Combustion optimization method for oil and gas field based on real-time component analysis of blowout exhaust gas

Abstract

The application discloses a combustion optimization method for oil and gas fields based on real-time component analysis of blowout exhaust gas, and belongs to the field of combustion optimization for oil and gas fields.S1, a sound spectrum-FTIR dual-source collaborative online monitoring system is built to obtain a standardized blowout exhaust gas component physical property data set;S2, flare flame timing characteristics and environmental meteorological condition data are collected to output a combustion state characteristic set and an instability early warning level identifier;S3, a parameterized combustion efficiency prediction model is hierarchically constructed to output real-time combustion efficiency, destruction removal efficiency and pollutant emission concentration prediction values;S4, a multi-constraint comprehensive target optimization function is constructed, and an improved adaptive particle swarm optimization algorithm is used for solving to output optimal control parameters and issue the optimal control parameters to field station actuators; andS5, dynamic iteration is performed.The above-mentioned combustion optimization method for oil and gas fields based on real-time component analysis of blowout exhaust gas realizes adaptive optimization control of flares under all working conditions, and takes into account combustion stability, combustion efficiency and pollutant emission compliance.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field combustion optimization technology, and in particular to an oil and gas field combustion optimization method based on real-time component analysis of venting exhaust gas. Background Technology

[0002] Oil and gas field development and fracturing operations generate large amounts of flared exhaust gases containing hydrocarbons and hydrogen sulfide, which are typically treated by closed-loop incineration using surface flares. However, with increasingly stringent environmental regulations, flares are required to maintain high combustion efficiency and strictly control SO2, CO, and NO emissions. x Pollutant emissions, along with the dynamic and irregular fluctuations in the composition and calorific value of the released exhaust gas with the formation's production, easily lead to problems such as flameout, incomplete combustion, and instantaneous exceedance of pollutant standards. Traditional fixed parameter control modes are difficult to adapt to complex and ever-changing on-site conditions. Summary of the Invention

[0003] The purpose of this invention is to provide a combustion optimization method for oil and gas fields based on real-time component analysis of venting exhaust gas, thereby solving the aforementioned technical problems.

[0004] To achieve the above objectives, this invention provides a method for optimizing combustion in oil and gas fields based on real-time component analysis of venting exhaust gases, comprising the following steps: S1. Establish a dual-source collaborative online monitoring system of acoustic spectrum-FTIR. Utilize the virial equation and a special correction model for acid gas to calculate in real time the true low calorific value, average molar mass, compressibility factor, and component fluctuation intensity index of the venting exhaust gas, thereby obtaining a standardized dataset of venting exhaust gas component properties. S2. Deploy an acoustic-TDLAS laser-meteorological multi-source synchronous sensor array to collect the temporal characteristics of the torch flame and environmental meteorological conditions. Use ARIMA-IGARCH joint temporal filtering combined with EWRMS exponential weighted statistical control chart to identify the precursors of local flame instability. Integrate the component physical property parameters output from S1 to construct a comprehensive combustion stability index and classify instability warning levels. Output the combustion state feature set and instability warning level identifier. S3 integrates the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and environmental meteorological parameters to construct a unified standardized prediction feature vector, establish a set of baseline efficiency calculation equations for perpetual flame interference correction, construct a parameterized combustion efficiency prediction model in layers, and output the predicted values ​​of combustion efficiency, destruction and removal efficiency, and pollutant emission concentration in real time. S4. Using the destruction and removal efficiency and pollutant emission concentration output by S3 as optimization targets, a multi-constraint comprehensive target optimization function is constructed by combining equipment range, safe and stable combustion, and environmental protection limits. The instability warning level output by S2 is read, and the component fluctuation intensity index output by S1 is introduced to adaptively and dynamically correct the optimization target weight and adjustment step size. An improved adaptive particle swarm algorithm is used to complete the multi-objective optimization solution, output the optimal control parameters and send them to the station execution mechanism. S5 collects the actual measured values ​​of flue gas, equipment valve position feedback, and full-process operation sequence data after the execution of S4 control. It conducts a quantitative evaluation of the optimization effect from four dimensions: safety, efficiency, economy, and environmental protection. Based on the incremental weighted least squares rolling correction of the S3 hierarchical parameterized combustion efficiency prediction model, it expands the operating condition characteristics and optimal control parameters to the optimization strategy library in pairs, and dynamically iterates and updates S1, S3, and S4 to form a fully closed-loop self-iterative combustion optimization link.

[0005] Therefore, the oil and gas field combustion optimization method based on real-time component analysis of venting exhaust gas, as described above, has the following beneficial effects: 1. High measurement accuracy and strong long-term stability: The dual-source linkage drift correction mechanism of FTIR spectroscopy and ultrasonic velocity is adopted. The ultrasonic inversion model error is iteratively corrected every 30 minutes, which can eliminate the measurement deviation caused by pipeline scaling, temperature and pressure shift and component drift, and realize high-precision continuous monitoring of exhaust gas properties at the second level. 2. Strong adaptability to operating conditions and low combustion efficiency prediction error: Based on the real-time low calorific value of the released exhaust gas, three independent sub-models are divided. Each interval is equipped with a dedicated weight vector and adaptive fluctuation and instability correction coefficients to distinguish the sensitivity differences of high, medium and low calorific value exhaust gas to flame disturbance and component fluctuation, which greatly improves the accuracy of combustion efficiency prediction under wide component fluctuation conditions. 3. Low false alarm rate for flame instability identification: Based on the industrial 3σ statistical control limit and 99.73% confidence level, flame anomalies can be identified. It can distinguish between normal airflow minor disturbances and real flameout and reignition instability events. It is linked with the Combustion Stability Index (CSI) dynamic correction efficiency prediction model to achieve accurate compensation for combustion efficiency attenuation under unstable conditions. 4. Pollutant emissions can be predicted in advance and meet environmental protection control requirements: Based on the conservation of waste gas components and materials, the total molar flow rate of flue gas is accurately calculated. Combined with the amount of residual pollutants generated and meteorological diffusion correction factors, the emission concentration of various flue gas pollutants is predicted in real time. Combustion parameters are optimized with legal environmental limits as hard constraints to ensure long-term emission compliance. 5. Fully automated operation without manual intervention: The system automatically collects operating data, determines the calorific value range, switches the corresponding sub-model, identifies flame instability, and calculates efficiency and pollutant concentration. When the exhaust gas composition fluctuates across ranges, the parameter set can be seamlessly switched to adapt to the on-site operating conditions where the exhaust gas composition fluctuates irregularly and frequently. 6. The model has self-iterative update capability: each calorific value interval independently stores working condition samples, and the incremental least squares method can be used to roll-correct the model weights and correction coefficients in intervals, adapting to the long-term slow changes in exhaust gas components at different stages of oil and gas field development. The model can maintain prediction accuracy without offline recalibration.

[0006] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0007] Figure 1 This is a flowchart of the oil and gas field combustion optimization method based on real-time component analysis of venting exhaust gas according to the present invention. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit 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 creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0009] It should be noted that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0010] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0011] like Figure 1 As shown, the oil and gas field combustion optimization method based on real-time component analysis of venting exhaust gas includes the following steps: S1. Establish a dual-source collaborative online monitoring system of acoustic spectrum-FTIR. Utilize the virial equation and a special correction model for acidic gas to calculate in real time the true low calorific value, average molar mass, compressibility factor, and component fluctuation intensity index of the venting exhaust gas, thereby obtaining a standardized dataset of venting exhaust gas component properties.

[0012] S2. Deploy an acoustic-TDLAS laser-meteorological multi-source synchronous sensor array to collect the temporal characteristics of the torch flame and environmental meteorological conditions. Use ARIMA-IGARCH joint temporal filtering combined with EWRMS exponential weighted statistical control chart to identify early signs of local flame instability. Integrate the component property parameters output from S1 to construct a comprehensive combustion stability index and classify instability warning levels. Output a combustion state feature set and instability warning level identifier.

[0013] S3 integrates the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and environmental meteorological parameters to construct a unified standardized prediction feature vector, establishes a baseline efficiency calculation equation set for permafrost interference correction, and constructs a parameterized combustion efficiency prediction model in layers to output predicted values ​​of combustion efficiency, destruction and removal efficiency, and pollutant emission concentration in real time.

[0014] S4. Using the destruction and removal efficiency and pollutant emission concentration output by S3 as optimization objectives, a multi-constraint comprehensive objective optimization function is constructed by combining equipment range, safe and stable combustion, and environmental protection limits. The instability warning level output by S2 is read, and the component fluctuation intensity index output by S1 is introduced to adaptively and dynamically correct the optimization objective weight and adjustment step size. An improved adaptive particle swarm algorithm is used to complete the multi-objective optimization solution, output the optimal control parameters, and send them to the station execution mechanism.

[0015] S5 collects the actual measured values ​​of flue gas, equipment valve position feedback, and full-process operation sequence data after the execution of S4 control. It conducts a quantitative evaluation of the optimization effect from four dimensions: safety, efficiency, economy, and environmental protection. Based on the incremental weighted least squares rolling correction of the S3 hierarchical parameterized combustion efficiency prediction model, it expands the operating condition characteristics and optimal control parameters to the optimization strategy library in pairs, and dynamically iterates and updates S1, S3, and S4 to form a fully closed-loop self-iterative combustion optimization link.

[0016] In step S1, acoustic spectrum-FTIR dual-source collaborative online monitoring equipment is deployed in the main venting straight pipe section downstream of the flare separator and upstream of the flare head in the oil and gas field. The acoustic spectrum-FTIR dual-source collaborative online monitoring equipment includes an integrated ultrasonic gas monitoring unit and an FTIR full-component spectral analysis unit. Among them, the ultrasonic gas monitoring unit is used to continuously collect the original dynamic operating condition data in the main venting straight pipe, including the volumetric flow rate of the venting exhaust gas. Pipeline static pressure Pipeline thermodynamic temperature and the speed of sound propagation of ultrasonic waves in gas The FTIR full-component spectral analysis unit is used to detect the true mole fraction of combustible heavy hydrocarbon components such as methane, ethane, propane, butane, and pentane, acidic combustible components such as hydrogen sulfide, and inert dilution components such as carbon dioxide and nitrogen in the venting exhaust gas, and outputs the standardized mole fraction sequence of each component. .

[0017] Step S1 specifically includes the following steps: S11, Based on the standardized mole fraction sequence of each component The average molar mass of the venting exhaust gas was inverted in real time using a first-order approximation model of the virial equation of state.

[0018] S111, Based on the standardized mole fraction sequence of each component Weighted calculation of the initial adiabatic index of the vented exhaust gas : ; In the formula: The equivalent adiabatic index of the exhaust gas; For the first The mole fraction of each gaseous component; For the first Standard adiabatic index of a pure component gas; This represents the total number of components detected in the emitted exhaust gas.

[0019] S112. Introduce the theoretical formula for the gas sound velocity from the first-order approximation of the virial equation of state, and establish the mapping relationship between the sound velocity and the average molar mass: ; In the formula: This is the universal gas constant; Let be the average molar mass of the emitted exhaust gas to be solved.

[0020] S113. By mathematically transforming the theoretical formula for the velocity of sound in gas in S12, the real-time formula for calculating the average molar mass is derived: ; By iteratively solving the average molar mass every 2 seconds using real-time updates of sound velocity, temperature, and adiabatic index, the rate of property updates under dynamic fluctuation conditions is significantly improved compared to traditional static chromatographic calculations. The final output is the real-time, accurate average molar mass of the emitted exhaust gas. .

[0021] S12, Based on the standardized mole fraction sequence of each component The Kay mixing rule was used to calculate the original pseudo-critical pressure of the venting exhaust gas. and pseudocritical temperature : ; ; In the formula: For the first Standard critical pressure of pure components; For the first Standard critical temperature for a pure component.

[0022] S13. Considering the characteristics of oil and gas field venting exhaust gases containing acidic components such as H2S and CO2, a Wichert-Aziz acidic gas correction model is introduced to correct the critical parameter offset error: .

[0023] In the formula: Correcting temperature offset for acidic gases; The mole fraction of hydrogen sulfide component is dimensionless. This represents the mole fraction of the carbon dioxide component.

[0024] S14, Original pseudo-critical pressure of exhaust gas based on S12 and pseudocritical temperature And the acid gas correction temperature offset of S13 Solve for the corrected pseudocritical temperature and pseudocritical pressure : ； .

[0025] S15. Calculate the relative pressure and relative temperature of the exhaust gas from the detonation jet: ; ; In the formula: The relative pressure of the gas; The relative temperature of the gas is denoted as .

[0026] S16, will , Input the Beggs-Brill iterative model and set the iteration convergence accuracy. Iteratively solve the real gas compressibility factor .

[0027] S17, Based on the standardized mole fraction sequence of each component and the real gas compressibility factor of S16 The ideal low calorific value and the actual low calorific value of the venting exhaust gas are calculated using a layered weighted method, distinguishing the calorific value contributions of combustible components, acidic combustible components, and inert components.

[0028] S171. Classify and define the calorific value of all components in the exhaust gas, dividing them into high-calorific-value hydrocarbon components, low-calorific-value acidic combustible components, and zero-calorific-value inert components. Calculate the ideal low-calorific-value under standard operating conditions by performing a weighted summation based on the mole fraction sequence in S113. ; In the formula: For the first under standard conditions Ideal low calorific value for a pure component; For the first The standard low calorific value of a pure component.

[0029] S172, Based on Real Gas Compressibility Factor High-pressure condition correction was performed on the ideal low heating value of S171 to obtain low heating value parameters that match the actual operating conditions of the pipeline in the field: ; In the formula: The low calorific value is based on the actual working conditions of the exhaust gas release site.

[0030] S173. Statistical solution of key component characteristic proportion parameters: ; ; In the formula: The total molar percentage of inert components in the exhaust gas; The total molar percentage of heavy hydrocarbon combustible components in the exhaust gas; The molar fraction of nitrogen in the exhaust gas; The mole fraction of water component in the exhaust gas; The molar fraction of ethane in the exhaust gas; The molar fraction of propane in the exhaust gas; The total mole fraction of butane in the exhaust gas; It represents the total mole fraction of pentane in the venting exhaust gas; used to characterize the degree of dilution and combustion activity level of the venting exhaust gas.

[0031] S18, Time-series real low-calorific-value sequence based on the output of S172 A sliding window coefficient of variation model was constructed to calculate the component fluctuation intensity index in real time and quantify the dynamic fluctuation level of the emitted exhaust gas.

[0032] S181. Construct a 5-minute fixed-length time-series sliding window and extract the continuous output of S172. Time series data, as a dataset for fluctuation feature analysis, balances the sensitivity of fluctuation identification with the ability to resist interference.

[0033] S182. Calculate the mean and standard deviation of the true low calorific value within the sliding window, respectively. ; ; In the formula: The average low heat value within the sliding window; The standard deviation of the lowest heating value within the sliding window; This represents the total number of data samples taken within the sliding window. for Real-time, accurate low calorific value.

[0034] S183. Based on the mean and standard deviation output from S182, calculate the component fluctuation intensity index: ; In the formula: The intensity index of fluctuation in the composition of the emitted exhaust gas.

[0035] S184. Classify the fluctuation level of the working condition based on the fluctuation intensity index output by S183: For stable operating conditions, For fluctuating operating conditions, This is a condition characterized by severe fluctuations.

[0036] S19. Integrate all output parameters from S11-S18 to construct a standardized emission exhaust gas component property dataset; the standardized emission exhaust gas component property dataset includes real-time average molar mass. Real gas compressibility factor Ideal low calorific value True low calorie value Inert component ratio The proportion of heavy hydrocarbon components Component fluctuation intensity index And operating condition fluctuation level.

[0037] Step S2 specifically includes the following steps: S21. Construct an acoustic-TDLAS laser-meteorological multi-source synchronous sensing array to collect the temporal characteristics of the torch flame and environmental meteorological conditions, and obtain the raw combustion time series dataset.

[0038] S211. To address the monitoring needs of oil and gas field flare vents in open-air environments with strong interference, smoke obstruction, and day-night uniformity, a standardized deployment of multi-source sensor arrays was completed. Four acoustic microphones were evenly deployed at 90° circumference around the flare tube, installed at the same height as the flare nozzle and 5m-8m away from the outer wall of the flare, to capture the acoustic characteristics of flame turbulence pulsation, localized flameout, and reignition throughout the entire area. A TDLAS tunable diode laser absorption spectroscopy optical path was horizontally installed in the core stable combustion zone of the perpetual flame 0.5m above the flare nozzle to monitor the temporal changes in core flame temperature using water vapor characteristic absorption lines. An integrated meteorological station was deployed in an open, unobstructed area of ​​the flare platform to simultaneously collect ambient wind speed, ambient temperature, and atmospheric static pressure parameters. All sensors were uniformly connected to the IEEE1588 precision clock synchronization system at the site, with a full-channel timing stamp error ≤10ms, achieving millisecond-level precise alignment of multi-source data.

[0039] S212. Perform pre-processing noise reduction on the original sensor signal acquired in S211 to remove power frequency interference, baseline drift and pulse noise anomalies; use notch filtering algorithm to eliminate power grid interference for 20kHz high-frequency acoustic time-series signal, and use moving average filtering to remove instantaneous pulse noise for TDLAS temperature time-series signal, while retaining the true fluctuation characteristics of flame.

[0040] S213. The preprocessed multi-source sensor data is time-series aggregated and resampled at a 1-second time granularity to unify the data output frequency and achieve a perfect second-level time series match with the component property dataset output in S1; the final output is a time-series aligned original combustion state dataset; the original combustion state dataset includes acoustic energy time series, standardized flame temperature time series data, and ambient wind speed. Ambient temperature and ambient atmospheric static pressure .

[0041] S22. Based on the standardized flame temperature time series data output from S21, the ARIMA(2,1,1)-IGARCH(1,1) joint time series filtering algorithm is used to complete the non-stationary time series detrending, autocorrelation elimination and time-varying variance fitting, and output a standardized stationary residual sequence.

[0042] S221. Select the flame temperature timing and acoustic energy timing after S21 polymerization as the original timing sequence to be processed. To address the inherent characteristics of non-stationarity, strong autocorrelation, and trend drift in torch combustion time series, first-order difference operations are employed to eliminate low-frequency trend terms. The formula for calculating the first-order time difference is as follows: ; In the formula: for The first-order difference sequence of the original time series to be processed; for The original timing sequence value to be processed at any given moment; for The original timing sequence value to be processed at any given moment.

[0043] S222. Construct an ARIMA(2,1,1) difference autoregressive moving average model, and perform time series fitting on the first-order difference sequence obtained in S221 to eliminate the autocorrelation characteristics of the sequence and extract the unbiased residual sequence; the fitting formula of the ARIMA(2,1,1) model is as follows: ; In the formula: This is the mean constant term of the differential time series, representing the time series baseline offset; is the first-order autoregressive coefficient, representing the linear influence weight of the difference data from the previous time step on the current data; is the second-order autoregressive coefficient, representing the weight of the linear influence of the difference data from the previous two time steps on the current data; The coefficient of the first-order moving average represents the correction weight of the residual at the previous time step on the current time series. The model fitting residuals from the previous time step; This represents the original fitting residuals output by the model at the current time. for The first-order difference sequence of the original time series to be processed; for The first-order difference sequence of the original time series to be processed.

[0044] S223, The residual sequence output by S222 The combustion time series exhibits typical characteristics such as time-varying fluctuations, heteroscedasticity, and fluctuation clustering. An IGARCH(1,1) integral generalized autoregressive conditional heteroscedasticity model is constructed to fit the real-time fluctuation variance of the residuals, capturing the drastic fluctuation characteristics of pre-flameout precursors. The IGARCH(1,1) time-varying variance calculation formula is as follows: ; Hard constraints of the model: ; In the formula: for The real-time conditional variance of the time-residual sequence characterizes the intensity of instantaneous fluctuations in the flame. It is the basic variance constant, which characterizes the inherent noise variance of the system; This is the weighting coefficient of the squared residuals for the current period, representing the contribution ratio of instantaneous fluctuations; The historical variance memory weighting coefficient represents the ability of historical fluctuations to have a lasting impact. for Real-time conditional variance of the time-residual sequence.

[0045] S224. Based on the original residuals from S222 and the real-time conditional variance from S223, time series standardization is performed to generate a stationary white noise sequence with zero mean, unit variance, and no autocorrelation. The time series standardization calculation formula is as follows: ; In the formula: for Time-standardized stationary residual sequence; for The time-residual conditional standard deviation, and .

[0046] This step ultimately outputs a standardized stationary residual sequence. This completely solves the shortcomings of traditional fixed threshold monitoring, such as high false alarm rate and inability to identify weak signs of instability.

[0047] S23. Based on the stationary residual sequence output by S22, the EWRMS exponentially weighted root mean square statistical control chart algorithm is used to realize the identification of early signs of local flame instability and the quantitative statistics of alarm frequency.

[0048] S231, the standardized stationary residual sequence output from S22 Batch aggregation noise reduction is performed to reduce single-point random noise interference and improve the stability of instability identification; and the batch mean aggregation calculation formula is as follows: ; In the formula: For the first The mean residual of each batch of data; This refers to the sampling capacity per batch.

[0049] S232. An exponentially weighted dynamic update mechanism is introduced to construct the EWRMS time-series volatility characteristic quantity, preserving the cumulative effect of historical volatility and accurately capturing the slowly evolving instability trend; and the iterative calculation formula of the EWRMS characteristic quantity is as follows: ; In the formula: For the first The batch's exponentially weighted root mean square volatility characteristic value indicates that the larger the value, the higher the degree of flame instability. The time-series smoothing weighting coefficient is fixed at 0.2, balancing sensitivity to instantaneous fluctuations with memory of historical trends. This is the index-weighted root mean square volatility characteristic of the previous batch.

[0050] S233. Based on big data of stable combustion conditions, calibrate statistical control limits, set an industrial threshold with a confidence level of 99.73%, and determine the unstable state. The formula for determining the control limit is as follows: ; Decision logic: When When a partial flameout or reignition instability event is detected in the current batch, a single instability alarm is triggered.

[0051] In the formula: This is the upper limit for statistical control; The threshold is set as the statistical control lower limit; the threshold is set based on the normal distribution 3σ criterion, and the corresponding false alarm rate of the system is only 0.27%, which meets the requirements of ultra-low false alarm in industry.

[0052] S234. Count the number of instability alarm triggers per unit time to quantify the severity of flame dynamic instability; and the formula for calculating the instability alarm frequency is as follows: ; In the formula: The frequency of instability alarms per second; The total number of instability alarms within a given time period; The duration is fixed at 1 second.

[0053] S24, combined with the component property parameters output from S1, the combustion instability state parameters output from S23, and the environmental meteorological condition parameters output from S21, calculate the momentum flux ratio. Combustion stability index : ; ; In the formula: This is the ratio of the momentum flux of the nozzle airflow to the ambient crosswind; the larger the value, the stronger the flame's resistance to crosswind interference. To determine the real-time density of the emitted exhaust gas, and ; The velocity of the exhaust gas released from the flare nozzle; The ambient air density, and , The average molar mass of dry air. The ambient air thermodynamic temperature; The comprehensive combustion stability index has a range of values. The closer the value is to 1, the better the combustion stability. This is the maximum alarm reference value for the critical flameout stage; The minimum calorific value threshold for safe combustion of the torch; The minimum momentum flux ratio for flame wind resistance safety; All weights are determined by the analytic hierarchy process (AHP), prioritizing the real-time instability of the flame, followed by the calorific value of the fuel, and finally the impact of environmental wind disturbance.

[0054] right The output results are adjusted for boundary constraints, limiting the numerical range to [value missing]. This eliminates the problem of calculation overflow under extreme working conditions and ensures the stability and reliability of the evaluation system.

[0055] S25, based on the output of S24 and S23 output Establish a four-level combustion instability early warning classification system and output standardized early warning level labels: Normal combustion: and The combustion rate is 10 times per second, with steady-state combustion, no localized flameout, and stable combustion efficiency.

[0056] Level 1 Warning (Mild Instability): and The number of times per second resulted in sporadic localized flameouts and a slight deterioration in the combustion state.

[0057] Level 2 Warning (Moderate Instability): and The combustion rate is 10 times per second, with frequent localized flameouts and a significant decrease in combustion efficiency.

[0058] Critical shutdown warning (severe instability): and At times per second, the flame is on the verge of complete extinguishment, posing a significant risk of escape of unburned hydrocarbons.

[0059] Output a unique instability warning level based on the real-time dual-indicator matching results. .

[0060] S26. Integrate all output parameters from S21 to S25 to construct a standardized flare combustion state feature set; the standardized flare combustion state feature set includes time-series fluctuation features. Frequency of instability alarms Momentum flux ratio Combustion stability index and instability warning level .

[0061] Step S3 specifically includes the following steps: S31. Perform feature cleaning and normalization on the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and the collected and stored environmental meteorological raw operating parameters to construct a unified standardized prediction feature vector.

[0062] S311. Using the unified IEEE 1588 time stamp as the matching benchmark, complete the precise time alignment of multi-source data; retrieve the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and the collected and stored environmental meteorological raw operating condition parameters, complete the data matching at a time granularity of 1 second, and remove abnormal time series samples with missing data, sensor fault marks, or time series offsets exceeding 10ms, to obtain time-synchronized complete multi-source raw feature samples.

[0063] S312. Perform extreme value standardization on all original features after time-series alignment in S311 to eliminate dimensional differences and unify the feature value range to... .

[0064] S313. Concatenate all the normalized features from S312 in a fixed order to construct a unified, standardized prediction feature vector with fixed dimensions. .

[0065] S32. Separate the main exhaust jet stream from the auxiliary fuel component flow of the pilot flame, establish a set of calculation equations for the pilot flame interference correction baseline efficiency, and eliminate the superimposed interference of the pilot flame fuel on the calculation of the basic combustion efficiency of the flare.

[0066] S321, Based on the volumetric flow rate of the exhaust gas output from S1 Average molar mass Real-time calculation of the total molar flow rate of the main exhaust combustible components : ; S322. Real-time collection of fuel volume flow rate of the pilot flame gas supply branch flow meter. Based on the fixed physical properties of the propellant fuel (pure methane), the molar flow rate of the auxiliary fuel for the propellant flame was calculated. : ; In the formula: Methane density under operating conditions in the gas supply pipeline for the permanent flame; Real-time volumetric flow rate of the long-burning fire branch; This represents the molar mass of pure methane.

[0067] S323. Calculate the theoretical complete combustion oxygen consumption of the main exhaust gas and the pilot flame fuel respectively, distinguish the combustion energy contribution of the two types of fuel, and construct a set of equations for the total combustion efficiency of the uninterrupted baseline theory: ; In the formula: The theoretical oxygen consumption for the complete combustion of the main exhaust gas; For the first The theoretical number of moles of oxygen required for the complete combustion of a unit mole of a component; The theoretical oxygen consumption for the complete combustion of methane in a continuous flame. This represents the theoretical oxygen consumption per mole of pure methane. This represents the total theoretical oxygen consumption of the system.

[0068] S324. Introducing a correction coefficient for interference from a persistent flame. By removing the contribution of fuel combustion from the propellant flame, a corrected theoretical efficiency benchmark value is obtained that only characterizes the combustion capacity of the exhaust gas itself. ; in, ; In the formula: The baseline theoretical combustion efficiency of the exhaust gas after eliminating interference from the propellant flame; The uncorrected, raw baseline combustion efficiency, including contributions from the propellant fuel, is calculated based on the matching degree between theoretical oxygen consumption and ambient oxygen supply.

[0069] S33, based on S1, outputs true low calorific value. The system divides the space into stratified intervals and constructs parameterized combustion efficiency prediction sub-models for different three-layer low calorific value stratified intervals to adapt to wide-component fluctuating exhaust gas conditions.

[0070] S331. Based on the safe combustion calorific value boundary of oil and gas field venting exhaust gas, divide the area into three low-calorific-value stratified zones: High calorific value stable range: It has a high proportion of heavy hydrocarbons and strong combustion activity.

[0071] Medium calorific value fluctuation range: The composition fluctuates frequently and is prone to local instability.

[0072] Low calorific value critical range: It has a high proportion of inert components and is extremely easy to extinguish, making it a high-risk operating condition.

[0073] S332, using the standardized predictive feature vector from S313 The reference theoretical combustion efficiency of the exhaust gas for input S324. Based on the fundamental offset term, linear parameterized combustion efficiency prediction models are established for the three low-calorific-value stratified intervals; and the general prediction formula for stratified parameterized combustion efficiency is as follows: ; In the formula: To determine the real-time predicted combustion efficiency for the lower three-layer low-calorific-value stratified range, the value range is... ; The optimal weighting coefficients are specific to the three low-calorific-value stratification intervals. Transpose of a vector, weight transpose vectors of different three-layer low-calorific-value stratification intervals Solved by interval-independent offline multivariate least squares calibration; This is the correction factor for flame instability; This is the component fluctuation correction coefficient.

[0074] S333, outputting stratified combustion efficiency to S332 By implementing boundary truncation constraints, model calculation overflow under extreme operating conditions is eliminated, and the final real-time combustion efficiency is obtained. : ; In the formula: The final real-time combustion efficiency after correction is given; the upper limit of 0.995 is the limit of perfectly ideal combustion, and the lower limit of 0.05 is the critical threshold for complete flameout. Values ​​exceeding the range are directly locked at the boundary values ​​to ensure the reasonableness of industrial calculation results.

[0075] S34, Final Real-Time Combustion Efficiency Based on S33 Output Coupled S1 exhaust gas component mole fraction sequence A multi-component destruction and removal efficiency calculation model was established to solve the destruction and removal efficiency of hydrocarbons and hydrogen sulfide respectively.

[0076] S341. Differentiate between three categories of target pollutants for disposal: heavy hydrocarbons hydrogen sulfide methane Obtain the molar flow rate of each pollutant inlet: ; In the formula: For the first Pollutant inlet molar flow rate; For the first mole fractions corresponding to pollutants; Values ​​include Three categories of controlled pollutants.

[0077] S342, combined with S33, final real-time combustion efficiency Calculate the outlet molar flow rate of residual pollutants after combustion and destruction: ; In the formula: For the first Molar flow rate of pollutants that escape from the outlet due to incomplete combustion.

[0078] S343. Define pollutant removal and destruction efficiency based on inlet and outlet pollutant molar flow rates. ; In the formula: For the first Efficiency of pollutant destruction and removal; the closer the value is to 100%, the more thorough the pollutant is destroyed by combustion. S344, Substituting S342 and simplifying, we get... .

[0079] By combining combustion efficiency, environmental meteorology, and component physical properties, a parameterized prediction model for multi-pollutant emission concentrations is constructed, and the predicted values ​​of flue gas pollutant emissions are output in real time.

[0080] S351, Total molar flow rate of main exhaust combustible components based on S1 S2 ambient wind speed and ambient atmospheric static pressure Calculate the flare gas dilution and diffusion correction factor: ; In the formula: The higher the wind speed and the greater the temperature difference between the flame and the environment, the higher the degree of flue gas dilution. This is the field diffusion calibration constant; The real-time flame core temperature is obtained from the TDLAS tunable diode laser absorption spectroscopy optical path.

[0081] S352. Based on the residual amount of pollutants after pollutant destruction, the total molar amount of flue gas generated, and the diffusion correction factor, predict the equivalent flue gas pollutant emission concentration at the chimney: ; In the formula: For the first Predicted emission concentrations of pollutants; The total molar flow rate of flue gas generated after complete combustion of fuel, and , , The original molar flow rate of the exhaust gas. This is the excess air-fuel ratio coefficient. This represents the theoretical molar consumption of oxygen.

[0082] S36. Integrate all calculated output parameters from S31-S35 to construct a combustion efficiency and pollutant concentration prediction feature dataset; the combustion efficiency and pollutant concentration prediction feature dataset contains standardized prediction feature vectors. Long-lasting flame interference correction reference efficiency Real-time combustion efficiency Methane destruction and removal efficiency Total hydrocarbon destruction and removal efficiency Hydrogen sulfide destruction and removal efficiency Methane predicted emission concentration Total hydrocarbon predicted emission concentration Predicted emission concentration of hydrogen sulfide Predicted emission concentrations of sulfur dioxide Predicted carbon monoxide emission concentration and predicted nitrogen oxide emission concentrations .

[0083] In step S332, the weight transpose vector The expression to be solved is as follows: ; in, ; ; In the formula: To correspond to the historical standardized prediction feature matrix of the three-layer low calorific value stratification interval, each row of the matrix stores one S313 output standardized prediction feature vector. ; for Transpose of; This is a sample set of actual combustion efficiency measured by on-site flue gas monitoring in this section; , These are the comprehensive combustion stability index and the component fluctuation intensity index for historical samples within the interval, respectively. The three-layer low-calorific-value stratified intervals are numbered. Represents the critical range of low calorific value. Represents the range of fluctuations in medium calorific value. This represents the stable range of high calorific value; for Basic calibration constants; A sensitive correction factor for instability in the three-layer low-calorific-value stratification interval; For the three-layer low calorific value stratification intervals Threshold; for Basic calibration constants; A sensitive factor for component fluctuations in the three-layer low-calorific-value stratified interval; This is the real-time component fluctuation intensity index; This is the threshold for the maximum permissible component fluctuation intensity at the site.

[0084] Step S4 specifically includes the following steps: S41. Read the full-dimensional prediction output data of S3, the inherent boundary parameters of the equipment, and the environmental protection national standard limit parameters to build a multi-objective optimization input benchmark set and constraint boundary set.

[0085] S411, Read the real-time destruction and removal efficiency output by S3. Predicted emission concentrations of multiple pollutants , , Real-time combustion efficiency Synchronously read the component fluctuation intensity index output by S1 True low calorie value Read the instability warning level output by S2. Combustion stability index All parameters are aligned with a 1-second timing granularity to ensure that the optimized input is synchronized with the on-site working conditions in real time.

[0086] S412. Input the inherent physical range constraints of the station's hardware equipment to construct a set of hard safety constraints for the equipment; fix and input the flow range of the combustion-supporting medium. Flare nozzle velocity range Flow rate of fuel for the long-lasting flame .

[0087] At the same time, the emission limits for sulfur dioxide, carbon monoxide, and nitrogen oxides are recorded. , , Enter the lowest stable combustion safety low calorific value of the torch. Minimum safe momentum flux ratio .

[0088] S413. Constructing an optimized vector of control variables .

[0089] In the formula: Let be the vector of control variables to be optimized; Real-time flow control of the combustion-supporting medium; The velocity of the exhaust gas released from the flare nozzle; Real-time flow control of fuel for the permanent flame; This is the matrix transpose symbol.

[0090] S42. Construct a multi-objective comprehensive optimization objective function that takes into account stable combustion efficiency, economical operation, and environmental protection and emission reduction, and superimpose multiple layers of hard constraints to form a complete optimization mathematical model.

[0091] S421. Establish three major optimization sub-objectives: maximize destruction and removal efficiency, minimize combustion medium energy consumption, and minimize comprehensive pollutant emissions. ; In the formula: For the target item of combustion efficiency deviation, the smaller the value, the better. The closer to the optimal target value; The system is preset with an optimal destruction efficiency target, fixed at 98%. For energy consumption and economic targets, the smaller the value, the lower the consumption of combustion medium and the better the operating economy. This refers to the maximum flow rate of the combustion-supporting medium. For environmental emission targets, the smaller the value, the lower the pollutant emissions; The weighted average emission concentration of three pollutants: sulfur dioxide, carbon monoxide, and nitrogen oxides. It is the weighted average limit of the national standards for pollutants.

[0092] S422. Construct the initial fixed-weight comprehensive optimization objective function: ; In the formula: Based on the comprehensive objective function value, The core optimization criterion; For efficiency, economy, and environmental protection, the initial fixed weighting coefficients are satisfied. The initial calibration values ​​are respectively Prioritize combustion safety and efficiency while also considering economic and environmental performance.

[0093] S423. Establish a complete set of hard constraint equations for the operating conditions: ; In the formula: The standard limit concentration for sulfur dioxide; This refers to the standard limit concentration of carbon monoxide. This refers to the standard limit concentration of nitrogen oxides.

[0094] S43, Instability warning level based on S2 output Matching the initial baseline adjustment step size: ; In the formula: The baseline adjustment step size represents the maximum correction ratio of the variable in a single iteration. The higher the instability level, the larger the baseline step size and the faster the system response speed.

[0095] S44, Introducing the component fluctuation intensity index from the output of S1 This enables adaptive dynamic correction of optimized weights and adjustment step sizes, adapting to conditions with drastic fluctuations in components.

[0096] S441, Constructing the Component Fluctuation Intensity Index The formula for dynamic adjustment of driving weights: ; In the formula: The efficiency, economic, and environmental weighting coefficients are dynamically adjusted. This is the weighting correction coefficient, fixed at 0.8; when When the load increases or operating conditions fluctuate more, the system automatically increases the weight of combustion safety and decreases the weight of energy consumption, prioritizing stable combustion; under stable operating conditions, it automatically reverts to the energy-saving optimization logic.

[0097] S442. Constructing an adaptive step size correction formula for wave intensity: ; In the formula: Adjust the step size for final adaptive optimization; This is the step size amplification factor, with a fixed value of 1.5.

[0098] S443. Based on the revised dynamic weights, update the comprehensive objective function to obtain the final adaptive optimization objective model: ; In the formula: It is the adaptive dynamic synthesis objective function.

[0099] S45. An improved adaptive particle swarm optimization algorithm is used to iteratively optimize the dynamic optimization model and solve for the global optimal control parameter vector.

[0100] S451. Initialize particle swarm population parameters and set basic iteration conditions: population size. Maximum number of iterations Convergence accuracy threshold Particle dimension and S41 optimization control variable vector Dimensions are consistent.

[0101] S452. Introduce a linearly decreasing inertia weight strategy to balance the algorithm's global search capability with its local fine-tuning optimization capability: ; In the formula: For the first Real-time inertia weights in each iteration; The initial maximum inertia weight is used to enhance the global large-scale search; To minimize the inertia weight at the end of the iteration, local fine-grained optimization is enhanced; This represents the current iteration number.

[0102] S453. Introduce a fixed learning factor to construct an improved formula for particle velocity and position update: ; In the formula: For the first The second iteration The velocity of each particle; For the first The second iteration The velocity of each particle; For the first The second iteration The position of each particle (the solution of the control variable); To fix the learning factor; for Uniform random numbers; For each individual particle, the historical optimal solution is provided. This is the globally optimal solution for the population. For the first The second iteration The position of each particle.

[0103] S454. The penalty function method is used to handle the multiple constraints in S423, penalizing out-of-bounds particles and guiding the population to converge to the feasible region; and the penalty fitness formula is as follows: ; In the formula: This is a fitness value with a penalty. The penalty coefficient is set to 1000; For the first The out-of-bounds deviation of a constraint condition, when the constraint is fully satisfied. .

[0104] S455. Set the iteration termination condition and complete the optimal solution output: when the change in the global optimal objective function value is less than the convergence accuracy for 20 consecutive iterations. or reaching the maximum number of iterations. When the iteration terminates, output the current global optimal control variable vector for the population: ; In the formula: To achieve the optimal flow rate of the combustion-supporting medium; To achieve the optimal flare nozzle velocity; The optimal fuel flow rate for the continuous flame.

[0105] S46. The optimal control variable vector output from S45 The signal is converted into a control valve opening command signal that the station can recognize. The control command is then sent in real time to the combustion medium control valve, the frequency conversion pressure regulating mechanism of the venting pipeline, and the pilot flame gas control valve via the Modbus TCP / IP industrial protocol. The valve position feedback and flow feedback signals of each actuator are collected in real time to verify the command execution deviation. If the deviation exceeds 5%, a secondary calibration is performed.

[0106] Step S5 specifically includes the following steps: S51. Time-series aligned acquisition of full-dimensional on-site measured true data and equipment operation feedback data after S4 regulation, constructing a closed-loop iterative original sample dataset.

[0107] Using the unified 1s time stamp of S1–S4 as a benchmark, the steady-state window period after the completion of S4 regulation is delayed to avoid interference from transient data distortion; the steady-state judgment time is set to 30s, and only stable operation data after regulation is stabilized are collected to ensure the authenticity, validity and steady-state consistency of the samples.

[0108] The measured true values ​​of flue gas after regulation are collected using an online flue gas analyzer at the site to obtain the baseline true values ​​parameters required for model calibration; among them, the measured true values ​​of flue gas include dry basis oxygen concentration. Dry basis carbon dioxide concentration Dry basis carbon monoxide concentration and dry basis total hydrocarbon concentration , serves as the sole truth label for the model correction of combustion efficiency and destruction / removal efficiency.

[0109] Collect real-time feedback data from S4 and simultaneously collect the actual operating flow rate of the combustion-supporting medium. Actual flow velocity at the torch nozzle Actual fuel flow rate of the pilot flame Feedback on the actual valve position opening of each regulating valve.

[0110] Synchronous backtracking retrieves the full-dimensional temporal feature parameters of S1, S2, and S3 corresponding to this round of optimization. , , , , , , , , , Construct a closed-loop iterative original sample library.

[0111] S52. Based on the measured true value data of flue gas from S51, the true baseline combustion efficiency and true destruction and removal efficiency are recalculated using the S32 perpetual flame interference correction algorithm, and used as the true value benchmark for evaluating the optimization effect.

[0112] S521 reuses the S32 shunt correction logic to strip the live flame interference from the measured flue gas components in S51, separating the combustion contribution of the exhaust gas and the live flame to obtain effective true flue gas component parameters: ; ; In the formula: The effective true carbon dioxide dry basis concentration after stripping away the interference of the propellant flame. The effective dry basis oxygen concentration after correction; This represents the atmospheric standard oxygen volume fraction.

[0113] S522. Based on the corrected effective flue gas composition, calculate the true value of the actual combustion efficiency after this round of regulation: ; In the formula: This represents the true value of the actual combustion efficiency on-site.

[0114] S523. Calculate the true value of the actual destruction and removal efficiency for this round of operation: ; In the formula: To determine the actual efficiency of destruction and removal under real-world working conditions; and All are global calibration linear fitting constants, and .

[0115] S53. Quantitatively evaluate the actual implementation effect of the S4 adaptive optimization strategy from four dimensions: safe and stable combustion, combustion efficiency, economical operation, and environmental protection and emission reduction. ; ; ; ; In the formula: For the improvement rate of combustion stability index; To optimize the pre-combustion stability index; The combustion stability index after S4 regulation; To optimize destruction efficiency and increase output; To optimize the actual destruction efficiency; This represents the actual destruction efficiency after S4 regulation; To reduce the unit consumption rate of combustion-supporting medium; This represents the historical average combustion-supporting consumption under a fixed-ratio strategy. To achieve the overall pollutant emission reduction rate; The average emission concentration of pollutants before optimization; This represents the average emission concentration of pollutants after S4 regulation.

[0116] S54. Based on the incremental weighted least squares method, online rolling correction of the S3 hierarchical parameterized prediction model is realized, and the model self-iterative upgrade is completed.

[0117] S541. Construct a time-weighted sample set, and the formula for calculating the sample weights is as follows: ; In the formula: For single-sample iteration weights, newer samples have higher weights; This is the time decay coefficient, with a fixed value of 0.02; This represents the time interval between the sample and the current time.

[0118] S542. Construct an incremental weighted least squares objective function to incrementally update the coefficients of the three-layer low-calorific-value stratified interval model in S3 interval by interval; and the incremental correction objective function is as follows: ; In the formula: The model residuals are weighted sum of squares; online parameter correction is achieved by minimizing the weighted residuals, without the need for offline retraining and without affecting online operation.

[0119] S543, Update the weight vector coefficients of the S33 hierarchical model interval by interval. Instability correction coefficient Fluctuation correction coefficient This ensures that different three-layer low calorific value stratified interval models continuously adapt to the latest field conditions, completely solving the problem of accuracy decay of traditional fixed models over time.

[0120] S544. Verify the correction accuracy using the Mean Absolute Error (MAE) as the verification index: If If the error is within acceptable limits, the correction is deemed effective, and the model is officially replaced and put online; if the error exceeds the limit, a second fine-tuning iteration correction is triggered.

[0121] S55. Integrate the steady-state operating condition characteristics, fluctuation characteristics, and combustion state characteristics of this round into an operating condition feature fingerprint vector. .

[0122] And bind the optimal control parameter vector output by S4 for this round. This forms a working condition fingerprint-optimal control paired sample unit.

[0123] S56. Calculate cosine similarity. To achieve deduplication and update iteration of the strategy library: when If the condition is determined to be of the same type, the old control strategy will be updated accordingly; when If a new operating condition is identified, it will be added to the inventory, and the strategy will continue to expand the scope of operating conditions covered.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A combustion optimization method for oil and gas fields based on real-time component analysis of venting exhaust gas, characterized in that: Includes the following steps: S1. Establish a dual-source collaborative online monitoring system of acoustic spectrum-FTIR. Utilize the virial equation and a special correction model for acid gas to calculate in real time the true low calorific value, average molar mass, compressibility factor, and component fluctuation intensity index of the venting exhaust gas, thereby obtaining a standardized dataset of venting exhaust gas component properties. S2. Deploy an acoustic-TDLAS laser-meteorological multi-source synchronous sensor array to collect the temporal characteristics of the torch flame and environmental meteorological conditions. Use ARIMA-IGARCH joint temporal filtering combined with EWRMS exponential weighted statistical control chart to identify the precursors of local flame instability. Integrate the component physical property parameters output from S1 to construct a comprehensive combustion stability index and classify instability warning levels. Output the combustion state feature set and instability warning level identifier. S3 integrates the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and environmental meteorological parameters to construct a unified standardized prediction feature vector, establish a set of baseline efficiency calculation equations for perpetual flame interference correction, construct a parameterized combustion efficiency prediction model in layers, and output the predicted values ​​of combustion efficiency, destruction and removal efficiency, and pollutant emission concentration in real time. S4. Using the destruction and removal efficiency and pollutant emission concentration output by S3 as optimization targets, a multi-constraint comprehensive target optimization function is constructed by combining equipment range, safe and stable combustion, and environmental protection limits. The instability warning level output by S2 is read, and the component fluctuation intensity index output by S1 is introduced to adaptively and dynamically correct the optimization target weight and adjustment step size. An improved adaptive particle swarm algorithm is used to complete the multi-objective optimization solution, output the optimal control parameters and send them to the station execution mechanism. S5 collects the actual measured values ​​of flue gas, equipment valve position feedback, and full-process operation sequence data after the execution of S4 control. It conducts a quantitative evaluation of the optimization effect from four dimensions: safety, efficiency, economy, and environmental protection. Based on the incremental weighted least squares rolling correction of the S3 hierarchical parameterized combustion efficiency prediction model, it expands the operating condition characteristics and optimal control parameters to the optimization strategy library in pairs, and dynamically iterates and updates S1, S3, and S4 to form a fully closed-loop self-iterative combustion optimization link.

2. The method for optimizing oil and gas field combustion based on real-time component analysis of venting exhaust gas according to claim 1, characterized in that: In step S1, acoustic spectrum-FTIR dual-source collaborative online monitoring equipment is deployed in the main venting straight pipe section downstream of the flare separator and upstream of the flare head in the oil and gas field. The acoustic spectrum-FTIR dual-source collaborative online monitoring equipment includes an integrated ultrasonic gas monitoring unit and an FTIR full-component spectral analysis unit. Among them, the ultrasonic gas monitoring unit is used to continuously collect the original dynamic operating condition data in the main venting straight pipe, including the volumetric flow rate of the venting exhaust gas. Pipeline static pressure Pipeline thermodynamic temperature and the speed of sound propagation of ultrasonic waves in gas The FTIR full-component spectral analysis unit is used to detect the true mole fraction values ​​of combustible heavy hydrocarbon components, hydrogen sulfide acidic combustible components, and carbon dioxide and nitrogen inert dilution components in the venting exhaust gas, and outputs the standardized mole fraction sequence of each component. .

3. The method for optimizing oil and gas field combustion based on real-time component analysis of venting exhaust gas according to claim 2, characterized in that: Step S1 specifically includes the following steps: S11, Based on the standardized mole fraction sequence of each component The average molar mass of the venting exhaust gas was inverted in real time using a first-order approximation model of the virial equation of state. S111, Based on the standardized mole fraction sequence of each component Weighted calculation of the initial adiabatic index of the vented exhaust gas : ； In the formula: The equivalent adiabatic index of the exhaust gas; For the first The mole fraction of each gaseous component; For the first Standard adiabatic index of a pure component gas; This represents the total number of components detected in the emitted exhaust gas. S112. Introduce the theoretical formula for the gas sound velocity from the first-order approximation of the virial equation of state, and establish the mapping relationship between the sound velocity and the average molar mass: ； In the formula: This is the universal gas constant; Let be the average molar mass of the emitted exhaust gas to be solved; S113. By mathematically transforming the theoretical formula for the velocity of sound in gas in S12, the real-time formula for calculating the average molar mass is derived: ； S12, Based on the standardized mole fraction sequence of each component The Kay mixing rule was used to calculate the original pseudo-critical pressure of the venting exhaust gas. and pseudocritical temperature : ； ； In the formula: For the first Standard critical pressure of pure components; For the first Standard critical temperature of pure components; S13. Considering the characteristics of oil and gas field venting exhaust gases containing acidic components such as H2S and CO2, a Wichert-Aziz acidic gas correction model is introduced to correct the critical parameter offset error: ； In the formula: Correcting temperature offset for acidic gases; The mole fraction of hydrogen sulfide component is dimensionless. The mole fraction of the carbon dioxide component; S14, Original pseudo-critical pressure of exhaust gas based on S12 and pseudocritical temperature And the acid gas correction temperature offset of S13 Solve for the corrected pseudocritical temperature and pseudocritical pressure : ； ； S15. Calculate the relative pressure and relative temperature of the exhaust gas from the detonation jet: ； ； In the formula: The relative pressure of the gas; The relative temperature of the gas; S16, will , Input the Beggs-Brill iterative model and set the iteration convergence accuracy. Iteratively solve the real gas compressibility factor ; S17, Based on the standardized mole fraction sequence of each component and the real gas compressibility factor of S16 The ideal low calorific value and the actual low calorific value of the venting exhaust gas are calculated by layer-weighted calculation, and the calorific value contribution of combustible components, acidic combustible components and inert components are distinguished. S171. Classify and define the calorific value of all components in the exhaust gas, dividing them into high-calorific-value hydrocarbon components, low-calorific-value acidic combustible components, and zero-calorific-value inert components. Calculate the ideal low-calorific-value under standard operating conditions by performing a weighted summation based on the mole fraction sequence in S113. ； In the formula: For the first under standard conditions Ideal low calorific value for a pure component; For the first The standard lower calorific value of a pure component; S172, Based on Real Gas Compressibility Factor High-pressure condition correction was performed on the ideal low heating value of S171 to obtain low heating value parameters that match the actual operating conditions of the pipeline in the field: ； In the formula: The low calorific value is based on the actual working conditions of the exhaust gas release site. S173. Statistical solution of key component characteristic proportion parameters: ； ； In the formula: The total molar percentage of inert components in the exhaust gas; The total molar percentage of heavy hydrocarbon combustible components in the exhaust gas; The molar fraction of nitrogen in the exhaust gas; The mole fraction of water component in the exhaust gas; The molar fraction of ethane in the exhaust gas; The molar fraction of propane in the exhaust gas; The total mole fraction of butane in the exhaust gas; The mole fraction of pentane in the exhaust gas; S18, Time-series real low-calorific-value sequence based on the output of S172 A sliding window variation coefficient model was constructed to calculate the component fluctuation intensity index in real time and quantify the dynamic fluctuation level of the emitted exhaust gas. S181. Construct a 5-minute fixed-length time-series sliding window and extract the continuous output of S172. Time series data, used as a dataset for fluctuation characteristic analysis; S182. Calculate the mean and standard deviation of the true low calorific value within the sliding window, respectively. ； ； In the formula: The average low heat value within the sliding window; The standard deviation of the lowest heating value within the sliding window; This represents the total number of data samples taken within the sliding window. for Real-time, accurate low calorific value; S183. Based on the mean and standard deviation output from S182, calculate the component fluctuation intensity index: ； In the formula: The intensity index of fluctuation in the composition of the emitted exhaust gas; S184. Classify the fluctuation level of the working condition based on the fluctuation intensity index output by S183: For stable operating conditions, For fluctuating operating conditions, This is a condition characterized by severe fluctuations. S19. Integrate all output parameters from S11-S18 to construct a standardized emission exhaust gas component property dataset; the standardized emission exhaust gas component property dataset includes real-time average molar mass. Real gas compressibility factor Ideal low calorific value True low calorie value Inert component ratio The proportion of heavy hydrocarbon components Component fluctuation intensity index And operating condition fluctuation level.

4. The oil and gas field combustion optimization method based on real-time component analysis of venting exhaust gas according to claim 3, characterized in that: Step S2 specifically includes the following steps: S21. Construct an acoustic-TDLAS laser-meteorological multi-source synchronous sensing array to collect torch flame temporal characteristics and environmental meteorological conditions data, and obtain the raw combustion time series dataset. S211. Four acoustic microphones are evenly distributed around the flare tube at a 90° circumference, installed at the same height as the flare nozzle and 5m-8m away from the outer wall of the flare, to capture the acoustic characteristics of flame turbulence pulsation, local flameout, and reignition throughout the entire area; a TDLAS tunable diode laser absorption spectroscopy optical path is horizontally installed in the core stable combustion zone of the permanent flame 0.5m above the flare nozzle to monitor the temporal changes in core flame temperature using the characteristic absorption spectral lines of water vapor; an integrated meteorological station is deployed in an open and unobstructed area of ​​the flare platform to simultaneously collect ambient wind speed, ambient temperature, and atmospheric static pressure parameters. S212. Perform pre-processing noise reduction on the raw sensor signal acquired in S211 to remove power frequency interference, baseline drift and pulse noise anomalies; use notch filtering algorithm to eliminate power grid interference for 20kHz high-frequency acoustic time-series signal, and use moving average filtering to remove instantaneous pulse noise for TDLAS temperature time-series signal, while retaining the true fluctuation characteristics of flame. S213. The preprocessed multi-source sensor data is time-series aggregated and resampled at a 1-second time granularity to unify the data output frequency and achieve a perfect second-level time series match with the component property dataset output in S1; the final output is a time-series aligned original combustion state dataset; the original combustion state dataset includes acoustic energy time series, standardized flame temperature time series data, and ambient wind speed. Ambient temperature and ambient atmospheric static pressure ; S22. Based on the standardized flame temperature time series data output from S21, the ARIMA(2,1,1)-IGARCH(1,1) joint time series filtering algorithm is used to complete the non-stationary time series detrending, autocorrelation elimination and time-varying variance fitting, and output a standardized stationary residual sequence. S221. Select the flame temperature timing and acoustic energy timing after S21 polymerization as the original timing sequence to be processed. The low-frequency trend term is eliminated using first-order difference operations; the formula for calculating the first-order difference in timing is as follows: ； In the formula: for The first-order difference sequence of the original time series to be processed; for The original timing sequence values ​​to be processed at each moment; for The original timing sequence values ​​to be processed at each moment; S222. Construct an ARIMA(2,1,1) difference autoregressive moving average model, and perform time series fitting on the first-order difference sequence obtained in S221 to eliminate the autocorrelation characteristics of the sequence and extract the unbiased residual sequence; the fitting formula of the ARIMA(2,1,1) model is as follows: ； In the formula: This is the mean constant term of the difference time series; These are the first-order autoregressive coefficients; These are the second-order autoregressive coefficients; The coefficient of the first-order moving average; The model fitting residuals from the previous time step; This represents the original fitting residuals output by the model at the current time. for The first-order difference sequence of the original time series to be processed; for The first-order difference sequence of the original time series to be processed; S223, The residual sequence output by S222 Combustion time series exhibit typical characteristics of time-varying fluctuations, heteroscedasticity, and fluctuation clustering. An IGARCH(1,1) integral generalized autoregressive conditional heteroscedasticity model is constructed to fit the real-time fluctuation variance of the residuals. The formula for calculating the IGARCH(1,1) time-varying variance is as follows: ； Hard constraints of the model: ; In the formula: for Real-time conditional variance of the time-residual sequence; The basic variance constant; The weighting coefficient for the squared residuals in the current period; The historical variance memory weighting coefficient; for Real-time conditional variance of the time-residual sequence; S224. Based on the original residuals from S222 and the real-time conditional variance from S223, time series standardization is performed to generate a stationary white noise sequence with zero mean, unit variance, and no autocorrelation. The time series standardization calculation formula is as follows: ； In the formula: for Time-standardized stationary residual sequences; for The time-residual conditional standard deviation, and ; S23. Based on the stationary residual sequence output by S22, the EWRMS exponentially weighted root mean square statistical control chart algorithm is used to realize the identification of early signs of local flame instability and the quantitative statistics of alarm frequency. S231, the standardized stationary residual sequence output from S22 Batch aggregation noise reduction is performed; and the batch mean aggregation calculation formula is as follows: ； In the formula: For the first The mean residual of each batch of data; This refers to the sampling capacity per batch. S232. An exponentially weighted dynamic update mechanism is introduced to construct the EWRMS time-series fluctuation characteristic; and the iterative calculation formula for the EWRMS characteristic is as follows: ； In the formula: For the first The batch's exponentially weighted root mean square volatility characteristic; These are the time-series smoothing weighting coefficients; This is the index-weighted root mean square volatility characteristic of the previous batch; S233. Based on big data of stable combustion conditions, calibrate statistical control limits, set an industrial threshold with a confidence level of 99.73%, and determine the unstable state. The formula for determining the control limit is as follows: ； Decision logic: When When a partial flameout or reignition instability event is detected in the current batch, a single instability alarm is triggered. In the formula: This is the upper limit for statistical control; This is the lower limit of statistical control. S234. Count the number of instability alarm triggers per unit time to quantify the severity of flame dynamic instability; and the formula for calculating the instability alarm frequency is as follows: ； In the formula: The frequency of instability alarms per second; The total number of instability alarms within a given time period; For statistical duration; S24, combined with the component property parameters output from S1, the combustion instability state parameters output from S23, and the environmental meteorological condition parameters output from S21, calculate the momentum flux ratio. Combustion stability index : ； ； In the formula: The ratio of the momentum flux of the nozzle airflow to that of the ambient crosswind; To determine the real-time density of the emitted exhaust gas, and ; The velocity of the exhaust gas released from the flare nozzle; The ambient air density, and , The average molar mass of dry air. The ambient air thermodynamic temperature; The overall combustion stability index; This is the maximum alarm reference value for the critical flameout stage; The minimum calorific value threshold for safe combustion of the torch; The minimum momentum flux ratio for flame wind resistance safety; All of these are the optimal weights determined by the analytic hierarchy process (AHP). right The output results are adjusted for boundary constraints, limiting the numerical range to [value missing]. ; S25, based on the output of S24 and S23 output Establish a four-level combustion instability early warning classification system and output standardized early warning level labels: Normal combustion: and times / second; Level 1 Warning: and times / second; Level 2 warning: and times / second; Critical shutdown warning: and times / second; Output a unique instability warning level based on the real-time dual-indicator matching results. ; S26. Integrate all output parameters from S21 to S25 to construct a standardized flare combustion state feature set; the standardized flare combustion state feature set includes time-series fluctuation features. Frequency of instability alarms Momentum flux ratio Combustion stability index and instability warning level .

5. The oil and gas field combustion optimization method based on real-time component analysis of venting exhaust gas according to claim 4, characterized in that: Step S3 specifically includes the following steps: S31. Perform feature cleaning and normalization on the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and the collected and stored environmental meteorological raw operating condition parameters to construct a unified standardized prediction feature vector. S311. Using the unified IEEE 1588 time stamp as the matching benchmark, complete the precise time alignment of multi-source data; retrieve the standardized emission exhaust gas component property dataset output by S1, the combustion state feature set output by S2, and the collected and stored environmental meteorological raw operating condition parameters, complete data matching at a time granularity of 1 second, and remove abnormal time series samples with missing data, sensor fault marks, or time series offsets exceeding 10ms to obtain time-synchronized complete multi-source raw feature samples; S312. Perform extreme value standardization on all original features after time-series alignment in S311 to eliminate dimensional differences and unify the feature value range to... ; S313. Concatenate all the normalized features from S312 in a fixed order to construct a unified, standardized prediction feature vector with fixed dimensions. ; S32. Separate the main exhaust jet stream from the auxiliary fuel component flow of the pilot flame, and establish a set of equations for calculating the baseline efficiency of the pilot flame interference correction. S321, Based on the volumetric flow rate of the exhaust gas output from S1 Average molar mass Real-time calculation of the total molar flow rate of the main exhaust combustible components : ； S322. Real-time collection of fuel volume flow rate of the pilot flame gas supply branch flow meter. Based on the fixed physical properties of the fuel for the pilot flame, the molar flow rate of the auxiliary fuel for the pilot flame was calculated. : ； In the formula: Methane density under operating conditions in the gas supply pipeline for the permanent flame; Real-time volumetric flow rate of the perpetual flame branch; This refers to the molar mass of pure methane. S323. Calculate the theoretical complete combustion oxygen consumption of the main exhaust gas and the pilot flame fuel respectively, distinguish the combustion energy contribution of the two types of fuel, and construct a set of equations for the total combustion efficiency of the uninterrupted baseline theory: ； In the formula: The theoretical oxygen consumption for the complete combustion of the main exhaust gas; For the first The theoretical number of moles of oxygen required for the complete combustion of a unit mole of a component; The theoretical oxygen consumption for the complete combustion of methane in a continuous flame. This represents the theoretical oxygen consumption per mole of pure methane. This represents the total theoretical oxygen consumption of the system. S324. Introducing a correction coefficient for interference from a persistent flame. By removing the contribution of fuel combustion from the propellant flame, a corrected theoretical efficiency benchmark value is obtained that only characterizes the combustion capacity of the exhaust gas itself. ； in, ； In the formula: The baseline theoretical combustion efficiency of the exhaust gas after eliminating interference from the propellant flame; The uncorrected, raw baseline combustion efficiency, including contributions from the propellant fuel, is calculated based on the matching degree between theoretical oxygen consumption and ambient oxygen supply. S33, based on S1, outputs true low calorific value. Divide the stratified intervals and construct parameterized combustion efficiency prediction sub-models for different three-layer low calorific value stratified intervals; S331. Based on the safe combustion calorific value boundary of oil and gas field venting exhaust gas, divide the area into three low-calorific-value stratified zones: High calorific value stable range: ; Medium calorific value fluctuation range: ; Low calorific value critical range: ; S332, using the standardized predictive feature vector from S313 The reference theoretical combustion efficiency of the exhaust gas for input S324. Based on the fundamental offset term, linear parameterized combustion efficiency prediction models are established for the three low-calorific-value stratified intervals; and the general prediction formula for stratified parameterized combustion efficiency is as follows: ； In the formula: To determine the real-time predicted combustion efficiency for the lower three-layer low-calorific-value stratified range, the value range is... ; The optimal weighting coefficients are specific to the three low-calorific-value stratification intervals. Transpose of a vector, weight transpose vectors of different three-layer low-calorific-value stratification intervals Solved by interval-independent offline multivariate least squares calibration; This is the correction factor for flame instability; This is the component fluctuation correction factor; S333, outputting stratified combustion efficiency to S332 By implementing boundary truncation constraints, model calculation overflow under extreme operating conditions is eliminated, and the final real-time combustion efficiency is obtained. : ； In the formula: The final real-time combustion efficiency after correction; S34, Final Real-Time Combustion Efficiency Based on S33 Output Coupled S1 exhaust gas component mole fraction sequence A multi-component destruction and removal efficiency calculation model was established to solve the destruction and removal efficiency of hydrocarbons and hydrogen sulfide respectively. S341. Differentiate between three categories of target pollutants for disposal: heavy hydrocarbons hydrogen sulfide methane Obtain the molar flow rate of each pollutant inlet: ； In the formula: For the first Pollutant inlet molar flow rate; For the first mole fractions corresponding to pollutants; Values ​​include Three categories of controlled pollutants; S342, combined with S33, final real-time combustion efficiency Calculate the outlet molar flow rate of residual pollutants after combustion and destruction: ； In the formula: For the first The molar flow rate of pollutants that escaped from incomplete combustion at the outlet; S343. Define pollutant destruction and removal efficiency based on inlet and outlet pollutant molar flow rates; ； In the formula: For the first Efficiency of pollutant removal and destruction; S344, Substituting S342 and simplifying, we get... ; S35, coupled with combustion efficiency, environmental meteorology, and component physical property parameters, constructs a parameterized prediction model for multi-pollutant emission concentrations, and outputs real-time predicted values ​​of flue gas pollutant emissions; S351, Total molar flow rate of main exhaust combustible components based on S1 S2 ambient wind speed and ambient atmospheric static pressure Calculate the flare gas dilution and diffusion correction factor: ； In the formula: This is a correction factor for flue gas dilution and diffusion. This is the field diffusion calibration constant; Real-time flame core temperature acquired by the TDLAS tunable diode laser absorption spectroscopy optical path; S352. Based on the residual amount of pollutants after pollutant destruction, the total molar amount of flue gas generated, and the diffusion correction factor, predict the equivalent flue gas pollutant emission concentration at the chimney: ； In the formula: For the first Predicted emission concentrations of pollutants; The total molar flow rate of flue gas generated after complete combustion of fuel, and , , The original molar flow rate of the exhaust gas. This is the excess air-fuel ratio coefficient. This represents the theoretical molar consumption of oxygen. S36. Integrate all calculated output parameters from S31-S35 to construct a feature dataset for predicting combustion efficiency and pollutant concentration; the feature dataset for predicting combustion efficiency and pollutant concentration includes standardized prediction feature vectors. Long-lasting flame interference correction reference efficiency Real-time combustion efficiency Methane destruction and removal efficiency Total hydrocarbon destruction and removal efficiency Hydrogen sulfide destruction and removal efficiency Methane predicted emission concentration Total hydrocarbon predicted emission concentration Predicted emission concentration of hydrogen sulfide Predicted emission concentrations of sulfur dioxide Predicted carbon monoxide emission concentration and predicted nitrogen oxide emission concentrations .

6. The method for optimizing oil and gas field combustion based on real-time component analysis of venting exhaust gas according to claim 5, characterized in that: In step S332, the weight transpose vector The expression to be solved is as follows: ； in, ； ； In the formula: To correspond to the historical standardized prediction feature matrix of the three-layer low calorific value stratification interval, each row of the matrix stores one S313 output standardized prediction feature vector. ; for Transpose of; This is a sample set of actual combustion efficiency measured by on-site flue gas monitoring in this section; , These are the comprehensive combustion stability index and the component fluctuation intensity index for historical samples within the interval, respectively. The three-layer low-calorific-value stratified intervals are numbered. Represents the critical range of low calorific value. Represents the range of fluctuations in medium calorific value. This represents the stable range of high calorific value; for Basic calibration constants; A sensitive correction factor for instability in the three-layer low-calorific-value stratification interval; For the three-layer low calorific value stratification intervals Threshold; for Basic calibration constants; A sensitive factor for component fluctuations in the three-layer low-calorific-value stratified interval; This is the real-time component fluctuation intensity index; This is the threshold for the maximum permissible component fluctuation intensity at the site.

7. The method for optimizing oil and gas field combustion based on real-time component analysis of venting exhaust gas according to claim 5 or 6, characterized in that: Step S4 specifically includes the following steps: S41. Read the full-dimensional prediction output data of S3, the inherent boundary parameters of the equipment, and the environmental protection national standard limit parameters to build a multi-objective optimization input benchmark set and constraint boundary set; S411, Read the real-time destruction and removal efficiency output by S3. Predicted emission concentrations of multiple pollutants , , Real-time combustion efficiency Synchronously read the component fluctuation intensity index output by S1 True low calorie value Read the instability warning level output by S2. Combustion stability index All parameters are aligned with a 1-second timing granularity to ensure that the optimized input is synchronized with the on-site working conditions in real time. S412. Input the inherent physical range constraints of the station's hardware equipment to construct a set of hard safety constraints for the equipment; fix and input the flow range of the combustion-supporting medium. Flare nozzle velocity range Flow rate of fuel for the long-lasting flame ; At the same time, the emission limits for sulfur dioxide, carbon monoxide, and nitrogen oxides are recorded. , , Enter the lowest stable combustion safety low calorific value of the torch. Minimum safe momentum flux ratio ; S413. Constructing an optimized vector of control variables ; In the formula: Let be the vector of control variables to be optimized; Real-time flow control of the combustion-supporting medium; The velocity of the exhaust gas released from the flare nozzle; Real-time flow control of fuel for the permanent flame; This is the matrix transpose symbol; S42. Construct a multi-objective comprehensive optimization objective function that takes into account stable combustion efficiency, economical operation, and environmental protection and emission reduction, and superimpose multiple layers of hard constraints to form a complete optimization mathematical model; S421. Establish three major optimization sub-objectives: maximize destruction and removal efficiency, minimize combustion medium energy consumption, and minimize comprehensive pollutant emissions. ； In the formula: For the target item of combustion efficiency deviation; The system is preset with an optimal destruction efficiency target; For energy consumption economic targets; This refers to the maximum flow rate of the combustion-supporting medium. For environmental emission targets; The weighted average emission concentration of three pollutants: sulfur dioxide, carbon monoxide, and nitrogen oxides. The weighted average limit of pollutants according to national standards; S422. Construct the initial fixed-weight comprehensive optimization objective function: ； In the formula: Based on the comprehensive objective function value, The core optimization criterion; For efficiency, economy, and environmental protection, the initial fixed weighting coefficients are satisfied. The initial calibration values ​​are respectively ; S423. Establish a complete set of hard constraint equations for the operating conditions: ； In the formula: The standard limit concentration for sulfur dioxide; This refers to the standard limit concentration of carbon monoxide. The standard limit concentration for nitrogen oxides; S43, Instability warning level based on S2 output Matching the initial baseline adjustment step size: ； In the formula: Adjust the step size based on the baseline; S44, Introducing the component fluctuation intensity index from the output of S1 This enables adaptive dynamic correction of optimized weights and adjustment step sizes; S441, Constructing the Component Fluctuation Intensity Index The formula for dynamic adjustment of driving weights: ； In the formula: The efficiency, economic, and environmental weighting coefficients are dynamically adjusted. This is the weighting adjustment factor; S442. Constructing an adaptive step size correction formula for wave intensity: ； In the formula: Adjust the step size for final adaptive optimization; This is the step size amplification factor; S443. Based on the revised dynamic weights, update the comprehensive objective function to obtain the final adaptive optimization objective model: ； In the formula: For adaptive dynamic synthesis objective function; S45. An improved adaptive particle swarm optimization algorithm is used to iteratively optimize the dynamic optimization model and solve for the global optimal control parameter vector. S451. Initialize particle swarm population parameters and set basic iteration conditions: population size. Maximum number of iterations Convergence accuracy threshold Particle dimension and S41 optimization control variable vector Consistent dimensions; S452, Introduce a linearly decreasing inertia weight strategy: ； In the formula: For the first Real-time inertia weights in each iteration; This is the initial maximum inertia weight; The minimum inertia weight at the end of the iteration; This represents the current iteration number; S453. Introduce a fixed learning factor to construct an improved formula for particle velocity and position update: ； In the formula: For the first The second iteration The velocity of each particle; For the first The second iteration The velocity of each particle; For the first The second iteration The position of each particle; To fix the learning factor; for Uniform random numbers; For each individual particle, the historical optimal solution is provided. This is the globally optimal solution for the population. For the first The second iteration The position of each particle; S454. The penalty function method is used to handle the multiple constraints in S423, penalizing out-of-bounds particles and guiding the population to converge to the feasible region; and the penalty fitness formula is as follows: ； In the formula: This is a fitness value with a penalty. This is the penalty coefficient; For the first The out-of-bounds deviation of a constraint condition, when the constraint is fully satisfied. ; S455. Set the iteration termination condition and complete the optimal solution output: when the change in the global optimal objective function value is less than the convergence accuracy for 20 consecutive iterations. or reaching the maximum number of iterations. When the iteration terminates, output the current global optimal control variable vector for the population: ； In the formula: To achieve the optimal flow rate of the combustion-supporting medium; To achieve the optimal flare nozzle velocity; To achieve the optimal fuel flow rate for the continuous flame; S46. The optimal control variable vector output from S45 The signal is converted into a control valve opening command signal that the station can recognize. The control command is then sent in real time to the combustion medium control valve, the frequency conversion pressure regulating mechanism of the venting pipeline, and the pilot flame gas control valve via the Modbus TCP / IP industrial protocol. The valve position feedback and flow feedback signals of each actuator are collected in real time to verify the command execution deviation. If the deviation exceeds 5%, a secondary calibration is performed.

8. The method for optimizing oil and gas field combustion based on real-time component analysis of venting exhaust gas according to claim 7, characterized in that: Step S5 specifically includes the following steps: S51. Time-series aligned acquisition of full-dimensional on-site measured true value data and equipment operation feedback data after S4 regulation, constructing a closed-loop iterative original sample dataset; Based on the unified 1s time stamp of S1–S4, the steady-state window period after the completion of S4 regulation is delayed and locked; the steady-state judgment time is set to 30s, and only the stable operation data after the regulation is stabilized is collected; The measured true values ​​of flue gas after regulation are collected using an online flue gas analyzer at the site to obtain the baseline true values ​​parameters required for model calibration; among them, the measured true values ​​of flue gas include dry basis oxygen concentration. Dry basis carbon dioxide concentration Dry basis carbon monoxide concentration and dry basis total hydrocarbon concentration , serving as the sole truth label for the model correction of combustion efficiency and destruction / removal efficiency; Collect real-time feedback data from S4 and simultaneously collect the actual operating flow rate of the combustion-supporting medium. Actual flow velocity at the torch nozzle Actual fuel flow rate of the pilot flame Feedback on the actual valve position and opening degree of each regulating valve; Synchronous backtracking retrieves the full-dimensional temporal feature parameters of S1, S2, and S3 corresponding to this round of optimization. , , , , , , , , , Construct a closed-loop iterative original sample library; S52. Based on the measured true value data of flue gas in S51, the true baseline combustion efficiency and true destruction and removal efficiency are recalculated using the S32 long-burner interference correction algorithm, and used as the true value benchmark for evaluating the optimization effect. S521 reuses the S32 shunt correction logic to strip the live flame interference from the measured flue gas components in S51, separating the combustion contribution of the exhaust gas and the live flame to obtain effective true flue gas component parameters: ； ； In the formula: The effective true carbon dioxide dry basis concentration after stripping away the interference of the propellant flame. The effective dry basis oxygen concentration after correction; This refers to the atmospheric standard oxygen volume fraction. S522. Based on the corrected effective flue gas composition, calculate the true value of the actual combustion efficiency after this round of regulation: ； In the formula: This represents the true value of the actual combustion efficiency on-site. S523. Calculate the true value of the actual destruction and removal efficiency for this round of operation: ； In the formula: To determine the actual efficiency of destruction and removal under real-world working conditions; and All are global calibration linear fitting constants, and ; S53. Quantitatively evaluate the actual implementation effect of the S4 adaptive optimization strategy from four dimensions: safe and stable combustion, combustion efficiency, economical operation, and environmental protection and emission reduction. ； ； ； ； In the formula: For the improvement rate of combustion stability index; To optimize the pre-combustion stability index; The combustion stability index after S4 regulation; To optimize destruction efficiency and increase output; To optimize the actual destruction efficiency; This represents the actual destruction efficiency after S4 regulation; To reduce the unit consumption rate of combustion-supporting medium; This represents the historical average combustion-supporting consumption under a fixed-ratio strategy. To achieve the overall pollutant emission reduction rate; The average emission concentration of pollutants before optimization; The average emission concentration of pollutants after S4 regulation; S54. Based on the incremental weighted least squares method, realize the online rolling correction of the S3 hierarchical parameterized prediction model and complete the model self-iterative upgrade; S541. Construct a time-weighted sample set, and the formula for calculating the sample weights is as follows: ； In the formula: For single-sample iterative weights; This is the time decay coefficient; This represents the time interval between the sample and the current time. S542. Construct an incremental weighted least squares objective function to incrementally update the coefficients of the three-layer low-calorific-value stratified interval model in S3 interval by interval; and the incremental correction objective function is as follows: ； In the formula: The weighted sum of squares of the model residuals; S543, Update the weight vector coefficients of the S33 hierarchical model interval by interval. Instability correction coefficient Fluctuation correction coefficient Complete the model iteration update; S544. Verify the correction accuracy using the Mean Absolute Error (MAE) as the verification index: If If the error exceeds the limit, the correction is deemed effective and the model is officially replaced and put online; if the error exceeds the limit, a second fine-tuning iteration correction is triggered. S55. Integrate the steady-state operating condition characteristics, fluctuation characteristics, and combustion state characteristics of this round into an operating condition feature fingerprint vector. ; And bind the optimal control parameter vector output by S4 for this round. This forms a pair of sample units for working condition fingerprinting and optimal control. S56. Calculate cosine similarity. To achieve deduplication and update iteration of the strategy library: when If the condition is determined to be of the same type, the old control strategy will be updated accordingly; when If a new operating condition is identified, it will be added to the inventory, and the strategy will continue to expand the scope of operating conditions covered.

Images