process flare combustion multi-source correlation data using multi-source monitoring sensors installed at the flare channel entrance, and generate flare combustion multi-source correlation monitoring data, wherein the flare combustion multi-source correlation monitoring data includes flare exhaust gas monitoring data, flare combustion external condition data, and flare combustion monitoring data. The exhaust gas combustion potential state interval analysis module is used to perform interval analysis of exhaust gas combustion potential state through flare exhaust gas monitoring data and flare combustion external condition data, and generate exhaust gas combustion potential state interval data. The combustion efficiency explicit analysis module is used to perform explicit analysis of flare combustion monitoring data and generate explicit flare combustion efficiency data. The combustion failure related feature analysis module is used to perform fuel-related feature analysis of combustion failure based on exhaust gas combustion potential state range data and flare combustion efficiency explicit data, and generate combustion failure related feature data. The flare combustion intelligent control module is used to design intelligent control parameters for flare combustion efficiency that are adjusted in real time based on combustion failure-related characteristic data; and to execute intelligent control operations for flare combustion based on the intelligent control parameters for flare combustion efficiency adjusted in real time.

[0014] The beneficial effects of this application are as follows: By setting up multi-source monitoring sensors at the flare channel entrance, this invention achieves comprehensive monitoring and processing of multi-source correlated data on flare combustion. The acquired flare exhaust gas monitoring data (including key operating parameters such as composition, concentration, flow rate, and flow), flare combustion external condition data (including environmental parameters such as ambient wind speed and air pressure), and flare combustion monitoring data (including state parameters such as flame morphology and combustion temperature) overcome the limitations of single monitoring indicators. The collaborative acquisition of multi-dimensional data provides comprehensive and accurate data support for subsequent combustion state analysis, efficiency assessment, and control parameter design, effectively solving the problems of one-sided combustion state judgment and inaccurate efficiency reflection caused by insufficient monitoring dimensions in existing technologies, laying the foundation for efficient and accurate control. Based on flare exhaust gas monitoring data and combustion external condition data, the invention conducts interval analysis of exhaust gas combustion potential state, and the analysis of operating condition type and combustion stability impact can clarify the basic impact of different exhaust gas operating conditions on the combustion process; the identification of combustible components and the design of dilution inhibition weights accurately quantify the combustibility of exhaust gas and the inhibitory effect of external factors; combined with the potential characteristic analysis of external environmental parameters, the accuracy of potential assessment is further improved. In particular, clearly dividing the combustion potential into two core regions, sufficient and constrained, allows for a clear definition of the core influencing factors on combustion stability under different operating conditions. When the potential is sufficient, combustion stability is mainly affected by the combustion organization method, while when the potential is constrained, it is limited by fuel supply. This provides a clear state basis for subsequent targeted control. It addresses the problems of unpredictable combustion stability and lack of targeted control strategies, and can proactively avoid risks such as combustion instability, flameout, or over-combustion, significantly improving the stability and safety of flare combustion. Morphological feature analysis and pattern recognition are performed on flare combustion monitoring data. Correlation deviation and efficiency response relationship analysis are conducted in conjunction with exhaust gas combustion potential state range data. The correlation deviation law and efficiency response characteristics between combustion morphology and efficiency under different potential constraints are analyzed. Based on this, a flare morphology-efficiency mapping index is designed, achieving explicit and quantitative assessment of combustion efficiency. Overcoming the limitations of relying on indirect indicators such as temperature and brightness to infer efficiency, this method establishes a clear quantitative mapping relationship between morphology and efficiency, allowing combustion efficiency to be presented directly and accurately, avoiding errors caused by indirect inference. Simultaneously, sensitivity analysis based on potential range constraints ensures that efficiency assessment results are fully adaptable to different combustion potential conditions, guaranteeing the scientific validity and reliability of the assessment results under various operating conditions. Precise explicit efficiency data provides a clear basis for understanding the combustion state. Relying on exhaust gas combustion potential range data and explicit flare combustion efficiency data, through detailed steps such as efficiency trend analysis, anomaly assessment, and failure precursor identification, accurate prediction and cause location of combustion failure are achieved. Efficiency trend feature analysis can capture subtle changes in efficiency and identify potential failure risks in advance; trend anomaly assessment uses quantitative threshold judgment to accurately define the degree of anomaly; and failure precursor feature analysis further identifies typical characteristic signals before failure.Crucially, this step, through correlation analysis of combustion potential range data and failure precursor characteristics, can clearly distinguish between two core characteristics: combustion failure - non-fuel limitation (corresponding to a state of sufficient potential, failure stemming from improper combustion organization) and combustion failure - fuel limitation (corresponding to a state of limited potential, failure stemming from insufficient fuel supply). This fundamentally solves the shortcomings of traditional technologies in diagnosing combustion failures, such as ambiguity and the inability to define the root cause of failure. It provides clear targeting for control strategies, making control interventions more targeted. Based on the identified failure causes, differentiated minimum intervention control parameters are designed to achieve personalized, precise, and real-time feedback control of flare combustion efficiency. For non-fuel-limited failures, by analyzing the relationship between parameters such as flame morphology and combustion temperature in combustion monitoring data and the adjustment of combustion organization, minimum intervention control parameters for combustion organization adjustment (such as flame anchoring parameters and other flare combustion structure-related parameters) are designed. Efficient combustion can be restored by optimizing the combustion structure without additional adjustments to fuel or air supply, minimizing unnecessary parameter adjustments. For fuel-limited failures, minimum intervention control parameters for fuel replenishment adjustment (such as auxiliary fuel replenishment amount, replenishment rate, and air content adjustment parameters) are designed. By precisely replenishing auxiliary fuel or adjusting air content, combustion potential is enhanced, avoiding energy waste and increased pollutant emissions caused by over-replenishment or air-fuel imbalance. Based on real-time updates of multi-source correlated monitoring data of flare combustion, control parameters are rapidly adjusted, and intelligent flare combustion control operations are immediately executed, demonstrating strong real-time response capabilities and effectively solving the problem of control lag. This ensures that the flare combustion system can quickly adapt and maintain stable and efficient operation under complex scenarios such as sudden changes in operating conditions and fluctuations in the external environment, further improving the system's safety, environmental friendliness, and energy utilization efficiency, fully meeting the stringent requirements of industrial production for flare systems.

[0015] Therefore, the real-time feedback-adjustable flare combustion efficiency control method and system of the present invention, by integrating multi-source correlated monitoring data of exhaust gas conditions, external environment, and combustion state, overcomes the shortcomings of existing technologies that have single monitoring dimensions and lack multi-source data correlation analysis. It can comprehensively capture the dynamic changes of the combustion system, achieving accurate judgment of combustion state and precise reflection of combustion efficiency. Through systematic analysis and interval division of exhaust gas combustion potential, it achieves early prediction and zoned control of combustion stability, solving the problem of insufficient targeting of existing control strategies and effectively avoiding combustion instability, flameout, or over-combustion. Furthermore, by establishing a clear mapping relationship between combustion mode and combustion efficiency... This system enables explicit and quantitative assessment of combustion efficiency, providing precise basis for adjusting control parameters and ensuring optimal combustion efficiency. By accurately distinguishing between fuel-limited and non-fuel-limited causes of combustion failure, it designs targeted minimum intervention control parameters, overcoming the shortcomings of existing technologies in combustion efficiency failure assessment and intervention methods. This reduces energy waste and pollutant emissions, and improves the stability of the combustion system. Based on real-time analysis of multi-source data, it designs intelligent control parameters and constructs a rapid response mechanism, solving the problem of control lag in existing technologies. It can respond promptly to sudden changes in operating conditions, ensuring that the flare combustion always maintains a stable and efficient state, thereby meeting the stringent requirements of industrial production for environmental compliance and operational safety. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the steps of an intelligent medication management method according to the present invention; Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0019] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a lightweight method for reconstructing 3D models of power facilities using curved surfaces. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of a method for lightweighting a 3D power facility model based on surface reconstruction according to the present invention. The method includes the following steps: Based on this, the present invention provides a method and system for controlling the flare combustion efficiency that can be adjusted in real time, in order to solve at least one of the above-mentioned technical problems.

[0020] To achieve the above objectives, a method for controlling flare combustion efficiency with real-time feedback adjustment includes the following steps: Step S1: Use the multi-source monitoring sensors set up at the flare channel entrance to perform multi-source correlation data monitoring and processing of flare combustion to generate multi-source correlation monitoring data of flare combustion, wherein the multi-source correlation monitoring data of flare combustion includes flare exhaust gas monitoring data, flare combustion external condition data and flare combustion monitoring data; In this embodiment of the invention, six sets of multi-source monitoring sensor components are evenly arranged circumferentially around the flare channel opening. The deployment ring surface of the components is parallel to the flare channel outlet plane, and the deployment height is flush with the center of the flare channel outlet. The spacing between adjacent components is strictly controlled to one-third of the flare channel diameter to ensure that there are no blind spots in the sensor monitoring range and that the signals do not interfere with each other. Each set of components integrates seven dedicated sensors, and the parameters and installation methods of each sensor are precisely matched: the exhaust gas component sensor adopts a near-infrared spectral sensor, with the probe facing the mainstream direction of the flare exhaust gas emission during installation, and the distance from the channel outlet is set to 50cm. It is configured with a spectral detection wavelength range of 1200-2500nm, a spectral resolution of 2nm, and an integration time of 50ms, and is used to detect the types of combustible components such as methane, ethane, and propane, as well as inert components such as nitrogen and carbon dioxide in the exhaust gas; the exhaust gas concentration sensor adopts a laser absorption sensor, which is coaxially installed with the exhaust gas component sensor. The emission direction passes through the main exhaust gas flow zone, equipped with a detection accuracy of ±0.1%VOL and a response time of ≤100ms, used to detect the volume fraction of each combustible component; the exhaust gas flow sensor is a differential pressure Pitoba tube sensor, embedded in the inner wall of the flare channel, installed 1m upstream of the channel outlet, configured with a measurement range of 0-100m³ / s and an accuracy of ±1.0%FS, obtaining the real-time exhaust gas velocity through the differential pressure measurement principle, and then converting it with the channel cross-sectional area to obtain the flow data; the environmental wind speed sensor is a three-dimensional ultrasonic wind speed sensor, installed 1m outside the flare channel outlet. A support frame at 5m, 2m higher than the passage exit, is configured with a measurement range of 0-30m / s, accuracy ±0.01m / s, and sampling frequency of 10Hz, for collecting horizontal and vertical wind speed data. A high-precision environmental pressure sensor, installed in a sealed protective box next to the wind speed sensor to avoid direct airflow impact, has a measurement range of 80-120kPa and accuracy ±0.1kPa, for collecting atmospheric pressure data. A flame pattern vision sensor, using a high-speed industrial camera with a high-temperature resistant quartz glass protective cover for the lens, is installed in the passageway. On the outer support of the torch, the lens axis forms a 30° angle with the central axis of the flame. It is configured with an image resolution of 1920×1080 pixels, a frame rate of 100fps, and an exposure time of 10μs, capturing flame contour images using high-speed imaging technology. The combustion temperature sensor is a thermocouple array sensor, containing 16 thermocouple probes arranged in a 4×4 matrix. The probes are embedded in high-temperature resistant ceramic protective sleeves and inserted into the flame monitoring area at a depth controlled to 20cm. It is configured with a measurement range of 0-1800℃ and an accuracy of ±1℃, used to acquire temperature data from different areas of the flame. All sensors start monitoring synchronously, with a uniform sampling frequency of 10Hz. The acquired raw data is transmitted to the data processing unit via shielded cables.The data processing unit first performs timestamp alignment on all raw data, eliminating response delay deviations between different sensors based on sensor-triggered synchronization signals, with time synchronization accuracy controlled within 1ms. Then, a Kalman filter algorithm is used to filter the data from each sensor, setting the filter gain coefficient to 0.05 to remove abnormal data caused by environmental vibration and electromagnetic interference. Subsequently, the filtered data is categorized and integrated, and packaged according to a preset data format: flare exhaust gas monitoring data includes exhaust gas component types, volume fractions of each component, exhaust gas flow rate, and time-series acquisition timestamps; flare combustion external condition data includes ambient horizontal wind speed, ambient vertical wind speed, ambient air pressure, and corresponding acquisition timestamps; flare combustion monitoring data includes flame contour image frames, temperature values ​​of different flame regions, and acquisition timestamps. Finally, this integration forms a structured and time-series unified multi-source correlation monitoring data for flare combustion, providing accurate and complete basic data support for subsequent analysis and processing steps.

[0021] Step S2: Perform range analysis of exhaust gas combustion potential state using flare exhaust gas monitoring data and flare combustion external condition data to generate exhaust gas combustion potential state range data; In this embodiment of the invention, flare exhaust gas monitoring data and flare combustion external condition data are extracted. First, the fluctuation amplitude of exhaust gas flow rate and concentration of each combustible component in the flare exhaust gas monitoring data is calculated for five consecutive sampling periods. Based on the fluctuation amplitude, the exhaust gas operating condition type is divided into stable operating condition, small fluctuation operating condition, and large fluctuation operating condition, generating exhaust gas operating condition type data. Based on the fluctuation characteristics of exhaust gas flow rate and concentration corresponding to different operating condition types, the influence of the stability of exhaust gas supply under each operating condition on the continuity and stability of the flare combustion flame is analyzed, generating exhaust gas operating condition combustion stability impact data. Feature matching technology is used to identify the exhaust gas component detection results, screening out all combustible components and determining their core attributes, generating exhaust gas combustible component data. Based on the calorific value and reactivity of each combustible component, different combustible components are assigned corresponding weight coefficients. Simultaneously, the dilution inhibition coefficient is calculated in conjunction with the concentration of inert components. The weight coefficients are then combined with the dilution inhibition coefficient. The dilution and suppression weighting parameter for exhaust gas combustibility is obtained by fusing control coefficients. This parameter is used to weight the concentration data of each combustible component to obtain exhaust gas combustibility characteristic data that characterizes the overall combustibility of the exhaust gas. Ambient wind speed and ambient air pressure data are extracted to analyze the degree of interference of wind speed on flame diffusion and the influence of air pressure on combustion reaction rate. The results of this influence are correlated and fused with the data on the influence of exhaust gas operating condition combustion stability and exhaust gas combustibility characteristic data to explore the coupling relationship among the three and generate exhaust gas combustion potential characteristic data. An exhaust gas combustion potential threshold range is set, and the exhaust gas combustion potential characteristic data is compared with the threshold range. When the data is in the range from the upper limit to the maximum value of the threshold, it is determined to be a state of sufficient exhaust gas combustion potential, and exhaust gas combustion potential state interval data is generated. When the data is in the range from the minimum value to the lower limit of the threshold, it is determined to be a state of limited exhaust gas combustion potential, and exhaust gas combustion potential state interval data is generated, thus completing the interval analysis of exhaust gas combustion potential state.

[0022] Step S3: Perform explicit analysis on the flare combustion monitoring data to generate explicit flare combustion efficiency data; In this embodiment of the invention, flame contour images and flame temperature data from flare combustion monitoring data are called. Edge extraction processing is performed on the flame contour images to extract flame shape parameters, height parameters, diffusion angle parameters, and jitter amplitude parameters. Flame temperature data is divided into regions to obtain temperature distribution data for the flame core area, transition area, and outer periphery, which are then integrated to form flare combustion morphology feature data. Based on the combination rules of flame shape parameters, height parameters, and other features, three typical modes are summarized: stable combustion mode, slightly fluctuating combustion mode, and violently fluctuating combustion mode. The extracted flare combustion morphology feature data is compared with the typical modes to determine the current flame morphology mode, generating flare combustion morphology mode data. Exhaust gas combustion potential state interval data is introduced. For both fully charged and constrained exhaust gas combustion potential states, the deviation between the flare combustion morphology mode and the standard combustion morphology mode under different states is analyzed. Feature parameters corresponding to the deviations are extracted to generate combustion potential states. The combustion morphology-related deviation characteristic data is used to determine the influence of each deviation parameter on the combustion efficiency. Based on this data, a correlation is established between the deviation parameter and the change in combustion efficiency, clarifying the response law of combustion efficiency under different degrees of deviation, and generating combustion morphology-related deviation-efficiency response relationship data. According to this response relationship data, deviation characteristic parameters and temperature distribution parameters that significantly affect combustion efficiency are selected to construct a flare morphology-efficiency mapping index including morphology deviation coefficient and temperature uniformity coefficient. This mapping index is then used to comprehensively calculate the flame morphology characteristic data and temperature distribution data in the flare combustion monitoring data to obtain quantitative data that can directly characterize combustion efficiency, completing the explicit analysis of flare combustion efficiency and generating explicit flare combustion efficiency data.

[0023] Step S4: Based on the exhaust gas combustion potential state range data and the flare combustion efficiency explicit data, perform fuel-related characteristic analysis of combustion failure and generate combustion failure-related characteristic data; In this embodiment of the invention, explicit data on flare combustion efficiency are collected for 20 consecutive sampling periods. The difference between data from adjacent periods is calculated, and the trend direction and rate of change of efficiency are extracted to generate explicit trend feature data of flare combustion efficiency. A threshold for the rate of change of efficiency and a threshold for the duration of the trend are set. The extracted rate of change of efficiency and the duration of the trend are compared with the corresponding thresholds. When the rate of change of efficiency exceeds the threshold and the duration reaches the set threshold, it is determined that the combustion efficiency trend is abnormal, and abnormal combustion efficiency trend assessment data is generated. Based on the assessment results of the abnormal combustion efficiency trend, the flare combustion morphology feature data and temperature distribution data before the anomaly occurred are reviewed to extract characteristic signals such as increased flame vibration amplitude and decreased core area temperature that appeared before the anomaly occurred. Generate combustion failure precursor characteristic data; correlate the exhaust gas combustion potential state range data generated in step S2. When the combustion failure precursor characteristic data corresponds to the exhaust gas combustion potential full state range data, considering the premise of sufficient fuel and air supply in this state, determine that the cause of failure is unrelated to fuel supply, focus on the characteristic parameters related to combustion organization mode for analysis, and generate combustion failure-non-fuel-limited characteristic data; when the combustion failure precursor characteristic data corresponds to the exhaust gas combustion potential limited state range data, considering the characteristic of insufficient combustibility of exhaust gas in this state, determine that the cause of failure is related to fuel supply, focus on the characteristic parameters related to fuel composition, concentration, etc. for analysis, and generate combustion failure-fuel-limited characteristic data, completing the fuel-related characteristic analysis of combustion failure.

[0024] Step S5: Design intelligent control parameters for flare combustion efficiency that are adjusted in real time based on combustion failure related characteristic data; execute intelligent flare combustion control operations according to the intelligent control parameters for flare combustion efficiency adjusted in real time.

[0025] In this embodiment of the invention, intelligent control parameters for flare combustion efficiency are designed and executed based on real-time feedback adjustments using combustion failure-related characteristic data. When the combustion failure-related characteristic data is combustion failure-non-fuel-limited characteristic data, correlation analysis is performed on flame morphology data and combustion temperature distribution data in the flare combustion monitoring data to clarify the intrinsic relationship between flame anchoring position and flame stability, and the correspondence between flame diffusion angle and temperature distribution uniformity. By analyzing these correlations, the adjustment direction of the combustion organization mode is determined. Based on the adjustment direction, the adjustment range and logic of the flame anchoring parameters are determined, i.e., the flame anchoring position is adjusted in the direction of improving temperature distribution uniformity, and the adjustment amplitude is determined according to the morphology deviation coefficient, generating the minimum intervention control parameters for combustion organization adjustment. When the combustion failure-related characteristic data is combustion failure-fuel-limited characteristic data, the difference between the concentration and flow rate of combustible components in the exhaust gas and the combustion demand is analyzed in conjunction with the exhaust gas combustion potential characteristic data to determine the direction of auxiliary fuel supply as supplementing combustible components with high calorific value. The relationship between air content and combustion reaction is analyzed in real time. The air content adjustment logic is determined to be based on the auxiliary fuel supply amount adjusted according to the stoichiometric ratio. Minimum intervention control parameters for fuel supply adjustment are generated based on the supply direction and adjustment logic. The minimum intervention control parameters for combustion organization adjustment and fuel supply adjustment are integrated to establish a parameter priority ranking mechanism. By analyzing the impact of the two types of parameter adjustments on combustion stability, the priority of flame anchoring parameter adjustment is set higher than that of auxiliary fuel supply parameter adjustment. When both types of parameters have adjustment needs at the same time, flame anchoring parameter adjustment is executed first, forming intelligent control parameters for flare combustion efficiency with real-time feedback adjustment. Based on these intelligent control parameters, the flame anchoring structure of the flare is adjusted through mechanical actuators to change the flame attachment position, or the supply amount and supply rate are adjusted through the valve adjustment mechanism of the auxiliary fuel supply device, and the output air volume is adjusted through the fan speed adjustment mechanism of the air supply device, so as to realize the real-time adjustment of the flare combustion state and ensure that the flare combustion efficiency is maintained in the optimal range.

[0026] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S2 is provided in this embodiment. Step S2 includes: Step S21: Analyze the exhaust gas condition status type of the flare exhaust gas monitoring data and generate exhaust gas condition status type data; In this embodiment of the invention, exhaust gas monitoring data from a flare is analyzed to determine the operational status of the exhaust gas. Focusing on the operational status characteristics of the exhaust gas transport process, four core parameters are selected: exhaust gas flow rate, methane concentration, ethane concentration, and propane concentration. Simultaneously, exhaust gas flow state characterization parameters (velocity gradient, Reynolds number) and transport state characterization parameters (flow rate fluctuation persistence, pressure stability) are incorporated as auxiliary analysis features to comprehensively cover the characteristic analysis needs of flow state, transport state, and exhaust gas concentration. The sampling period is set to 10Hz, and the analysis duration covers five consecutive sampling periods, constructing an operational status recognition model based on a long short-term memory network. The model's input layer dimension is set to 8 (corresponding to the four core concentration and flow rate parameters, two flow state parameters, and two transport state parameters), the hidden layer has two layers with 64 neurons per layer, and the output layer dimension is 3 (corresponding to three types: stable operation, small fluctuation operation, and large fluctuation operation). The activation function is ReLU, the optimizer is Adam, the learning rate is set to 0.001, the number of iterations is set to 500, and an early stopping mechanism is implemented to prevent overfitting, with an early stopping patience of 20. The velocity gradient is calculated using the velocity difference between adjacent sampling periods; the Reynolds number is calculated based on the exhaust gas density, viscosity, and channel diameter; the persistence of flow fluctuations is determined by statistically analyzing the number of consecutive fluctuation periods; and pressure stability is quantified using the standard deviation of exhaust gas pressure data. The continuously collected core parameters and time-series auxiliary feature data are divided into training and testing sets in an 8:2 ratio. The training set data is labeled with corresponding operating condition labels (stable operating condition corresponds to uniform flow, smooth transport, and constant concentration; small fluctuation operating condition corresponds to slight flow disturbance, short-term transport fluctuation, and small concentration change; large fluctuation operating condition corresponds to violently turbulent flow, continuous transport fluctuation, and significant concentration fluctuation) and then input into the model for training. Training stops when the accuracy of the testing set reaches 98%. The time-series data of the core parameters and auxiliary features to be analyzed are input into the trained model. The model, through learning the time-series fluctuation characteristics and the coupling correlation between flow, transport, and concentration, outputs the operating condition type determination result. The determination result and the quantitative analysis data of each state feature are integrated to generate exhaust gas operating condition state type data.

[0027] Step S22: Analyze the impact of exhaust gas operating conditions on flare combustion stability based on exhaust gas operating condition data, and generate exhaust gas operating condition combustion stability impact data. In this embodiment of the invention, the impact of exhaust gas operating conditions on flare combustion stability is analyzed using exhaust gas operating condition type data. A flame continuity threshold of 95% and a flame height fluctuation allowable range of ±10% are configured. A condition-stability correlation prediction model based on random forest is constructed. The model input features include operating condition type encoding, exhaust gas flow fluctuation coefficient, and standard deviation of combustible component concentration. The output is a quantitative value of the combustion stability impact (range 0-1, with larger values ​​indicating a more positive impact). 1000 sets of historical data under different operating conditions are selected as training samples, labeled with the flame continuity percentage, flame height fluctuation value, and stability impact level under each condition. The training samples are input into the model for training. The number of decision trees is set to 100, the maximum depth to 10 layers, and the minimum number of sample splits to 2. Model parameters are optimized using 5-fold cross-validation. The exhaust gas operating condition type data and the corresponding exhaust gas parameters under the operating condition are input into the trained model. The model outputs a stability influence quantification value. At the same time, the results are calibrated by combining the measured data of flame continuity ratio and flame height fluctuation value. The quantification value, influence characteristics and calibration results are integrated to generate exhaust gas operating condition combustion stability influence data.

[0028] Step S23: Identify and process the combustible components of the flare gas monitoring data to generate combustible component data of the flare gas; In this embodiment of the invention, combustible components of flare gas are identified from flare gas monitoring data. The spectral detection wavelength range is configured as 1200-2500 nm, the spectral resolution is set to 2 nm, and the integration time is 50 ms. Near-infrared absorption spectra of the flare gas are acquired using a spectral acquisition device. A component identification model based on a convolutional neural network is constructed. The model input is one-dimensional spectral data of the spectrum (the dimension is consistent with the number of spectral detection points). Three convolutional layers are set, with 32, 64, and 128 convolutional kernels in each layer, and a kernel size of 3×1. Max pooling is used in the pooling layers (pooling kernel size 2×1). Two fully connected layers are set (with 256 and 128 neurons respectively). The output layer is the category probability of common combustible components and inert components. A standard spectral database containing combustible components such as methane, ethane, propane, butane, and ethylene, as well as inert components such as nitrogen, carbon dioxide, and oxygen, was established. 5000 sets of spectral data were selected from this database as training samples, labeled with their corresponding component categories, and input into the model for training. The learning rate was set to 0.002, and the number of iterations was 300. The cross-entropy loss function was used to optimize the model. The collected exhaust gas spectral data was input into the trained model, and the model output the probability of the presence of each component. A probability threshold of 0.95 was set; when the probability of a component was greater than or equal to the threshold, that component was determined to be present in the exhaust gas. The model's built-in characteristic peak intensity calibration module eliminated matrix interference, determined the presence status of each combustible component, and integrated the identified combustible component types, characteristic peak positions, and peak intensity information to generate exhaust gas combustible component data.

[0029] Step S24: Design the dilution and suppression weights for exhaust gas flammability based on the exhaust gas flammability component data, generate exhaust gas flammability dilution and suppression weight parameters, and perform exhaust gas flammability characteristic analysis on the exhaust gas flammability component data through the exhaust gas flammability dilution and suppression weight parameters to generate exhaust gas flammability characteristic data. In this embodiment of the invention, dilution suppression weights for exhaust gas flammability are designed based on exhaust gas flammability component data. A weight prediction model based on gradient boosting trees is constructed. The model input features include the concentration of each flammable component, its calorific value, and the total concentration of inert components. The output is the exhaust gas flammability dilution suppression weight parameter (value range 0-1, the larger the value, the weaker the dilution suppression effect). The calorific value of each flammable component is measured using a calorific value testing device. 2000 sets of weight calibration data under different component concentration ratios are collected as training samples and divided into training and testing sets at an 8:2 ratio. The number of decision trees is set to 80, the learning rate to 0.01, and the maximum depth to 8 layers. The training set is input into the model for training, and the model performance is verified using the testing set. Training is completed when the prediction error is less than 2%. The concentration of each flammable component, its calorific value, and the total concentration of inert components in the exhaust gas flammability component data are input into the trained model. The model outputs the exhaust gas flammability dilution suppression weight parameter based on the learned component characteristics and dilution suppression rules. Based on the weighting parameter, a weighted fusion model is constructed. The concentration of each combustible component and its corresponding calorific value are weighted and summed, and then multiplied by the weighting parameter of the combustibility dilution and inhibition of exhaust gas to obtain a quantitative value that characterizes the overall combustibility of exhaust gas. This quantitative value is the combustibility characteristic data of exhaust gas.

[0030] Step S25: Based on the external conditions data of flare combustion and the data on the impact of exhaust gas combustion stability on exhaust gas operating conditions and the combustibility characteristics data of exhaust gas, perform exhaust gas combustion potential characteristic analysis to generate exhaust gas combustion potential characteristic data. In this embodiment of the invention, the combustion potential characteristics of exhaust gas are analyzed based on the impact of external conditions data of flare combustion on the combustion stability of exhaust gas and the combustibility characteristics data of exhaust gas, generating exhaust gas combustion potential characteristic data. The external conditions data of flare combustion are refined into four core parameters: ambient wind speed (horizontal and vertical wind speeds), ambient air pressure, ambient temperature, and ambient humidity. Each parameter is accompanied by a time-series acquisition marker to ensure temporal consistency with the exhaust gas operating conditions data and combustibility characteristic data. A multi-factor coupled prediction model based on a deep neural network was constructed. The input layer dimension was set to 8 (corresponding to the quantified values ​​of the influence of horizontal wind speed, vertical wind speed, ambient air pressure, ambient temperature, ambient humidity, the quantified values ​​of the combustion stability of exhaust gas under operating conditions, exhaust gas combustibility characteristics, and exhaust gas flow rate level). The feature encoding layer mapped each input parameter to a 64-dimensional feature space through a fully connected layer. The cross-fusion layer used an attention mechanism to calculate the coupling weights between each feature. The deep mapping layer had three hidden layers with 128, 64, and 32 neurons in each layer, respectively. The gradient vanishing problem was solved through residual connections. The output layer dimension was 1, and the sigmoid activation function was used to ensure that the output combustion potential feature data values ​​were strictly controlled within the range of 0-1, with larger values ​​indicating stronger combustion potential. The model optimizer was RMSprop, with a learning rate of 0.001 and 400 iterations. L2 regularization (regularization coefficient 0.001) was added to suppress overfitting, and a gradient clipping mechanism was configured to set the upper limit of the gradient norm to 1.0 to avoid gradient explosion during training. During the data preparation phase, 3000 sets of combustion potential calibration data were collected under different external conditions, operating stability, and combustibility characteristics. Each set of data included complete external condition parameters, quantitative values ​​of the impact of exhaust gas operating conditions on combustion stability, exhaust gas combustibility characteristic data, and corresponding measured values ​​of combustion potential (obtained through actual measurement using a combustion heat release rate detection device). The calibration data were divided into training and validation sets in a 7:3 ratio, with the training set data first undergoing normalization. During training, a phased training strategy was adopted: In the first phase (the first 150 iterations), the coupling weights of the cross-fusion layer were fixed, and only the feature encoding layer and deep mapping layer were trained, allowing the model to initially learn the mapping relationship between single factors and combustion potential; in the second phase (iterations 151-300), the coupling weight constraints were released, allowing the model to autonomously learn the dynamic coupling rules between multiple factors; in the third phase (iterations 301-400), all layer parameters were fine-tuned, and the training effect was monitored using the prediction accuracy of the validation set as an indicator. An early stopping mechanism was triggered when the prediction accuracy of the validation set showed no improvement for 20 consecutive iterations.After training, the model underwent comprehensive performance validation: 5-fold cross-validation was used to calculate the average prediction error, requiring an average error of less than 2%; 100 sets of test data under extreme external conditions such as high wind speed (≥8m / s), low air pressure (≤90kPa), and high humidity (≥80%) were selected for specific validation to ensure the model's prediction accuracy under extreme conditions. Horizontal wind speed, vertical wind speed, ambient air pressure, ambient temperature, ambient humidity, as well as the quantified values ​​of the influence of exhaust gas combustion stability and exhaust gas combustibility characteristics from the external conditions of the flare combustion were simultaneously input into the trained model. The model outputs preliminary quantified values ​​of combustion potential characteristics based on the learned multi-factor coupling laws; subsequently, it was combined with historical data from similar processes. A correction model is established based on the measured combustion potential data of the flare. The correction model is constructed based on multinomial regression. The correction coefficient is dynamically adjusted according to the similarity between the current external conditions and historical operating conditions. When the similarity is greater than or equal to the threshold, the historical average deviation is used for correction. When the similarity is less than the threshold, the statistical distribution of the model prediction error is used for probability correction to ensure that the deviation between the corrected combustion potential characteristic data and the measured value is less than 3%. Finally, the corrected quantitative values ​​and time series labels are integrated to generate exhaust gas combustion potential characteristic data. The exhaust gas combustion potential characteristic data is a quantitative data characterizing the strength of the combustion reaction capability of the flare exhaust gas under specific external conditions (ambient wind speed, air pressure, etc.) and operating conditions (flow, transport, concentration characteristics).

[0031] Step S26: Perform exhaust gas combustion potential state interval discrimination processing based on exhaust gas combustion potential characteristic data to generate exhaust gas combustion potential state interval data.

[0032] In this embodiment of the invention, the exhaust gas combustion potential feature data is processed by interval discrimination of exhaust gas combustion potential state, and an interval discrimination model based on support vector machine is constructed, supplemented by a threshold comparison mechanism to improve the reliability of the results. The model input is exhaust gas combustion potential feature data, ambient wind speed level, and exhaust gas flow rate level, and the output is the potential state category (fully developed state, restricted state, transitional state). Through a large amount of experimental data, the effective range of exhaust gas combustion potential feature data is determined to be 0-1.0. The interval division benchmark thresholds are configured as follows: upper threshold 0.7 and lower threshold 0.3. 800 sets of different potential feature data and corresponding operating condition parameters are selected as training samples. After labeling the state categories, they are input into the model for training. The kernel function is set as radial basis function, the penalty coefficient is set to 10, and the gamma value is set to 0.1. The exhaust gas combustion potential feature data generated in step S25, along with the corresponding environmental and exhaust gas parameters, are input into the trained model. The model outputs preliminary state discrimination results. Simultaneously, the potential feature data is compared with a benchmark threshold. When the model output matches the threshold comparison result, the state category is directly determined. When the results are inconsistent, historical similar data is retrieved for secondary verification. When the feature data is in the 0.7-1.0 range, it is determined to be a state of sufficient exhaust gas combustion potential. The range, mean, and fluctuation characteristics of the data within this range are recorded to generate sufficient exhaust gas combustion potential state interval data. When the feature data is in the 0-0.3 range, it is determined to be a state of limited exhaust gas combustion potential. Relevant data is recorded to generate limited exhaust gas combustion potential state interval data. When the feature data is in the 0.3-0.7 range, it is determined to be a transitional state, and only records are made for filing, completing the interval discrimination processing.

[0033] Furthermore, the exhaust gas combustion potential state range data mentioned in step S26 includes exhaust gas combustion potential full state range data and exhaust gas combustion potential limited state range data.

[0034] Furthermore, step S3 includes the following steps: Step S31: Analyze the flare combustion morphology characteristics based on the flare combustion monitoring data to generate flare combustion morphology characteristic data; In this embodiment of the invention, the combustion morphology characteristics of the torch are analyzed based on the torch combustion monitoring data. Flame contour images and flame temperature data are selected as the core data sources for analysis, with an image resolution of 1920×1080 pixels and a temperature acquisition accuracy of ±1℃. An improved Canny edge detection algorithm is used to process the flame contour image, setting a high threshold of 200 and a low threshold of 100. Gaussian filtering (with a standard deviation of 1.4) is used to smooth image noise and extract the flame contour boundary. Based on the contour boundary, the flame shape parameters (including flame aspect ratio and contour irregularity), height parameters (vertical distance between the flame tip and the torch outlet), diffusion angle parameters (the angle between the maximum flame diffusion point and the outlet center), and jitter amplitude parameters (the standard deviation of the flame center displacement in 10 consecutive frames) are calculated. A region growing algorithm was used to divide the flame temperature data into regions. The seed point was set as the pixel with the highest temperature, and the growth threshold was set to 50℃. The flame was divided into a core region (temperature ≥ 1200℃), a transition region (800℃ ≤ temperature < 1200℃), and a peripheral region (temperature < 800℃). The average temperature, temperature standard deviation, and area ratio of each region were calculated. The extracted morphological parameters and temperature distribution parameters were integrated to generate torch combustion morphological feature data.

[0035] Step S32: Analyze the torch combustion morphology pattern based on the torch combustion morphology characteristic data to generate torch combustion morphology pattern data; In this embodiment of the invention, torch combustion morphology pattern analysis is performed based on torch combustion morphology feature data, and a morphology pattern recognition model based on a convolutional neural network is constructed. The model input is the feature matrix of the flame contour image (dimension 64×64). Two convolutional layers are set (with 32 and 64 convolutional kernels respectively, and a kernel size of 3×3). The pooling layer uses average pooling (pooling kernel size 2×2), and one fully connected layer is set (with 128 neurons). The output layer contains three morphology patterns (stable combustion pattern, slightly fluctuating combustion pattern, and violently fluctuating combustion pattern). The activation function is Softmax, the optimizer is Adam, the learning rate is set to 0.001, and the number of iterations is set to 300. 1500 sets of torch combustion morphology feature data and corresponding morphology pattern labels under different combustion states are collected as training samples and divided into training set and test set in an 8:2 ratio. The training set is input into the model for training, and training stops when the recognition accuracy of the test set reaches 97%. The extracted torch combustion morphology feature data is input into the trained model. The model outputs the morphology pattern determination result based on the learned morphology feature rules. At the same time, a feature matching mechanism is used to verify the result. The morphology feature data is compared with the standard feature thresholds of three typical patterns. When the model output matches the feature matching result, the morphology pattern corresponding to the current flame is determined, and torch combustion morphology pattern data is generated.

[0036] Step S33: Based on the exhaust gas combustion potential state range data and the flare combustion morphology pattern data, perform combustion morphology correlation deviation and combustion efficiency response relationship analysis under the combustion potential state range constraint, and generate combustion morphology correlation deviation-efficiency response relationship data. In this embodiment of the invention, a correlation analysis model based on graph neural network is constructed to analyze the correlation deviation of combustion morphology and combustion efficiency response under the constraint of combustion potential state interval data and flare combustion morphology pattern data. The model inputs are the exhaust gas combustion potential state code (fully constrained state code is 1, constrained state code is 0), the flare combustion morphology pattern code, and morphological feature parameters. The outputs are the combustion morphology correlation deviation coefficient and the combustion efficiency response value. The model nodes are set as potential state nodes, morphology pattern nodes, and feature parameter nodes. The edge weights are the correlation strength between each node, and the key correlation features are strengthened through an attention mechanism. 2000 sets of historical data under different potential states and different morphology patterns are selected as training samples. The deviation degree of the corresponding morphology from the standard morphology and the change value of combustion efficiency are labeled and input into the model for training. The number of iterations is set to 400 and the learning rate is 0.002. The model parameters are optimized through 5-fold cross-validation. Data on exhaust gas combustion potential state range and flare combustion morphology patterns are input into the trained model. The model first calculates the correlation deviation characteristic parameters (including morphological parameter deviation and temperature distribution deviation) between the current morphology pattern and the standard morphology pattern under the corresponding potential state. Then, it analyzes the variation law of combustion efficiency under different degrees of deviation. The sensitivity of each deviation parameter to combustion efficiency is verified through gradient perturbation experiments. The weights of sensitive parameters are determined, and the correlation deviation characteristics and efficiency response law are integrated to generate combustion morphology correlation deviation-efficiency response relationship data.

[0037] Step S34: Design a mapping index for flare combustion mode and combustion efficiency based on the combustion mode correlation deviation-efficiency response relationship data, and generate a flare mode-efficiency mapping index. In this embodiment of the invention, a mapping index for flare combustion morphology and combustion efficiency is designed based on combustion morphology-efficiency response relationship data. The composition and weights of the mapping index are determined using an analytic hierarchy process (AHP). Parameters significantly affecting combustion efficiency are selected from the combustion morphology-efficiency response relationship data, including the flame morphology deviation coefficient (the degree of deviation from the aspect ratio and diffusion angle), temperature uniformity coefficient (the proportion of temperature difference between the core and peripheral areas), and vibration stability coefficient (the normalized value of vibration amplitude). A hierarchical model is constructed, with the target layer representing the combustion efficiency mapping value, the criterion layer representing the three selected core parameters, and the scheme layer representing the specific quantitative values ​​of each parameter. The weights of the criterion layer are determined using an expert scoring method, with the morphology deviation coefficient weight set to 0.4, the temperature uniformity coefficient weight set to 0.35, and the vibration stability coefficient weight set to 0.25. The calculation expression for the torch shape-efficiency mapping index is constructed based on weights. The expression is the weighted sum of the quantified values ​​of each core parameter and their corresponding weights. The quantified values ​​are normalized and mapped to the 0-1 range to ensure that the value range of the mapping index is 0-1. The larger the value, the higher the combustion efficiency, thus generating the torch shape-efficiency mapping index.

[0038] Step S35: Perform explicit analysis of flare combustion efficiency on the flare combustion monitoring data based on the flare shape-efficiency mapping index to generate explicit flare combustion efficiency data.

[0039] In this embodiment of the invention, the flare combustion efficiency is made explicit based on the flare shape-efficiency mapping index. The data sampling interval is configured to be 100ms, and the analysis period is one sampling period. Flame contour images and flame temperature data from the flare combustion monitoring data are retrieved, and real-time quantified values ​​of the flame shape deviation coefficient, temperature uniformity coefficient, and jitter stability coefficient are extracted according to the processing standard in step S31. These three quantified values ​​are substituted into the calculation expression of the flare shape-efficiency mapping index, and the real-time quantified combustion efficiency value is obtained through weighted summation. A data verification mechanism is set up to compare the calculated quantified combustion efficiency value with the measured combustion efficiency value under the same historical conditions. When the deviation is less than 3%, the quantified value is directly output; when the deviation is greater than or equal to 3%, the quantified value is corrected using the correction coefficient of the mapping index (trained based on historical deviation data). The corrected quantified combustion efficiency value is used as the final result to complete the explicit analysis of flare combustion efficiency and generate explicit flare combustion efficiency data.

[0040] Furthermore, step S33 includes the following steps: Step S331: Analyze the correlation deviation characteristics between the flare combustion mode and the exhaust gas combustion potential state interval using the exhaust gas combustion potential state interval data, and generate combustion mode correlation deviation characteristic data of the combustion potential state. In this embodiment of the invention, the correlation deviation characteristics between the flare combustion morphology pattern and the exhaust gas combustion potential state interval are analyzed using exhaust gas combustion potential state interval data. First, a standard morphology pattern database is constructed, which stores the corresponding stable combustion standard morphology parameter sets according to the exhaust gas combustion potential fully developed state and the restricted state. The standard parameters for the fully developed state include a standard aspect ratio of 1.8, a standard diffusion angle of 35°, a standard jitter amplitude of ≤5mm, and a standard temperature distribution (core area ≥40%, transition area 40%-50%, and peripheral area ≤10%). The standard parameters for the restricted state include a standard aspect ratio of 1.2, a standard diffusion angle of 25°, a standard jitter amplitude of ≤8mm, and a standard temperature distribution (core area ≥30%, transition area 45%-55%, and peripheral area ≤15%). All standard parameters are accompanied by operating condition adaptation range markings. A bi-branch correlation deviation analysis model based on an attention mechanism was constructed, with the two branches corresponding to the full state and the restricted state, respectively. The model inputs are the exhaust gas combustion potential state code (full state code is 1, restricted state code is 0), morphological parameters (aspect ratio, diffusion angle, jitter amplitude) and temperature distribution data (area percentage of each region, average temperature) from the flare combustion morphology pattern data. Two hidden layers were set, with 64 neurons in each layer. The activation function was ReLU, the optimizer was Adam, the learning rate was set to 0.001, and the number of iterations was set to 250. The attention mechanism focuses on the coupling correlation features between the potential state and the morphological parameters, and the weight allocation ranges from 0.1 to 0.8. The data on the exhaust gas combustion potential state range and the flare combustion morphology pattern data are synchronously input into the model. The model matches the standard morphological parameter set of the corresponding branch according to the state code. By comparing each parameter, the absolute deviation and deviation ratio of the current morphological parameter and the standard parameter, as well as the difference in the regional ratio of the current temperature distribution and the standard temperature distribution, are calculated. Then, the deviation parameters are weighted and filtered through an attention mechanism, and key deviation parameters (morphological parameter deviation value and temperature distribution deviation value) with a weight ≥ 0.3 are retained. At the same time, the potential state adaptation label corresponding to each deviation parameter is marked. The quantitative value of the key deviation parameter, the deviation ratio, and the state adaptation label are integrated to generate combustion morphology-related deviation feature data of the combustion potential state. The combustion morphology-related deviation feature data of the combustion potential state reflects that when the flame morphology deviates from the expected exhaust gas potential, there is a decrease in flare combustion efficiency.

[0041] Step S332: Perform combustion efficiency sensitivity analysis on the combustion mode correlation deviation feature data of combustion potential state to generate combustion mode correlation deviation feature efficiency sensitivity data. In this embodiment of the invention, a sensitivity analysis of combustion efficiency based on the combustion morphology-related deviation feature data of combustion potential states is performed. A gradient boosting regression-based interval sensitivity analysis model is constructed. The model is divided into two independently trained regression branches according to the fully potential state and the restricted state. The input is the combustion morphology-related deviation feature data of the combustion potential state (including the quantified values ​​of key deviation parameters, deviation ratio, and state adaptation label), and the output is the combustion efficiency sensitivity coefficient corresponding to each deviation parameter (the value ranges from 0 to 1, and the larger the value, the more significant the impact of the parameter on combustion efficiency). 2000 sets of historical data under different potential states are selected as training samples, including 1000 sets of samples each for the fully potential state and the restricted state. Each set of samples is labeled with key deviation parameters and corresponding measured changes in combustion efficiency. The samples are divided into training and testing sets in an 8:2 ratio. The number of decision trees is set to 100, the learning rate is 0.01, the maximum depth is 8 layers, and the minimum number of leaf nodes is 5. The model parameters are optimized through 5-fold cross-validation. Combustion morphology-related deviation characteristic data of combustion potential state are input into the model branch corresponding to the potential state. The model outputs an initial sensitivity coefficient by analyzing the fluctuation range of combustion efficiency when the gradient of each deviation parameter changes. At the same time, the result is verified by the control variable method. Other deviation parameters are fixed as standard values, and only the target deviation parameter is gradually adjusted by 5%, 10%, and 15% to monitor the rate of change of combustion efficiency. The coefficient value is corrected according to the deviation between the rate of change and the initial sensitivity coefficient. After correction, normalization is used to ensure that the coefficient range meets the requirements. The sensitivity coefficients of each deviation parameter and the verification deviation value are integrated to generate combustion morphology-related deviation characteristic efficiency sensitivity data. The combustion morphology-related deviation characteristic efficiency sensitivity data reflects the sensitivity characteristics of the impact on flare combustion efficiency when the matching status between combustion potential state and combustion morphology deviates.

[0042] Step S333: Based on the efficiency sensitivity data of combustion mode correlation deviation characteristics, analyze the relationship between combustion mode correlation deviation and combustion efficiency response under the constraint of combustion potential state interval, and generate combustion mode correlation deviation-efficiency response relationship data.

[0043] In this embodiment of the invention, the relationship between combustion mode correlation deviation and combustion efficiency response is analyzed based on the efficiency sensitivity data of combustion mode correlation deviation characteristics under the constraint of combustion potential state interval. A response relationship model based on a two-branch neural network is constructed, with the two branches strictly corresponding to the fully developed and constrained states of exhaust gas combustion potential, respectively, to ensure the accuracy of the interval constraint. The model input includes the efficiency sensitivity data of combustion mode correlation deviation characteristics (sensitivity coefficients of each deviation parameter), the quantized values ​​of each deviation parameter, the potential state code, and real-time exhaust gas flow data. The output includes the combustion efficiency response value (range 0-1, with larger values ​​indicating higher combustion efficiency) and the response delay time (unit: sampling period). The model's input layer dimension is set to 7, the feature encoding layer maps the input parameters to a 64-dimensional feature space, and two hidden layers are set (with 128 and 64 neurons respectively). Residual connections are used to address the vanishing gradient problem. The output layer dimension is 2, the activation function is Sigmoid, the optimizer is RMSprop, the learning rate is set to 0.001, and the number of iterations is set to 400. L2 regularization (regularization coefficient 0.001) is added to suppress overfitting, and an adaptive learning rate adjustment mechanism is configured. When the validation set error increases for 15 consecutive iterations, the learning rate is automatically halved. 3000 sets of combustion efficiency response data (1500 sets each for fully developed and constrained states) corresponding to different potential states and different combinations of deviation parameters are collected as training samples. Each set of samples contains complete input parameters and corresponding measured values ​​of combustion efficiency and response delay. These are divided into training and validation sets in a 7:3 ratio. The training set is input into the model for training after min-max standardization. Training stops when the validation set prediction accuracy reaches 96%. The efficiency sensitivity data of combustion morphology-related deviation characteristics, the quantified values ​​of each deviation parameter, the potential state code, and the real-time exhaust gas flow data are input into the corresponding branch model. The model first selects the core influencing parameters based on the sensitivity coefficient, and then analyzes the dynamic change law of combustion efficiency under different deviations of the core parameters. The corresponding mapping relationship between the combination of deviation parameters and the combustion efficiency response value and response delay time is established. The influence threshold under the potential state constraint is clarified (morphology deviation threshold ≤15% under the full state and morphology deviation threshold ≤20% under the restricted state). The mapping relationship, influence threshold, and response delay data are integrated to generate combustion morphology-related deviation-efficiency response relationship data. The combustion morphology-related deviation-efficiency response relationship data reflects the relationship between the deviation of the matching status between the combustion potential state and the combustion morphology and the influence on the flare combustion efficiency.

[0044] Furthermore, step S4 includes the following steps: Step S41: Perform trend feature analysis on the explicit data of flare combustion efficiency to generate explicit trend feature data of flare combustion efficiency; In this embodiment of the invention, trend feature analysis is performed on the explicit data of torch combustion efficiency. The data sampling frequency is configured to be 10Hz, and the analysis duration covers 20 consecutive sampling periods. A trend feature extraction model based on a long short-term memory network is constructed. The input layer dimension of the model is set to 1 (corresponding to the explicit combustion efficiency data), the hidden layer is set to 2 layers, and the number of neurons in each layer is 64. The output layer dimension is 3 (corresponding to the three features of trend direction, rate of change, and fluctuation amplitude). The activation function is ReLU, the optimizer is Adam, the learning rate is set to 0.001, the number of iterations is set to 300, and an early stopping mechanism is set to prevent overfitting. The early stopping patience is set to 20. 2000 sets of historical explicit combustion efficiency time-series data are collected as training samples, labeled with corresponding trend feature labels, and then input into the model for training. Training stops when the feature extraction accuracy of the test set reaches 97%. The explicit data of torch combustion efficiency from 20 consecutive sampling periods are input into the trained model. The model learns the temporal variation pattern, calculates the difference between adjacent period data, extracts the trend direction of efficiency change (increasing, stable, decreasing), the rate of change (the amount of efficiency change per unit time), and the fluctuation amplitude (the standard deviation of efficiency data in consecutive periods), and integrates the three feature parameters to generate explicit trend feature data of torch combustion efficiency.

[0045] Step S42: Based on the characteristics of the apparent trend of flare combustion efficiency, conduct combustion efficiency trend anomaly assessment processing to generate combustion efficiency trend anomaly assessment data; In this embodiment of the invention, combustion efficiency trend anomaly assessment processing is performed based on the explicit trend characteristics of flare combustion efficiency. Trend anomaly judgment thresholds are configured as follows: the rate of decrease threshold is 0.02 / sampling period, the anomaly duration threshold is 3 sampling periods, and the fluctuation amplitude threshold is 0.05. An anomaly assessment model based on support vector machines is constructed. The model input consists of the trend direction, rate of change, and fluctuation amplitude from the explicit trend characteristic data of flare combustion efficiency, and the output is the anomaly assessment result (normal, slightly abnormal, severely abnormal). 1500 sets of historical data under different trend characteristics are selected as training samples, labeled with corresponding anomaly level labels, and input into the model for training. The kernel function is set to a radial basis function, the penalty coefficient is set to 10, and the gamma value is set to 0.1. The extracted explicit trend feature data of flare combustion efficiency is input into the trained model, and the model outputs preliminary anomaly assessment results. At the same time, the trend feature data is compared with preset thresholds. When the trend direction is downward, the rate of change exceeds the rate of change threshold and the duration reaches the abnormal duration threshold, or the fluctuation amplitude exceeds the fluctuation amplitude threshold, it is judged as an anomaly. Combining the model output with the threshold comparison results, when the two are consistent, the anomaly level is determined. When they are inconsistent, historical similar data is retrieved for secondary verification. The anomaly level, anomaly start time, and anomaly features are integrated to generate combustion efficiency trend anomaly assessment data.

[0046] Step S43: Analyze the characteristics of combustion failure precursors based on the abnormal combustion efficiency trend assessment data, and generate combustion failure precursor characteristic data; In this embodiment of the invention, combustion failure precursor feature analysis is performed based on combustion efficiency trend anomaly assessment data. The precursor feature extraction window is configured as the 10 sampling periods before the occurrence of the anomaly, and a precursor feature recognition model based on a convolutional neural network is constructed. The model input is a feature matrix (10×8 dimension) composed of torch combustion morphology feature data (including parameters such as flame aspect ratio, diffusion angle, shaking amplitude, and temperature distribution) from the 10 sampling periods before the occurrence of the anomaly. Two convolutional layers are set (32 and 64 convolutional kernels respectively, with a kernel size of 3×3), max pooling is used in the pooling layer (pooling kernel size of 2×2), and one fully connected layer is set (128 neurons). The output layer is a common failure precursor feature category (increased flame shaking amplitude, decreased core temperature, abnormally widened diffusion angle, etc.). 1200 sets of historical data under abnormal operating conditions are collected as training samples, labeled with corresponding precursor feature labels, and then input into the model for training. The learning rate is set to 0.002, the number of iterations is 250, and the cross-entropy loss function is used to optimize the model. Based on the abnormal assessment data of combustion efficiency trend, the time of the anomaly is determined. The combustion morphology and temperature distribution data of the torch in the 10 sampling periods before the anomaly occurred are traced back and input into the trained model. The model outputs the identified precursor features. Through the feature verification mechanism, the identified features are matched with the anomaly development pattern to screen out the core precursor features directly related to the decline in combustion efficiency. The types, occurrence time and change amplitude of the core precursor features are integrated to generate combustion failure precursor feature data.

[0047] Step S44: Analyze combustion failure-related features of combustion failure precursor features using exhaust gas combustion potential state range data to generate combustion failure-related feature data.

[0048] In this embodiment of the invention, combustion failure-related feature analysis is performed on combustion failure precursor feature data using exhaust gas combustion potential state interval data. An attention-based correlation analysis model is constructed. The model inputs are exhaust gas combustion potential state codes (fully charged state code is 1, constrained state code is 0) and core precursor feature parameters from the combustion failure precursor feature data. The outputs are combustion failure-related feature categories and specific feature parameters. The model has two hidden layers with 64 neurons per layer, using ReLU as the activation function, Adam as the optimizer, a learning rate of 0.001, and 200 iterations. 1000 sets of historical data under different potential states and different failure precursors are selected as training samples, labeled with corresponding failure-related feature tags, and input into the model for training. Training stops when the classification accuracy on the test set reaches 98%. The exhaust gas combustion potential state range data and combustion failure precursor feature data are simultaneously input into the trained model. When the model identifies a potential state code of 1 (sufficient state), it determines that the cause of failure is unrelated to fuel supply. It then focuses on in-depth analysis of precursor features related to combustion organization (such as flame anchoring deviation, abnormal diffusion angle, etc.), extracts the quantitative values ​​and variation patterns of these features, and generates combustion failure-non-fuel-limited feature data. When the model identifies a potential state code of 0 (limited state), it determines that the cause of failure is related to fuel supply. It then focuses on in-depth analysis of fuel-related precursor features (such as sudden drop in core temperature, shrinkage of combustion range, etc.), extracts the quantitative values ​​and variation patterns of these features, and generates combustion failure-fuel-limited feature data.

[0049] Furthermore, the combustion failure-related feature data in step S44 includes combustion failure-non-fuel-limited feature data or combustion failure-fuel-limited feature data.

[0050] Furthermore, step S44 includes the following steps: When the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential full state interval data, the non-fuel-limited feature analysis of combustion failure is performed on the combustion failure precursor feature data to generate combustion failure-non-fuel-limited feature data; or, when the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential limited state interval data, the fuel-limited feature analysis of combustion failure is performed on the combustion failure precursor feature data to generate combustion failure-fuel-limited feature data.

[0051] In this embodiment of the invention, when the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential fully developed state interval data, non-fuel-constrained feature analysis of combustion failure is performed on the combustion failure precursor feature data. The feature analysis window is configured to be the 10 sampling periods before the anomaly occurs, and a non-fuel-constrained feature extraction model based on an attention mechanism is constructed. The model input consists of the core precursor feature parameters in the combustion failure precursor feature data (including flame anchoring offset, diffusion angle fluctuation value, flame center jitter amplitude, and the rate of change of area ratio of each temperature region). Two hidden layers are set, with 64 neurons in each layer. The activation function is ReLU, the optimizer is Adam, the learning rate is set to 0.001, and the number of iterations is set to 200. 800 sets of historical failure condition data under the fully developed exhaust gas combustion potential state are selected as training samples. After labeling with non-fuel-constrained feature tags (such as combustion organization imbalance features and flame stability imbalance features), the data are input into the model for training. Training stops when the feature extraction accuracy of the test set reaches 98%. The model is trained by inputting the early signs of combustion failure into the model. The model learns the failure characteristic patterns under the full potential state, focuses on in-depth analysis of combustion organization-related features, calculates the percentage deviation of flame anchoring offset from the standard anchoring position, the peak value and duration of diffusion angle fluctuations, and the variation amplitude of the area proportion of each temperature region. It then identifies core characteristic parameters that significantly affect combustion failure, integrates the quantified values, trends, and influence weights of these parameters, and generates combustion failure-non-fuel-limited characteristic data. When the exhaust gas combustion potential state interval data corresponding to the early signs of combustion failure characteristic data is the exhaust gas combustion potential limited state interval data, fuel-limited characteristic analysis of combustion failure is performed on the early signs of combustion failure characteristic data. The feature analysis window is configured as the 10 sampling periods before the anomaly occurs, and a fuel-limited characteristic identification model based on gradient boosting trees is constructed. The model input consists of core early signs characteristic parameters from the early signs of combustion failure characteristic data (including the rate of temperature decrease in the flame core area, the duration of the core area temperature below 1200℃, the rate of combustion range reduction, and the degree of flame brightness decay). The number of decision trees is set to 80, the learning rate to 0.01, and the maximum depth to 8 layers. 800 sets of historical data on failure conditions under limited combustion potential of exhaust gas were selected as training samples. After labeling fuel-limited feature tags (such as insufficient supply of combustible components and incomplete combustion reaction), the data were input into the model for training. The model parameters were optimized through 5-fold cross-validation. Training was stopped when the feature recognition accuracy of the test set reached 97%.The model is trained by inputting the early signs of combustion failure into the model. The model learns the failure characteristics under the confined potential state, focuses on fuel supply-related characteristics for in-depth analysis, calculates the deviation between the core temperature drop rate and the standard temperature drop rate, the quantified value of the rate of reduction in combustion range, and the normalized value of the degree of flame brightness decay, and screens out the core characteristic parameters that have a significant impact on combustion failure. The quantified values, trends and influence weights of these parameters are integrated to generate combustion failure-fuel confinement characteristic data.

[0052] Furthermore, step S5 includes the following steps: Step S51: When the combustion failure related characteristic data is combustion failure-non-fuel-limited characteristic data, analyze the combustion organization regulation relationship of the flare combustion monitoring data to generate combustion organization regulation relationship data; based on the combustion organization regulation relationship data and the combustion failure-non-fuel-limited characteristic data, analyze the minimum intervention control parameters for combustion organization regulation to generate the minimum intervention control parameters for combustion organization regulation. In this embodiment of the invention, when the combustion failure-related feature data is combustion failure-non-fuel-limited feature data, combustion organization regulation relationship analysis is performed on the flare combustion monitoring data. The analysis period is configured as 10 consecutive sampling periods, and a regulation relationship model based on graph neural network is constructed. The model input includes flame morphology parameters (flame anchoring offset, diffusion angle, jitter amplitude), temperature distribution parameters (temperature values ​​in each region, temperature uniformity), and combustion organization regulation parameters (flame anchoring structure position, combustion nozzle angle) from the flare combustion monitoring data. Nodes are set as morphology parameter nodes, temperature parameter nodes, and regulation parameter nodes, and the edge weights are the correlation strength between each node. Key regulation correlation features are strengthened through an attention mechanism. 1500 sets of historical data under non-fuel-limited conditions are selected as training samples, and the variation patterns of morphology and temperature parameters under different regulation parameters are labeled. These are input into the model for training, with 300 iterations and a learning rate of 0.001. The model parameters are optimized through 5-fold cross-validation. The flare combustion monitoring data is input into the trained model, and the model outputs the correlation patterns between the flame anchoring structure position and flame stability, and between the nozzle angle and temperature distribution uniformity, generating combustion organization regulation relationship data. Based on this data and the combustion failure-non-fuel-limited characteristic data, a minimum intervention control parameter optimization model based on gradient descent is constructed. The objective function is set as minimizing the adjustment amount of the regulating parameter and eliminating combustion failure characteristics, with the constraint that the variation range of the regulating parameter does not exceed ±10% of the rated value. The core failure parameters (flame anchoring offset deviation ratio and diffusion angle fluctuation peak) in the combustion failure-non-fuel-limited characteristic data are used as the optimization target input model. The model determines the minimum intervention adjustment amount through iterative calculation. Specifically, the adjustment distance of the anchoring structure is determined based on the flame anchoring offset deviation ratio, and the adjustment angle of the nozzle angle is determined based on the diffusion angle fluctuation peak. The adjustment distance, adjustment angle, and adjustment rate are integrated to generate the minimum intervention control parameter for combustion organization regulation.

[0053] Step S52: When the combustion failure related characteristic data is combustion failure-fuel limitation characteristic data, analyze the minimum intervention control parameters for fuel supply regulation based on the exhaust gas combustion potential characteristic data and the combustion failure-fuel limitation characteristic data, and generate the minimum intervention control parameters for fuel supply regulation. In this embodiment of the invention, when the combustion failure-related feature data is combustion failure-fuel limitation feature data, the minimum intervention control parameter analysis for fuel replenishment regulation is performed based on the exhaust gas combustion potential feature data and the combustion failure-fuel limitation feature data. The replenishment regulation response time threshold is configured to be 2 sampling periods, and a replenishment parameter optimization model based on a dual-branch neural network is constructed. The two branches correspond to the calculation of auxiliary fuel replenishment amount and the calculation of air content regulation, respectively. The model inputs are exhaust gas combustion potential feature data (quantified value of combustion potential), combustion failure-fuel limitation feature data (core area temperature decrease rate, combustion range reduction rate), and exhaust gas parameters (combustible component concentration, exhaust gas flow rate). The outputs are auxiliary fuel replenishment amount and air regulation amount. The model input layer dimension is set to 6, the hidden layer is set to 2 layers (the number of neurons is 128 and 64 respectively), the output layer dimension is 2, the activation function is Sigmoid, the optimizer is RMSprop, the learning rate is set to 0.001, the number of iterations is set to 400, and L2 regularization (regularization coefficient 0.001) is added to suppress overfitting. 2000 sets of historical data under fuel-constrained operating conditions were selected as training samples. Optimal fuel supply and air conditioning rates were labeled under different potential states and failure degrees, and input into the model for training. Training stopped when the prediction accuracy on the validation set reached 96%. Exhaust gas combustion potential characteristic data and combustion failure-fuel-constrained characteristic data were input into the trained model. The model first calculates the basic fuel supply based on the difference between the quantified combustion potential value and the standard potential value. Then, it adjusts the fuel supply based on the core area temperature decrease rate and the combustion range reduction rate. Simultaneously, it calculates the required air conditioning rate based on the auxiliary fuel supply according to the stoichiometric ratio to ensure thorough fuel-air mixing and combustion. The auxiliary fuel supply amount, supply rate, air conditioning amount, and regulation rate are integrated to generate the minimum intervention control parameters for fuel supply regulation.

[0054] Step S53: Design intelligent control parameters for flare combustion efficiency that are adjusted in real time by adjusting the minimum intervention control parameters through combustion organization and / or fuel replenishment; In this embodiment of the invention, intelligent control parameters for flare combustion efficiency are designed with real-time feedback adjustment by using minimum intervention control parameters for combustion organization adjustment and / or fuel replenishment adjustment. A parameter fusion model based on multi-objective optimization is constructed, with the optimization objectives being optimal combustion efficiency, minimum intervention adjustment, and fastest control response. Parameter priority weights are configured as follows: combustion efficiency optimization weight is set to 0.5, minimum intervention adjustment weight is set to 0.3, and fastest control response weight is set to 0.2. When only the minimum intervention control parameter for combustion organization adjustment exists, this parameter is directly used as the basic control parameter. A dynamic correction coefficient is set based on real-time explicit combustion efficiency data. The correction coefficient is determined based on the deviation between the current efficiency value and the optimal efficiency value; the larger the deviation, the larger the correction coefficient. The basic control parameter is adjusted through the correction coefficient to generate intelligent control parameters for flare combustion efficiency with real-time feedback adjustment. When only the minimum intervention control parameter for fuel replenishment adjustment exists, a predictive correction term is set based on this parameter and combined with the real-time changing trend of exhaust gas combustion potential characteristic data. The predictive correction term is determined based on the rate of change of potential data to avoid lag in the replenishment parameter. The basic parameter and the predictive correction term are integrated to generate intelligent control parameters. When both types of parameters exist simultaneously, they are input into a multi-objective optimization model. The model calculates the optimal combination ratio of each parameter through weight allocation, sets the priority of flame anchoring parameter adjustment to be higher than that of auxiliary fuel supply parameter adjustment, and prioritizes the adjustment of combustion organization adjustment parameters. If the efficiency still does not meet the standard after the combustion organization adjustment parameter adjustment, the fuel supply adjustment parameter adjustment is then initiated. The optimized parameter combination and execution order are integrated to generate intelligent control parameters for flare combustion efficiency that are adjusted in real time.

[0055] Step S54: Execute intelligent flare combustion control operation based on the intelligent control parameters of flare combustion efficiency adjusted according to real-time feedback.

[0056] In this embodiment of the invention, intelligent control of flare combustion is performed based on intelligent control parameters for flare combustion efficiency adjusted in real time. The control execution frequency is configured to be consistent with the sensor sampling frequency (10Hz), constructing a control execution system based on closed-loop feedback, including a mechanical execution module, a parameter monitoring module, and a deviation correction module. The mechanical execution module consists of a flame anchoring structure adjustment mechanism, an auxiliary fuel supply valve adjustment mechanism, and an air supply fan speed adjustment mechanism, with the response delay of each mechanism controlled within 50ms. The parameter monitoring module collects the actual execution values ​​of the control parameters (actual adjustment distance of the anchoring structure, actual fuel replenishment, and actual air supply) and flare combustion status data (flame morphology, combustion temperature, and combustion efficiency) in real time. The target value in the intelligent control parameters is compared with the actual execution value collected by the parameter monitoring module. When the deviation is greater than 2%, the output of the execution mechanism is adjusted through the deviation correction module to ensure that the actual execution value accurately matches the target value. For combustion organization regulation and control, the stepper motor of the flame anchoring structure adjustment mechanism adjusts the anchoring position, gradually correcting according to the adjustment distance and rate in the intelligent control parameters, while high-speed imaging technology monitors flame shape changes in real time. For fuel supply regulation and control, the electric adjustment mechanism of the auxiliary fuel supply valve adjusts the valve opening, supplying fuel precisely according to the supply amount and rate, while the frequency conversion adjustment mechanism of the air supply fan adjusts the speed to match the amount of air required for fuel combustion. During the control operation, the real-time combustion efficiency explicit data is continuously compared with the optimal efficiency threshold. When the efficiency value remains within the optimal range (0.9-1.0) for three consecutive sampling cycles, the current control parameters are maintained. When the efficiency value deviates from the optimal range, the parameter redesign process in step S53 is triggered to obtain the updated intelligent control parameters and continue the control operation, forming a closed-loop control to ensure that the flare combustion is always maintained in a stable and efficient state.

[0057] This specification provides a control system for flare combustion efficiency that can be adjusted in real time, used to execute the control method for flare combustion efficiency that can be adjusted in real time as described above. The control system for flare combustion efficiency that can be adjusted in real time includes: The flare combustion multi-source correlation monitoring module is used to monitor and process flare combustion multi-source correlation data using multi-source monitoring sensors installed at the flare channel entrance, and generate flare combustion multi-source correlation monitoring data, wherein the flare combustion multi-source correlation monitoring data includes flare exhaust gas monitoring data, flare combustion external condition data, and flare combustion monitoring data. The exhaust gas combustion potential state interval analysis module is used to perform interval analysis of exhaust gas combustion potential state through flare exhaust gas monitoring data and flare combustion external condition data, and generate exhaust gas combustion potential state interval data. The combustion efficiency explicit analysis module is used to perform explicit analysis of flare combustion monitoring data and generate explicit flare combustion efficiency data. The combustion failure related feature analysis module is used to perform fuel-related feature analysis of combustion failure based on exhaust gas combustion potential state range data and flare combustion efficiency explicit data, and generate combustion failure related feature data. The flare combustion intelligent control module is used to design intelligent control parameters for flare combustion efficiency that are adjusted in real time based on combustion failure-related characteristic data; and to execute intelligent control operations for flare combustion based on the intelligent control parameters for flare combustion efficiency adjusted in real time.

[0058] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0059] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for controlling the combustion efficiency of a flare with real-time feedback adjustment, characterized in that, Includes the following steps: Step S1: Use the multi-source monitoring sensors set up at the flare channel entrance to perform multi-source correlation data monitoring and processing of flare combustion to generate multi-source correlation monitoring data of flare combustion, wherein the multi-source correlation monitoring data of flare combustion includes flare exhaust gas monitoring data, flare combustion external condition data and flare combustion monitoring data; Step S2: Perform range analysis of exhaust gas combustion potential state using flare exhaust gas monitoring data and flare combustion external condition data to generate exhaust gas combustion potential state range data; Step S3: Perform explicit analysis on the flare combustion monitoring data to generate explicit flare combustion efficiency data; Step S4: Based on the exhaust gas combustion potential state range data and the flare combustion efficiency explicit data, perform fuel-related characteristic analysis of combustion failure and generate combustion failure-related characteristic data; Step S5: Design intelligent control parameters for flare combustion efficiency that are adjusted in real time based on combustion failure related characteristic data; execute intelligent flare combustion control operations according to the intelligent control parameters for flare combustion efficiency adjusted in real time.

2. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Analyze the exhaust gas condition status type of the flare exhaust gas monitoring data and generate exhaust gas condition status type data; Step S22: Analyze the impact of exhaust gas operating conditions on flare combustion stability based on exhaust gas operating condition data, and generate exhaust gas operating condition combustion stability impact data. Step S23: Identify and process the combustible components of the flare gas monitoring data to generate combustible component data of the flare gas; Step S24: Design the dilution and suppression weights for exhaust gas flammability based on the exhaust gas flammability component data, generate exhaust gas flammability dilution and suppression weight parameters, and perform exhaust gas flammability characteristic analysis on the exhaust gas flammability component data through the exhaust gas flammability dilution and suppression weight parameters to generate exhaust gas flammability characteristic data. Step S25: Based on the external conditions data of flare combustion and the data on the impact of exhaust gas combustion stability on exhaust gas operating conditions and the combustibility characteristics data of exhaust gas, perform exhaust gas combustion potential characteristic analysis to generate exhaust gas combustion potential characteristic data. Step S26: Perform exhaust gas combustion potential state interval discrimination processing based on exhaust gas combustion potential characteristic data to generate exhaust gas combustion potential state interval data.

3. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 2, characterized in that, The exhaust gas combustion potential state range data mentioned in step S26 includes exhaust gas combustion potential full state range data and exhaust gas combustion potential limited state range data.

4. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Analyze the flare combustion morphology characteristics based on the flare combustion monitoring data to generate flare combustion morphology characteristic data; Step S32: Analyze the torch combustion morphology pattern based on the torch combustion morphology characteristic data to generate torch combustion morphology pattern data; Step S33: Based on the exhaust gas combustion potential state range data and the flare combustion morphology pattern data, perform combustion morphology correlation deviation and combustion efficiency response relationship analysis under the combustion potential state range constraint, and generate combustion morphology correlation deviation-efficiency response relationship data. Step S34: Design a mapping index for flare combustion mode and combustion efficiency based on the combustion mode correlation deviation-efficiency response relationship data, and generate a flare mode-efficiency mapping index. Step S35: Perform explicit analysis of flare combustion efficiency on the flare combustion monitoring data based on the flare shape-efficiency mapping index to generate explicit flare combustion efficiency data.

5. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 4, characterized in that, Step S33 includes the following steps: Step S331: Analyze the correlation deviation characteristics between the flare combustion mode and the exhaust gas combustion potential state interval using the exhaust gas combustion potential state interval data, and generate combustion mode correlation deviation characteristic data of the combustion potential state. Step S332: Perform combustion efficiency sensitivity analysis on the combustion mode correlation deviation feature data of combustion potential state to generate combustion mode correlation deviation feature efficiency sensitivity data. Step S333: Based on the efficiency sensitivity data of combustion mode correlation deviation characteristics, analyze the relationship between combustion mode correlation deviation and combustion efficiency response under the constraint of combustion potential state interval, and generate combustion mode correlation deviation-efficiency response relationship data.

6. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 3, characterized in that, Step S4 includes the following steps: Step S41: Perform trend feature analysis on the explicit data of flare combustion efficiency to generate explicit trend feature data of flare combustion efficiency; Step S42: Based on the characteristics of the apparent trend of flare combustion efficiency, conduct combustion efficiency trend anomaly assessment processing to generate combustion efficiency trend anomaly assessment data; Step S43: Analyze the characteristics of combustion failure precursors based on the abnormal combustion efficiency trend assessment data, and generate combustion failure precursor characteristic data; Step S44: Analyze combustion failure-related features of combustion failure precursor features using exhaust gas combustion potential state range data to generate combustion failure-related feature data.

7. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 6, characterized in that, The combustion failure-related feature data mentioned in step S44 includes combustion failure-non-fuel-limited feature data or combustion failure-fuel-limited feature data.

8. The method for controlling flare combustion efficiency with real-time feedback adjustment according to claim 7, characterized in that, Step S44 includes the following steps: When the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential full state interval data, the non-fuel-limited feature analysis of combustion failure is performed on the combustion failure precursor feature data to generate combustion failure-non-fuel-limited feature data; or, when the exhaust gas combustion potential state interval data corresponding to the combustion failure precursor feature data is the exhaust gas combustion potential limited state interval data, the fuel-limited feature analysis of combustion failure is performed on the combustion failure precursor feature data to generate combustion failure-fuel-limited feature data.

9. The method for controlling the flare combustion efficiency with real-time feedback adjustment according to claim 7, characterized in that, Step S5 includes the following steps: Step S51: When the combustion failure related characteristic data is combustion failure-non-fuel-limited characteristic data, analyze the combustion organization regulation relationship of the flare combustion monitoring data to generate combustion organization regulation relationship data; based on the combustion organization regulation relationship data and the combustion failure-non-fuel-limited characteristic data, analyze the minimum intervention control parameters for combustion organization regulation to generate the minimum intervention control parameters for combustion organization regulation. Step S52: When the combustion failure related characteristic data is combustion failure-fuel limitation characteristic data, analyze the minimum intervention control parameters for fuel supply regulation based on the exhaust gas combustion potential characteristic data and the combustion failure-fuel limitation characteristic data, and generate the minimum intervention control parameters for fuel supply regulation. Step S53: Design intelligent control parameters for flare combustion efficiency that are adjusted in real time by adjusting the minimum intervention control parameters through combustion organization and / or fuel replenishment; Step S54: Execute intelligent flare combustion control operation based on the intelligent control parameters of flare combustion efficiency adjusted according to real-time feedback.

10. A control system for flare combustion efficiency that can be adjusted in real time, characterized in that, For executing the control method for adjusting the flare combustion efficiency in real time as described in claim 1, the control system for adjusting the flare combustion efficiency in real time includes: The flare combustion multi-source correlation monitoring module is used to monitor and process flare combustion multi-source correlation data using multi-source monitoring sensors installed at the flare channel entrance, and generate flare combustion multi-source correlation monitoring data, wherein the flare combustion multi-source correlation monitoring data includes flare exhaust gas monitoring data, flare combustion external condition data, and flare combustion monitoring data. The exhaust gas combustion potential state interval analysis module is used to perform interval analysis of exhaust gas combustion potential state through flare exhaust gas monitoring data and flare combustion external condition data, and generate exhaust gas combustion potential state interval data. The combustion efficiency explicit analysis module is used to perform explicit analysis of flare combustion monitoring data and generate explicit flare combustion efficiency data. The combustion failure related feature analysis module is used to perform fuel-related feature analysis of combustion failure based on exhaust gas combustion potential state range data and flare combustion efficiency explicit data, and generate combustion failure related feature data. The flare combustion intelligent control module is used to design intelligent control parameters for flare combustion efficiency that are adjusted in real time based on combustion failure-related characteristic data; and to execute intelligent control operations for flare combustion based on the intelligent control parameters for flare combustion efficiency adjusted in real time.

Images