Carbon capture process intelligent diagnosis control method and system based on ai decision

Abstract

The application discloses a carbon capture process intelligent diagnosis control method and system based on AI decision, and particularly relates to the technical field of industrial process intelligent control, and comprises the following steps: obtaining control intentions on the edge computing side and the cloud side and aligning to generate a synchronous control sequence; calculating six types of conflict factors to form a standardized conflict factor vector; generating a limited adaptive weight vector in combination with data freshness, source reliability and conflict duration; calculating an uncalibrated risk value through bounded gain aggregation calculation; determining a coupling risk coefficient according to an uncalibrated risk-coupling risk control table and outputting a graded disposal suggestion; the application calculates six types of conflict factors through time alignment and unit unification, improves the conflict identification precision of the carbon capture process; realizes dynamic adaptability of risk assessment through the generation of a limited adaptive weight; generates an uncalibrated risk value based on bounded gain aggregation, enhances the executable of risk quantification and regulation, and guarantees the safe and stable operation of the carbon capture process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of industrial processes, and more particularly, to an AI decision-based intelligent diagnosis control method and system for a carbon capture process. BACKGROUND

[0002] A carbon capture process involves multiple core units such as an absorption tower, a desorption tower, a regeneration heat exchanger, a solvent circulating pump, a compressor, and a steam supply system. There are multi-dimensional process variables such as temperature, pressure, flow rate, and concentration between the units, and these variables have significant nonlinearity and strong coupling characteristics. In actual industrial scenarios, in order to reduce energy consumption, ensure process stability, and improve carbon dioxide capture rate, enterprises gradually introduce edge computing and cloud optimization collaborative operation architecture. The edge computing side is responsible for local rapid response control, and the cloud side relies on historical data and global optimization strategy to execute decision instructions. The collaborative operation of the two can theoretically balance local stability and global economy, but also makes the control level more complex in terms of interaction and conflict risk, and puts forward higher intelligent requirements for diagnosis and regulation.

[0003] Existing carbon capture process control methods mainly rely on local controllers or single optimization platforms for decision-making, lack of synchronization comparison mechanism between edge and cloud, and cannot form a unified time reference and measurement unit system. Due to the lack of data alignment and metadata supplement, the data collected by the edge side is often difficult to be directly used for cloud decision-making due to communication delay, time drift, and high proportion of missing data, resulting in deviation in the basis of control decision-making. In addition, traditional conflict identification methods usually only judge in one dimension, such as only detecting whether the adjustment direction is consistent, without comprehensively quantifying multiple factors such as direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree, and boundary approximation degree, and cannot fully reflect the essence of control conflict. Further, the existing risk assessment mechanism lacks dynamic weighting capability for data freshness, source credibility, and conflict duration, and cannot give effective early warning and grading suggestions in the early stage of control conflict, and finally can only rely on manual experience for intervention, resulting in delayed regulation response and may cause energy waste, device oscillation, and even safety accidents. Therefore, an AI decision-based intelligent diagnosis control method and system for a carbon capture process is proposed to solve the above problems. SUMMARY

[0004] To achieve the above object, the present application provides the following technical scheme: a carbon capture process intelligent diagnosis control method based on AI decision, comprising the following steps: step one, obtaining control intention data of the edge computing side and the cloud side in the same time period, aligning the data time base of the control intention and unifying the measurement units, and generating a synchronous control sequence for comparative analysis; step two, calculating six types of conflict factors based on the synchronous control sequence, the six types of conflict factors including: direction reverse degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approaching degree, wherein the direction reverse degree is used to measure the instruction direction angle and the proportion of opposite signs, the amplitude asymmetry degree is used to measure the relative difference of the adjustment amplitude on both sides, the phase mismatch degree is used to measure the time center offset and the period offset, the energy impact degree is used to measure the expected power step or steam load step strength, the sensitive amplification degree is used to measure the amplification effect of the key process quantity under the current working condition sensitivity, and the boundary approaching degree is used to measure the reverse value of the remaining margin of the preset constraint; then, the standardization processing is performed to obtain a standardized conflict factor vector; step three, calculating an adaptive weight vector according to the data freshness, source reliability and conflict duration, and setting a weight upper limit according to the device safety level to obtain a limited adaptive weight vector; step four, performing weighted superposition on the standardized conflict factor vector and the limited adaptive weight vector by using a bounded gain aggregation to obtain an uncalibrated risk value; step five, looking up the uncalibrated risk value according to a preset uncalibrated risk-coupling risk control table to obtain a corresponding coupling risk coefficient, and outputting a disposal suggestion based on a preset grading interval corresponding to the coupling risk coefficient to guide the carbon capture process control system to implement dynamic regulation and control.

[0005] In a preferred embodiment, the synchronous control sequence is obtained by the following steps: first, identifying valve steps, pump speed transitions, set value changes and alarm confirmations on the edge computing side and the cloud side through preset event determination rules, and screening effective anchor points by carbon element conservation and energy balance two types of invariants, solving the piecewise linear time mapping with global non-decreasing and slope upper and lower bounds for the anchor point set passing through the constraints, to obtain the initial estimation of clock drift and communication delay; implementing limited dynamic time warping and phase correction on the initial time mapping to maintain the monotonicity of the mapping function and limit the local slope, while adding a small phase item within a limited lag range according to the cross-correlation peak, and outputting the aligned time base and the alignment residual sequence containing mean, variance and maximum deviation; completing the state-dependent conversion between volume flow and mass flow, steam flow and heat load, and different concentration expressions based on process property relationships, unifying the measurement units, and adding source reliability, alignment residual confidence band, version summary identifier and integrity check summary for each time, and finally forming a synchronous control sequence carrying complete metadata.

[0006] In a preferred embodiment, in the process of forming the synchronization control sequence carrying complete metadata, the density and enthalpy are calculated according to temperature, pressure and solvent composition, the volume flow is converted into mass flow, the steam flow is converted into heat load, the mass fraction and volume fraction are converted into mole fraction, the pressure, temperature and concentration are unified into a pre-designed quantity unit system; a source reliability is added to each time point, the source reliability is calculated by jointly weighting the instrument's latest calibration date, online self-checking result and missing proportion; an alignment residual confidence band is added to each time point, the upper and lower boundaries of the alignment residual confidence band are given by the interval estimation of the alignment residual sequence; a version summary identifier and an integrity check summary are added to each time point, the version summary identifier is composed of the model name, model version and training time; the integrity check summary is composed of a deterministic summary string of the configuration parameter list, and the synchronization control sequence carrying metadata is generated after summarization.

[0007] In a preferred embodiment, the specific steps of step two include: based on the synchronization control sequence, extracting the control intention increment of the edge computing side and the cloud side in the same control period, pairing one by one according to the control channel, obtaining the paired data set of the two sides in different pre-set quantization increments, and recording the rated range, rated energy capacity and constraint limit value of each channel.

[0008] In a preferred embodiment, the six types of conflict factors are calculated on the paired data set, specifically: the calculation of the direction reversal degree is to calculate the direction angle of the two side increments of the same channel first, the direction angle takes the angle value from zero to one hundred and eighty degrees, then calculate the proportion of the channels with opposite signs in all channels, the direction reversal degree is equal to the weighted average of the direction angle divided by one hundred and eighty and the proportion; the calculation of the amplitude asymmetry degree is to take the absolute value of the difference between the absolute values of the two side increments of the same channel, then divide by the sum of the absolute values of the two side increments plus a pre-set minimum positive lower limit value as the denominator, and finally take the arithmetic average of all channels; the calculation of the phase mismatch degree is to calculate the cross-correlation function of the two side control intentions within the specified maximum time lag range, take the absolute value of the time lag that produces the maximum correlation and normalize it with the maximum time lag, and at the same time calculate the time center difference of the two side control intentions and normalize it with the control period of the channel, the phase mismatch degree is equal to the weighted average of the two normalized results; the calculation of the energy impact degree is to convert the control intention increments of each channel through the energy conversion coefficient under the current working condition into expected electric power step and expected thermal load step, take the absolute value of the sum of the two, and normalize it with the sum of the rated electric power and rated thermal load of the corresponding unit; the calculation of the sensitive amplification degree is to obtain the local first-order sensitivity coefficient of each input channel to the key process quantity at the current operating point, multiply the control intention increments of each channel by the corresponding sensitivity coefficient and sum them up to get the expected change amplitude of each key process quantity, take the maximum value of all key process quantities and normalize it with the allowed change bandwidth of the process quantity; the calculation of the boundary approximation degree is to calculate the remaining margin of each controlled constraint after executing the two side control intentions, if the constraint is an upper limit, subtract the expected value from the upper limit, if the constraint is a lower limit, subtract the lower limit from the expected value, negative values are counted as zero, then normalize with the safety bandwidth of the corresponding constraint and subtract the normalized result, finally take the maximum value of all constraints as the boundary approximation degree.

[0009] In a preferred embodiment, the standardized conflict factor vector refers to: the six types of conflict factors calculated based on the synchronous control sequence are respectively subjected to pre-set standard processing, each conflict factor is mapped to the standardized range of zero to one, then the standardized conflict factor vector is obtained by summarizing according to the pre-set arrangement order.

[0010] In a preferred embodiment, the restricted adaptive weight vector is obtained by the following steps: calculating a data freshness score, specifically, calculating a time lag value, a communication delay value and a missing proportion for each data channel for six types of conflict factors on a unified time reference, the time lag value being the current unified time minus the latest valid timestamp of the channel, the communication delay value being the communication delay estimation obtained by the aforementioned time mapping, and the missing proportion being the proportion of invalid samples to all samples within a fixed evaluation window; linearly reverse mapping the time lag value according to a preset maximum acceptable time difference to obtain a timeliness component, linearly reverse mapping the communication delay value according to a preset maximum acceptable delay to obtain a channel delay component, and obtaining an integrity component according to a one-minus relationship, and then taking a geometric mean of the timeliness component, the channel delay component and the integrity component to obtain a data freshness score in the range of zero to one; calculating a conflict duration, specifically, constructing an instantaneous conflict indicator within the same evaluation window, the instantaneous conflict indicator being equal to the maximum value of the product of a structural conflict strength and an amplification strength and a boundary strength, the structural conflict strength being the arithmetic mean of a direction reversal degree, an amplitude asymmetry degree and a phase mismatch degree, the amplification strength being the arithmetic mean of an energy impact degree and a sensitive amplification degree, and the boundary strength being a boundary approximation degree; setting a conflict threshold and identifying a time period in which the indicator continuously exceeds the threshold, calculating the proportion of time exceeding the threshold and calculating the average value of the indicator when exceeding the threshold, and taking the product of the proportion of time and the average value as an unweighted conflict duration, and then weighting according to a linearly increasing weight from near to far in time to obtain a conflict duration in the range of zero to one; forming a restricted adaptive weight vector, specifically, multiplying the source credibility weight and the data freshness score for each component of the six types of conflict factors, and multiplying the result by an exponential function of the conflict duration, the exponential being given by a sensitive coefficient corresponding to the device security level, the sensitive coefficient being determined values corresponding to the device security levels of one, two, three and four, respectively; convexly combining the obtained current weight and the last period weight according to a fixed smoothing coefficient to obtain a smoothed weight, and truncating according to the weight upper limit set by the device security level, the weight upper limit of one, two, three and four being determined values in order from small to large, and finally outputting the restricted adaptive weight vector.

[0011] In a preferred embodiment, the uncalibrated risk value is obtained by the following steps: establishing a gain limiting function for each of the direction reversal degree, the amplitude asymmetry degree, the phase mismatch degree, the energy impact degree, the sensitive amplification degree and the boundary approximation degree, the gain limiting function adopts a smooth saturation form, i.e. taking the gain coefficient multiplied by the index value as the numerator, taking one plus the coefficient multiplied by the index value as the denominator, and then taking the smaller value of the ratio and the corresponding gain upper limit, the growth coefficient of each factor and the determined value of the gain upper limit are preset according to the safety level of the device; inputting the standardized conflict factor vector into the gain limiting function to obtain a gain limited factor vector, and multiplying each component of the gain limited factor vector by the limited adaptive weight vector to form a weighted component set, performing a truncated arithmetic average on the weighted component set to obtain an aggregated base value, and the truncation ratio is a fixed value between the preset minimum ratio and the maximum ratio; performing double threshold shaping and convex combination smoothing on the aggregated base value obtained by the truncated arithmetic average to obtain the uncalibrated risk value.

[0012] In a preferred embodiment, the step of performing double threshold shaping and convex combination smoothing on the aggregated base value obtained by the truncated arithmetic average includes: setting a low threshold and a high threshold, the low threshold and the high threshold are fixed numerical values and the low threshold is less than the high threshold; outputting zero when the aggregated base value does not exceed the low threshold; when the aggregated base value is between the low threshold and the high threshold, outputting a proportional value equal to the aggregated base value minus the low threshold, and then divided by the high threshold minus the low threshold; when the aggregated base value exceeds the high threshold, outputting a value equal to one minus an exponential function, the argument of the exponential function is the aggregated base value minus the product of the high threshold and a fixed convergence coefficient, the fixed convergence coefficient is preset according to the safety level of the device; forming a convex combination of the current cycle output and the output of the last evaluation cycle according to a fixed smoothing coefficient, the fixed smoothing coefficient is a determined value in the range of zero to one, and limiting the convex combination result in the range of zero to one to obtain the uncalibrated risk value.

[0013] In a preferred embodiment, the AI decision-based carbon capture process intelligent diagnosis control system specifically comprises: an alignment and unification module, which acquires control intention data of the edge computing side and the cloud side in the same time period, aligns the data time base of the control intention and unifies the units of measurement, and generates a synchronous control sequence for comparative analysis; a conflict quantification module, which calculates six types of conflict factors based on the synchronous control sequence, the six types of conflict factors including direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree, wherein the direction reversal degree is used to measure the proportion of the instruction direction angle and the sign opposite, the amplitude asymmetry degree is used to measure the relative difference of the adjustment amplitude of the two sides, the phase mismatch degree is used to measure the time center offset and the period offset, the energy impact degree is used to measure the expected power step or steam load step strength, the sensitive amplification degree is used to measure the amplification effect of the key process quantity under the current working condition sensitivity, and the boundary approximation degree is used to measure the reverse value of the remaining margin of the preset constraint; then the standardization processing is performed to obtain a standardized conflict factor vector; a weight generation module, which calculates an adaptive weight vector according to the data freshness, the source reliability and the conflict duration, and sets a weight upper limit according to the device safety level to obtain a limited adaptive weight vector; a gain aggregation module, which performs weighted superposition on the standardized conflict factor vector and the limited adaptive weight vector by using a bounded gain aggregation to obtain an uncalibrated risk value; and a mapping disposal module, which looks up the corresponding coupling risk coefficient according to a preset uncalibrated risk-coupling risk control table based on the uncalibrated risk value, and outputs a disposal suggestion based on a preset grading interval corresponding to the coupling risk coefficient to guide the carbon capture process control system to implement dynamic regulation and control.

[0014] The technical effects and advantages of the present application are as follows: by acquiring the control intention data of the edge computing side and the cloud side in the same time period, the present application performs time base alignment and unit unification, ensures that the subsequent analysis is completed under a unified standard, and thus effectively avoids misjudgment caused by time deviation or unit inconsistency between different data sources. On this basis, the present application calculates six types of conflict factors including direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree, and performs standardization processing to form a standardized conflict factor vector that can be directly quantified. This method can comprehensively analyze from multiple dimensions including instruction direction, amplitude difference, time offset, energy impact, sensitivity amplification and constraint approximation, and compared with the traditional method based on only a single index, can significantly improve the accuracy and reliability of the carbon capture process conflict identification.

[0015] The application solves the problem of fixed weight distribution in traditional risk assessment and the inability to adapt to working condition changes by calculating an adaptive weight vector based on data freshness, source reliability and conflict duration, and generating a restricted adaptive weight vector under the constraint of device security level. The application dynamically adjusts the weight during calculation, so that factors with fresher data, more reliable sources and longer conflict duration have a higher proportion in risk assessment, thereby ensuring that the calculation result can reflect the real-time running state. This mechanism ensures high consistency between risk assessment and the actual running environment of the device, and improves the agility and adaptability of the carbon capture process control strategy.

[0016] The application obtains an uncalibrated risk value by weighting and superimposing the standardized conflict factor vector and the restricted adaptive weight vector in a bounded gain aggregation manner, and looks up the uncalibrated risk-coupling risk reference table to output a coupling risk coefficient and a corresponding disposal suggestion. This method not only converts complex conflict characteristics into quantifiable and interpretable risk coefficients, but also directly provides corresponding regulation suggestions, providing clear basis for dynamic regulation of the carbon capture process control system. Through this closed-loop mechanism from conflict identification to risk mapping to disposal output, the application can seamlessly connect risk assessment and regulation, thereby effectively reducing system operation risk and improving process control stability. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to facilitate understanding of those skilled in the art, the application will be further described below with reference to the accompanying drawings;

[0018] Figure 1 The principle diagram of the AI decision-based carbon capture process intelligent diagnosis control method in the application.

[0019] Figure 2 The principle diagram of the AI decision-based carbon capture process intelligent diagnosis control system in the application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0021] REFERENCE Figure 1 - Figure 2 The following embodiments are obtained:

[0022] Embodiment 1: An AI decision-based carbon capture process intelligent diagnosis control method, comprising the following steps:

[0023] Step one, obtain the control intention data of the edge computing side and the cloud side in the same time period, align the data time reference of the control intention, and unify the measurement units to generate a synchronized comparison sequence for comparative analysis; this step is used to ensure that data from different computing sides can be accurately compared and analyzed under the same time sequence and the same measurement system. By aligning the time reference, the time error caused by clock drift, network delay or data collection period difference between the edge computing node and the cloud platform can be eliminated, ensuring that the behavior of the control intention at the same time point is comparable. By unifying the measurement units, the numerical deviation caused by the different dimension systems adopted by different computing sides can be avoided, such as volume flow and mass flow, steam flow and heat load, concentration representation, etc. The synchronized comparison sequence finally generated by this step not only contains the aligned control intention data, but also combines metadata such as source credibility and time residual, so that the subsequent conflict analysis is based on traceable, verifiable and unambiguous basis.

[0024] Step two, calculate six types of conflict factors based on the synchronized comparison sequence, including direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree, wherein the direction reversal degree is used to measure the proportion of the instruction direction angle and the opposite sign, the amplitude asymmetry degree is used to measure the relative difference of the adjustment amplitude of the two sides, the phase mismatch degree is used to measure the time center offset and period offset, the energy impact degree is used to measure the expected power step or steam load step strength, the sensitive amplification degree is used to measure the amplification effect of the key process quantity under the current working condition sensitivity, and the boundary approximation degree is used to measure the reverse value of the remaining margin of the preset constraint; then, the standardization processing is performed to obtain a standardized conflict factor vector; this step is used to quantify the difference in control behavior between the edge computing side and the cloud side into comparable indicators. The direction reversal degree reveals whether the adjustment direction of the two sides exists completely opposite trend, if such conflict frequently occurs, it will cause the equipment to produce reciprocating adjustment in a short time. The amplitude asymmetry degree reflects the significant difference of the adjustment amplitude of the two sides, which can identify the over-adjustment or insufficient adjustment caused by model parameter drift. The phase mismatch degree can find the coordination imbalance caused by lag response or advance execution by analyzing the offset of the two sides in time. The energy impact degree directly quantifies the influence of the adjustment action on the power and heat load level, avoiding the impact of short-time energy fluctuation on the system operation. The sensitive amplification degree considers the amplification effect of the adjustment action on the key process quantity under different working conditions, which can be used to identify the risk amplification effect. The boundary approximation degree measures whether the control behavior pushes the system to the constraint limit, which is an important early warning factor for preventing overrunning. After the six types of conflict factors are standardized, a standardized conflict factor vector is formed, which provides a unified input basis for subsequent risk aggregation and adaptive weight allocation.

[0025] Step three, calculate the adaptive weight vector according to the data freshness, source reliability and conflict duration, and set the weight upper limit according to the device security level to obtain the restricted adaptive weight vector; this step is used to dynamically adjust the influence weight of each conflict factor in the risk aggregation process. Data freshness reflects the timeliness and integrity of input data, high freshness means that the data can more accurately represent the current state. Source reliability is used to measure the reliability of the data source, which can prevent distorted results caused by abnormal sensors or model version errors from being amplified. Conflict duration is used to measure the stability of the conflict phenomenon in the time dimension. If the conflict is only short-term noise, its influence weight should be reduced; if the conflict duration is long and the intensity is stable, its weight should be increased. Finally, by setting the weight upper limit according to the device security level, it can avoid a single conflict factor being amplified indefinitely due to short-term abnormalities, and ensure the stability and controllability of the risk assessment results.

[0026] Step four, use bounded gain aggregation to weight and superimpose the standardized conflict factor vector and the restricted adaptive weight vector to obtain the uncalibrated risk value; this step fuses multiple conflict factors according to their weights through the gain aggregation mechanism, and avoids abnormal amplification of risk index caused by extreme values through the gain limiting function. The core of bounded gain aggregation is to map the conflict factor and its weight to a controlled gain response, so that the contribution of each factor to the final risk value is within a predictable range. During the aggregation process, a truncated average or smoothing mechanism can be introduced to suppress the influence of short-term abnormal values, so as to ensure that the uncalibrated risk value can accurately reflect the comprehensive degree of multi-source conflicts and not distorted by local abnormalities.

[0027] Step five, look up the uncalibrated risk value according to the pre-set uncalibrated risk-coupling risk table to obtain the corresponding coupling risk coefficient, and output the handling suggestion based on the pre-set hierarchical interval corresponding to the coupling risk coefficient to guide the carbon capture process control system to implement dynamic regulation and control. This step realizes the direct association between risk results and executable control strategies by mapping the uncalibrated risk value to the coupling risk coefficient. Through the pre-set correspondence between uncalibrated risk and coupling risk, complex multi-dimensional conflict information can be converted into a risk quantization result in the range of zero to one. The coupling risk coefficient is divided into multiple hierarchical intervals, each interval corresponds to different handling suggestions, such as parameter fine-tuning, instruction freezing, model shadow running, manual intervention or emergency derating operation. Through this mapping relationship, the system can automatically trigger the dynamic regulation and control strategy that matches the current state according to the risk level, realize the closed-loop management from risk assessment to decision execution, and improve the safety, stability and operation efficiency of the carbon capture process.

[0028] The synchronization control sequence is obtained by the following steps: first, the valve step, pump speed transition, set value change and alarm confirmation are identified on the edge computing side and the cloud side through preset event determination rules. The determination rule of the valve step is that the change amount of the valve opening in a single control period exceeds one percent of the valve range and the duration is not less than two seconds, for example, the valve opening changes from twenty percent to thirty-five percent and remains for three seconds, then it is determined as a valid valve step event. The determination rule of the pump speed transition is that the circulating pump speed changes more than three percent of the rated speed within five seconds and the direction is consistent, for example, the circulating pump speed increases from one thousand revolutions per minute to one thousand one hundred revolutions per minute without reverse fluctuation, then it is determined as a valid pump speed transition event. The determination rule of the set value change is that the change amount of the target set value detected in the control loop exceeds one percent of the control amount range and is a single direction adjustment, for example, the target value of the solvent circulation amount is adjusted from one hundred twenty tons per hour to one hundred twenty-five tons per hour, then it is determined as a valid set value change event. The determination rule of the alarm confirmation is that when the alarm state in the control system changes from unconfirmed to confirmed, it is recorded, for example, after the carbon dioxide compressor outlet pressure limit alarm is confirmed by the operator, it is cleared, then a valid alarm confirmation event is generated.

[0029] After identifying these events, all events are mapped as time anchor points, and valid anchor points are screened by two types of invariant constraints of carbon element conservation and energy balance. The carbon element conservation constraint is that the change of carbon dioxide capture amount at the time anchor point needs to satisfy the material balance deviation between the calculated results of flue gas flow, flue gas carbon dioxide concentration, solvent circulation amount and reboiler heat load, etc. process amount, which is not more than two percent, for example, when the difference between the calculated capture amount and the measured value is less than one percent, the anchor point is determined as a valid anchor point. The energy balance constraint is that the difference between the steam heat supply and the analytical column heat load must be within the allowable deviation range, for example, the analytical column heat load is ten megawatts and the steam heat supply is ten point one megawatts, then the deviation is one percent, which meets the screening condition of valid anchor points. For the set of valid anchor points screened by the constraints, the least square fitting method is used to solve the piecewise linear time mapping function with global non-decreasing property and slope upper and lower bounds. The time mapping function requires that the time cannot be reversed in the global range, and the slope range is limited to zero point nine to one point one, so as to avoid too large alignment error caused by time stretching or compression. Through the mapping process, the initial estimation of clock drift and communication delay can be obtained, for example, the calculation result shows that there is five hundred milliseconds of clock offset and one hundred milliseconds of communication delay between the edge computing node and the cloud.

[0030] After obtaining the initial time mapping, a constrained dynamic time warping and phase correction operation is performed. Dynamic time warping is used to adjust the time mapping function in a small range to align the time axis of the edge computing side and the cloud side, while maintaining the monotonicity of the time mapping function and limiting the local slope to no more than one point one or below zero point nine. In the warping process, the cross-correlation function is used to calculate the correlation peak of the edge control intent and the cloud control intent within the maximum allowed lag of two seconds, and when the correlation peak is detected at a lag of zero point five seconds, the phase offset is set to zero point five seconds and is superimposed into the time mapping function for correction. The final output of the time alignment result not only provides a unified time reference, but also generates an alignment residual sequence containing the mean, variance and maximum deviation, for example, after alignment, the residual mean is zero point one second, the residual variance is zero point zero two seconds, and the maximum deviation is zero point three seconds.

[0031] After completing the time alignment, the state-dependent conversion is performed based on the process property relationship to unify the measurement units of different data. The volume flow is converted to mass flow by measuring the temperature, pressure and solvent composition, for example, the solvent has a density of one thousand kilograms per cubic meter at room temperature, and the volume flow is one hundred twenty cubic meters per hour, then the converted mass flow is twelve thousand tons per hour. When converting the steam flow to the heat load, the steam enthalpy difference calculation is used, for example, the steam mass flow is thirty tons per hour, and the enthalpy difference is two thousand five hundred kilojoules per kilogram, then the heat load is seventy-five megawatts. The solvent concentration is converted from mass fraction and volume fraction to mole fraction when necessary, for example, the solvent mass fraction is thirty percent, and the molar mass is forty-five grams per mole, then the converted mole fraction is thirty-two percent. The pressure, temperature and concentration are also unified to the standard measurement unit system to ensure that the subsequent conflict factor calculation has no numerical ambiguity.

[0032] To ensure the traceability and reliability of data, after completing unit conversion, the source credibility, alignment residual confidence band, version summary identifier and integrity check summary are attached to each time point. The source credibility is calculated by weighting the most recent calibration date of the instrument, online self-test results and missing data ratio. For example, if the most recent calibration time of an instrument is within a week and the online self-test passes completely, and the missing data ratio is less than one percent, the credibility can reach 0.95. The alignment residual confidence band is estimated according to the mean and variance of the alignment residual sequence. For example, the residual confidence interval is 0.05 seconds to 0.15 seconds. The version summary identifier includes the model name, model version number and training time. For example, the model name is absorption tower regulation model, the version is 1.0, and the training time is December 2024. The integrity check summary is calculated based on the hash function of the configuration parameter list to prevent the configuration parameters from being tampered with or lost. Finally, the data after time alignment, unit conversion and metadata labeling are integrated to form a synchronous control sequence carrying complete metadata. This sequence not only accurately represents the control intention of the edge computing side and the cloud side at the same time, but also provides the credibility, error range and model information related to the data, providing a reliable input basis for subsequent conflict factor calculation, weight generation and risk aggregation.

[0033] In the process of forming the synchronization control sequence carrying complete metadata, first, the density and enthalpy value are calculated according to the temperature, pressure and solvent composition. The calculation of density is based on the physical equation of the solvent. For example, when the solvent is a single amine solution, the density can be calculated by using the functional relationship of temperature, pressure and solvent composition. When the temperature is 35 degrees Celsius, the pressure is normal pressure, and the mass fraction of the solvent is 30%, the density can be calculated as 10,050 kg / m3. The calculation of enthalpy value uses thermodynamic state equation. For example, under the same conditions, the enthalpy value is 150,000 J / kg, which is used for subsequent heat conversion. After the calculation of density and enthalpy value is completed, the volume flow is converted into mass flow. The unit of volume flow is m3 / h, and the mass flow is obtained by multiplying the density. For example, the solvent flow is 120 m3 / h, and the density is 10,050 kg / m3, so the converted mass flow is 120,600 kg / h. Then, the steam flow is converted into heat load. The calculation of heat load is obtained by multiplying the steam mass flow and the enthalpy difference. For example, the steam mass flow is 30 t / h, and the enthalpy difference is 2,500 J / kg, so the converted heat load is 75 MW. Next, the mass fraction and volume fraction are converted into mole fraction. Mole fraction is used to ensure that the concentration of different components can be expressed uniformly. For example, the mass fraction of the solvent is 30%, the molar mass is 45 g / mol, and the density of the solvent solution is 1,000 kg / m3. After conversion, the mole fraction of the solvent is 32%. For gaseous components, the volume fraction can be converted by combining the standard state molar volume to ensure consistency of concentration data in different phases. After the above conversion is completed, the pressure, temperature and concentration are unified to the pre-designed quantity unit system. For example, the pressure is unified to kPa, the temperature is unified to Celsius, and the concentration is unified to mole fraction, to ensure consistency and comparability of data in subsequent calculations.

[0034] After completing the unit conversion, the source credibility is added to each time point. Source credibility is an important indicator to measure the reliability of the data at that time, which is calculated by the weighted sum of the latest calibration date, online self-checking results and missing data ratio. For example, if the pressure sensor has completed traceable calibration within the past seven days, the online self-checking shows that its output is stable and there is no offset, and the missing data ratio is less than one percent, then the source credibility score of the sensor can reach 0.95. If the calibration period of the temperature sensor exceeds 30 days or the missing data ratio exceeds 5%, the source credibility will be reduced to less than 0.8. The calculation formula of source credibility can use weighted average method, set the calibration date weight to 0.4, the online self-checking result weight to 0.4, and the missing data ratio weight to 0.2, to ensure that the credibility can reflect the health status of the sensor and also reflect the availability of real-time data. Then the alignment residual confidence band is added to each time point. The alignment residual confidence band is obtained based on the interval estimation of the residual sequence after time alignment, for example, by calculating the mean, variance and percentile of the residual sequence, the upper and lower confidence boundaries are given. If the residual mean of a certain period is 0.1 seconds and the residual variance is 0.02 seconds, then the confidence interval can be calculated as 0.05 seconds to 0.15 seconds. The confidence band can provide error boundaries for subsequent risk analysis, to judge the reliability of time alignment. The version summary identifier and integrity check summary are added to each time point. The version summary identifier includes the model name, model version and training time, for example, the name of the absorption tower control model is absorption tower regulation model, the version is 1.0, and the training time is December 2024. The integrity check summary is generated by a hash algorithm based on the model and configuration parameter list, for example, using the SHA-256 algorithm to summarize the configuration file including model hyperparameters, input variable definition, boundary conditions, etc., to ensure that any changes can be quickly detected. This summary string can effectively prevent data inconsistency caused by model version confusion or configuration file tampering.

[0035] Finally, time alignment, unit conversion, source credibility, alignment residual confidence band, version summary identifier and integrity check summary are summarized to form a synchronous control sequence carrying complete metadata. For example, the record of a certain time point in the synchronous control sequence includes time 10:36, solvent mass flow 126,000 kg / h, steam heat load 75 MW, solvent mole fraction 32%, source credibility 0.95, alignment residual confidence band 0.05-0.15 seconds, model version absorption tower regulation model 1.0, and integrity check summary a 64-bit hexadecimal string. This synchronous control sequence provides a traceable, verifiable and consistent data basis for subsequent conflict factor calculation, adaptive weight generation and risk aggregation, ensuring that the intelligent diagnosis and control of the carbon capture process have high reliability and interpretability.

[0036] The specific steps of step two include: first, based on the synchronization control sequence, extracting the control intention increment of the edge computing side and the cloud side in the same control cycle, pairing each control channel one by one to obtain the paired data set of the two sides in different preset quantization increments, and recording the rated range, rated energy capacity and constraint limit value of each channel. The control intention increment refers to the numerical change of the adjusting variable within the adjacent two sampling time points. For example, in the circulating pump speed control channel, when the speed in the last sampling period is 1000 rpm and the speed in the current period is 1050 rpm, the control intention increment is 50 rpm. When pairing each channel, the time consistency of the edge computing side and the cloud side is realized through the synchronization control sequence, and the channel identifier is matched to ensure that the same device and variable can be one-to-one corresponding. For each channel, the rated range is recorded, for example, the valve opening range is zero to one hundred percent; the rated energy capacity is recorded, for example, the heat load capacity of the reboiler is one hundred megawatts; and the constraint limit value is recorded, for example, the temperature at the top of the analytical tower should not exceed one hundred and twenty degrees Celsius. These data are used for subsequent normalization calculation of conflict factors.

[0037] After obtaining the paired data set, six types of conflict factors are calculated. Specifically, first, the direction reversal degree is calculated. The direction angle of the two side increments of the same channel is calculated first, and the direction angle takes an angle value of zero to one hundred and eighty degrees. The direction angle is equal to the dot product of the two increment vectors divided by the product of the modulus of the two increment vectors, and then the inverse cosine is taken. For example, in the solvent circulating pump channel, if the edge computing side increment is positive one hundred kilograms per hour and the cloud side increment is negative fifty kilograms per hour, the angle between them is close to one hundred and eighty degrees. Then, the proportion of channels with opposite signs to the total channels is calculated. For example, among the ten channels, four channels have opposite directions, so the proportion is zero point four. Finally, the direction reversal degree is equal to the weighted average of the direction angle divided by one hundred and eighty and the proportion of opposite signs, if the weights are zero point six and zero point four respectively, then the direction reversal degree is zero point six times the normalized value of the angle plus zero point four times the proportion of opposite signs.

[0038] Second, the amplitude asymmetry degree is calculated. The absolute value of the difference between the absolute values of the two side increments of the same channel is taken, and the sum of the absolute values of the two side increments plus a pre-set minimum positive lower limit value is taken as the denominator to divide. For example, in the solvent flow control channel, the edge computing side increment is twenty tons per hour, the cloud side increment is thirty tons per hour, the difference is ten tons per hour, and the sum of the absolute values is fifty tons per hour. If the lower limit value is zero point one ton per hour, then the amplitude asymmetry degree is ten divided by fifty point one, which is about zero point one nine nine. Finally, the arithmetic mean of all channels is taken as the global amplitude asymmetry degree.

[0039] Third, the calculation of phase mismatch degree is to calculate the cross-correlation function of control intentions on both sides within the specified maximum time lag range, take the absolute value of the time lag that produces the maximum correlation, and normalize it with the maximum time lag, for example, in the flue gas compressor power control channel, there is a zero-point five-second lag between the edge calculation side and the cloud side control instruction, and the maximum allowed lag is two seconds, so the normalized lag value is zero-point two-five; At the same time, the time center difference of the control intentions on both sides is calculated and normalized with the control cycle of the channel, for example, the time center difference is zero-point two seconds, and the control cycle is one second, so the result is zero-point two; The phase mismatch degree is equal to the weighted average of the two normalized results, and the weight can be set according to the dynamic characteristics of the actual regulated object, for example, the lag weight is zero-point seven, and the time center difference weight is zero-point three, so the phase mismatch degree is zero-point seven times zero-point two-five plus zero-point three times zero-point two, and the result is about zero-point two three-five.

[0040] Fourth, the calculation of energy impact degree is to convert the control intention increment of each channel into expected electric power step and expected heat load step through the energy conversion coefficient under the current working condition, for example, in the reboiler channel, if the steam flow increment is two tons per hour, and the enthalpy difference is two thousand five hundred kilojoules per kilogram, then the heat load step is one thousand three hundred eighty-nine kilowatts; In the circulating pump channel, if the speed increases by fifty revolutions per minute, the corresponding power step is ten kilowatts; After summing up the energy steps of all channels, normalize it with the sum of the rated electric power and rated heat load of the corresponding unit, for example, the total energy step is one thousand four hundred kilowatts, and the rated capacity is ten thousand kilowatts, then the energy impact degree is zero-point one four.

[0041] Fifth, the calculation of sensitive amplification degree is to obtain the local first-order sensitivity coefficient of the key process quantity to each input channel at the current operating point, for example, the sensitivity coefficient of the analytical tower top temperature to the solvent flow is zero-point zero two degrees Celsius per ton per hour, and the sensitivity coefficient of the reboiler steam quantity is zero-point zero five degrees Celsius per ton per hour; Multiply the solvent flow increment of twenty tons per hour by the sensitivity coefficient to get zero-point four degrees Celsius, and multiply the reboiler steam quantity increment of five tons per hour by the sensitivity coefficient to get zero-point two five degrees Celsius, and the sum of the two is zero-point six five degrees Celsius; Take the allowed change bandwidth of one degree Celsius, and the sensitive amplification degree is zero-point six five.

[0042] Sixth, the calculation of the boundary approximation degree is to calculate the remaining margin of each controlled constraint after the execution of the control intention on both sides, if the constraint is an upper limit, then subtract the expected value from the upper limit, if the constraint is a lower limit, then subtract the lower limit from the expected value, and negative values are counted as zero, for example, the analytical tower pressure upper limit is two hundred kilopascals, and the expected pressure is one hundred and ninety kilopascals, then the remaining margin is ten kilopascals, the safety bandwidth is fifty kilopascals, the normalized value is zero point two, and the normalized result is one minus the normalized result, which is zero point eight; if there are multiple constraints, then the maximum value of all constraints is taken as the boundary approximation degree. Finally, the six types of conflict factors calculated are standardized and mapped, so that the values of each conflict factor are in the interval of zero to one, and a standardized conflict factor vector is generated as the basis for subsequent weight calculation and risk aggregation, ensuring that the conflict diagnosis in the carbon capture process has comparability, traceability and stability.

[0043] The specific steps of step two include: first, based on the synchronization control sequence, the control intention increment of the edge computing side and the cloud side in the same control period is extracted, and the control channels are paired one by one to obtain paired data sets of different preset quantization increments on both sides, and the rated range, rated energy capacity and constraint limit value of each channel are recorded. Control intention increment refers to the numerical change of the adjusting variable within the adjacent two sampling time points, for example, in the circulating pump speed control channel, when the speed is one thousand revolutions per minute in the last sampling period, the speed in this period is one thousand and fifty revolutions per minute, and the control intention increment is fifty revolutions per minute. When pairing each channel, the time consistency of the edge computing side and the cloud side is realized through the synchronization control sequence, and the channels are matched by channel identification to ensure that the same equipment and variables can be matched one by one. For each channel, the rated range is recorded, for example, the valve opening range is zero to one hundred percent; the rated energy capacity is recorded, for example, the heat load capacity of the reboiler is one hundred megawatts; and the constraint limit value is recorded, for example, the temperature at the top of the analytical tower should not exceed one hundred and twenty degrees Celsius, which is used for subsequent normalization calculation of conflict factors.

[0044] After obtaining the paired data set, six types of conflict factors are calculated, specifically: first, the calculation of the direction reversal degree is to calculate the direction angle of the increment of the same channel on both sides first, the direction angle takes an angle value of zero to one hundred and eighty degrees, the direction angle is equal to the dot product of the two increment vectors divided by the product of the modulus of the two increment vectors, and then takes the inverse cosine, for example, in the solvent circulating pump channel, if the edge computing side increment is positive one hundred kilograms per hour and the cloud side increment is negative fifty kilograms per hour, then the angle between them is close to one hundred and eighty degrees; then calculate the proportion of channels with opposite signs in all channels, for example, four of the ten channels have opposite directions, then the proportion is zero point four; finally, the direction reversal degree is equal to the weighted average of the direction angle divided by one hundred and eighty and the proportion of opposite signs, if the weights are zero point six and zero point four respectively, then the direction reversal degree is zero point six times the angle normalized value plus zero point four times the opposite sign proportion.

[0045] Second, the calculation of amplitude asymmetry is the absolute value of the difference between the absolute values of the increments on both sides of the same channel, and the sum of the absolute values of the increments on both sides plus a pre-set minimum positive lower limit value as the denominator. For example, in the solvent flow control channel, the edge calculation side increment is twenty tons per hour, and the cloud side increment is thirty tons per hour, the difference is ten tons per hour, the sum of the absolute values of the two is fifty tons per hour, and if the lower limit value is zero point one ton per hour, the amplitude asymmetry is ten divided by fifty point one, the result is about zero point one nine nine; finally, take the arithmetic average of all channels as the global amplitude asymmetry.

[0046] Third, the calculation of phase mismatch degree is to calculate the cross-correlation function of the control intentions on both sides within a specified maximum time lag range, take the absolute value of the time lag that produces the maximum correlation, and normalize it with the maximum time lag. For example, in the flue gas compressor power control channel, there is a lag of zero point five seconds between the control instructions of the edge calculation side and the cloud side, and the maximum allowed lag is two seconds, so the normalized lag value is zero point two five; at the same time, the time center difference of the control intentions on both sides is calculated and normalized with the control period of the channel, for example, the time center difference is zero point two seconds and the control period is one second, so the result is zero point two; the phase mismatch degree is equal to the weighted average of the two normalized results, and the weight can be set according to the dynamic characteristics of the actual regulated object. For example, the lag weight is zero point seven and the time center difference weight is zero point three, so the phase mismatch degree is zero point seven times zero point two five plus zero point three times zero point two, the result is about zero point two three five.

[0047] Fourth, the calculation of energy impact degree is to convert the control intention increments of each channel into expected power steps and expected heat load steps through the energy conversion coefficient under the current operating condition. For example, in the reboiler channel, if the steam flow increment is two tons per hour and the enthalpy difference is two thousand five hundred kilojoules per kilogram, the heat load step is one thousand three hundred eighty-nine kilowatts; in the circulating pump channel, if the speed increases by fifty revolutions per minute, the corresponding power step is ten kilowatts. After summing up the energy steps of all channels, normalize it with the sum of the rated electric power and rated heat load of the corresponding unit. For example, the total energy step is one thousand four hundred kilowatts, and the rated capacity is ten thousand kilowatts, so the energy impact degree is zero point one four.

[0048] Fifth, the calculation of sensitive amplification degree is to obtain the local first-order sensitivity coefficient of the key process quantity to each input channel at the current operating point. For example, the sensitivity coefficient of the analytical tower top temperature to the solvent flow is zero point zero two degrees Celsius per ton per hour, and the sensitivity coefficient to the reboiler steam quantity is zero point zero five degrees Celsius per ton per hour. Multiplying the solvent flow increment of twenty tons per hour by the sensitivity coefficient gives zero point four degrees Celsius, and multiplying the reboiler steam quantity increment of five tons per hour by the sensitivity coefficient gives zero point two five degrees Celsius. The sum of the two is zero point six five degrees Celsius, and the allowed change bandwidth is one degree Celsius, so the sensitive amplification degree is zero point six five.

[0049] Sixth, the calculation of the boundary approximation degree is to calculate the remaining margin of each controlled constraint after the execution of the control intention on both sides, if the constraint is an upper limit, then subtract the expected value from the upper limit, if the constraint is a lower limit, then subtract the lower limit from the expected value, and negative values are counted as zero, for example, the upper limit of the analytical tower pressure is two hundred kilopascals, the expected pressure is one hundred and ninety kilopascals, then the remaining margin is ten kilopascals, the safety bandwidth is fifty kilopascals, the normalized value is zero point two, and one minus the normalized result is zero point eight; if there are multiple constraints, then the maximum value of all constraints is taken as the boundary approximation degree.

[0050] The definitions of key parameters in the calculation formula of "six conflict factors" and the working condition association rules are supplemented: the acquisition method of the weighted coefficient in the direction reversal degree: weight item one is the direction angle normalization value, and the direction angle θ is in the range of [0, 180] degrees; weight item two is the opposite sign channel proportion, defined as the number of opposite sign channels divided by the total number of channels; the weighted average coefficient can be set as 0.5:0.5, or adjusted according to the direction change sensitivity of the control system response characteristic curve; this characteristic curve can be established through a standard step response test; if the specific working condition has a more sensitive situation to the direction polarity change (such as a steam regulating valve), the direction angle weight can be increased to 0.7 to reflect the importance of direction conflict. The maximum lag range in phase mismatch degree: the maximum lag time window Δt_max is 3 to 5 control periods, with a default value of 3×T_c, where T_c is the length of a single control period (such as 5s); if the system response has an inertial amplification effect (such as a high-temperature heat exchanger), Δt_max can be taken as 5×T_c; this value is set according to the definition of the identification time window of hysteresis in the "Process Control System Engineering Design Standard" (GB / T38898-2020). The truncation proportion in the truncated arithmetic mean: the upper and lower limits are fixed at 15% and 30% respectively, with a default value of 20%, which is derived from the extreme value exclusion strategy in the "Statistical Quality Control Engineering Handbook" for multi-factor aggregation; it can be adjusted according to industry experience based on historical conflict data distribution curve. Working condition related parameters and data source explanation: energy conversion coefficient in energy impact degree: control intention increment (such as opening Δu) is converted into equivalent electric power P_e and heat load Q_h, using a linear approximation conversion model: P_e=α_e*Δu, Q_h=α_h*Δu, where α_e and α_h are the electric energy and heat energy conversion coefficients corresponding to the unit opening. Coefficients α_e and α_h can be extracted from device energy efficiency evaluation data or gain values in control models, such as flow-load conversion modules in DCS systems or P&ID design parameters; if there is no direct record, a linear regression model can be obtained through operation data fitting (refer to IEC61512-3 batch processing energy mapping specification). The method of obtaining sensitivity coefficient in sensitive amplification degree: the sensitivity coefficient is defined as the local partial derivative of the key process quantity with respect to the input variable (∂y / ∂u), which is calculated from historical operation data using linearization method at operating point; a rolling window (recommended 10-30 points) is used to fit a local first-order linear model to obtain it; the key process quantities include carbon dioxide absorption rate, solvent temperature, tower pressure, etc., which need to meet the "process safety relevance" and "regulation accessibility" principles; if a neural network or fuzzy system is used as a control agent model, the sensitivity coefficient can be directly extracted through back propagation or gradient tracking.To avoid relying on trial and error adjustment, the default setting values and adjustment ranges of all key parameters of the present application are recommended to be derived and verified according to the following standards or literature: national standards and industry specifications: GB / T38898-2020 "Process control system engineering design standard", HG / T20571-2015 "Chemical control system design specification", NB / T20384-2021 "Carbon dioxide capture engineering design guide", IEC61512-3 "Batch processing control system-energy efficiency data interface". Finally, the six types of conflict factors calculated are standardized and mapped, so that each conflict factor value is in the interval of zero to one, and a standardized conflict factor vector is generated as the basis for subsequent weight calculation and risk aggregation, ensuring that the conflict diagnosis in the carbon capture process has comparability, traceability and stability.

[0051] The standardized conflict factor vector refers to the six types of conflict factors calculated based on the synchronous control sequence are respectively subjected to preset standard processing, each conflict factor is mapped to the standardized range of zero to one, and then is summarized according to the preset arrangement order to obtain the standardized conflict factor vector. The direction reversal degree is standardized: the direction reversal degree represents the degree of difference in control intention direction between the edge computing side and the cloud side in the same channel, and its value is derived from the weighted result of the direction angle of the two instructions and the opposite sign ratio. To avoid the dimensional difference between different channels affecting risk calculation, the direction reversal degree is mapped to the range of zero to one. For example, under a certain working condition, the control instruction angle of the edge computing side and the cloud side is ninety degrees, corresponding to a normalized value of zero point five, and if three of the ten channels have opposite signs, the opposite sign ratio is zero point three, and the weighted direction reversal degree is zero point forty-eight. The result is standardized by linear mapping from the minimum value zero to the maximum value one, and the final standardized direction reversal degree is zero point forty-eight. The amplitude asymmetry degree is standardized: the amplitude asymmetry degree reflects the relative proportion of the difference in adjustment amplitude between the two sides in the same channel. The amplitude asymmetry degree obtained after calculation is directly mapped to the preset minimum value zero and maximum value one. For example, in the solvent circulating pump channel, the edge computing side increment is twenty tons per hour, the cloud side increment is thirty tons per hour, the difference between the two is ten tons per hour, the absolute value sum is fifty tons per hour, and the normalized result is zero point two, which is directly mapped to the standardized range of zero to one.

[0052] The phase mismatch degree is standardized: the phase mismatch degree is used to measure the time offset degree of the control instructions on the edge computing side and the cloud side, including the weighted average of the maximum correlation lag and the time center difference. Taking a certain control channel as an example, if the maximum correlation lag is 0.5 seconds and the upper limit of the lag is 2 seconds, the lag part is normalized to 0.25; the time center difference is 0.1 seconds, and the control period is 1 second, which is normalized to 0.1. The weighted average of the two is 0.17, which is mapped to the standardized range of 0 to 1 as the standardized phase mismatch degree input. The energy impact degree is standardized: the energy impact degree is used to measure the impact amplitude of the control instruction adjustment on the system power and thermal load. Under a certain working condition, if the reboiler steam flow increment is converted to 1,200 kW, the circulating pump power increment is converted to 20 kW, and the sum is 1,220 kW, and the rated energy capacity is 10,000 kW, the normalized energy impact degree is 0.122. After mapping the value to the standardized range of 0 to 1, it is directly used for subsequent calculation.

[0053] The sensitive amplification degree is standardized: the sensitive amplification degree is used to measure the influence intensity of the control instruction on the key process quantity at the current operating point, for example, the sensitivity coefficient of the top temperature of the analytical column is 0.05 degrees Celsius per ton per hour, when the solvent flow increment is 20 tons per hour, the expected temperature change is 1 degree Celsius, and the allowed temperature fluctuation bandwidth is 2 degrees Celsius. The original value of the sensitive amplification degree is 0.5, which is mapped to the standardized range of 0 to 1 and used for risk calculation. The boundary approximation degree is standardized: the boundary approximation degree is used to measure whether the control instruction pushes the process quantity to the constraint limit. For example, the upper limit of the top pressure of the analytical column is 200 kPa, the expected pressure is 180 kPa, the remaining margin is 20 kPa, and the safety bandwidth is 50 kPa. The normalized result is 0.6, and the boundary approximation degree is 0.4 obtained by subtracting 1 from the value. The value is directly mapped to the range of 0 to 1 and involved in the calculation. Finally, the standardized direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree are summarized in a fixed order to form a standardized conflict factor vector. For example, the standardized results of the six types of conflict factors at a certain time are 0.48, 0.2, 0.17, 0.122, 0.5 and 0.4, respectively. The standardized conflict factor vector formed finally is (0.48, 0.2, 0.17, 0.122, 0.5, 0.4). This vector will be used as the core input for subsequent adaptive weight calculation and bounded gain aggregation, ensuring that the basic data for risk calculation has uniform dimensions, strong comparability and is suitable for dynamic analysis under multiple working conditions.

[0054] The restricted adaptive weight vector is obtained by the following steps: first, a data freshness score is calculated. The data freshness score is used to measure the comprehensive quality of each type of data for the six types of conflict factors in terms of time validity, communication delay and integrity. Under a unified time reference, the time lag value, communication delay value and missing rate are calculated for each data channel. The time lag value refers to the time difference between the current unified time and the latest valid timestamp of the channel, for example, the current unified time is ten o'clock zero minutes, and the latest valid timestamp is nine o'clock fifty-nine minutes and fifty seconds, then the time lag value is ten seconds. The communication delay value is derived from the communication delay estimation in the time mapping stage, for example, the data transmission delay from the edge computing side to the cloud side is zero point three seconds, then the communication delay value is zero point three seconds. The missing rate is the ratio of the number of invalid samples to the total number of samples within a fixed evaluation window, for example, within a five-minute evaluation window, there are three hundred sampling points, fifteen of which are invalid, then the missing rate is fifteen divided by three hundred, equal to zero point zero five.

[0055] After obtaining the time lag value, the communication delay value and the missing rate, they are respectively standardized. The time lag value is linearly inversely mapped according to the preset maximum acceptable time difference, for example, the maximum acceptable time difference is thirty seconds, and the time lag value of ten seconds is mapped to one minus ten divided by thirty, and the timeliness component is obtained as zero point six seven. The communication delay value is linearly inversely mapped according to the maximum acceptable delay, for example, the maximum acceptable delay is one second, and zero point three seconds are mapped to one minus zero point three divided by one, and the channel delay component is obtained as zero point seven. The missing rate is converted into the integrity component through a minus mapping relationship, for example, the missing rate is zero point zero five, and the integrity component is one minus zero point zero five, and the integrity component is obtained as zero point nine five. The timeliness component, the channel delay component and the integrity component are taken as the geometric mean, and the data freshness score is obtained as the cube root of (zero point six seven times zero point seven times zero point nine five), which is approximately equal to zero point seven seven. The value is used for subsequent weight calculation. Secondly, the conflict duration is calculated, which is used to measure the continuity and intensity of the conflict in time. Within the same evaluation window, the instantaneous conflict indicator is constructed, which is equal to the maximum value of the structural conflict strength and the amplification strength multiplied by the boundary strength. The structural conflict strength is the arithmetic mean of the direction reverse degree, the amplitude asymmetry degree and the phase mismatch degree, for example, they are zero point four, zero point three and zero point two respectively, then the structural conflict strength is (zero point four plus zero point three plus zero point two) divided by three, equal to zero point three three. The amplification strength is the arithmetic mean of the energy impact degree and the sensitive amplification degree, for example, they are zero point five and zero point four respectively, then the amplification strength is (zero point five plus zero point four) divided by two, equal to zero point four five. The boundary strength is the boundary approximation degree, for example, it is zero point six, then the maximum value of the instantaneous conflict indicator is zero point one five, which is the product of the structural conflict strength and the amplification strength, and the boundary strength zero point six, and finally the instantaneous conflict indicator is zero point six.

[0056] A conflict threshold is set, for example, the conflict threshold is zero point five, if the instantaneous conflict indicator exceeds zero point five for fifty consecutive seconds, the proportion is zero point five in the evaluation window of total one hundred seconds, and the average value of the instantaneous conflict indicator when exceeding the threshold is zero point six, then the unweighted conflict duration is zero point five times zero point six, equal to zero point three. Again, the result is applied linearly increasing weight from near to far by time, for example, the weight of the last thirty seconds is one, the weight of the middle thirty seconds is zero point eight, and the weight of the farthest forty seconds is zero point six. After weighting, superposition and normalization, the conflict duration is zero point three five, and is mapped to the range of zero to one.

[0057] In the construction process of the limited adaptive weight vector, the time maximum lag value, that is, the upper limit of the time lag parameter for measuring control data synchronization, is set to one to five seconds by default. This value comes from the actual sampling period and feedback response time of the carbon capture process control system. In the industrial coal-fired power plant or natural gas capture scene, the typical feedback control period is two to three seconds, so it is recommended to set the maximum acceptable lag time to three seconds. If the control system has a high-speed feedback mechanism (such as the edge side running at a millisecond-level sampling frequency), the maximum lag time threshold can be calculated by multiplying the sampling period by ten to twenty times. Secondly, the calibration period of the weight update, that is, the length of the time window of the limited adaptive weight vector update, is set to a fixed period in the range of ten minutes to thirty minutes by default. This period setting aims to balance the response sensitivity to conflict trend changes and the stability of system disturbances, and its specific value is recommended to be determined based on the following three types of indicators: one is the conflict factor fluctuation frequency, two is the device switching frequency, and three is the response inertia of the control execution structure. If the conflict factor fluctuates frequently (for example, the conflict duration exceeds three minutes every five minutes), the calibration period should be shortened; if the control system executor has large inertia (such as solvent regenerator temperature control), the calibration period can be appropriately extended to avoid repeated adjustments.

[0058] For various weight coefficients involved in the weight composition, such as the weight distribution of timeliness, delay and completeness in the data freshness score, the time proportion and intensity weighting factor in the conflict duration, it is recommended to use the normalized three-factor equal-weight weighting principle, or to weight and optimize according to the key target of device operation. For example, in a steam-driven capture system, it is recommended to increase the sensitive weight of the conflict duration factor corresponding to the energy impact degree to enhance the response capability to the sudden change of thermal load. For the setting of the weight upper limit and the index sensitive coefficient according to the safety level, it is recommended to map the parameters according to the definition of the safety level in the industry actual standard. For example, referring to the constraint strength of the first to fourth safety levels in the process safety management guide, the corresponding weight upper limits are set to 20%, 35%, 50% and 70%, and the index sensitive coefficients are one, one and a half, two and three, respectively. This setting ensures that in the high safety level scene, the change of the conflict factor has a stronger driving effect on the maximum weight composition, so that the adjustment suggestion is triggered earlier. And if there are uncertain factors in the actual industrial environment that require dynamic adjustment of the above parameters, a rolling self-tuning mechanism can be built based on process modeling results, historical data distribution and bias evaluation indicators to periodically correct the above empirical parameters. The self-tuning mechanism can refer to the parameter updating strategy of statistical process control methods, fuzzy weighted optimization methods or Kalman filter models, so that the entire weight calculation process has a self-adaptive adjustment capability while having a definite boundary, further improving the usability and robustness of the method in complex working conditions.

[0059] Source credibility is used to measure the reliability of the data source in terms of calibration status, device self-checking and data continuity, and is an important factor in limited adaptive weight vector calculation. The calculation of source credibility consists of the following parts: the latest calibration date of the instrument is used to reflect whether the measurement accuracy of the data source is reliable. First, record the date of the last completed measurement calibration for each measuring point, and calculate the time interval by subtracting the current unified time from the date. For example, a certain pressure transmitter completed calibration one hundred and twenty days before the current date, so the time interval is one hundred and twenty days. Then, linearly decay map according to the time interval and the preset maximum valid calibration period. If the maximum valid calibration period is one hundred and eighty days, then the calibration component is equal to one minus the time interval divided by one hundred and eighty, for example, the calibration component of one hundred and twenty days is one minus one hundred and twenty divided by one hundred and eighty, equal to zero point three three. If the time interval exceeds one hundred and eighty days, the calibration component takes zero, indicating that the data has no calibration credibility. In actual application, for example, if a flowmeter is only thirty days away from the last calibration, the calibration component is one minus thirty divided by one hundred and eighty, equal to zero point eight three, indicating that the data from the flowmeter source is highly credible in terms of calibration.

[0060] The online self-check result is used to reflect the self-diagnosis status of the data acquisition device. The device generates a self-check score through the self-check function, with the score ranging from zero to one, zero indicating that the device is completely abnormal, and one indicating that the device is completely normal. The self-check score can be obtained by weighting based on the following indicators: signal stability: for example, analyze the standard deviation of the flow signal within one minute, if the standard deviation is less than one percent of the rated range, then the full score of one is obtained, otherwise it is decreased in proportion. Hardware status: the device self-check function can return a health status flag, for example, the power voltage is normal and the score is one, and the power is low and the score is decreased in proportion to the deviation. Historical failure rate: for example, the number of self-check exceptions recorded in the past seven days, if it is zero, the score is one, if there is an exception, the score is deducted in proportion. For example, the self-check score of a certain temperature sensor is signal stability of 0.95, hardware status of 1, and historical failure rate of 0.85, and the average is obtained by weighting the three indicators equally. The self-check score is (0.95+1+0.85) divided by 3, which is equal to 0.93.

[0061] The missing proportion is used to reflect the data continuity within a fixed evaluation window. The missing proportion is equal to the number of invalid samples divided by the total number of samples, for example, in a five-minute evaluation window, a total of 300 samples are collected, and six samples are invalid due to communication packet loss, then the missing proportion is six divided by 300, which is equal to 0.02. The missing proportion is mapped to one minus the missing proportion, for example, one minus 0.02 is equal to 0.98, indicating that the data continuity is very high. If the missing proportion exceeds 20%, the component takes zero, indicating that the data continuity is lost and does not have reliability. The calibration component, online self-check score and missing proportion component are combined and calculated by weighted geometric mean, and the weight can be set according to the process safety requirements, for example, the weight of the three components is one third. For example, the calibration component of a certain flowmeter is 0.83, the self-check score is 0.93, and the missing proportion component is 0.98, then the source credibility is equal to the cube root of (0.83*0.93*0.98), which is approximately equal to 0.91.

[0062] The restricted adaptive weight vector is formed. For each component of the six conflict factors, first multiply the source credibility weight by the data freshness score, for example, the source credibility weight is 0.9, the data freshness score is 0.777, and the result is 0.693. Then multiply the result by the exponential function of the conflict duration, the index is determined by the sensitivity coefficient corresponding to the device security level, for example, the device security level is level 2, and the sensitivity coefficient is 1, then the exponential function of the conflict duration 0.35 is 0.35, and the result is 0.35, and then multiply 0.693, the result is 0.243. The current weight and the last period weight are convexly combined according to the fixed smoothing coefficient, for example, the last period weight is 0.2, and the smoothing coefficient is 0.7, then the weighted result is 0.7 times 0.243 plus 0.3 times 0.2, which is equal to 0.2341. Finally, the smoothing weight is truncated according to the weight upper limit corresponding to the device security level, for example, the weight upper limit of level 2 is 0.5, and the current weight 0.2341 is lower than the upper limit, so it does not need to be truncated, and the final restricted adaptive weight vector is the weighted result set of the six conflict factors, for example, (0.2341, 0.2, 0.21, 0.22, 0.25, 0.23).

[0063] The restricted adaptive weight vector ensures that in the conflict risk assessment of the carbon capture process, the data quality, conflict intensity and time continuity can be considered, and the differentiated weight upper limit of the device security level is combined to realize the dynamic adjustment of the risk weight, so that the evaluation result has real-time performance and safety, and provides a high-reliability basic weight input for subsequent bounded gain aggregation and risk value calculation.

[0064] The uncalibrated risk value is used to quantitatively evaluate the risk before calculating the coupling risk coefficient. Its generation process includes gain limitation, weighted aggregation and double threshold shaping, and the final result is obtained by convex combination smoothing. The specific calculation steps are as follows: construction and application of gain limitation function: gain limitation functions are established for direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree. The gain limitation function adopts a smooth saturation form, that is, the index value of the conflict factor is multiplied by the corresponding gain coefficient as the numerator, and one plus the gain coefficient multiplied by the index value is taken as the denominator to form a fraction expression, and the smaller value of the result and the corresponding gain upper limit is taken, so as to avoid the nonlinear amplification of a single conflict factor to the overall risk in extreme cases. For example, when the direction reversal degree is 0.6 and the growth coefficient is 2, the calculation result is 2 times 0.6 divided by 1 plus 2 times 0.6, which is 0.554. If the gain upper limit is 0.7, the gain limited direction reversal degree is 0.554.

[0065] The determination of the growth coefficient and the upper limit of the gain takes values according to the classification of the device safety level. For example, when the safety level is level one, the growth coefficient and the upper limit of the gain take lower values to ensure that the risk quantification is more conservative. When the safety level is level four, the growth coefficient and the upper limit of the gain take higher values to improve the sensitivity to risk signals. The six components of the standardized conflict factor vector are respectively input into the gain limiting function to obtain a gain limited factor vector, thereby providing a basis for subsequent weighted calculation.

[0066] Formation of the weighted component set and the truncated arithmetic mean: the gain limited factor vector and the limited adaptive weight vector are multiplied component by component to obtain a weighted component set. The set can reflect the actual contribution of each conflict factor in the comprehensive risk quantification. For example, when the energy impact degree is 0.5 and the corresponding weight is 0.8, the weighted result is 0.4. The truncated arithmetic mean is performed on the weighted component set, and the truncated proportion is a fixed value between the preset minimum proportion and the maximum proportion. For example, the minimum proportion is 10%, and the maximum proportion is 20%. The truncated arithmetic mean is used to eliminate the influence of abnormal values. For example, when the six weighted components are 0.4, 0.38, 0.42, 0.7, 0.69 and 0.2, if the truncated proportion is 10%, the maximum value 0.7 and the minimum value 0.2 are removed, and the average of the remaining four values is taken to obtain an aggregated base value of 0.47.

[0067] Double threshold shaping and convex combination smoothing: double threshold shaping and convex combination smoothing are performed on the aggregated base value to ensure stable output of the uncalibrated risk value. The low threshold and the high threshold are set: the low threshold and the high threshold are fixed values, and the low threshold is less than the high threshold. For example, the low threshold is 0.2, and the high threshold is 0.7. Low threshold interval processing: when the aggregated base value does not exceed the low threshold, the output is zero. For example, when the aggregated base value is 0.15, the output is 0.0. Threshold interval proportion mapping: when the aggregated base value is between the low threshold and the high threshold, the output value is equal to the aggregated base value minus the low threshold, and then divided by the high threshold minus the low threshold. For example, when the aggregated base value is 0.5, the result is (0.5-0.2) divided by (0.7-0.2), which is equal to 0.6. High threshold interval processing: when the aggregated base value exceeds the high threshold, the output value is equal to one minus the exponential function, and the argument of the exponential function is the product of the aggregated base value minus the high threshold and a fixed convergence coefficient. For example, the convergence coefficient is 5, and when the aggregated base value is 0.8, the exponential term is 5 times 0.1, which is equal to 0.5, and the exponential function is about 0.607, and the output value is one minus 0.607, which is equal to 0.393. The output of the current period and the output of the last evaluation period are combined into a convex combination according to a fixed smoothing coefficient. For example, the smoothing coefficient is 0.7, and the final output is equal to 0.7 times the current period result plus 0.3 times the last period result. The smoothed result is limited in the range of zero to one, which is the uncalibrated risk value. For example, if the uncalibrated risk value of the last period is 0.5, and the current period calculation value is 0.6, then the final uncalibrated risk value is 0.7 times 0.6 plus 0.3 times 0.5, which is equal to 0.57.

[0068] It needs to be noted that the growth coefficient is used to control the response slope of the conflict factor, which is defined as the growth rate of the gain contribution per unit increase of the conflict factor. In practical applications, the determination of the growth coefficient needs to consider system sensitivity, control rhythm, and safety level: safety level is level one (high safety sensitivity): the growth coefficient is recommended to be set to a small value, such as zero point five, to avoid triggering a high risk evaluation for a small conflict factor value; safety level is level two or three: the growth coefficient is recommended to be set to one to one point five, to maintain moderate linear response; safety level is level four (low safety sensitivity): the growth coefficient can be set to two to three, to allow rapid amplification of the conflict indicator. The coefficient can be fitted by the system disturbance caused by historical conflict events, and the best response slope is found by using the least mean square error principle.

[0069] The upper limit of the gain is used to constrain the maximum influence ability of a single conflict factor, to prevent an abnormal value factor from having an overwhelming influence in weight aggregation. It is recommended to be set in conjunction with the tolerance bandwidth or load margin of the device: high safety level scenarios (such as supercritical capture devices) can be set to zero point six to zero point seven five; medium safety level scenarios (such as standard amine liquid capture processes) are recommended to be set to zero point eight; low safety level scenarios can be relaxed to zero point nine to one. The upper limit of the gain value can be set by referring to the constraint margin index in the system process design (such as the maximum steam load safety threshold or the maximum control loop overshoot ratio), to ensure consistency with the physical boundary.

[0070] The tailing ratio controls the influence of the abnormal weighting factor on the aggregated base value, which is recommended to be selected in the following range: when the system disturbance is frequent or the data quality is easy to fluctuate, the tailing ratio is recommended to be set to twenty to thirty percent; if the system has a stable operating environment, it can be set to ten to twenty percent; extreme conditions (such as high load operation or switching stage) can be increased to forty percent for noise suppression. The ratio can be dynamically set by statistical analysis of the skewness and kurtosis of the conflict factor aggregation distribution, or the optimal tailing interval can be inversely deduced according to the proportion of the disturbance contribution of historical abnormal events to the indicator.

[0071] Low threshold and high threshold: the low threshold should be set to the warning lower limit of the acceptable fluctuation of the aggregated base value, and the recommended range is zero point one to zero point two; the high threshold is set to the actual risk inflection point of the conflict response, and the recommended range is zero point five to zero point six five. Convergence coefficient: used to control the convergence rate of the output value to one in high risk situations; it is recommended to be set to one to two for high safety level to improve response sensitivity; it is recommended to be set to zero point five to one for medium and low safety level scenarios. The convergence coefficient can be determined by numerical simulation fitting based on the functional relationship between conflict influence delay and response cost, to determine the asymptotic speed of the exponential function to the upper limit value.

[0072] The mixing ratio of the smoothing coefficient in the convex combination is used for the weighted fusion between the current risk value and the historical risk value. It is suggested that the default range is set to be from 0.6 to 0.8; the smoothing coefficient can be appropriately increased to maintain the response stability when the working condition fluctuates sharply; and the smoothing coefficient can be reduced to 0.5 to enhance the sensitivity in the steady state operation stage. The coefficient can be automatically adjusted by constructing a function model of the historical risk value change standard deviation and the process variable change frequency. The parameter setting has a clear engineering correspondence, and can be obtained through industry experience, industrial data backtracking analysis or controller parameter setting method.

[0073] The uncalibrated risk value is looked up in the preset uncalibrated risk-coupling risk table to obtain the corresponding coupling risk coefficient, and a disposal suggestion is output based on the preset grading interval corresponding to the coupling risk coefficient, to guide the dynamic regulation and control of the carbon capture process control system. The specific process is as follows:

[0074] Establishing the uncalibrated risk-coupling risk table: first, based on a large amount of historical operation data, accident records and intervention effects, a mapping relationship between the uncalibrated risk value and the coupling risk coefficient is established. The uncalibrated risk value is a continuous value in the range of zero to one, and the coupling risk coefficient is also a continuous value in the range of zero to one, which is formed by combining piecewise linear interpolation and nonlinear correction. For example, the uncalibrated risk value is divided into a low risk interval (zero to zero point three), a medium risk interval (zero point three to zero point seven) and a high risk interval (zero point seven to one), and the response curve of the coupling risk coefficient is fitted in each interval using historical data. In the low risk interval, an approximate linear mapping is adopted, for example, when the uncalibrated risk value is 0.2, the coupling risk coefficient is 0.21; in the medium risk interval, a nonlinear amplification coefficient is introduced, for example, when the uncalibrated risk value is 0.5, the coupling risk coefficient is 0.6; and in the high risk interval, a saturation convergence curve is adopted, for example, when the uncalibrated risk value is 0.9, the coupling risk coefficient is 0.95. The table is dynamically updated by combining off-line training and on-line verification, and when new accident cases or intervention effect data are added, the table can be adaptively adjusted to ensure that the mapping of the coupling risk coefficient is consistent with the running characteristics of the current device.

[0075] Grading according to the coupling risk coefficient: the coupling risk coefficient found is compared with the preset risk grading interval to determine the risk level and provide a basis for subsequent disposal suggestions.

[0076] The risk grading interval is divided in the range of zero to one, for example: low risk interval: zero to zero point three, indicating that the device is in normal operation state, only the current control strategy needs to be maintained; medium risk interval: zero point three to zero point seven, indicating that there is potential fluctuation risk in the device, parameter pre-adjustment or load balance is needed; high risk interval: zero point seven to one, indicating that the device operation has significant risk, preventive regulation measures or forced intervention strategy need to be triggered. For example, when the uncalibrated risk value is zero point six, the coupling risk coefficient obtained by the comparison table is zero point six five, which is in the medium risk interval, then the system will mark the risk level as medium and call the corresponding medium risk disposal strategy.

[0077] Output disposal suggestion and execute dynamic regulation: based on the risk level and the coupling risk coefficient, generate disposal suggestion for the carbon capture process, and pass the disposal suggestion to the carbon capture process control system to guide it to execute dynamic regulation.

[0078] The disposal suggestion includes: when the risk level is low risk, it is suggested to keep the current operation parameters unchanged, only periodic monitoring is needed, for example, maintaining the current absorption tower temperature at one hundred and twenty degrees Celsius, and the regeneration tower pressure at zero point two megapascal; when the risk level is medium risk, it is suggested to make slight adjustment, for example, reducing the absorption tower liquid circulation flow by five percent, or reducing the regeneration tower reboiler heat load by three percent, to reduce the possibility of subsequent risk rising;

[0079] When the risk level is high risk, it is suggested to immediately take forced intervention measures, for example, reducing the steam flow of the carbon capture device by twenty percent, or directly triggering the bypass mode, to avoid equipment damage or safety accidents caused by too high risk. For example, when the coupling risk coefficient calculated by the system is zero point eight, the automatically generated disposal suggestion is to reduce the regeneration tower heat load by ten percent and at the same time reduce the solvent flow by five percent, the adjustment instruction is issued to the field execution unit through the process control system, and finally the parameter adjustment is completed within two minutes, ensuring that the risk falls back to the medium risk interval from the high risk interval. This step makes the risk quantification process have explainability and executability through the mapping of uncalibrated risk value and coupling risk coefficient, ensuring that the carbon capture process control system can dynamically adjust the operation parameters according to the actual risk level, effectively improving the safety and stability of the system.

[0080] Embodiment 2: The carbon capture process intelligent diagnosis control system based on AI decision, specifically includes:

[0081] The alignment and unification module obtains the control intention data of the edge computing side and the cloud side in the same time period, aligns the data time reference of the control intention and unifies the measurement units, and generates a synchronous comparison sequence for comparative analysis;

[0082] The conflict quantification module calculates six types of conflict factors based on the synchronous reference sequence, the six types of conflict factors including direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree, wherein the direction reversal degree is used to measure the proportion of the instruction direction angle and the sign opposite, the amplitude asymmetry degree is used to measure the relative difference of the adjustment amplitude on both sides, the phase mismatch degree is used to measure the time center offset and the period offset, the energy impact degree is used to measure the expected power step or steam load step strength, the sensitive amplification degree is used to measure the amplification effect of the key process quantity under the current working condition sensitivity, and the boundary approximation degree is used to measure the reverse value of the remaining margin of the preset constraint; then, the standardization processing is performed to obtain a standardized conflict factor vector;

[0083] The weight generation module calculates an adaptive weight vector according to the data freshness, the source credibility and the conflict duration, and sets a weight upper limit according to the device security level to obtain a limited adaptive weight vector;

[0084] The gain aggregation module performs weighted superposition on the standardized conflict factor vector and the limited adaptive weight vector by using a bounded gain aggregation to obtain an uncalibrated risk value;

[0085] The mapping treatment module looks up the uncalibrated risk value according to a preset uncalibrated risk-coupling risk reference table to obtain a corresponding coupling risk coefficient, and outputs a treatment suggestion based on a preset grading interval corresponding to the coupling risk coefficient to guide the carbon capture process control system to implement dynamic regulation and control.

[0086] The above formulas are all dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0087] It should be understood that the size of the sequence number of each process in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0088] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0090] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for AI decision based intelligent diagnostic control of carbon capture process, characterized in that, The method comprises the following steps: Step one, obtain the control intention data of the edge computing side and the cloud side in the same time period, align the data time base of the control intention and unify the measurement units, and generate a synchronous control sequence for comparative analysis; Step two, calculate six types of conflict factors based on the synchronous control sequence, including direction reversal degree, amplitude asymmetry degree, phase mismatch degree, energy impact degree, sensitive amplification degree and boundary approximation degree, wherein the direction reversal degree is used to measure the proportion of instruction direction angle and sign opposite, the amplitude asymmetry degree is used to measure the relative difference of adjustment amplitude on both sides, the phase mismatch degree is used to measure the time center offset and period offset, the energy impact degree is used to measure the expected power step or steam load step strength, the sensitive amplification degree is used to measure the amplification effect of key process quantity under current working condition sensitivity, and the boundary approximation degree is used to measure the reverse value of the remaining margin of the preset constraint; then, the standardization processing is performed to obtain a standardized conflict factor vector; Step three, calculate an adaptive weight vector according to data freshness, source credibility and conflict duration, and set a weight upper limit according to the device safety level to obtain a limited adaptive weight vector; Step four, perform weighted superposition on the standardized conflict factor vector and the limited adaptive weight vector by using a bounded gain aggregation to obtain an uncalibrated risk value; Step five, look up the uncalibrated risk value according to a preset uncalibrated risk-coupling risk correspondence table to obtain a corresponding coupling risk coefficient, and output a handling suggestion based on a preset grading interval corresponding to the coupling risk coefficient to guide the carbon capture process control system to implement dynamic regulation and control.

2. The AI decision based intelligent diagnostic control method for carbon capture process according to claim 1, wherein, The synchronous control sequence is obtained by the following steps: First, identify valve steps, pump speed transitions, set value changes and alarm confirmations on the edge computing side and the cloud side through preset event determination rules, and filter effective anchor points by carbon element conservation and energy balance two types of invariants, solve the piecewise linear time mapping with global non-decreasing and slope upper and lower bounds for the anchor point set passing through the constraints, so as to obtain the initial estimation of clock drift and communication delay; Implement limited dynamic time warping and phase correction on the initial time mapping to maintain the monotonicity of the mapping function and limit the local slope, and add a small phase item within a limited lag range according to the cross-correlation peak, thereby unifying the aligned time base and the aligned residual sequence containing mean, variance and maximum deviation; Based on the process property relationship, complete the state-dependent conversion between volume flow and mass flow, steam flow and heat load, and different concentration expressions, unify the measurement units, and add source credibility, version summary identifier and integrity check summary to each time point, and finally form a synchronous control sequence carrying complete metadata.

3. The AI decision based intelligent diagnostic control method for carbon capture process according to claim 2, wherein, In the process of forming the synchronous control sequence carrying complete metadata, the density and enthalpy value are calculated according to the temperature, pressure and solvent composition, the volume flow is converted into mass flow, the steam flow is converted into heat load, the mass fraction and volume fraction are converted into mole fraction, and the pressure, temperature and concentration are unified to the pre-designed measurement unit system; The source credibility is calculated by the latest calibration date of the instrument, the online self-checking result and the missing rate; The version summary identifier is composed of the model name, the model version and the training time; The integrity check summary is composed of the deterministic summary string of the configuration parameter list, and the synchronized control sequence carrying the metadata is generated after summarization.

4. The AI decision based intelligent diagnostic control method for carbon capture process according to claim 3, wherein, The specific steps of step two include: Based on the synchronized control sequence, the control intention increments of the edge computing side and the cloud side in the same control period are extracted, and the paired data sets of the two sides in different preset quantization increments are obtained according to the control channels, and the rated range, the rated energy capacity and the constraint limit value of each channel are recorded; Six types of conflict factors are calculated on the paired data set, which are: The calculation of the direction reverse degree is to calculate the direction angle of the increments of the same channel on the two sides first, and the direction angle takes the angle value from zero to one hundred and eighty degrees, then the proportion of the channels with opposite signs in all channels is calculated, and the direction reverse degree is equal to the weighted average of the direction angle divided by one hundred and eighty and the proportion; The calculation of the amplitude asymmetry degree is to take the absolute value of the difference between the absolute values of the increments on the two sides of the same channel, and then divide by the sum of the absolute values of the increments on the two sides plus the pre-set minimum positive lower limit value as the denominator, and finally take the arithmetic average of all channels; The calculation of the phase mismatch degree is to calculate the cross-correlation function of the control intentions on the two sides within the specified maximum time lag range, take the absolute value of the time lag that produces the maximum correlation and normalize it with the maximum time lag, and at the same time, calculate the time center difference of the control intentions on the two sides and normalize it with the control period of the channel, and the phase mismatch degree is equal to the weighted average of the two normalized results; The calculation of the energy impact degree is to convert the control intention increments of each channel into expected electric power steps and expected heat load steps through the energy conversion coefficient under the current working condition, take the absolute value of the sum of the two, and normalize it with the sum of the rated electric power and the rated heat load of the corresponding unit; The calculation of the sensitive amplification degree is to obtain the local first-order sensitivity coefficients of the key process quantities to each input channel at the current operating point, multiply the control intention increments of each channel by the corresponding sensitivity coefficients, and then sum them up to get the expected change amplitude of each key process quantity, take the maximum value of all key process quantities and normalize it with the allowed change bandwidth of the process quantity; The calculation of the boundary approximation degree is to calculate the remaining margin of each controlled constraint after executing the control intentions on the two sides, if the constraint is an upper limit, subtract the expected value from the upper limit, if the constraint is a lower limit, subtract the lower limit from the expected value, and negative values are counted as zero, then normalize it with the safety bandwidth of the corresponding constraint and take one minus the normalized result, finally take the maximum value of all constraints as the boundary approximation degree.

5. The AI decision based intelligent diagnostic control method for carbon capture process as claimed in claim 3, wherein, The standardized conflict factor vector refers to: The six types of conflict factors calculated based on the synchronized control sequence are subjected to a pre-set standardization process, each conflict factor is mapped to a standardized range of zero to one, and then the standardized conflict factor vector is obtained by summarizing according to the pre-set arrangement order.

6. The AI decision based intelligent diagnostic control method for carbon capture process according to claim 5, wherein, The restricted adaptive weight vector is obtained by the following steps: The data freshness score is calculated as follows: in a unified time reference, the time lag value, the communication delay value and the missing proportion are calculated for each data channel of the six types of conflict factors, the time lag value is the current unified time minus the latest valid timestamp of the channel, the communication delay value is the communication delay estimation obtained by the aforementioned time mapping, and the missing proportion is the proportion of invalid samples in all samples within a fixed evaluation window; the timeliness component is obtained by linearly mapping the time lag value in reverse according to a preset maximum acceptable time difference, the channel delay component is obtained by linearly mapping the communication delay value in reverse according to a preset maximum acceptable delay, the integrity component is obtained by taking the missing proportion according to a minus relationship, and the geometric mean of the timeliness component, the channel delay component and the integrity component is taken to obtain a data freshness score in the range of zero to one; The conflict duration is calculated as follows: in the same evaluation window, the instantaneous conflict indicator is constructed, the instantaneous conflict indicator is equal to the maximum value of the structural conflict strength and the amplification strength multiplied by the boundary strength, the structural conflict strength is the arithmetic mean of the direction reversal degree, the amplitude asymmetry degree and the phase mismatch degree, the amplification strength is the arithmetic mean of the energy impact degree and the sensitive amplification degree, and the boundary strength is the boundary approximation degree; The conflict threshold is set, and the time period during which the indicator continuously exceeds the threshold is identified, the time proportion of the time during which the threshold is exceeded is calculated, and the average value of the indicator when the threshold is exceeded is calculated, the product of the time proportion and the average value is taken as the unweighted conflict duration, and the conflict duration in the range of zero to one is obtained by weighting according to a linearly increasing weight from near to far; The restricted adaptive weight vector is formed as follows: for each component of the six types of conflict factors, the source credibility weight is multiplied by the data freshness score, and the result is multiplied by an exponential function of the conflict duration, the exponential is given by the sensitive coefficient corresponding to the device security level, and the sensitive coefficient is determined according to the device security level, the device security level is one, two, three or four, and the weight upper limit is set according to the device security level, the weight upper limit of one, two, three and four is determined in the order from small to large, and finally the restricted adaptive weight vector is output.

7. The AI decision based intelligent diagnostic control method for carbon capture process according to claim 5, wherein, The uncalibrated risk value is obtained by the following steps: Gain limiting functions are established for the direction reversal degree, the amplitude asymmetry degree, the phase mismatch degree, the energy impact degree, the sensitive amplification degree and the boundary approximation degree, the gain limiting function adopts a smooth saturation form, that is, the ratio of the gain coefficient multiplied by the index value as the numerator and one plus the coefficient multiplied by the index value as the denominator, and the smaller value is taken with the corresponding gain upper limit, the determined values of the growth coefficient and the gain upper limit of each factor are preset according to the device security level; The standardized conflict factor vector is input into the gain limiting function to obtain a gain limited factor vector, and multiplied by the restricted adaptive weight vector to form a weighted component set, and the aggregated base value is obtained by performing a truncated arithmetic mean on the weighted component set, and the truncation proportion is a fixed value between the preset minimum proportion and the maximum proportion; The aggregate base value obtained by truncating the arithmetic mean is subjected to double threshold shaping and convex combination smoothing to obtain the uncalibrated risk value.

8. The AI decision based intelligent diagnostic control method for carbon capture process according to claim 7, wherein, The steps for performing double-threshold shaping and convex combination smoothing on the aggregate base value obtained from the truncated arithmetic mean include: Set a low threshold and a high threshold, with the low threshold and high threshold taking fixed values ​​and the low threshold being smaller than the high threshold. The output value is zero when the polymer base value does not exceed the low threshold. When the base value of the polymer is between the low threshold and the high threshold, the output is equal to the polymer base value minus the low threshold, and then divided by the high threshold minus the low threshold. When the aggregation base value exceeds the high threshold, the output is equal to a function minus an exponential function. The independent variable of the exponential function is the aggregation base value minus the product of the high threshold and the fixed convergence coefficient. The fixed convergence coefficient is preset according to the device safety level. The output of this period and the output of the previous evaluation period are combined into a convex combination with a fixed smoothing coefficient. The fixed smoothing coefficient is a definite value in the range of zero to one, and the result of the convex combination is restricted to the range of zero to one to obtain the uncalibrated risk value.

9. The AI decision based intelligent diagnostic control system for carbon capture process according to any one of claims 1-8, characterized in that, Specifically, it includes: The alignment and unification module acquires control intent data from the edge computing side and the cloud side within the same time period, aligns the data time base of the control intent and unifies the units of measurement, and generates a synchronous control sequence for comparative analysis. The conflict quantification module calculates six types of conflict factors based on the synchronous comparison sequence. These six conflict factors include direction reversal, amplitude asymmetry, phase mismatch, energy impact, sensitivity amplification, and boundary approximation. Direction reversal measures the angle between command directions and the proportion of opposite signs; amplitude asymmetry measures the relative difference in adjustment amplitudes on both sides; phase mismatch measures the time center offset and period offset; energy impact measures the intensity of expected power step or steam load step; sensitivity amplification measures the amplification effect on key process quantities under the current operating condition sensitivity; and boundary approximation measures the reverse value of the remaining margin of the preset constraint. Then, the data is summarized and standardized to obtain a standardized conflict factor vector. The weight generation module calculates an adaptive weight vector based on data freshness, source credibility, and conflict persistence, and sets a weight upper limit according to the device security level to obtain a restricted adaptive weight vector. The gain aggregation module uses bounded gain aggregation to weight and superimpose the standardized conflict factor vector and the restricted adaptive weight vector to obtain the uncalibrated risk value. The mapping and handling module looks up the uncalibrated risk value according to the preset uncalibrated risk-coupled risk lookup table, obtains the corresponding coupled risk coefficient, and outputs handling suggestions based on the preset coupled risk coefficient corresponding to the graded interval, guiding the carbon capture process control system to implement dynamic regulation.

Images