Iron and steel enterprise multi-process carbon emission online monitoring and early warning platform based on Internet of Things

Abstract

The invention discloses an iron and steel enterprise multi-process carbon emission on-line monitoring and early warning platform based on the Internet of Things, and relates to the technical field of carbon emission abnormity monitoring and alarming in the industrial production process, and the platform comprises a sensor network which is disposed at key equipment nodes of iron making, steel making and steel rolling processes, and is used for collecting carbon emission data and process parameters in real time; and the edge computing node is in communication connection with the sensor network and is used for carrying out filtering and time alignment processing on the acquired data. According to the iron and steel enterprise multi-process carbon emission online monitoring and early warning platform based on the Internet of Things, the problems of false alarm and missing alarm caused by the characteristic difference of iron and steel multi-process carbon emission are fundamentally solved through process exclusive carbon behavior map modeling and dynamic threshold generation mechanism; the stepped threshold value of the electric furnace process covers the power transition emission fluctuation, and the converter oxygen blowing intensity and the floating binding of the threshold value upper limit avoid the instantaneous peak value false alarm.

Description

Technical Field

[0001] This invention relates to the field of carbon emission anomaly monitoring and alarm technology in industrial production processes, specifically to an online monitoring and early warning platform for multi-process carbon emissions in steel enterprises based on the Internet of Things. Background Technology

[0002] The IoT-based online monitoring and early warning platform for carbon emissions across multiple processes in steel enterprises relies on sensor networks to collect real-time carbon emission data from core processes such as ironmaking, steelmaking, and rolling. It achieves dynamic perception of emission trends through edge computing and cloud collaboration. While existing technologies can ensure data integrity and real-time transmission, their alarm systems have significant shortcomings: platforms generally use a globally unified static threshold model to trigger warnings, failing to effectively distinguish the inherent mechanisms and dynamic characteristics of carbon emissions from different processes. The blast furnace process primarily consumes fossil fuels, exhibiting short-duration, high-intensity pulse-like emissions; the electric arc furnace process relies on electricity, with emissions exhibiting a stepped, cumulative characteristic; and emissions from the heating furnace in the rolling mill are deeply coupled with the temperature curve and billet specifications. This heterogeneity between processes causes the single fixed threshold model to be severely out of touch with reality. For example, normal peak values ​​during the electric arc furnace charging period are often misjudged as abnormal, triggering invalid alarms and disrupting production; and gradual risks such as slow leakage from blast furnace gas pipelines are missed because they do not exceed the global static threshold. Meanwhile, during process switching or fluctuations in process parameters, the system lacks an adaptive mechanism to dynamically adjust its judgment logic, requiring frequent manual intervention, which reduces response time and increases the operational burden. The existing solution ignores the nonlinear correlation between process characteristics, changes in process status, and carbon emissions, resulting in severely insufficient early warning accuracy and making it difficult to support refined carbon management decisions in steel enterprises. The urgent problem to be solved is: how to overcome the insufficient adaptability of static threshold models caused by differences in carbon emission mechanisms across multiple processes, in order to achieve accurate and reliable dynamic early warning. Summary of the Invention

[0003] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises based on the Internet of Things, comprising:

[0004] Sensor networks are deployed at key equipment nodes in the ironmaking, steelmaking, and steel rolling processes to collect carbon emission data and process parameters in real time.

[0005] Edge computing nodes communicate with sensor networks to filter and time-align the collected data.

[0006] The cloud analytics module connects to the edge computing node network and includes:

[0007] The carbon behavior mapping modeling unit analyzes the process mechanism and constructs a dynamic threshold margin range;

[0008] A dual-modal residual collaborative verification unit performs theoretical-measured residual analysis and cross-process migration verification.

[0009] The early warning circuit breaker control unit dynamically adjusts the threshold sensitivity and drives model evolution.

[0010] Preferably, the carbon behavior mapping modeling unit performs the following operations:

[0011] For the blast furnace process, a transfer function is established for blast temperature, coke ratio, and carbon emissions;

[0012] For the electric furnace process, a mapping relationship between arc power, molten steel composition and discharge rate is established;

[0013] Based on real-time process parameters, a narrow-amplitude pulse-type threshold range or a stepped threshold range is generated.

[0014] Preferably, the generation response to changes in the process state of the dynamic threshold margin range includes: adjusting the upper limit of the threshold when the oxygen blowing intensity of the converter changes; and synchronously correcting the threshold reference value when the steel rolling heating temperature curve shifts.

[0015] Preferably, the dual-modal residual collaborative verification unit includes:

[0016] The theoretical-measured residual analysis module calculates the residual between real-time carbon emission data and theoretical values ​​of carbon behavior graphs;

[0017] Short-term process fluctuation filter: When the residual peak value matches the spatiotemporal match of a feeding event or oxygen blowing event, the alarm is blocked.

[0018] The gradual leakage tracker performs time-accumulated analysis on continuous low-amplitude residuals and triggers an alarm when the accumulated amount exceeds the limit.

[0019] Preferably, the dual-modal residual collaborative verification unit further includes:

[0020] The cross-process migration inspection module inputs the abnormal characteristics of the current process into the process similar process model for deduction;

[0021] When the consistency between the simulation results and the current residual characteristics exceeds a preset threshold, an anomaly is confirmed and an alarm is activated.

[0022] Preferably, the process similarity model satisfies one of the following conditions: the thermal efficiency deviation between the rolling mill heating furnace and the steelmaking baking furnace is less than a set value; the pressure balance coefficient between the blast furnace gas system and the coking process is on the same order of magnitude.

[0023] Preferably, the early warning circuit breaker control unit performs the following:

[0024] For processes with continuous false alarms, increase the floating range of the threshold margin interval;

[0025] For processes that are missed in reporting, the fluctuation range of the threshold margin interval is reduced.

[0026] Preferably, the model evolution is achieved through closed-loop feedback, including:

[0027] The discrepancy between the manual review conclusions and the system decision data is transmitted in reverse to the carbon behavior mapping modeling unit;

[0028] Dynamically update the weight parameters of the transfer function and mapping relationship.

[0029] Preferably, it also includes an alarm terminal, which is connected to the cloud analysis module to receive verified abnormal events and risk tracing reports.

[0030] An early warning method for an online monitoring and early warning platform for carbon emissions from multiple processes in a steel enterprise based on the Internet of Things (IoT), comprising:

[0031] S1: Respond to the process switching command and load the carbon behavior spectrum model of the corresponding process;

[0032] S2: Confirm the anomaly through dual residual verification and execute the circuit breaker control strategy;

[0033] S3: Update model parameters based on human feedback data.

[0034] This invention provides an online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises based on the Internet of Things (IoT). It has the following beneficial effects:

[0035] This IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in the steel industry fundamentally solves the problem of false alarms and missed alarms caused by differences in carbon emission characteristics across multiple steel processes through process-specific carbon behavior graph modeling and dynamic threshold generation mechanisms: the narrow-amplitude pulse-type threshold for the blast furnace process encompasses normal airflow disturbances, the stepped threshold for the electric arc furnace process covers power transition emission fluctuations, and the floating binding of the converter oxygen blowing intensity with the upper limit of the threshold avoids false alarms due to instantaneous peaks; the strong spatiotemporal matching rules in the dual residual verification filter short-term process interferences, the weighted cumulative analysis and equipment vibration spectrum linkage capture gradual leakage anomalies, and the rigid process similarity criterion for cross-process migration verification ensures the reliability of anomaly judgment.

[0036] This IoT-based online carbon emission monitoring and early warning platform for multi-process steel enterprises employs a circuit breaker mechanism to dynamically balance early warning parameters: false alarm events are categorized and thresholds are expanded, missed alarm events are categorized and narrowed, and adjustments are followed by mandatory shift locking to prevent oscillations. Manual feedback drives a closed-loop model evolution, including: progressive adjustment of the blast furnace transfer function's blast temperature sensitivity coefficient; insertion of compensation nodes into the missed alarm interval of the electric arc furnace curve; and activation of new versions only after full verification that they surpass historical baselines. Sub-unit-level positioning of alarm terminal equipment and coding of process causes significantly increase handling efficiency, while visualization of the multi-parameter correlation matrix in risk tracing reports supports decision optimization. This forms a fully autonomous "monitoring-early warning-handling-evolution" system, greatly reducing maintenance intensity and improving carbon management efficiency. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the module interaction of the Internet of Things-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to the present invention;

[0038] Figure 2 This is a flowchart illustrating the Internet of Things-based online monitoring and early warning method for carbon emissions from multiple processes in steel enterprises according to the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Please see Figure 1 and Figure 2 This invention provides a technical solution: an online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises based on the Internet of Things, comprising:

[0041] Sensor networks are deployed at key equipment nodes in the ironmaking, steelmaking, and steel rolling processes to collect carbon emission data and process parameters in real time.

[0042] Edge computing nodes communicate with sensor networks to filter and time-align the collected data.

[0043] The cloud analytics module connects to the edge computing node network and includes:

[0044] The carbon behavior mapping modeling unit analyzes the process mechanism and constructs a dynamic threshold margin range;

[0045] A dual-modal residual collaborative verification unit performs theoretical-measured residual analysis and cross-process migration verification.

[0046] The early warning circuit breaker control unit dynamically adjusts the threshold sensitivity and drives model evolution.

[0047] It should be further explained that, during the specific implementation process, when the IoT-based online monitoring and early warning platform for carbon emissions across multiple processes in steel enterprises is launched, the sensor network deployed in key equipment such as blast furnaces, converter steelmaking, and rolling mill heating furnaces continuously collects carbon emission concentration data and related process parameters, including blast temperature, coke ratio, arc power, and billet temperature curves. After receiving the raw data, the edge computing nodes perform preprocessing: timestamp alignment eliminates transmission delay differences, and sliding window filtering smooths high-frequency noise. After the preprocessed data is uploaded to the cloud analysis module, the core early warning process unfolds according to the following logic:

[0048] The carbon behavior mapping modeling unit constructs dedicated dynamic baselines for the carbon emission mechanisms of different processes. For the ironmaking process, based on the coupling relationship between blast furnace blast temperature and coke ratio, a transfer function model of carbon emissions as a function of blast temperature is established, automatically lowering the carbon emission baseline value when the blast temperature increases. For the electric arc furnace steelmaking process, a stepped cumulative emission curve is generated based on the mapping relationship between arc power fluctuations and molten steel carbon content. Based on this dynamic baseline, a nonlinear margin adaptation algorithm generates threshold ranges that fluctuate according to real-time process conditions: when the converter oxygen blowing intensity increases, the upper limit of the threshold is simultaneously raised to accommodate short-term emission peaks; when the temperature curve of the rolling mill heating furnace deviates from the standard operating conditions, the threshold baseline value is corrected according to a preset proportional coefficient to ensure that the range width matches the process fluctuation range.

[0049] The dual-modal residual collaborative verification unit performs two levels of anti-interference screening, including: First, calculating the residual sequence between real-time carbon emission data and theoretical values ​​of carbon behavior maps. When the residual instantaneously exceeds the dynamic threshold, a short-term process fluctuation filter is activated. If the system detects that the peak time point matches the feeder signal or oxygen lance blowing record, and the spatial location is in the raw material area or blowing area, it is determined to be a normal fluctuation and the alarm is blocked. When the residual shows a continuous low-amplitude positive deviation, the gradual leakage tracker starts a time accumulation counter. If the accumulated amount exceeds the set limit, a leakage alarm is triggered.

[0050] Secondly, cross-process migration verification is initiated for preliminary abnormal events: the current abnormal characteristics of blast furnace gas pressure are extracted and projected onto a coking dry quenching model with similar processes for simulation and deduction. Only when the consistency between the two in terms of pressure decay slope and emission increment trend reaches the verification threshold is it confirmed as a real abnormality.

[0051] The robustness of the early warning circuit breaker control unit dynamic maintenance system includes: when a certain process experiences three consecutive false alarms, automatically widening the floating range of the threshold margin interval for that process; and when the system records a missed alarm event that has been manually verified, immediately compressing the threshold interval range for the corresponding process.

[0052] All discrepancies between manually reviewed conclusions and system decisions are fed back to the carbon behavior mapping modeling unit. The transfer function weight parameters are then corrected using a backpropagation algorithm, achieving a closed-loop evolution of the early warning logic. Finally, confirmed anomalies are pushed to the workshop terminal, simultaneously generating a source tracing report containing the location of the abnormal process and an analysis of the process's causes.

[0053] The carbon behavior mapping modeling unit performs the following operations:

[0054] For the blast furnace process, a transfer function is established for blast temperature, coke ratio, and carbon emissions;

[0055] For the electric furnace process, a mapping relationship between arc power, molten steel composition and discharge rate is established;

[0056] Based on real-time process parameters, a narrow-amplitude pulse-type threshold range or a stepped threshold range is generated.

[0057] It should be further explained that, in the specific implementation process, when the carbon behavior mapping modeling unit of the cloud analysis module is started, the corresponding modeling engine is first activated according to the process type identifier. For the blast furnace process of ironmaking, the system calls the preset blast temperature-coke ratio coupled analyzer, which includes: collecting real-time hot blast stove blast temperature and coke load data, establishing a transfer function of carbon emissions with respect to blast temperature changes based on the kinetic equation of coke gasification reaction in the blast furnace, and when the blast temperature rises to the set range, due to the increase in combustion efficiency, the carbon emission benchmark value per unit of molten iron produced decreases according to a nonlinear curve; at the same time, the slope of the function is dynamically corrected according to the measured data of coke sulfur content.

[0058] For electric arc furnace steelmaking processes, an arc power-molten steel composition tracker is activated: it monitors electrode voltage and current and molten steel spectral analysis values ​​in real time, calculates effective smelting power through an arc thermal efficiency model, and constructs a mapping curve of emission rate with power step increase by combining data from molten steel carbon content sensors; when scrap steel impurities exceed the standard, it automatically increases the emission rate compensation coefficient in the low power range.

[0059] After completing the mechanism modeling, the nonlinear margin adaptation algorithm generates a process-specific threshold range based on the real-time process status. The blast furnace process adopts a narrow-amplitude pulse-type threshold: a fixed narrow margin range is set under normal furnace conditions, centered on the dynamic baseline output by the transfer function; when the hot blast pressure fluctuation exceeds the warning value, the algorithm expands the upper limit of the range proportionally based on the pressure deviation to accommodate the instantaneous pulse emission caused by airflow disturbance.

[0060] The electric arc furnace process adopts a stepped threshold, including: maintaining a margin range during the stable arc power stage; when the power jumps, such as when the melting period transitions to the oxidation period, the system identifies the characteristics of molten steel splashing and activates a transition protection mode, including: temporarily raising the upper limit of the threshold to a set multiple of the standard value to cover the sharp increase in carbon emissions caused by molten steel splashing, and automatically restoring the baseline range after the power stabilizes.

[0061] To address the unique characteristics of the steel rolling process, the modeling unit links the heating furnace temperature curve with the billet specification database: when a thick slab is detected entering the furnace, the cumulative emission period is extended based on the billet heat capacity calculation model, and a lag-type threshold range is generated simultaneously; if the furnace temperature control deviation continuously exceeds the limit, a proportional downward adjustment mechanism for the threshold baseline value is triggered to avoid false alarms caused by inaccurate temperature control. All dynamic threshold parameters are written to the process profile library in real time for the verification unit to access.

[0062] The dynamic threshold margin range is generated in response to changes in process conditions, including: adjusting the upper limit of the threshold when the oxygen blowing intensity in the converter changes; and simultaneously correcting the baseline threshold value when the rolling heating temperature curve deviates. It should be further noted that in the specific implementation process, when the cloud analysis module detects a change in process parameters, the carbon behavior mapping modeling unit immediately initiates the dynamic threshold refresh process. For the converter steelmaking process, the system continuously monitors the oxygen lance blowing intensity data: when the oxygen blowing intensity exceeds the standard operating condition setting range, the upper limit of the threshold range is increased proportionally according to the intensity deviation, allowing for the inclusion of the instantaneous surge in carbon emissions unique to the oxygen blowing stage; if the oxygen blowing intensity drops sharply and the duration exceeds the safety window, the lower limit of the threshold is simultaneously lowered to capture possible abnormal gas recovery. Simultaneously, the oxygen blowing end time is recorded, and the expanded threshold range is maintained for a set delay after oxygen shutdown to cover residual decarburization reactions in the molten steel, after which the baseline margin is automatically restored.

[0063] For the heating furnace temperature curve in the steel rolling process, the modeling unit compares the measured temperature with the standard process curve in real time. When an overall furnace temperature deviation is detected, such as a delayed heating due to overly dense billet arrangement, the threshold benchmark value is adjusted downwards based on the temperature integral deviation value using a proportional coefficient to avoid false alarms caused by reduced heating efficiency. When local temperature fluctuations exceed limits, such as burner blockage causing temperature oscillations, a segmented compensation mechanism is triggered, including temporarily expanding the threshold margin at the emission monitoring points corresponding to the abnormal temperature zone, while maintaining the original range in the stable temperature zone. After the temperature curve returns to normal, the threshold benchmark value gradually returns to the standard value in the form of a ramp function to prevent false alarms caused by sudden changes.

[0064] In process switching scenarios, the system predicts the type of process to be activated based on production line scheduling signals: when blast furnace tapping is completed and electric arc furnace charging begins, the electric arc furnace stepped threshold model is preloaded; if the measured carbon emission pattern after the process switch deviates significantly from the model's prediction, such as detecting blast furnace pulse characteristics in electric arc furnace mode, an emergency threshold calibration procedure is initiated, including: dynamically reconstructing threshold parameters based on emission data from the initial set time period after the switch, until the matching degree reaches the standard, and then switching to normal monitoring. All threshold adjustment records are written to the operation log for analysis by the circuit breaker control unit.

[0065] The dual-modal residual co-validation unit includes:

[0066] The theoretical-measured residual analysis module calculates the residual between real-time carbon emission data and theoretical values ​​of carbon behavior graphs;

[0067] Short-term process fluctuation filter: When the residual peak value matches the spatiotemporal match of a feeding event or oxygen blowing event, the alarm is blocked.

[0068] The gradual leakage tracker performs time-accumulated analysis on continuous low-amplitude residuals and triggers an alarm when the accumulated amount exceeds the limit.

[0069] It should be further explained that, in the specific implementation process, after the dual-modal residual collaborative verification unit is activated, the theoretical-measured residual analysis module receives the dynamic theoretical emission curve output by the carbon behavior spectrum modeling unit in real time and performs residual calculation with the measured carbon emission data from the sensors. When a positive peak appears in the residual sequence and its amplitude exceeds the upper limit of the dynamic threshold, the system immediately activates the short-term process fluctuation filter: it retrieves the production event logs within a set time period before and after the current moment. If it identifies that the feeding signal of the feeder overlaps with the residual peak time, and the feeding position is in the same process unit as the carbon emission monitoring point, it is marked as raw material volatilization interference; if the oxygen lance blowing record shows that the blowing intensity reaches a high load state at the peak time, and the residual spatial distribution matches the converter flue gas flow direction, it is determined to be an oxygen blowing reaction disturbance. If any of the above conditions are met, the alarm is immediately masked, and the peak value is included in the normal fluctuation mode of the process profile library.

[0070] For negative residual abnormalities, the module performs reverse verification, including: when the residual is continuously below the lower limit of the threshold, checking the opening data of the gas recovery valve, including: if the opening increases synchronously and the duration matches, it is determined that the recovery system is absorbing carbon normally; otherwise, the equipment fault pre-inspection process is triggered.

[0071] When the residual shows a sustained low-amplitude positive deviation, the gradual leakage tracker initiates time-cumulative analysis: residual values ​​are sampled at fixed intervals. When the number of consecutive samples exceeds a set threshold and the deviation remains stable, the cumulative counter is weighted and accumulated according to the deviation magnitude. If the accumulated amount exceeds a preset limit, and low-frequency abnormal components are detected in the equipment vibration spectrum, a leakage alarm is generated. During the tracking process, the system periodically compares the residual trend with the equipment inspection records: if recent manual inspection confirms no leakage risk, the accumulated amount limit is automatically increased to enhance immunity; otherwise, the limit is decreased to increase sensitivity.

[0072] All unmasked abnormal events are transferred to the cross-process migration inspection module, and the filtering decision criteria are written into the verification log for the circuit breaker unit to call.

[0073] The dual-modal residual co-validation unit also includes:

[0074] The cross-process migration inspection module inputs the abnormal characteristics of the current process into the process similar process model for deduction;

[0075] When the consistency between the simulation results and the current residual characteristics exceeds a preset threshold, an anomaly is confirmed and an alarm is activated.

[0076] It should be further explained that, in the specific implementation process, when the theoretical-measured residual analysis module outputs an initial abnormal event, the cross-process migration verification module is immediately activated. The system first searches the process similarity map library for candidate models associated with the current process: for an anomaly in the rolling mill heating furnace, it automatically matches the ladle baking heat efficiency model of the steelmaking process; for an alarm in the blast furnace gas system, it associates the dry quenching pressure balance model of the coking process. The matching criteria strictly follow the preset similarity criteria, including: the rolling mill heating furnace and the ladle baking heat exchanger must meet the requirement that the thermal efficiency deviation is less than the historical statistical threshold, and the blast furnace and coking models must have pressure balance coefficients of the same engineering magnitude.

[0077] After model matching is completed, the module extracts the current abnormal feature vector: for abnormal heating furnace temperature distribution, the longitudinal temperature gradient curve of the furnace and the emission increment rate are extracted; for blast furnace gas pressure fluctuations, the pressure decay time constant and the slope of carbon monoxide concentration change are collected. The feature vector is projected onto the target similar process model for extrapolation: after the abnormal heating furnace data is input into the ladle baking machine model, the emission response mode of the baking machine under the same temperature gradient is simulated; after the blast furnace gas features are input into the dry quenching coke model, the carbon release trajectory under the pressure decay scenario is reconstructed.

[0078] The inferred results are compared with the current measured residual characteristics using two indicators: for the heating furnace scenario, the trend of emission increment caused by the temperature gradient must match the set threshold, and the rate of change must be consistent; for the blast furnace scenario, the difference in pressure decay time constant must be within the allowable error range, and the signs of the carbon monoxide concentration change slopes must be the same. Only when both indicators meet the standards is it confirmed as a real anomaly and an alarm is activated. If a single indicator fails to match, a manual review process is initiated, and the discrepancy data is marked as transfer learning samples to expand the discrimination dimensions of the similarity map library.

[0079] All cross-process verification conclusions are fed back to the early warning and circuit breaker control unit, driving the dynamic optimization of similar model matching rules.

[0080] A process similarity model must meet one of the following conditions: the thermal efficiency deviation between the rolling mill heating furnace and the steelmaking ladle baker is less than a set value; the pressure balance coefficients of the blast furnace gas system and the coking process are on the same order of magnitude. It should be further noted that during implementation, when the cross-process migration verification module initiates the model matching process, the system selects candidate models based on a preset rigid criterion for process similarity. For the matching scenario between the rolling mill heating furnace and the steelmaking ladle baker, the module calls upon the thermal efficiency statistics database for the past three months: calculating the absolute deviation between the average thermal efficiency of the heating furnace and the baseline thermal efficiency of the baker. Only when this deviation remains consistently below the upper limit of historical operating condition fluctuations and does not show a trend of divergence is it confirmed as a valid similar model. After successful matching, the system monitors the heating furnace exhaust temperature and the baker gas calorific value data in real time. If the instantaneous thermal efficiency deviation exceeds the dynamic tolerance band during operation, the width of which adjusts proportionally with the furnace load rate, the migration verification is paused and the thermal efficiency recalibration process is triggered.

[0081] For the matching of the blast furnace gas system and the coking dry quenching model, the module rigorously verifies the consistency of the pressure balance coefficient: extracting the inlet pressure of the blast furnace gravity dust collector and the circulating gas pressure of the coking dry quenching, and calculating the ratio of their pressure balance coefficients. Only when this ratio is within the acceptable engineering range, such as between 0.8 and 1.2, and the overlap of the dominant frequency distribution of pressure pulsations exceeds a set threshold, is it determined to be a system of the same magnitude. During operation, if a sudden change in the pressure coefficient is detected due to the start-up or shutdown of the blast furnace TRT generator unit, the system automatically activates a buffer correction mode, including: freezing the migration verification for a set period after unit switching, and resuming verification after the pressure stabilizes again.

[0082] An exception handling mechanism is activated under special operating conditions: when the heating furnace is rolling ultra-thick slabs, the thermal efficiency deviation requirement is temporarily relaxed, and historical production records of the same specification are loaded as an auxiliary verification dataset; when the blast furnace is undergoing maintenance, the system switches to the backup coke oven model and activates the simplified verification process. All matching operations are logged, including deviation values, magnitude ratios, and exception handling flags, for analysis and optimization by the model evolution unit.

[0083] The warning circuit breaker control unit executes:

[0084] For processes with continuous false alarms, increase the floating range of the threshold margin interval;

[0085] For processes that are missed in reporting, the fluctuation range of the threshold margin interval is reduced.

[0086] It should be further explained that, in the specific implementation process, the early warning circuit breaker control unit continuously monitors the alarm decision records of each process. When a specific process generates three consecutive false alarms, the system automatically triggers the threshold margin range expansion procedure. The criteria for judging false alarms are: manual verification confirms that it is a normal process fluctuation. For false alarm events in the converter steelmaking process, differentiated expansion is performed according to the type of false alarm cause: if it is caused by an instantaneous peak during the oxygen blowing period, the upper limit of the threshold fluctuation range is increased proportionally; if it is caused by volatilization interference during the charging period, the upper and lower limits are expanded simultaneously to form a symmetrical inclusive range.

[0087] The expansion range is dynamically calculated based on the historical false alarm frequency, including: for each new false alarm of the same type, the floating range increases by a fixed percentage of the base value, but the total expansion amount does not exceed the safety tolerance threshold. When the system receives a manually confirmed missed event, such as a slow leak in a blast furnace gas pipeline that has not been detected, the fuse control unit immediately activates the threshold tightening mechanism.

[0088] For gradual leakage underreporting, the narrowing of the threshold margin range is performed in two steps: first, the current lower limit of the threshold is lowered to the set proportion of the emission value corresponding to the underreporting event, and the set duration is continuously monitored; if no false alarms occur during this period, the fluctuation range is further compressed to the optimized value. For sudden event underreporting, such as a sudden drop in emissions due to electric furnace electrode breakage, the emergency tightening mode is directly activated, including: narrowing the entire threshold range to the compression coefficient of the benchmark value, while suspending cross-process verification procedures to prevent chain misjudgments.

[0089] All adjustment operations adhere to stability constraints: within the same process, the fluctuation range after threshold expansion must be maintained for at least three production shifts before further adjustments are allowed; after a tightening operation, a single shift is forcibly locked to avoid frequent oscillations. Adjustment parameters are synchronized in real time to the carbon behavior mapping modeling unit, triggering collaborative updates of the dynamic baseline. The system statistically analyzes the ratio of false alarm rate to false negative rate monthly. When this ratio exceeds the equilibrium range, all process thresholds are automatically reset to the initial baseline, and a global model calibration process is initiated.

[0090] Model evolution is achieved through closed-loop feedback, including:

[0091] The discrepancy between the manual review conclusions and the system decision data is transmitted in reverse to the carbon behavior mapping modeling unit;

[0092] Dynamically update the weight parameters of the transfer function and mapping relationship.

[0093] It should be further explained that, in the specific implementation process, when the difference between the manual review conclusion and the system decision is input through the HMI interface, the early warning circuit breaker control unit immediately initiates the model evolution process. The system first parses the data type of the difference: for false alarm events in the blast furnace process, it extracts the "normal blast temperature fluctuation range" data packet confirmed by the review and transmits it in reverse to the blast temperature-coke ratio transfer function corrector of the carbon behavior graph modeling unit.

[0094] The corrector progressively adjusts the sensitivity coefficient of the carbon emission baseline value in the transfer function to wind temperature changes based on the actual wind temperature deviation. This includes: if multiple checks show false alarms during wind temperature fluctuations, the sensitivity coefficient is reduced to smooth the baseline curve; conversely, if the response to missed events is insufficient, the coefficient is increased to enhance sensitivity. The change in the slope of the adjusted function is constrained by process safety limits to prevent overcorrection from causing model instability.

[0095] To address the evolution of the mapping relationship in electric arc furnace processes, the system focuses on the segmented correction of the arc power-emission rate curve: when a systematic underreporting occurs in a manually marked power range, the modeling unit inserts a compensation node in that range, calculates the compensation slope based on the emission gap value of historical underreporting events, and forms a locally upturned correction curve; if a power transition point experiences consecutive false alarms, the curve within a set bandwidth before and after that point is smoothed and filtered to eliminate unnecessary abrupt changes. All correction operations follow the "small steps, quick progress" principle, including: the magnitude of a single adjustment does not exceed a set proportion of the maximum slope of the curve, but continuous iteration is allowed until the error converges.

[0096] After parameter calibration, the dynamic threshold generation logic is updated synchronously: changes in the transfer function slope trigger the repositioning of the baseline value in the threshold margin interval; and the smoothing of the electric furnace curve leads to the widening of the step-like threshold transition band. The updated model parameters and threshold rules are packaged into a version snapshot and automatically cross-validated with the best version from the past three months. A full replacement is activated only when the new version's combined false negative and false negative metrics on the validation dataset are better than the historical baseline; otherwise, it rolls back to the previous version and triggers an expert consultation process. Version iteration information is synchronized in real time to the dual-modal residual validation unit to ensure that the screening logic and model evolution are consistent.

[0097] It also includes an alarm terminal, which communicates with the cloud analysis module to receive verified abnormal events and risk tracing reports. It should be further explained that, in the specific implementation process, after the cloud analysis module completes dual-modal residual collaborative verification, the abnormal events confirmed by the circuit breaker control unit are encapsulated into structured alarm packages. The alarm terminal receives this data packet in real time through a dedicated data channel and parses it, which contains three core layers: an abnormal process location layer accurate to equipment sub-units such as blast furnace hot blast valve groups or electric furnace electrode areas; a process cause layer labeled with specific cause codes such as "oxygen blowing intensity exceeding limits" or "heating furnace temperature deviation"; and a risk level layer that automatically classifies warning color codes based on the residual exceedance magnitude and duration. The terminal interface dynamically renders a 3D map of the steel plant area, overlaying pulsed halo alarms at the corresponding equipment locations, while simultaneously pushing voice broadcasts strictly following the "process-location-cause" three-part template: "Steelmaking process converter area, abnormal oxygen blowing intensity, orange warning."

[0098] The risk tracing report is dynamically generated by the model evolution unit, and each report is bound to a unique code for the alarm event. The main body of the report includes a carbon behavior graph comparison view: the left side shows the deviation trajectory between the measured emission curve and the dynamic threshold range during the abnormal period, and the right side juxtaposes the theoretical model prediction curve, with the difference area highlighted by a heat map.

[0099] The root cause analysis section calls upon the coupling records of process parameters, including: when a converter alarm occurs, a time-series linkage chart of oxygen blowing intensity, oxygen lance height, and silicon content in molten iron is displayed; for blast furnace leakage events, a correlation matrix of gas pressure, dust collector differential pressure, and fan frequency is attached. The report footer embeds a model evolution traceability chain: marking the version number of the carbon behavior graph used in this determination and linking to the historical optimization records of that version, such as "Compensation and correction based on the 2024-Q2 missed event".

[0100] The terminal implements a tiered response mechanism: alarms below orange trigger the job order system, automatically assigning equipment inspection tasks to the responsible personnel in the corresponding areas; red alarms simultaneously activate the emergency control interface, sending a pressure reduction command to the gas recovery system or injecting a speed reduction signal into the production line PLC. All operations are logged to generate a decision-making closed-loop log, which is then transmitted back to the cloud for circuit breaker rule optimization.

[0101] An early warning method for an online monitoring and early warning platform for carbon emissions from multiple processes in a steel enterprise based on the Internet of Things (IoT), comprising:

[0102] S1: Respond to the process switching command and load the carbon behavior spectrum model of the corresponding process;

[0103] S2: Confirm the anomaly through dual residual verification and execute the circuit breaker control strategy;

[0104] S3: Update model parameters based on human feedback data.

[0105] It should be further explained that, in the specific implementation process, when the production line scheduling system sends a process switching instruction, the early warning method immediately terminates the current monitoring thread and calls the carbon behavior spectrum model of the target process from the process profile library. The loading process performs triple verification: checking the compatibility between the process type identifier and the model version; verifying that the real-time sensor layout matches the model input dimension; and confirming the clock synchronization status of the edge computing nodes. After the verification passes, the dynamic threshold margin range is preloaded into the memory buffer to ensure seamless monitoring.

[0106] The dual residual verification process is activated in stages: First, theoretical-measured residual analysis is performed. When the emission data of the steel rolling furnace continuously exceeds the lower limit of the stepped threshold, a short-term process fluctuation filter is activated. This includes: if the system is currently in the waiting-for-materials-and-holding stage and the furnace door opening and closing records are normal, it is attributed to a delay in thermal inertia release, triggering a delayed alarm and marking it for observation; otherwise, a cross-process migration test is triggered, projecting the abnormal characteristics onto the steelmaking baking furnace model. The migration test requires both the emission increment trend consistency and thermal efficiency deviation to meet the standards. If either indicator fails, the alarm output is frozen and transferred to the manual review queue.

[0107] The circuit breaker control strategy is dynamically embedded in the decision chain: During the monitoring cycle of the converter oxygen blowing stage, for each false alarm event, the upper limit of the threshold fluctuation range is automatically expanded, but the total expansion is constrained by the daily production plan. This includes: if the production schedule is for low-carbon steel, the maximum expansion range is compressed to maintain sensitivity; for blast furnace gas system missed events, the optimal threshold compression coefficient is calculated back based on the leakage rate prediction model, and a gradual narrowing is implemented during the equipment maintenance window. All manual review conclusions are fed back in real time. When an event is marked as "electrode fracture not detected", the method prioritizes calling the historical fault sample set of the electric furnace arc-emission mapping relationship, and locally covers the standard model parameters within a set time period until maintenance is completed.

[0108] It should be further explained that, during the specific implementation process, when the IoT-based online monitoring and early warning platform for carbon emissions across multiple processes in steel enterprises is activated, the sensor network deployed in key equipment such as blast furnaces, converters, and rolling mill heating furnaces continuously collects carbon emission concentration data and related process parameters, including blast temperature, coke ratio, arc power, and billet temperature curves. Edge computing nodes perform timestamp alignment and sliding window filtering on the raw data to eliminate transmission delays and high-frequency noise. The preprocessed data is then uploaded to the cloud analysis module, triggering the core early warning process.

[0109] Process-specific dynamic modeling includes the following: For the ironmaking process, the system establishes a transfer function for carbon emissions based on blast temperature and coke ratio parameters. When the hot blast stove blast temperature rises to a specific range, the function automatically lowers the carbon emission baseline value due to increased combustion efficiency; if the sulfur content of the coke exceeds the standard, the slope of the function is dynamically corrected. For the electric arc furnace process, a mapping curve between arc power fluctuations and steel carbon content on emission rates is constructed. When excessive impurities in scrap steel are detected, the emission compensation coefficient in the low-power range is increased. Based on this mechanism model, a nonlinear margin adaptation algorithm generates dynamic thresholds: when the converter oxygen blowing intensity increases, the upper limit of the threshold range is increased proportionally according to the oxygen blowing deviation; when the temperature curve of the rolling mill heating furnace deviates, the threshold baseline value is lowered based on the temperature integral deviation ratio. During process switching, the system preloads the target process's dedicated threshold model and verifies the sensor layout and clock synchronization status.

[0110] Dual anti-interference verification includes the following: calculating the residual sequence between real-time emission data and the theoretical model. When a momentary peak appears in the residual, the short-term process fluctuation filter retrieves the production event log for three-dimensional matching: if the residual peak time point completely coincides with the feeder's feeding signal, and the feeding position and the monitoring point are in the same process unit, the alarm is masked; if the spatial distribution of the residual is consistent with the converter flue gas flow direction model, and the oxygen blowing intensity record reaches a high load state, it is also judged as a normal disturbance. For continuous low-amplitude positive residuals, the gradual leakage tracker starts weighted time cumulative analysis: sampling residual values ​​at fixed periods, when the number of consecutive samplings exceeds the limit and the deviation from the amplitude is stable, it is weighted and accumulated according to the amplitude value; when the accumulated amount exceeds the set limit and there are low-frequency abnormal components in the equipment vibration spectrum, a leakage alarm is triggered. For preliminary anomalies, the cross-process migration verification module projects the feature vectors onto the process similarity model: for anomalies in the rolling mill heating furnace, the model needs to be matched with the steelmaking baking furnace model, requiring that the thermal efficiency deviation between the two be less than the historical fluctuation limit; for anomalies in blast furnace gas, the coking model is associated, verifying that the pressure balance coefficient ratio is in the range of 0.8 to 1.2 and that the pulsation frequency coincides. Only when the anomaly features are extrapolated in the similarity model in a direction consistent with the current residual trend and the magnitude of change matches, is it confirmed as a true anomaly.

[0111] The circuit breaker evolution closed-loop control includes the following: When a specific process generates three consecutive false alarms requiring manual verification, the system classifies and expands the threshold: for oxygen blowing interference, only the upper limit of the threshold is expanded; for material feeding volatilization interference, both the upper and lower limits are expanded simultaneously. The expansion range increases with the number of false alarms of the same type, but the total amount is constrained by safety tolerance. For missed alarm events, the threshold is compressed step by step for gradual leakage: first, the lower limit is lowered to the set proportion of the missed emission value, and the fluctuation range is further narrowed after monitoring stabilizes; for sudden failures, the entire range is narrowed urgently and cross-process verification is suspended. After all adjustment operations, a production shift is forcibly locked.

[0112] Manual review results drive model evolution: blast furnace false alarms trigger a gradual adjustment of the blast temperature sensitivity coefficient in the transfer function, while missed alarms increase sensitivity; the electric furnace model performs curve smoothing in areas with frequent false alarms and inserts compensation nodes in areas with missed alarms. New model versions are activated only if their overall performance is superior to the baseline in historical data verification. Finally, verified anomalies are pushed to the alarm terminal, located to the equipment subunit on the 3D plant map, labeled with process cause codes, and linked to a traceability report containing map comparison and parameter correlation matrices. Alarms below orange level automatically dispatch inspection work orders, while red alarms trigger pressure reduction commands or production line speed control.

[0113] A method for online monitoring and early warning of carbon emissions from multiple processes in steel enterprises based on the Internet of Things includes the following steps:

[0114] Step S1: The sensor network collects carbon emission data and process parameters such as air temperature, coke ratio, arc power, and billet temperature curve in real time from the ironmaking, steelmaking, and rolling processes.

[0115] Step S2: The edge computing nodes perform timestamp alignment and sliding window filtering on the raw data to eliminate transmission delay and high-frequency noise;

[0116] Step S3: Upload the preprocessed data to the cloud analysis module to identify the current process type and load the corresponding carbon behavior spectrum model;

[0117] Step S4: Construct process-specific dynamic thresholds, including:

[0118] Blast furnace process: Based on the transfer function that automatically lowers the carbon emission baseline value when the blast temperature rises, combined with the slope of the coke ratio dynamic correction function;

[0119] Electric furnace process: Establish a mapping curve between electric arc power and emission rate, and increase the compensation coefficient in the low power range when the impurities in the scrap steel exceed the standard;

[0120] Converter process: The upper limit of the oxygen blowing intensity deviation threshold is increased proportionally; Rolling process: The baseline value of the temperature integral deviation trigger threshold is adjusted downward proportionally.

[0121] Step S5: Perform dual residual verification, including:

[0122] Theoretical-measured residual analysis: residual spikes must perfectly match the spatiotemporal timing of the feeding event to suppress alarms; continuous low-amplitude residuals trigger weighted cumulative analysis, superimposed with equipment vibration spectrum for corroboration;

[0123] Cross-process migration inspection: For steel rolling anomalies, the thermal efficiency deviation of the steelmaking baking oven model must be less than the historical fluctuation limit; for blast furnace anomalies, the pressure balance coefficient ratio of the coking model must be in the range of 0.8 to 1.2 and the pulsation main frequency must coincide.

[0124] Step S6: Adaptive control of circuit breaker, including:

[0125] In the event of consecutive false alarms: oxygen blowing interference only expands the upper limit of the threshold, while feeding interference expands both the upper and lower limits simultaneously, with the expansion range increasing with the number of similar false alarms;

[0126] When a missed report is confirmed: for gradual leakage, the threshold is reduced step by step; for sudden failure, the threshold is narrowed urgently and cross-process verification is suspended.

[0127] Manual review results drive model evolution: blast furnace blast temperature sensitivity coefficient is gradually adjusted, and compensation nodes are inserted into the electric furnace underreporting interval;

[0128] Step S7: Verified abnormal events are pushed to the terminal, located in the equipment sub-unit on the 3D plant map, and the process cause code is marked;

[0129] Step S8: Generate a risk tracing report, including the emission curve deviation trajectory, multi-parameter correlation matrix, and model evolution version chain;

[0130] Step S9: Orange alarm automatically dispatches equipment inspection work orders; red alarm triggers gas pressure reduction or production line speed reduction control.

[0131] Step S10: The conclusions of manual processing are fed back to the cloud to drive the closed-loop optimization of the threshold circuit breaker rules and carbon behavior spectrum model.

[0132] By using process-specific carbon behavior mapping modeling and dynamic threshold generation mechanisms, the problem of false alarms and missed alarms caused by the differences in carbon emission characteristics across multiple steel processes is fundamentally solved: the narrow-amplitude pulsed threshold of the blast furnace process encompasses normal airflow disturbances, the stepped threshold of the electric arc furnace process covers power transition emission fluctuations, and the floating binding of the converter oxygen blowing intensity and the upper limit of the threshold avoids false alarms due to instantaneous peaks; the strong spatiotemporal matching rules in the dual residual verification filter short-term process interferences, the weighted cumulative analysis and equipment vibration spectrum linkage capture gradual leakage anomalies, and the rigid process similarity criteria for cross-process migration inspection ensure the reliability of anomaly judgment.

[0133] The circuit breaker mechanism achieves dynamic balancing of early warning parameters: false alarm events are categorized and thresholds are expanded, while missed events are categorized and narrowed, with forced shift locking after adjustment to prevent oscillations; manual feedback drives a closed-loop evolution of the model, including: progressive adjustment of the blast furnace transfer function's blast temperature sensitivity coefficient, insertion of compensation nodes into the missed interval of the electric furnace curve, and activation of new versions only after full verification that they are superior to historical baselines; sub-unit-level positioning of alarm terminal equipment and coding of process causes greatly increase handling efficiency, while visualization of the multi-parameter correlation matrix in risk tracing reports supports decision optimization. This forms a fully autonomous "monitoring-early warning-handling-evolution" system, significantly reducing maintenance intensity and improving carbon management efficiency.

[0134] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0135] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises based on the Internet of Things, characterized in that, include: Sensor networks are deployed at key equipment nodes in the ironmaking, steelmaking, and steel rolling processes to collect carbon emission data and process parameters in real time. Edge computing nodes communicate with sensor networks to filter and time-align the collected data. The cloud analytics module connects to the edge computing node network and includes: A carbon behavior mapping modeling unit is used to analyze the process mechanism and construct a dynamic threshold margin range. A dual-modal residual collaborative verification unit performs theoretical-measured residual analysis and cross-process migration verification. The early warning circuit breaker control unit dynamically adjusts the threshold sensitivity and drives model evolution.

2. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 1, characterized in that: The carbon behavior mapping modeling unit performs the following operations: For the blast furnace process, a transfer function is established for blast temperature, coke ratio, and carbon emissions; For the electric furnace process, a mapping relationship between arc power, molten steel composition and discharge rate is established; Based on real-time process parameters, a narrow-amplitude pulse-type threshold range or a stepped threshold range is generated.

3. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 2, characterized in that: The generation response to process state changes in the dynamic threshold margin range includes: adjusting the upper limit of the threshold when the oxygen blowing intensity of the converter changes; and synchronously correcting the threshold reference value when the steel rolling heating temperature curve deviates.

4. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 1, characterized in that: The dual-modal residual collaborative verification unit includes: The theoretical-measured residual analysis module calculates the residual between real-time carbon emission data and theoretical values ​​of carbon behavior graphs; Short-term process fluctuation filter: When the residual peak value matches the spatiotemporal match of a feeding event or oxygen blowing event, the alarm is blocked. The gradual leakage tracker performs time-accumulated analysis on continuous low-amplitude residuals and triggers an alarm when the accumulated amount exceeds the limit.

5. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 4, characterized in that: The dual-modal residual collaborative verification unit also includes: The cross-process migration inspection module inputs the abnormal characteristics of the current process into the process similar process model for deduction; When the consistency between the simulation results and the current residual characteristics exceeds a preset threshold, an anomaly is confirmed and an alarm is activated.

6. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 5, characterized in that: The process similarity model satisfies one of the following conditions: the thermal efficiency deviation between the rolling mill heating furnace and the steelmaking baking furnace is less than the set value; the pressure balance coefficient between the blast furnace gas system and the coking process is on the same order of magnitude.

7. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 1, characterized in that: The early warning circuit breaker control unit performs the following: For processes with continuous false alarms, increase the floating range of the threshold margin interval; For processes that are missed in reporting, the fluctuation range of the threshold margin interval is reduced.

8. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 1, characterized in that: The model evolution is achieved through closed-loop feedback, including: The discrepancy between the manual review conclusions and the system decision data is transmitted in reverse to the carbon behavior mapping modeling unit; Dynamically update the weight parameters of the transfer function and mapping relationship.

9. The IoT-based online monitoring and early warning platform for carbon emissions from multiple processes in steel enterprises according to claim 1, characterized in that: It also includes an alarm terminal, which communicates with the cloud analysis module to receive verified abnormal events and risk tracing reports.

10. A warning method based on the platform according to any one of claims 1-9, characterized in that, include: S1: Data acquisition steps: Through a sensor network deployed at key equipment nodes in the ironmaking, steelmaking, and rolling processes of steel enterprises, carbon emission data and related process parameters of each process are collected in real time. S2: Data preprocessing step: The edge computing node communicates with the sensor network, receives the collected data, and performs filtering and time alignment processing on it; S3: Carbon Behavior Mapping Modeling Steps: The cloud analysis module receives the data processed by the edge computing nodes, analyzes the process mechanism of each process, and constructs the dynamic threshold margin range of carbon emissions for each process based on the analysis results and the processed data. S4: Dual-modal residual collaborative verification steps: The dual-modal residual collaborative verification unit of the cloud analysis module performs residual analysis on the carbon emission data of each process, comparing theoretical and measured values, and performs cross-process migration verification. S5: Early warning and control steps: The early warning and circuit breaker control unit of the cloud analysis module dynamically adjusts the threshold sensitivity of the dynamic threshold margin range based on the results of the dual-modal residual collaborative verification. When carbon emission data is detected to exceed the dynamic threshold margin range, an early warning is issued, and circuit breaker control is activated if necessary. The carbon behavior map model is also driven to evolve based on continuous monitoring data. S6: Alarm Terminal Interaction Steps: The alarm terminal establishes a communication connection with the cloud analysis module, receives abnormal event information and corresponding risk tracing reports verified by the dual-modal residual collaborative verification unit, and displays and prompts the received information.

Images