A method and system for on-line evaluation of synergistic performance of pollution reduction and carbon reduction in short-process steel production of an electric furnace

Abstract

This invention discloses an online evaluation method and system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace (EAF) short-process steel production, relating to the field of industrial process control technology. The method includes: constructing a furnace-level carbon pollution data chain using furnace ID as the core index; identifying internal furnace operating condition stages; normalizing operating conditions through a similar furnace benchmark library; identifying synergistic or conflicting carbon pollution states; generating a carbon pollution sensitivity factor matrix to characterize the marginal impact of adjustable variables; and finally generating an optimized pollution reduction and carbon reduction path under multiple constraints and outputting it in a closed loop to the industrial control system. This invention decentralizes the evaluation granularity to the furnace-level and operating condition stage levels, eliminating the interference of objective operating condition differences on the evaluation results. It can accurately identify conflicting states such as pollution reduction with carbon increase and carbon reduction with pollution increase, and transform the evaluation results into executable industrial control parameters, realizing online synergistic evaluation and closed-loop optimization of pollution reduction and carbon reduction in the EAF short-process.

Description

Technical Field

[0001] This invention relates to the field of industrial process control technology, and more specifically to an online evaluation method and system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production. Background Technology

[0002] Electric arc furnace (EAF) short-process steel production is characterized by a short process, a high proportion of scrap steel recycling, and relatively low direct fossil energy consumption, making it an important path for the steel industry's low-carbon transformation. However, the actual pollution reduction and carbon reduction effect of the EAF short-process is not constant, and it is affected by factors such as scrap steel quality, EAF power supply system, oxygen / natural gas injection, furnace door operation, dust removal system operation, green electricity ratio, time-of-use electricity pricing, production cycle time, and product steel grade.

[0003] In existing technologies, steel companies typically conduct carbon emission accounting and pollutant emission monitoring separately. Carbon emission accounting is often based on enterprise, process, or monthly boundaries, while pollutant emission monitoring is often based on discharge outlets, hours, or days. The data boundaries, time granularity, and evaluation objects of the two are inconsistent, making it difficult to reflect the synergistic relationship between carbon and pollution in specific furnace cycles or operating conditions. In addition, existing green performance evaluation or carbon footprint management methods are usually based on indicator systems, weight calculations, report statistics, and comprehensive scoring. The results are mostly used for management decisions or compliance disclosures and are difficult to translate into executable industrial control parameters such as electric arc furnace power supply curves, oxygen injection volume, dust removal fan frequency, scrap steel ratio, and green electricity scheduling.

[0004] Specifically, existing technologies suffer from the following technical problems: The granularity of accounting is coarse; current enterprise-level or process-level accounting cannot pinpoint specific furnace batches, stages, and control parameters; carbon emissions and pollutant emissions are calculated separately, making it impossible to identify conflicting operating conditions such as "pollution reduction with increased carbon" or "carbon reduction with increased pollution"; different steel grades, scrap steel quality, furnace age, tapping temperature, and production load lead to incomparable per-ton steel indicators, resulting in distorted evaluation results; the time granularity differs among MES, DCS, EMS, CEMS, material, quality, and transportation data, making it difficult to form a reliable data chain; traditional evaluation results remain at the level of scores, grades, or reports, and cannot be converted into control parameters for electric arc furnace production processes; carbon pollution coupling effects exist between dust removal air volume, power supply curves, injection parameters, scrap steel ratio, and energy consumption periods, and multi-objective collaborative optimization methods are lacking.

[0005] Therefore, how to propose an online evaluation method and system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production, and overcome the shortcomings of existing technologies, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides an online evaluation method and system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production. This method delves into the evaluation criteria at the furnace sub-level and stage operating condition level, placing carbon emissions and pollutant emissions on the same time axis and within the same furnace data chain. It eliminates objective operating condition differences through normalization based on similar operating conditions, identifies and locates synergistic or conflicting states of pollution reduction and carbon reduction through carbon pollution conflict identification, and further generates control paths that can be fed back to production, energy, and environmental protection systems. This addresses the problems of coarse evaluation granularity, disconnected carbon pollution evaluation, incomparable results, and inability to implement control in existing technologies. To achieve the above objectives, the present invention adopts the following technical solution: An online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production includes: S1. Acquire multi-source operational data during the electric arc furnace short-process steel production process; S2. Using the furnace ID as the core index, perform time synchronization, furnace binding, and data trust verification on the multi-source operation data to construct a furnace-level carbon pollution data chain; S3. Based on the electric furnace power curve, oxygen flow rate, flue gas parameters and equipment status signals, identify and classify the working conditions within the furnace. S4. Construct a benchmark library of similar heats based on steel grade, scrap steel quality, furnace age, tapping temperature and production load, and calculate the normalized carbon emission intensity and normalized pollutant emission intensity of the current heat based on the benchmark library of similar heats. S5. Identify the carbon-pollution synergy or conflict state of the current furnace based on the normalized carbon emission intensity and normalized pollutant emission intensity. S6. Based on the marginal impact of adjustable control variables on carbon emissions and pollutant emissions, generate a carbon pollution sensitivity factor matrix; S7. Under the constraints of steel quality, production cycle time, equipment capacity, emission limits and energy costs, and combining the carbon pollution synergy or conflict state with the carbon pollution sensitivity factor matrix, generate an optimized path for pollution reduction and carbon reduction. S8. Output the pollution reduction and carbon reduction optimization path to the production scheduling system, energy management system, electric furnace control system or environmental protection control system to form a closed-loop control.

[0007] Optionally, the multi-source operational data includes: production data, electric furnace operation data, energy data, environmental protection data, material data, quality data, and transportation data; the furnace-level carbon pollution data chain uses the furnace ID as the primary key to establish an association model of furnace-operating stage-equipment-discharge port-energy flow-material flow-transportation flow-product batch, so that each furnace of steel corresponds to a set of carbon pollution profiles.

[0008] Optionally, the data reliability verification includes consistency verification of power consumption and power supply time, consistency verification of flue gas flow rate and dust removal fan frequency, consistency verification of material input and steel output, identification of abnormal constant values ​​of pollutant concentration, identification of missing data intervals, and consistency verification of furnace number across systems.

[0009] Optionally, the operating stages in S3 include at least one of the following: charging stage, melting stage, oxidation stage, reduction stage, tapping stage, and refining stage; the identification rules for each operating stage are as follows: When the scrap steel addition signal is valid, the furnace cover or furnace door is open, and the active power of the electric furnace is lower than the low power standby threshold, the corresponding time window is identified as the charging stage. When the furnace lid is closed, if the active power of the electric furnace is continuously higher than the effective power supply judgment threshold, the power curve enters the arc climbing stage or the high power melting stage, and the flue gas temperature continues to rise, the corresponding time window is identified as the melting stage. When the oxygen injection flow rate is continuously higher than the oxidation blowing flow rate threshold, and the CO volume fraction is higher than the decarbonization reaction judgment threshold or the CO volume fraction change rate changes from positive to negative, the corresponding time window is identified as the oxidation stage. When the oxygen injection flow rate is lower than the oxygen flow rate threshold during the reduction period, the CO volume fraction decreases by more than a preset ratio relative to the peak value during the oxidation period, and the active power of the electric furnace enters the low power heat preservation or intermittent power supply range, the corresponding time window is identified as the reduction stage. When the tapping tilting signal is valid and the electric furnace active power is lower than the tapping low power threshold, the corresponding time window is identified as the tapping stage. When the tapping stage is over, the ladle arrives at the refining station outside the furnace, and the refining start signal is valid or the active power of the refining furnace is higher than the refining power supply judgment threshold, the corresponding time window is identified as the refining stage.

[0010] Optionally, S4 includes: S41. Construct a feature vector for furnace operation conditions. The feature vector includes steel grade, scrap grade, scrap inclusion index, furnace age, target tapping temperature, production load, smelting cycle, green electricity ratio, and quality constraint features. S42. Select a set of candidate heats from historical heats that meet the requirements of being on the same production line, having comparable steel grades, meeting quality standards, and having reliable data. Calculate the comprehensive operating condition distance between the current heat and the candidate heats and generate a similarity coefficient. S43. Select a preset number of furnaces according to their similarity from high to low to form a set of similar furnaces; S44. Based on the carbon emission intensity and pollutant emission intensity of similar furnace sets, a benchmark carbon emission intensity and a benchmark pollutant emission intensity are generated by weighted calculation. S45. The ratio of the actual carbon emission intensity of the current furnace batch to the benchmark carbon emission intensity is taken as the normalized carbon emission intensity, and the ratio of the actual pollutant emission intensity of the current furnace batch to the benchmark pollutant emission intensity is taken as the normalized pollutant emission intensity.

[0011] Optionally, the carbon pollution synergy or conflict state includes: synergistic optimization zone, pollution reduction and carbon increase zone, carbon reduction and pollution increase zone, and dual high-risk zone, and the determination rule is as follows: When the normalized carbon emission intensity is ≤1 and the normalized pollutant emission intensity is ≤1, it is identified as a synergistic optimization zone; when the normalized carbon emission intensity is >1 and the normalized pollutant emission intensity is ≤1, it is identified as a pollution reduction and carbon increase zone; when the normalized carbon emission intensity is ≤1 and the normalized pollutant emission intensity is >1, it is identified as a carbon reduction and pollution increase zone; when the normalized carbon emission intensity is >1 and the normalized pollutant emission intensity is >1, it is identified as a dual high-risk zone.

[0012] Optionally, the adjustable control variables include at least one of the following: scrap steel ratio, power transmission curve, oxygen injection rate, natural gas injection rate, dust removal air volume, green electricity usage period, and technological upgrading measures; the carbon pollution sensitivity factor matrix is ​​used to describe the marginal impact of each adjustable control variable on carbon emission intensity, pollutant emission intensity, production cost, quality risk, and production cycle time; the carbon pollution sensitivity factor matrix is ​​calculated and generated by local weighted regression, digital twin perturbation method, or historical sample finite difference method.

[0013] Optionally, in S7, a multi-objective optimization model is established with carbon emission intensity per ton of steel, pollutant emission intensity, comprehensive cost, quality risk, and production cycle disturbance as optimization objectives; the constraints include steel grade quality and tapping temperature constraints, scrap steel inventory and batching constraints, equipment capacity constraints, emission limits and environmental protection facility constraints, production cycle and process connection constraints, electricity price and green electricity availability constraints, and safety boundary and interlocking constraints; the pollution reduction and carbon reduction optimization path includes at least one of the following: low-carbon priority path, pollution reduction priority path, cost balance path, and quality conservative path.

[0014] Optionally, S8 includes: decomposing the optimized path into control commands corresponding to different systems and outputting them to the MES, EMS, DCS and environmental control system respectively; collecting the actual execution data and performance results of each system, calculating the control deviation and performance deviation, and dynamically correcting the benchmark library of similar furnaces, the sensitive factor matrix and the optimization model; and selecting the output mode of automatic execution or execution after manual confirmation based on path risk, control amplitude and data reliability.

[0015] Optionally, an online evaluation system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production includes: The multi-source data acquisition module is used to acquire multi-source operational data during the electric arc furnace short-process steel production process; The boiler secondary data chain construction module is used to perform time synchronization, boiler binding and data trust verification of multi-source operation data with boiler ID as the core index, forming a boiler secondary carbon pollution data chain. The operating condition stage identification module is used to identify the operating condition stage of a furnace based on the electric furnace power curve, oxygen flow rate, flue gas parameters, and equipment status signals. The Similar Heat Trial Benchmark Library module is used to construct a similar heat trial benchmark library based on steel grade, scrap steel quality, furnace age, tapping temperature and production load, and to calculate normalized carbon emission intensity and normalized pollutant emission intensity. The carbon pollution conflict identification module is used to identify the synergistic or conflicting states of carbon pollution based on the normalization results. The sensitivity factor matrix generation module is used to generate a carbon pollution sensitivity factor matrix based on the marginal impact of adjustable control variables on carbon emissions and pollutant emissions. The multi-constraint path optimization module is used to generate pollution reduction and carbon reduction optimization paths under constraints such as steel grade quality, production cycle time, equipment capacity, emission limits and energy costs. The closed-loop control feedback module is used to output the optimized path to the production scheduling, energy management, electric furnace control or environmental control system, and dynamically correct the model based on the execution feedback.

[0016] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an online evaluation method and system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production, which has the following beneficial effects: This invention decentralizes the evaluation criteria from the enterprise, monthly, and process levels down to the furnace and operational stage levels, pinpointing performance differences at specific furnace and stage levels. By using a benchmark library of similar furnaces for operational condition normalization, it eliminates interference from differences in objective conditions such as steel grades, scrap quality, and furnace age, avoiding misjudgments of per-ton steel indicators. A four-quadrant carbon pollution identification mechanism is established to accurately identify pollution reduction with increased carbon emissions, carbon reduction with increased pollution emissions, high-risk conditions, and collaborative optimization states, resolving the conflict in optimization directions caused by separate evaluations of carbon pollution. A sensitivity factor matrix establishes a quantitative correlation between evaluation indicators and control parameters, outputting executable control paths such as scrap ratio, power supply curves, injection parameters, dust removal air volume, and energy consumption periods, realizing the implementation from evaluation to control. Multi-source consistency verification, anomaly identification, and credibility assignment ensure the reliability of accounting and evaluation after the fusion of multi-source heterogeneous data. This invention systematically solves the technical problems of difficulty in simultaneously evaluating carbon emissions and pollutant emissions, difficulty in locating the causes of operating conditions, and difficulty in converting them into control parameters in short-process electric furnace production by combining a combination of furnace secondary data chain construction, electric furnace operating condition stage identification, similar operating condition normalization, carbon pollution conflict identification, sensitive factor matrix, multi-constraint path optimization, and closed-loop control feedback. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a diagram of the overall architecture of the present invention.

[0019] Figure 2 This is a flowchart of an online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to the present invention.

[0020] Figure 3 This is a schematic diagram of the furnace secondary carbon pollution data link structure of the present invention.

[0021] Figure 4 This is a schematic diagram of the carbon pollution conflict identification matrix of the present invention.

[0022] Figure 5 This is a schematic diagram of the carbon pollution sensitivity factor matrix of the present invention.

[0023] Figure 6 This is a schematic diagram of the closed-loop carbon pollution synergistic control of the present invention. Detailed Implementation

[0024] 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.

[0025] This invention discloses an online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production, such as... Figure 2 As shown, it includes: S1. Acquire multi-source operational data during the electric arc furnace short-process steel production process; S2. Using the furnace ID as the core index, perform time synchronization, furnace binding, and data trust verification on the multi-source operation data to construct a furnace-level carbon pollution data chain; S3. Based on the electric furnace power curve, oxygen flow rate, flue gas parameters and equipment status signals, identify and classify the working conditions within the furnace. S4. Construct a benchmark library of similar heats based on steel grade, scrap steel quality, furnace age, tapping temperature and production load, and calculate the normalized carbon emission intensity and normalized pollutant emission intensity of the current heat based on the benchmark library of similar heats. S5. Identify the carbon-pollution synergy or conflict state of the current furnace based on the normalized carbon emission intensity and normalized pollutant emission intensity. S6. Based on the marginal impact of adjustable control variables on carbon emissions and pollutant emissions, generate a carbon pollution sensitivity factor matrix; S7. Under the constraints of steel quality, production cycle time, equipment capacity, emission limits and energy costs, and combining the carbon pollution synergy or conflict state with the carbon pollution sensitivity factor matrix, generate an optimized path for pollution reduction and carbon reduction. S8. Output the pollution reduction and carbon reduction optimization path to the production scheduling system, energy management system, electric furnace control system or environmental protection control system to form a closed-loop control.

[0026] In a specific implementation, an online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production is provided, such as... Figure 1 As shown, it specifically includes: Step 1: Acquisition of multi-source data and construction of a furnace secondary carbon pollution data chain 1. Acquisition of multi-source operational data The system acquires multi-source operational data for the electric furnace short-process production from MES, DCS, EMS, CEMS, LIMS, WMS, access control and transportation system, and environmental management platform. The data acquisition frequency can be set to second-level, minute-level, furnace-level, shift-level, or daily-level based on the data type. The sources and uses of various data types are shown in Table 1 below. Table 1

[0027] The detailed parameter definitions for each type of data are as follows: (1) MES data parameters MES primarily provides production execution data such as production plan, heat number, steel grade, output, process cycle time, and product batch. Specific parameters include: Heat number information: heat_id is a unique production identifier for each heat of steel in an electric arc furnace and is the core index for multi-source data binding; Production plan: plan number, order number, corresponding to production task, customer order or production schedule number; Steel grade information: steel grade, grade, and target composition, which are the steel grade and chemical composition requirements that the current heat or product needs to achieve; Production information: planned output, actual steel output, and qualified output, used to calculate electricity consumption per ton of steel, carbon emissions per ton of steel, and pollutant emissions per ton of steel; Process information: Electric furnace, refining, continuous casting, and rolling process codes, which represent the process route the product has gone through; Time information: charging start time, power supply start time, tapping time, refining end time, used for furnace boundary identification and multi-system data time alignment; Production cycle time: smelting cycle, refining cycle, continuous casting cycle, used to evaluate production efficiency and path optimization constraints; Product batches: billet batches, coil batches, bar and wire rod batches, used to transfer the carbon pollution performance of each furnace to the product-level carbon pollution profile; Process route: electric furnace—refining—continuous casting—rolling path, used to determine the allocation boundary of carbon emissions and pollutant emissions; Team information: shifts and work teams, used for production management traceability and abnormal furnace analysis.

[0028] (2) DCS data parameters

[0029] DCS primarily provides on-site process control data for electric furnaces, refining processes, dust removal, pulse-jet cleaning, and flue gas control. Specific parameters include: Electric furnace electrical parameters: three-phase voltage, three-phase current, active power, reactive power, and power factor, which reflect the power supply status of the electric furnace and are used to identify the melting stage, calculate power consumption, and optimize the power supply curve; Power supply parameters: power supply time, power curve, electrode adjustment signal, which characterize the power input method in the electric furnace smelting process; Electrode control parameters: electrode lifting position, electrode consumption, and electrode current fluctuation, used to determine arc stability, abnormal power consumption, and electrode carbon emissions; Furnace status parameters: furnace cover opening and closing, furnace door opening and closing, and steel tapping tilting signal, used to divide the working conditions such as charging, melting, and tapping; Injection parameters: oxygen flow rate, natural gas flow rate, carbon powder injection quantity, and injection pressure, used to analyze melting efficiency, oxidation reaction, carbon emissions, and NOx generation risk; Flue gas parameters: furnace gas temperature, flue gas temperature, furnace pressure, CO concentration, used to determine the reaction state inside the furnace, flue gas load, and dust removal requirements; Dust removal control parameters: fan frequency, air volume, valve opening, and bag pressure difference, used to evaluate the operating status of the dust removal system and the power consumption of the fan; Refining parameters: LF power, stirring gas flow rate, refining time, and molten steel temperature, used to calculate energy consumption and quality control constraints in the refining process; Safety interlock parameters: alarm signals, interlock status, equipment fault status, used to identify abnormal operating conditions and eliminate abnormal data; Control setpoints: target power, target flow rate, target air volume, and temperature setpoints, used to generate control parameter feedback after path optimization.

[0030] (3) EMS data parameters

[0031] EMS primarily provides data on energy consumption and supply, including electricity, gas, steam, oxygen, compressed air, green electricity, and waste heat. Specific parameters include: Electricity consumption: total electricity consumption, electric furnace electricity consumption, refining electricity consumption, rolling electricity consumption, and dust removal electricity consumption, used to calculate indirect carbon emissions from purchased electricity and electricity consumption per ton of steel; Time-of-use electricity data: time-of-use electricity consumption, peak and off-peak electricity consumption, and time-of-use electricity price, used to optimize electric furnace production periods and reduce energy costs; Green electricity data: green electricity usage, green electricity ratio, and electricity volume corresponding to green electricity certificates, used to adjust the carbon emission factor of purchased electricity; On-site power generation: photovoltaic power generation, power supply from self-owned power plant, and waste heat power generation, used to distinguish power sources and calculate carbon emissions; Fuel consumption: Consumption of natural gas, coke oven gas, converter gas, diesel, etc., used to calculate carbon emissions from direct combustion; Industrial gases: consumption of oxygen, nitrogen, argon, and compressed air, used to evaluate energy consumption in processes such as spraying, stirring, and cutting; Steam heat: Steam consumption, hot water consumption, and heat recovery, used for evaluating carbon emissions from purchased heat and waste heat utilization; Energy medium pressure: oxygen pressure, gas pressure, compressed air pressure, used to assess the stability of energy supply and the feasibility of process control; Energy costs: electricity price, gas price, steam price, and oxygen unit price, used as cost constraints in path optimization; Energy balance: energy input, output, loss, and recovery, used to verify the reliability of energy data.

[0032] (4) CEMS data parameters

[0033] CEMS primarily provides continuous monitoring data on pollutant concentrations, flue gas flow rates, temperature, pressure, and oxygen content at organized emission outlets. Specific parameters include: Particulate matter data: PM concentration, particulate matter emission rate, used to calculate particulate matter emissions and particulate matter emission intensity per ton of steel; Gaseous pollutants: SO2 concentration, NOx concentration, CO concentration, used to calculate the emission levels of sulfur dioxide, nitrogen oxides, and carbon monoxide; Carbon-related gases: CO2 concentration and CO concentration, which can be used to help determine carbon emissions, combustion status, and in-furnace reaction status; Flue gas flow rate: standard dry flow rate, operating condition flow rate, flue gas velocity, used to convert pollutant concentration into emissions; Flue gas condition parameters: flue gas temperature, pressure, humidity, and oxygen content, used for standard condition conversion, concentration calculation, and data correction; Emission outlet information: outlet number, outlet type, and corresponding equipment, used to bind emission data to electric furnace, refining furnace, rolling mill, or dust removal system; Monitoring time: sampling time, minute average, hourly average, and daily average, used to match the furnace time window; Data status: valid status, calibration status, maintenance status, and anomaly status, used to determine whether CEMS data can be used for performance evaluation; Information on exceeding standards: time of exceeding standards, factor of exceeding standards, and multiple of exceeding standards, used to identify high-risk or abnormal environmental conditions; Total emissions: The cumulative amount of pollutants emitted, used for process-level, furnace-level, or daily emission accounting.

[0034] (5) LIMS data parameters

[0035] LIMS primarily provides testing and analysis data for scrap steel, molten steel, slag, and finished steel. Specific parameters include: Scrap steel testing data: scrap steel grade, inclusion content, oil content, and galvanized material ratio, used to determine the impact of scrap steel quality on power consumption, pollutants, and metal recovery rate; Steel composition: The content of elements such as C, Si, Mn, P, S, Cr, Ni, and Mo is used to determine whether the target composition constraints of the steel grade are met; Finished product composition: Chemical composition of finished steel, used for product quality traceability and furnace batch normalization; Slag composition: CaO, SiO2, FeO, MgO, etc., used to evaluate slag-making system, desulfurization and dephosphorization effect and material process emissions; Temperature monitoring: tapping temperature, refining temperature, and pouring temperature are used to determine the rationality of thermal regime and power consumption; Mechanical properties: yield strength, tensile strength, elongation, impact energy, used for product quality constraints in path optimization; Defect detection: cracks, inclusions, segregation, surface defects, used to identify the impact of process adjustments on quality risks; Inspection batch: Sample number, furnace number, product batch number, used to bind to MES furnace number and product batch; Detection time: Sampling time and testing time, used for data time-series matching; Judgment results: qualified, unqualified, re-inspected, or accepted with concession, used for quality risk assessment.

[0036] (6) WMS data parameters

[0037] WMS primarily provides inventory, inbound / outbound, batch, and storage location data for scrap steel, auxiliary materials, alloys, finished products, and semi-finished products. Specific parameters include: Scrap steel inventory: scrap steel categories, grades, and inventory levels, used for scrap steel ratio optimization and inventory constraints; Source of scrap steel: supplier, place of origin, batch number, used to determine the quality stability of scrap steel and carbon emissions during transportation; Scrap steel storage location: location of the material yard, stockpile number, and storage area number, used for material tracking and batching execution; Warehouse entry data: entry time, entry weight, and acceptance results, used to confirm the actual arrival of scrap steel and auxiliary materials; Outbound data: outbound time, outbound weight, and furnace number used to link material consumption to furnace number; Auxiliary material inventory: Inventory levels of lime, dolomite, fluorite, carbon powder, etc., used for slag-making processes and material carbon emission calculations; Alloy inventory: Inventory levels of ferrosilicon, ferromanganese, ferrochrome, ferronickel, etc., used for steel composition constraints and cost optimization; Finished goods inventory: Inventory of billets, coils, bars, and wire rods, used for product batch and order delivery management; Inventory costs: unit price, batch cost, and procurement cost, used as cost constraints in path optimization; Inventory data: discrepancies between physical inventory and records, inventory time, and adjustment amounts, used for verifying the reliability of material data.

[0038] (7) Access control and transportation system data parameters

[0039] The access control transportation system primarily provides data on vehicle entry and exit, transported materials, vehicle energy type, transport weight, and queue waiting times for clean transportation evaluation and carbon emission calculation. Specific parameters include: Vehicle identification: license plate number, vehicle number, carrier, used for vehicle identification and logistics tracking; Vehicle types: heavy trucks, semi-trailers, tank trucks, loaders, etc., used to determine transport emission factors; Energy types: diesel, LNG, electric, hydrogen, used to calculate transportation carbon emissions and the proportion of clean transportation; Emission stages: China V, China VI, new energy, etc., used to evaluate the level of vehicle pollutant emissions; Inbound and outbound times: Inbound time and outbound time are used to calculate dwell time, queuing time, and transportation efficiency; Transported materials: scrap steel, auxiliary materials, alloys, finished steel, slag, etc., used to determine the attribution of the transportation link; Transport weight: gross weight, tare weight, net weight, used to calculate carbon emissions per unit of material transport; Transportation distance: external transportation distance and internal transfer distance, used for estimating carbon emissions and pollutant emissions from transportation; Loading and unloading status: loading time, unloading time, and waiting time, used to evaluate logistics efficiency and vehicle idling emissions; Clean transportation ratio: The proportion of transportation volume by new energy vehicles and the proportion of rail or water transport, used for the clean transportation performance evaluation of short-process steel enterprises.

[0040] (8) Environmental management and control platform data parameters

[0041] Environmental management platforms typically integrate data from CEMS, fugitive emissions, plant boundary air quality monitoring stations, video surveillance, treatment facility operation, and ultra-low emission control. Specific parameters include: Organized emission data: emission concentration, emission amount, and emission rate, sourced from CEMS or online monitoring systems, used for pollutant performance evaluation; Unorganized emission data: monitoring of particulate matter and dust in the plant area, and alarms for unorganized emissions, used to assess the pollution risk in the material yard, transfer, loading and unloading, and furnace front areas; Factory boundary monitoring data: PM2.5, PM10, TSP, SO2, NO2, CO, used to assess factory boundary environmental quality and diffusion impact; Operational data of environmental protection facilities: dust collector operating status, desulfurization and denitrification operating status, and fan status, used to determine whether the environmental protection facilities are operating normally; Video surveillance data: material yard enclosure, vehicle washing, road dust, and exhaust plumes, used to help identify fugitive emissions and environmental anomalies; Exceedance alarm data: exceeding time, exceeding factor, alarm level, used to identify high-risk furnace batches or abnormal operating conditions; Energy consumption of environmental protection equipment: power consumption of dust removal fans, power consumption of desulfurization and denitrification, and power consumption of water pumps, used to identify the state of "pollution reduction and carbon increase"; Environmental protection logbook: records of facility start-up and shutdown, chemical consumption, and maintenance, used to verify the authenticity of the operation of the treatment facilities; Discharge permit data: permitted emission concentration, permitted emission amount, and discharge outlet number, used as emission constraints for path optimization; Ultra-low emission data: organized emissions, unorganized emissions, clean transportation, and monitoring indicators, used for performance evaluation of ultra-low emissions in steel enterprises.

[0042] 2. Time synchronization, furnace batch binding, and trusted verification

[0043] Data from different frequencies and sources are uniformly mapped to the furnace timeline. The furnace timeline includes at least the following nodes: charging start, power supply start, melting and clearing, oxidation end, tapping, and refining end.

[0044] For equipment data such as electric furnaces and refining furnaces that operate on a per-furnace basis, the data is directly sliced ​​and bound according to the furnace start time, power supply start time, tapping time, and equipment status signals. For continuously operating equipment such as dust removal systems, energy stations, continuous casting, and rolling mills, the data is allocated and bound based on time windows, equipment status, flue gas flow, damper opening, power load, material flow direction, and production plan.

[0045] The system assigns a credibility identifier to each piece of bound data. The credibility identifier can be determined based on the data source, sampling frequency, abnormal status, degree of manual input, missing data compensation method, and multi-source consistency verification results.

[0046] Data reliability verification includes: consistency verification of power consumption and power supply time, consistency verification of flue gas flow rate and dust removal fan frequency, consistency verification of material input and steel output, identification of abnormal CEMS concentration constant value, identification of missing data intervals, and consistency verification of furnace number across systems.

[0047] 3. Furnace secondary carbon pollution data link structure

[0048] Using the furnace ID as the primary key, a correlation model is established between furnace number, operating condition stage, equipment, discharge port, energy flow, material flow, transportation flow, and product batch. The data chain structure is as follows: Figure 3 As shown, this data model provides a complete carbon pollution profile for each heat of steel, enabling traceable correlation between pollutant emissions, carbon emissions, production processes, and quality outcomes.

[0049] Step 2: Identification of Electric Furnace Operating Conditions

[0050] Based on power curves, oxygen flow rate, natural gas flow rate, furnace door opening and closing, furnace cover opening and closing, flue gas temperature, CO concentration, tapping tilting signal, and scrap steel addition signal, the heat cycle is divided into charging stage, melting stage, oxidation stage, reduction stage, tapping stage, and refining stage. Multi-signal fusion criteria are used to segment and identify the electric arc furnace heat cycle process. The specific identification rules are as follows: Charging phase: When the scrap steel addition signal or the basket feeding signal is valid, and the furnace cover or furnace door is open, and the electric furnace active power is lower than the low power standby threshold, the corresponding time window is identified as the charging period. Melting stage: When the furnace lid is closed, the active power of the electric furnace is continuously higher than the effective power supply judgment threshold, and the power curve enters the arc climbing stage or the high power melting stage, while the flue gas temperature shows a continuous upward trend, the corresponding time window is identified as the melting period. Oxidation stage: When the oxygen injection flow rate is continuously higher than the oxidation blowing flow rate threshold, and the CO volume fraction is higher than the decarbonization reaction judgment threshold or the CO volume fraction change rate changes from positive to negative, the corresponding time window is identified as the oxidation period. Reduction phase: When the oxygen injection flow rate is lower than the oxygen flow rate threshold during the reduction period, the CO volume fraction decreases by more than a preset ratio relative to the peak value during the oxidation period, and the active power of the electric furnace enters the low power heat preservation or intermittent power supply range, the corresponding time window will be identified as the reduction period. Steel tapping stage: When the steel tapping tilting signal or the EBT tapping port opening signal is valid, and the electric furnace active power is lower than the low power threshold for steel tapping, the corresponding time window will be identified as the steel tapping period; Refining stage: When the steel ladle arrives at the ladle refining station after the steel tapping period ends, and the MES refining start signal is valid or the active power of the LF refining furnace is higher than the refining power supply judgment threshold, the corresponding time window will be identified as the ladle refining period.

[0051] The present invention further corrects the stage identification results through a furnace state machine: when the duration of a candidate stage is less than the preset minimum duration, the state switch is not triggered; when multiple stage identification conditions are met at the same time, the current stage is determined according to the priority of the steel tapping tilt signal, scrap steel addition signal, furnace cover opening and closing signal, effective power supply status, oxygen injection flow rate, CO volume fraction, flue gas temperature and natural gas burner flow rate.

[0052] The corresponding relationships between each stage are shown in Table 2 below: Table 2

[0053] Step 3: Construction of a benchmark library for similar furnace batches and normalization of operating conditions

[0054] Operating condition normalization is used to eliminate the interference of differences in objective production conditions on the evaluation results, so that the evaluation results can reflect the true process control performance.

[0055] 1. Data structure of the benchmark library for similar furnace batches

[0056] The system uses the furnace number (heat_id) as the primary key to associate data from MES, DCS, EMS, CEMS, LIMS, WMS, access control and transportation systems, and environmental management platforms according to timestamps, equipment codes, material batches, and product batches, forming historical furnace sample data. Each historical furnace sample includes at least the following: operating characteristics, process curves, carbon emission intensity, pollutant emission intensity, and data reliability.

[0057] (1) Feature vector of furnace operation condition

[0058] For the i-th furnace, the system constructs a furnace operating condition feature vector x_i: x_i=[g_i,s_i,I_i^scrap,A_i,T_i^tap,L_i,τ_i,ρ_i^green,q_i]; Where g_i represents the steel grade or steel class; s_i represents the scrap grade and the proportion vector of each grade of scrap; I_i^{scrap} represents the scrap inclusion index; A_i represents the furnace age or furnace lining service status; T_i^{tap} represents the target tapping temperature; L_i represents the production load; τ_i represents the smelting cycle; ρ_i^green represents the green electricity ratio; and q_i represents the quality constraint characteristics.

[0059] (2) Scrap steel inclusion index

[0060] The scrap steel inclusion index is used to transform factors in scrap steel that are detrimental to smelting efficiency, pollution control, and quality stability into continuous variables. The calculation method is as follows: I_i^scrap=Σ_(r=1)^Rλ_r·z_(i,r); Where z_(i,r) represents the normalized value of the r-th adverse factor, which includes at least one of the risks of exceeding limits for residual elements such as oil, mud, galvanized material, thin material, non-metallic inclusions, copper and tin; λ_r is the corresponding weight.

[0061] (3) Feature standardization

[0062] For continuous variables such as target tapping temperature, furnace age, production load, smelting cycle, and green electricity ratio, the system uses standard deviation normalization or range normalization: _(i,m)=(x_(i,m)-μ_m) / σ_m; _(i,m)=(x_(i,m)-x_m^min) / (x_m^max-x_m^min); Where μ_m and σ_m are the mean and standard deviation of the m-th feature in the historical furnace batches, respectively; x_m^min and x_m^max are the minimum and maximum values ​​of the feature, respectively.

[0063] 2. Similar Furnace Search

[0064] (1) Selection of candidate furnace set

[0065] For the current heat i, the system first filters the candidate heat set Ωi from the historical heat database. Candidate heats must meet the following conditions: they must be from the same or comparable production line, use the same steel grade or belong to the same steel family, pass quality assessment, and have their data verified as reliable. Ω_i={j|line_j=line_i,g_j∈G(g_i),Q_j=1,M_j>0,D_j=1}; Wherein, line_j represents the production line; G(g_i) represents the set of steels that are the same as or belong to the same steel class as the current steel grade; Q_j represents the quality qualification mark; M_j represents the actual steel output or qualified steel output; and D_j represents the data reliability verification mark.

[0066] The system removes furnaces that are under maintenance, cold furnaces, furnaces that are drying, furnaces that are involved in accidents, furnaces with invalid CEMS, furnaces with abnormal steel output, furnaces with abnormal smelting cycles, and furnaces whose electricity consumption or pollutant emissions deviate significantly from the reasonable range of the process.

[0067] (2) Distance calculation for mixed working conditions

[0068] Since the furnace characteristics include continuous variables, categorical variables, ranking variables, vector variables, and process curves, the system uses a weighted mixed distance to describe the difference in operating conditions between the current furnace and candidate furnaces: d_ij^static=[Σ_(m=1)^Mw_m·d_m(x_(i,m),x_(j,m))] / [Σ_(m=1)^Mw_m]; Where w_m is the weight of the m-th feature; d_m is the distance function of the m-th feature.

[0069] For continuous variables, standardized absolute distance is used: d_m(x_(i,m),x_(j,m))=|(x_(i,m)-x_(j,m)) / R_m|; Where R_m is the range, standard deviation, or engineering allowable fluctuation scale of this feature.

[0070] For categorical variables, the category matching distance is used: if x_(i,m)=x_(j,m), then d_m=0; if they belong to the same steel class or adjacent scrap grades, then d_m=0.5; if they are not comparable, then d_m=1.

[0071] For the scrap steel proportioning vector, a half-L1 distance is used: d_s(s_i,s_j)=1 / 2·Σ_(r=1)^R|s_(i,r)-s_(j,r)|.

[0072] (3) Process curve similarity correction

[0073] After each furnace cycle, the system can further incorporate process similarity data from power curves, oxygen flow rate curves, and flue gas temperature curves. Since different furnace cycles have different smelting periods, the time axis is first normalized, and then dynamic time warping distance is used for comparison. d_ij^curve=η_P·DTW( _i, _j)+η_O·DTW( _i, _j)+η_T·DTW( _i^fg, _j^fg); in, _i represents the time-normalized active power curve. _i represents the oxygen injection flow rate curve. _i^fg represents the flue gas temperature curve, DTW represents the dynamic time warping distance, and η_P, η_O, and η_T represent the weights of each process curve, respectively.

[0074] (4) Overall working condition distance and similarity coefficient

[0075] D_ij=α·d_ij^static+β·d_ij^curve+γ·d_ij^quality, and α+β+γ=1; Wherein, d_ij^quality represents the difference in quality control characteristics, including component control level, target P / S content, target endpoint carbon content, quality judgment level, etc.

[0076] The system generates a similarity coefficient based on the distance under comprehensive operating conditions: S_ij=exp[-D_ij^2 / (2σ^2)]·R_j·exp(-λ·Δt_j); Where σ represents the distance scale parameter; R_j represents the data reliability of historical furnace j; Δt_j represents the time interval between historical furnace j and the current furnace; exp(-λ·Δt_j) is the time decay coefficient, used to reduce the impact of older furnaces on the current benchmark.

[0077] (5) Determination of the set of similar furnace batches

[0078] The system sorts furnaces by similarity S_ij from high to low, selects the top K furnaces, or selects furnaces with similarity not lower than the threshold S_min, to form a set N_i of similar furnaces. N_i=TopK_(j∈Ω_i)(S_ij); N_i={j∈Ω_i|S_ij≥S_min}; When the number of similar furnace runs is less than the preset minimum sample size K_min, the system expands the search scope in the following order: from the same steel grade to the same steel class; from the same scrap steel grade combination to similar scrap steel grade combinations; from the same electric furnace production line to the same type of electric furnace production line; from the same target tapping temperature range to adjacent temperature ranges; and if necessary, clustering benchmarks or regression benchmarks are used for supplementation.

[0079] 3. Normalized intensity calculation

[0080] For the current furnace cycle i, let CIi be the actual carbon emission intensity and CIbase,i be the baseline carbon emission intensity for similar furnace cycles. Then the normalized carbon emission intensity is: NCIorm,i = CIbase,iCIi

[0081] Let Pi be the actual pollutant emission intensity, and PIbase,i be the baseline pollutant emission intensity for similar furnace batches. Then the normalized pollutant emission intensity is: NPInorm,i = PIbase,iPIi

[0082] When NCIrm,i is less than 1, it means that the carbon performance of the current furnace is better than the benchmark under similar operating conditions; when NPCIrm,i is less than 1, it means that the pollutant performance of the current furnace is better than the benchmark under similar operating conditions.

[0083] 4. Dynamic updates to the benchmark library

[0084] The benchmark library for similar furnace runs can be dynamically updated as production progresses. Furnace runs with low data reliability, equipment malfunctions, substandard quality, or accidental shutdowns can have their sample weight reduced or be removed from the benchmark library. Furnace runs that have been verified to have reached a state of collaborative optimization can be used as preferred samples to update the benchmark path.

[0085] Step 4: Calculation of carbon emissions and pollutant emission intensity

[0086] 5.1 Calculation of carbon emission intensity

[0087] Carbon emissions include at least indirect emissions from purchased electricity, direct emissions from fuel, emissions from material processes, emissions from electrode consumption, and emissions from transportation.

[0088] (1) Calculation of secondary carbon emissions from the furnace

[0089] For the j-th historical furnace batch, the furnace batch carbon emissions can be expressed as: E_j^CO2=E_j^elec+E_j^fuel+E_j^material+E_j^electrode+E_j^transport-E_j^recovery; Wherein, E_j^elec represents the indirect carbon emissions corresponding to purchased electricity or on-site electricity; E_j^fuel represents the direct carbon emissions corresponding to the combustion of fuels such as natural gas, coal gas, and diesel; E_j^material represents the carbon emissions from the material processing of materials such as lime, dolomite, and carbon powder; E_j^electrode represents the carbon emissions from electrode consumption; E_j^transport represents the carbon emissions from the transportation process; and E_j^recovery represents the emission reductions calculated from waste heat recovery, green electricity substitution, or energy recovery.

[0090] The carbon emissions associated with purchased electricity are calculated based on the time-varying electricity carbon factor and the proportion of green electricity: E_j^elec=Σ_(t∈T_j)P_j(t)·Δt·[(1-ρ_j(t))EF_grid(t)+ρ_j(t)EF_green(t)]; Where P_j(t) is the power consumption of furnace j at time t; Δt is the sampling time interval; ρ_j(t) is the green electricity ratio at time t; EF_grid(t) is the grid electricity carbon emission factor at time t; and EF_green(t) is the emission factor corresponding to green electricity or the verified substitution factor.

[0091] Carbon emissions from fuel combustion can be expressed as: E_j^fuel=Σ_(f=1)^FAD_(j,f)·NCV_f·CC_f·OF_f·44 / 12; Where AD_(j,f) represents the activity data for fuel of type f; NCV_f represents the lower heating value; CC_f represents the carbon content per unit calorific value; and OF_f represents the carbon oxidation rate.

[0092] (2) Furnace secondary carbon emission intensity

[0093] CI_j = E_j^CO2 / M_j;

[0094] Where M_j represents the qualified steel output or actual steel output of heat j. The unit of CI_j can be tCO2 / t steel or kgCO2 / t steel.

[0095] (3) Generation of baseline carbon emission intensity

[0096] The system generates weights based on the similarity of each furnace in the similar furnace set Ni: ω_ij = S_ij / Σ_(j∈N_i)S_ij; To mitigate the impact of abnormal furnace runs, the system can truncate or Windsorize the carbon emission intensity of similar furnace runs: CI_j =clip(CI_j,Q_5%^CI,Q_95%^CI); Among them, Q_5%^CI and Q_95%^CI are the 5th percentile and 95th percentile of carbon emission intensity in the set of similar furnaces, respectively.

[0097] The average baseline carbon emission intensity for similar operating conditions in the current furnace run is: CI_base,i=Σ_(j∈N_i)ω_ij·CI_j ; In another implementation, the system uses weighted quantiles to generate an advanced benchmark carbon emission intensity: CI_adv,i=Q_p^w({CI_j}_{j∈N_i},{ω_ij}_{j∈N_i}); Where Q_p^w represents the weighted p-quantile. Since lower carbon emission intensity is better, p is preferably taken as 0.20-0.35 to represent the optimal furnace level under similar operating conditions.

[0098] 5. Calculation of pollutant emission intensity

[0099] Pollutant emissions include at least one or more of particulate matter, SO2, and NOx.

[0100] (1) Calculation of organized pollutant emissions

[0101] For the j-th furnace, the p-th exhaust outlet, and the k-th pollutant, the system calculates the organized emissions based on the CEMS-converted concentration or standard dry concentration, the standard dry flow rate of the flue gas, and the sampling time: E_(j,k,p)^org=Σ_(t∈T_j)[C_(k,p)(t)·Q_p(t)·Δt·κ_p(t)] / 10^6; Where C_(k,p)(t) is the concentration of pollutant of type k at outlet p at time t, in mg / Nm³; Q_p(t) is the standard dry flow rate of flue gas, in Nm³ / h; Δt is the sampling time, in h; κ_p(t) is the conversion correction factor for temperature, pressure, humidity, oxygen content, and standard conditions; 10^6 is used to convert mg to kg. Pollutant k may include particulate matter, SO2, NOx, CO, etc.

[0102] (2) Furnace allocation for multiple furnaces sharing the same exhaust outlet

[0103] When a single exhaust outlet or dust removal system corresponds to multiple furnaces or multiple processes, the system generates an allocation weight based on the contribution of flue gas from each furnace, power load, dust removal air volume, and furnace door opening status. _(j,p)=[Σ_(t∈T_j)(a1·Q_j^offgas(t)+a2·P_j(t)+a3·F_j^fan(t)+a4·I_j^door(t))] / [ Σ_(h∈H_p)Σ_(t∈T_h)(a1·Q_h^offgas(t)+a2·P_h(t)+a3·F_h^fan(t)+a4·I_h^door(t))]; Where Q_j^offgas(t) represents the amount of furnace gas or flue gas in furnace j; P_j(t) represents the active power of the electric furnace; F_j^fan(t) represents the frequency or air volume of the dust removal fan; I_j^door(t) represents the furnace door opening status; a1 to a4 are the apportionment weight coefficients; H_p represents the set of furnaces related to the discharge port p.

[0104] After allocation, the organized emissions from furnace j at outlet p are: E_(j,k,p)^org,alloc= _(j,p)·E_(p,k)^org.

[0105] (3) Furnace secondary pollutant emission intensity

[0106] The total emissions of Class k pollutants from the secondary stage of the furnace can be expressed as: E_(j,k)^poll=E_(j,k)^org+E_(j,k)^unorg+E_(j,k)^transport; Where E_(j,k)^unorg represents the converted amount of fugitive emissions from material yard, loading and unloading, transfer, furnace front operation, etc.; E_(j,k)^transport represents the pollutant emissions corresponding to internal and external transportation.

[0107] PI_(j,k)=E_(j,k)^poll / M_j; The unit of PI_(j,k) can be kg / t steel or g / t steel.

[0108] (4) Generation of baseline pollutant emission intensity

[0109] The system first performs outlier processing on the pollutant emission intensity of similar furnace batches: PI_(j,k) =clip(PI_(j,k),Q_5%,k^PI,Q_95%,k^PI); The average baseline emission intensity of pollutant class k under similar operating conditions is: PI_base,i,k=Σ_(j∈N_i)ω_ij·PI_(j,k) ; The advanced baseline emission intensity for pollutant category k is: PI_adv,i,k=Q_p^w({PI_(j,k)}_{j∈N_i},{ω_ij}_{j∈N_i}), p=0.20-0.35; If a comprehensive pollutant baseline intensity needs to be generated, θ_k can be determined based on the discharge permit limit, ultra-low emission limit, enterprise internal control limit, or health impact weight, and then calculated: PI_base,i^comprehensive=Σ_(k=1)^Kθ_k·[PI_base,i,k / L_k], and Σ_(k=1)^Kθ_k=1; Where L_k represents the permitted emission limit, ultra-low emission limit, or internal control limit for pollutant of category k.

[0110] Step 5: Identification of Carbon Pollution Synergy / Conflict Status

[0111] Based on normalized carbon emission intensity and normalized pollutant emission intensity, the current furnace cycle or operating condition is identified as belonging to a synergistic optimization zone, a pollution reduction and carbon increase zone, a carbon reduction and pollution increase zone, or a dual high-risk zone. The identification matrix is ​​as follows: Figure 4 As shown.

[0112] After obtaining the actual carbon emission intensity CI_i, actual pollutant emission intensity PI_(i,k), baseline carbon emission intensity CI_base,i, and baseline pollutant emission intensity PI_base,i,k for the current furnace cycle, the system generates normalized performance indicators: NCI_i = CI_i / CI_base,i; NPI_(i,k)=PI_(i,k) / PI_base,i,k; NPI_i=Σ_(k=1)^Kθ_k·NPI_(i,k); When NCI_i < 1, it indicates that the carbon emission intensity of the current furnace is better than the benchmark under similar operating conditions; when NCI_i > 1, it indicates that it is worse than the benchmark under similar operating conditions. When NPI_i < 1, it indicates that the pollutant emission performance of the current furnace is better than the benchmark under similar operating conditions; when NPI_i > 1, it indicates that it is worse than the benchmark under similar operating conditions.

[0113] The discrimination rules for the four types of states are shown in Table 3 below: Table 3

[0114] Different states correspond to different subsequent control strategies: Collaborative optimization zone: solidified into the preferred furnace batch sample, serving as a template for subsequent paths; Pollution reduction and carbon increase zone: Optimize dust removal air volume, fan frequency, energy period and waste heat recovery; Carbon reduction and pollution increase zone: Improve capture efficiency, adjust furnace door operation and dust control strategies; High-risk areas: Jointly adjust scrap steel ratio, power supply system, pulse jet parameters and dust removal parameters.

[0115] Step Six: Generation of Carbon Pollution Sensitivity Factor Matrix

[0116] For adjustable variables such as scrap steel ratio, power supply curve, oxygen injection rate, natural gas injection rate, dust removal air volume, and green electricity usage period, their marginal impacts on carbon emission intensity, pollutant emission intensity, cost, quality risk, and production cycle time are calculated, forming a sensitivity factor matrix, such as... Figure 5 As shown, the arrows only indicate how the sensitivity factor matrix is ​​expressed.

[0117] Mark_mn=↓, s_mn<-ε_n

[0118] Mark_mn=—,|s_mn|≤ε_n

[0119] Mark_mn=↑, s_mn>ε_n

[0120] For indicators such as CI, PI, Cost, QR, and CT, which are considered "the lower the better," ↓ indicates that increasing the variable is beneficial to lowering the indicator; ↑ indicates that increasing the variable may lead to an increase in the indicator; — indicates that the marginal effect is not significant.

[0121] The final sensitivity factor matrix output by the system may include a numerical matrix, a direction matrix, a confidence matrix, and a constraint feasibility matrix.

[0122] M_i={S_i,Mark_i,Rel_i,Feas_i}

[0123] Where S_i is the numerical sensitivity factor matrix, Mark_i is the direction matrix, Rel_i is the confidence matrix, and Feas_i is the feasibility matrix for whether the variable perturbation satisfies the engineering constraints.

[0124] The actual values ​​are dynamically generated based on on-site data from similar furnace batches using locally weighted regression, finite difference, or digital twin models. This matrix serves as a bridge between evaluation and control, upgrading evaluation methods to evaluation-driven industrial process optimization and control methods.

[0125] 1. Parameterized representation of adjustable variables

[0126] The parameterized definitions of adjustable variables are shown in Table 4 below: Table 4

[0127] The current adjustable control variable vector for each furnace is: U_i=[u_scrap,u_power,u_O2,u_NG,u_dust,u_green]^T; The feature vector of the current furnace operation condition is: Z_i=[steel_grade,scrap_grade,impurity_index,furnace_age,T_tap^target,load,τ,ρ_green]^T; The sensitivity factor is calculated as the rate of change of each response index when a small perturbation is applied to U_i under similar operating conditions Z_i.

[0128] 2. Definition of Sensitivity Factor Matrix

[0129] The sensitivity matrix is ​​used to describe the marginal impact of each adjustable variable on each response indicator. Let the number of adjustable variables be M and the number of response indicators be N, then the sensitivity matrix for the i-th furnace is defined as follows: M_i=[s_mn]_(M×N); s_mn= _n / _m|_(U=U_i,Z=Z_i); In the formula, _m is the normalized representation of the m-th adjustable variable. _n is the normalized representation of the nth response index. Response indices include CI_i, PI_i,k, Cost_i, QR_i, and CT_i.

[0130] _m=(u_m-u_m^0) / Δu_m^eng; _n = y_n / B_n; Where u_m^0 is the actual control variable value of the current furnace, Δu_m^eng is the allowable adjustment step size of the project, and B_n is the benchmark scale of the response index, which can be the benchmark strength of similar furnaces, the enterprise's internal control limit, the emission permit limit, or the cost benchmark.

[0131] When s_mn < 0, it means that increasing this control variable will decrease the corresponding response indicator. For indicators such as carbon emission intensity, pollutant emission intensity, cost, quality risk, and production cycle time, where lower values ​​are better, negative sensitivity factors usually indicate a beneficial effect, while positive sensitivity factors usually indicate a detrimental effect.

[0132] 3. Sensitivity Factor Calculation Method

[0133] (1) Local weighted regression method based on similar furnace batches

[0134] The system retrieves a set of furnaces Ni similar to the current furnace i from a similar furnace benchmark database and performs local weighted regression based on the similarity weight ωij. This method can estimate the marginal impact of adjustable variables on each response index under similar operating conditions.

[0135] y_j,n=β_0,n+Σ_mβ_mn·(u_j,m-u_i,m)+Σ_lγ_ln·(z_j,l-z_i,l)+ε_j,n; _n=(X_i^TW_iX_i+λI)^(-1)X_i^TW_iy_n; W_i=diag(ω_ij),j∈N_i; In the formula, y_j,n represents the nth response index of similar furnace j; u_j,m represents the mth adjustable variable of similar furnace j; z_j,l represents the lth operating condition characteristic of similar furnace j; β_mn is the local marginal influence coefficient of the mth adjustable variable on the nth response index; λ is the regularization coefficient, used to improve stability under small sample or collinear conditions.

[0136] The final sensitivity factor is: s_mn= _mn·Δu_m^eng / B_n.

[0137] When s_mn < 0, it means that increasing this control variable will decrease the corresponding response indicator. For indicators such as carbon emission intensity, pollutant emission intensity, cost, quality risk, and production cycle time, where "lower is better," negative sensitivity factors usually indicate a beneficial effect, while positive sensitivity factors usually indicate a detrimental effect.

[0138] (2) Computational methods based on finite difference or digital twin models

[0139] When an enterprise has an electric furnace energy consumption model, flue gas emission model, dust removal system model, or digital twin simulation model, it can use the finite difference method for engineering disturbances to calculate the sensitivity factor. s_mn^FD=[f_n(U_i+Δu_m^eng·e_m,Z_i)-f_n(U_i-Δu_m^eng·e_m,Z_i)] / (2B_n); In the formula, f_n(·) is the prediction model or simulation model of the nth response index, and e_m is the unit perturbation vector in the direction of the mth control variable. For variables that cannot be perturbed bidirectionally, one-sided finite difference can be used.

[0140] s_mn^FD=[f_n(U_i+Δu_m^eng·e_m,Z_i)-f_n(U_i,Z_i)] / B_n; 4. The perturbation rules for different adjustable variables are shown in Table 5: Table 5

[0141] 5. Reliability and output format of sensitivity factors

[0142] The system further generates a direction matrix, an influence level matrix, and a confidence matrix: Direction_mn = sign(s_mn); Level_mn = |s_mn|; Rel_mn=R_data·R_model·R_feasible; Where R_data represents the reliability of similar furnace data, R_model represents the reliability of model fitting, and R_feasible represents the feasibility of engineering perturbations.

[0143] The final sensitivity factor matrix output by the system includes a numerical matrix, a direction matrix, a confidence matrix, and a constraint feasibility matrix.

[0144] Step 7: Optimization of Multi-Constraint Pollution Reduction and Carbon Degradation Pathways

[0145] Under constraints such as steel grade quality, tapping temperature, scrap steel inventory, equipment capacity, emission limits, production cycle time, electricity price, green electricity availability, and safety boundaries, the system generates at least one pollution reduction and carbon reduction path with carbon emissions per ton of steel, pollutant emissions, cost, quality risks, and cycle time disturbances as optimization objectives.

[0146] 1. Optimize variable definitions

[0147] For the current furnace batch i, the adjustable control variable vector u_i includes at least one of the following: scrap steel ratio, power supply curve parameters, oxygen injection rate, natural gas injection rate, dust removal air volume, green electricity usage period, and priority of technical upgrade measures: u_i=[u_scrap,u_power,u_O2,u_NG,u_dust,u_green,u_tech]; Among them, the scrap steel proportion variable satisfies the proportion constraint: u_scrap=[r_1,…,r_R],Σ_{r=1}^{R}r_r=1,r_r^min≤r_r≤r_r^max; When the proportion of a certain type of scrap steel is increased, the system preferentially adjusts other scrap steel categories according to the proportion conservation rule to ensure that the total scrap steel input ratio remains unchanged.

[0148] r_a'=r_a+Δr,r_b'=r_b-Δr·r_b / (1-r_a),b≠a; The meanings of each variable are shown in Table 6: Table 6

[0149] 2. Multi-objective optimization function

[0150] The system establishes a multi-objective optimization model with carbon emissions per ton of steel, pollutant emissions, cost, quality risk, and production cycle disturbance as optimization objectives. For candidate path u, the objective function vector is defined as: min_uF_i(u)=[CI_i(u),PI_i(u),Cost_i(u),QR_i(u),TD_i(u)]; Where CI_i(u) is the carbon emission intensity per ton of steel, PI_i(u) is the comprehensive pollutant emission intensity, Cost_i(u) is the comprehensive cost per unit of steel, QR_i(u) is the quality risk index, and TD_i(u) is the production cycle disturbance index.

[0151] (1) Carbon emission intensity target per ton of steel

[0152] CI_i(u) = E_i^CO2(u) / M_i(u); E_i^CO2(u)=E_i^elec(u)+E_i^fuel(u)+E_i^material(u)+E_i^electrode(u)+E_i^transport(u)-E_i^recovery(u); Where M_i(u) is the current batch of qualified steel production; E_i^elec, E_i^fuel, E_i^material, E_i^electrode, E_i^transport and E_i^recovery represent the carbon emissions or emission reductions corresponding to electricity, fuel, materials, electrodes, transportation and recycling, respectively.

[0153] Carbon emissions from purchased electricity can be calculated based on time-varying electricity carbon factor and the proportion of green electricity: E_i^elec(u)=Σ_{t∈T_i}P_i(t,u)·Δt·[(1-ρ_i^green(t,u))EF_grid(t)+ρ_i^green(t,u)EF_green(t)]; (2) Pollutant emission intensity targets PI_{i,k}(u)=E_{i,k}^{poll}(u) / M_i(u); E_{i,k}^{poll}(u)=E_{i,k}^{org}(u)+E_{i,k}^{unorg}(u)+E_{i,k}^{transport}(u); Where k represents particulate matter, SO2, NOx, or other pollutant categories. Organized emissions can be calculated from pollutant concentration, flue gas flow rate, and time integral: E_{i,k}^{org}(u)=Σ_{t∈T_i}C_k(t,u)·Q(t,u)·Δt·κ(t) / 10^6; When it is necessary to determine the comprehensive pollutant emission intensity, it can be dimensionless based on emission limits or environmental management weights: PI_i(u)=Σ_{k=1}^{K}θ_k·PI_{i,k}(u) / L_k,Σθ_k=1; (3) Cost target Cost_i(u)=C_i^scrap(u)+C_i^energy(u)+C_i^aux(u)+C_i^env(u)+C_i^quality(u)+C_i^delay(u); Wherein, C_i^scrap is the cost of materials such as scrap steel and alloys, C_i^energy is the cost of energy media such as electricity, natural gas, and oxygen, C_i^aux is the cost of auxiliary materials, C_i^env is the cost of environmental protection and operation, C_i^quality is the cost of rework, downgrading, or concession acceptance due to quality risks, and C_i^delay is the cost of cycle time disturbance and delayed delivery.

[0154] C_i^energy(u)=Σ_tP_i(t,u)·Δt·Price_e(t)+Σ_fAD_{i,f}(u)·Price_f; (4) Quality risk objectives Quality risk targets are used to transform steel composition, target tapping temperature, residual elements, inclusions, and endpoint control deviations into binding or punitive indicators.

[0155] QR_i(u)=w_T·max(0,|T_i^tap(u)-T_i^target|-ε_T) / ε_T+Σ_ew_e·max(0,|A_{i,e}(u)-A_{e}^{target}|-ε_e) / ε_e+Σ_hw_h·max(0,R_{i,h}(u)-R_h^max) / R_h^max; Where T_i^tap is the predicted tapping temperature, A_{i,e} is the predicted value of the e-th compositional element, R_{i,h} is the risk index of the h-th type of residual elements or inclusions, and ε_T and ε_e are the allowable deviations between temperature and composition.

[0156] (5) Production cycle disturbance target

[0157] TD_i(u)=w_τ·|τ_i(u)-τ_i^plan| / τ_i^plan+w_wait·Wait_i(u) / τ_i^plan+w_cast·CastMismatch_i(u)

[0158] Where τ_i(u) is the predicted smelting cycle under the candidate path, τ_i^plan is the planned smelting cycle, Wait_i represents the waiting time caused by the shift of energy consumption period, waiting for green electricity or process connection, and CastMismatch_i represents the penalty item for mismatch with continuous casting times, refining cycle time or order delivery.

[0159] 3. Constraint System

[0160] When generating paths, the system expresses all types of constraints as a unified constraint set: G_i(u)≤0,H_i(u)=0,u∈Ω_i; (1) Constraints on steel grade quality and tapping temperature A_e^min≤A_{i,e}(u)≤A_e^max,e=1,…,E; T_tap^min≤T_i^tap(u)≤T_tap^max; Res_h(u_scrap)≤Res_h^max,h=1,…,H; The above constraints ensure that the candidate paths meet the requirements of steel grade target composition, tapping temperature window, upper limit of residual elements, and quality control.

[0161] (2) Scrap steel inventory and batching constraints

[0162] m_{i,r}(u)=M_i^scrap·r_r,0≤m_{i,r}(u)≤Stock_r; Σ_{r=1}^{R}r_r=1,r_r^min≤r_r≤r_r^max; Σ_rm_{i,r}(u)·Cu_r≤Cu^max,Σ_rm_{i,r}(u)·Sn_r≤Sn^max; Where Stock_r represents the inventory of the r-th type of scrap steel, and Cu_r and Sn_r represent the residual element levels of copper, tin and other elements in this type of scrap steel, respectively.

[0163] (3) Equipment capacity constraints

[0164] 0≤P_i(t,u)≤P_TR^max,|dP_i(t,u) / dt|≤Ramp_P^max; Q_O2^min≤Q_O2(t,u)≤Q_O2^max, Q_NG^min≤Q_NG(t,u)≤Q_NG^max; F_dust^min≤F_dust(t,u)≤F_dust^max,Δp_bag(t,u)≤Δp_bag^max; These constraints are used to ensure that the operating boundaries of equipment such as electric furnace transformer capacity, electrode current, oxygen pipeline capacity, natural gas burner capacity, dust removal fan capacity, and bag filter pressure differential are not exceeded.

[0165] (4) Emission limits and environmental protection facility constraints

[0166] C_k(t,u)≤L_k^conc,E_{i,k}^{poll}(u)≤L_k^mass; F_dust(t,u)≥F_min(stage(t)), η_dust(t,u)≥η_min; Where L_k^conc is the emission concentration limit of the k-th pollutant, L_k^mass is the total control or internal control limit, F_min(stage(t)) is the minimum dust removal air volume required for different operating stages, and η_dust is the dust removal efficiency.

[0167] (5) Production cycle time and process connection constraints

[0168] τ_i^min≤τ_i(u)≤τ_i^max; Start_i^min≤Start_i(u)≤Start_i^max; CastWindow_i^start≤Tap_i(u)≤CastWindow_i^end; The above constraints ensure that the electric furnace, ladle refining, continuous casting and rolling processes remain in a seamless state, avoiding delays in casting, waiting for ladle, or order delays caused by waiting during low-carbon periods or path adjustments.

[0169] (6) Electricity price, green electricity availability and low-carbon time constraints

[0170] 0≤E_green(t,u)≤G_avail(t); E_total(t,u)=E_grid(t,u)+E_green(t,u)+E_self(t,u); t_start(u)∈T_allowed,Δt_lowcarbon∈[Δt_min,Δt_max]; Where G_avail(t) is the amount of green electricity available during time period t, T_allowed is the allowed production period, and Δt_lowcarbon is the time offset to shift to the low-carbon electricity period.

[0171] (7) Safety boundary and interlocking constraints

[0172] I_door(t)=1 P_i(t,u)≤P_safe; T_fg(t,u)≤T_fg^max,CO(t,u)≤CO^safe; Q_NG(t,u)>0 I_ignition(t)=1 and F_dust(t,u)≥F_safe; Safety boundaries are used to prevent dangerous conditions such as high-power power supply with the furnace door open, excessive flue gas temperature, excessive CO, unignited natural gas injection, and insufficient dust removal capacity.

[0173] 4. Optimize the solution and path type

[0174] Since the dimensions of each objective are different, the objective function is first normalized: N_n(u)=F_n(u) / F_{base,n},n∈{CI,PI,Cost,QR,TD}; Where F_{base,n} can be the calculated baseline value of the current furnace, the baseline value of similar furnaces, or the enterprise's internal control baseline. The system can solve for candidate paths using at least one of the following methods: weighted summation, Pareto sorting, multi-objective evolutionary algorithm, sequential quadratic programming, or mixed integer programming.

[0175] The system can solve for candidate paths using at least one of the following methods: weighted summation, Pareto sorting, multi-objective evolutionary algorithm, sequential quadratic programming, or mixed integer programming. For the weighted summation method: min_uJ_i(u)=Σ_nλ_n·N_n(u)+μ·Penalty_i(u); Penalty_i(u)=Σ_amax(0,G_{i,a}(u))²+Σ_b|H_{i,b}(u)|; Where λ_n is the weight of each optimization objective, Penalty_i(u) is the constraint violation penalty term, and μ is the penalty coefficient.

[0176] For multi-scheme scenarios, generate a set of Pareto non-dominated paths: Pareto_i = {u | There does not exist v ∈ Ω_i such that F_n(v) ≤ F_n(u) holds for all n and at least one objective is strictly better}; 5. How the sensitivity factor matrix is ​​used in step seven: Step six outputs a sensitivity factor matrix S=[s_mn], which describes the marginal impact of the m-th adjustable variable on the n-th target indicator. Step seven, before solving, prioritizes screening effective variables based on the sensitivity factor matrix and generates search directions.

[0177] s_{mn}= N_n(u) / u_m≈[N_n(u+Δu_m)-N_n(u)] / u_m; Where u_m is the dimensionless form of the m-th variable. For variables such as scrap steel ratio, power supply curve, injection volume, dust removal air volume, and green electricity period, the system uses a perturbation method that conforms to process constraints to calculate the marginal impact.

[0178] Direction_{mn}=sign(s_{mn}),Level_{mn}=|s_{mn}|; When a variable has a favorable influence on the target indicator and a high level of influence, that variable is given priority in the candidate path set. The S800 can use linearization approximation to quickly generate initial candidate paths: N_n(u+Δu)≈N_n(u)+Σ_ms_{mn}· u_m; min_{Δu}Σ_nλ_n[N_n(u)+Σ_ms_{mn}· u_m]+μ·Penalty_i(u+Δu); The linearization results described above are only used for preliminary screening of candidate paths; the final recommended path needs to be re-substituted into the carbon emission intensity and pollutant emission intensity calculation model from step four, and verified in conjunction with a digital twin or prediction model.

[0179] 6. Output Forms of Pollution Reduction and Carbon Degradation Pathways

[0180] The system outputs a pollution reduction and carbon reduction path, which can be a single path or multiple candidate paths with different preferences. The path results should include at least the adjustment amount of control variables, the execution period, the constraint satisfaction status, the expected pollution reduction and carbon reduction effect, and risk warnings.

[0181] Path_i^q={Δr_r,ΔP(t),ΔQ_O2(t),ΔQ_NG(t),ΔF_dust(t),Δt_lowcarbon,z_tech,ΔF_i^q,Risk_i^q}; Where q represents the q-th candidate path, and ΔF_i^q represents the change in the target metric of this path relative to the current baseline: ΔF_i^q=[CI_i(u^q)-CI_i(u^0),PI_i(u^q)-PI_i(u^0),Cost_i(u^q)-Cost_i(u^0),QR_i(u^q)-QR_i(u^0),TD_i(u^q)-TD_i(u^0)]; To illustrate the contribution of each variable to the path effect, the system can calculate the variable-target contribution: Contribution_{mn}^q=|s_{mn}· _m^q| / Σ_m|s_{mn}· _m^q| There is no path that is strictly better for all conditions and at least one objective. The system can generate candidate paths with different preferences, as shown in Table 7 below: Table 7

[0182] 7. Path optimization execution process

[0183] Obtain the production, energy, environmental protection, material, quality and transportation data for the current furnace batch, and call step four to generate the current baseline index F_i(u0).

[0184] Step six is ​​called to generate the sensitivity factor matrix S=[s_mn], which identifies adjustable variables that are highly sensitive to carbon emissions, pollutants, costs, quality, and cycle time.

[0185] A constraint set is constructed based on steel grade quality, tapping temperature, inventory, equipment capacity, emission limits, production cycle time, electricity price, available green electricity, and safety boundaries.

[0186] Candidate variables are generated based on the sensitivity factor matrix to adjust the direction, and candidate paths are generated within the feasible region.

[0187] Perform multi-objective optimization on the candidate paths to obtain the Pareto candidate path set or the weighted optimal path.

[0188] Substitute the candidate paths back into the calculation model of step four to verify the carbon emission intensity and pollutant emission intensity per ton of steel, and check all constraints.

[0189] Output at least one executable path to the production scheduling system, energy management system, or environmental control system, and write back the field feedback data after execution to update the models in steps four and six.

[0190] Step 8: Closed-loop control feedback output

[0191] The optimized path outputs to MES, EMS, DCS, or environmental control systems. Output content includes scrap steel ratio recommendations, power supply curve adjustments, oxygen / natural gas injection parameters, dust collector fan frequency, green energy consumption periods, energy storage charging and discharging strategies, and technological upgrade priorities. The closed-loop process is as follows: Figure 6 As shown.

[0192] Overall mathematical model of closed-loop control output

[0193] For the current furnace i, the system defines the executable path output from step seven as a control variable vector: u_i=[u_scrap,i,P_i^set(t),q_O2,i^set(t),q_NG,i^set(t),Q_dust,i^set(t),z_green,i(t),t_start,i,a_tech,i]; Wherein, u_scrap,i represents the scrap steel ratio vector; P_i^set(t) represents the electric arc furnace power supply setting curve; q_O2,i^set(t) represents the oxygen injection flow rate setting curve; q_NG,i^set(t) represents the natural gas injection or burner flow rate setting curve; Q_dust,i^set(t) represents the dust removal air volume setting curve; z_green,i(t) represents the green electricity usage ratio or green electricity allocation identifier; t_start,i represents the planned start-up time of the furnace; and a_tech,i represents the technical improvement measure or control strategy number.

[0194] The system forms the target prediction function based on the S500 computation layer and the S700 sensitivity factor matrix layer: _i(u_i)=[ I_i(u_i),PÎ_i(u_i), ost_i(u_i), isk_i(u_i), elay_i(u_i)]^T in, I_i represents the predicted carbon emission intensity per ton of steel, and PÎ_i represents the predicted comprehensive pollutant emission intensity. ost_i represents the predicted production cost. isk_i represents the predicted quality risk. elay_i represents the predicted beat perturbation.

[0195] To unify different objective units, a normalized objective function is constructed: J_1(u_i)= I_i(u_i) / CI_base,i; J_2(u_i)=Σ_kθ_k·PÎ_i,k(u_i) / PI_base,i,k; J_3(u_i)= ost_i(u_i) / Cost_base,i; J_4(u_i)= isk_i(u_i) / Risk_base,i; J_5(u_i)= elay_i(u_i) / τ_base,i; Given the constraints Ω_i, the system generates either a Pareto path set or a comprehensive recommended path. U_i^Pareto={u_i∈Ω_i|There does not exist u_i′∈Ω_i such that J_n(u_i′)≤J_n(u_i) and at least one of them is strictly less than}; u_i^ =argmin_{u_i∈Ω_i}Σ_nη_nJ_n(u_i)+λ_s·V_safety(u_i)+λ_c·V_control(u_i); Where η_n is the target weight, V_safety represents the safety boundary violation penalty term, V_control represents the control smoothness penalty term, and λ_s and λ_c are the penalty weights.

[0196] 1. Subsystem output format

[0197] (1) Production execution instructions output to MES

[0198] The MES receives information such as furnace cycle plan, scrap steel batching plan, target tapping temperature, and production cycle time. Instructions include: Cmd_MES,i={heat_id_i,grade_i,t_start,i^ ,u_scrap,i^ ,T_tap,i^ ,τ_plan,i^ ,route_i,batch_id_i}; Scrap steel proportioning instructions are converted into various types of scrap steel feeding quality: M_scrap,i,r^ =r_i,r^ ·M_scrap,i^total,r=1,2,…,R; Σ_rr_i,r^ =1,0≤M_scrap,i,r^ ≤Inv_r; Before executing the MES, the steel quality and production cycle constraints are checked. If they are not met, the system is marked as requiring manual confirmation.

[0199] c_l^min(grade_i)≤ _l(u_i^ )≤c_l^max(grade_i),l∈E_grade; T_tap^min≤T_tap,i^ ≤T_tap^max; τ_plan,i^ ≤τ_max,i,|t_start,i^ -t_start,i^plan|≤Δt_allow; If the above constraints are not met, the MES side will not automatically issue the execution, but will mark the path as requiring manual confirmation or return to step seven for re-optimization.

[0200] (2) Energy dispatch instructions output to EMS

[0201] The EMS receives information such as furnace power consumption plan, time-of-use power demand, green electricity usage ratio, and energy consumption period. Instructions include: Cmd_EMS,i={E_i^ (t),P_i^set(t),z_green,i^ (t), t_start, i^ ,t_end,i^ Cost_energy, i^ }; The time-of-use electricity consumption is calculated according to the power transmission curve: E_i^ (t)=P_i^set(t)·Δt; The allocation of green electricity meets the constraints on the availability of green electricity: E_green,i^ (t)=z_green,i^ (t)·E_i^ (t) Σ_iE_green,i^ (t)≤G_avail(t),0≤z_green,i^ (t)≤1; Electricity purchase cost and indirect carbon emissions from electricity can be expressed as follows: Cost_power,i=Σ_tE_i^ (t)·[z_green,i^ (t)·Price_green(t)+(1-z_green,i^ [(t))·Price_grid(t)]; E_CO2,power,i=Σ_tE_i^ (t)·[z_green,i^ (t)·EF_green(t)+(1-z_green,i^ [(t))·EF_grid(t)]; EMS uses the above formula to determine whether the current furnace batch meets the requirements for electricity price, green electricity availability, transformer capacity, maximum demand, and electricity safety boundary.

[0202] (3) Process control setpoints output to DCS

[0203] The DCS side receives parameters such as the electric furnace power supply curve, oxygen injection curve, and natural gas injection curve. To avoid process instability caused by sudden changes in setpoints, amplitude limiting and slope constraints are used to generate actual issued commands.

[0204] Power transmission command issued: P_cmd(t)=sat(P_act(t)+κ_P[P_i^set (t)-P_act(t)],P_min,P_max,R_P); Oxygen injection flow rate command issued: q_O2,cmd(t)=sat(q_O2,act(t)+κ_O[q_O2,i^set (t)-q_O2,act(t)],q_O2^min,q_O2^max,R_O); The natural gas injection or burner flow rate instruction is as follows: q_NG,cmd(t)=sat(q_NG,act(t)+κ_G[q_NG,i^set (t)-q_NG,act(t)],q_NG^min,q_NG^max,R_G); Where sat(·) represents the amplitude limiting function that simultaneously satisfies the upper and lower limit constraints and the unit time rate of change constraint; κ_P, κ_O, and κ_G represent the tracking coefficients for power transmission, oxygen injection, and natural gas injection, respectively; and R_P, R_O, and R_G represent the maximum climbing rate or the maximum rate of change, respectively.

[0205] The limiting function can be expressed as: sat(x,x_min,x_max,R)=min(max(x,x_min),x_max), and |x(t)-x(t-Δt)| / Δt≤R; DCS execution must also meet equipment capacity and safety interlock boundaries: P_min≤P_cmd(t)≤P_max,|dP_cmd(t) / dt|≤R_P; q_O2^min≤q_O2,cmd(t)≤q_O2^max, q_NG^min≤q_NG, cmd(t)≤q_NG^max; Interlock_safety(t) = 1 Cmd_DCS(t) = Hold or SafeMode; (4) Control commands for treatment facilities output to the environmental protection control system The environmental control system receives commands including dust removal air volume, fan frequency, valve opening degree, and emission constraint targets. Cmd_ENV,i={Q_dust,i^set (t),f_fan,i^ (t),α_valve,i^ (t),C_k^target,L_k^permit}; Dust collector airflow closed-loop control command combined with real-time concentration feedback correction: Q_dust,cmd(t)=sat(Q_dust,i^set (t)+K_PM·[C_PM(t)-C_PM^target]_++K_door·I_door(t)+K_flow·ΔQ_fg(t),Q_min,Q_max,R_Q); Where, [x]_+=max(x,0); C_PM(t) represents the real-time concentration of particulate matter; C_PM^target represents the internal control target concentration of the enterprise; I_door(t) represents the furnace door opening status; ΔQ_fg(t) represents the flue gas volume deviation; K_PM, K_door, and K_flow are feedback correction coefficients.

[0206] If pollutants such as NOx, SO2, and CO are considered simultaneously, it can be expanded to: Q_dust,cmd(t)=sat(Q_dust,i^set (t)+Σ_kK_k·[C_k(t)-C_k^target]_++K_door·I_door(t),Q_min,Q_max,R_Q); The environmental control system should meet emission limits and fan capacity boundaries during operation. _k(t,u_i^ )≤L_k^permit·μ_k,0<μ_k≤1; Q_min≤Q_dust,cmd(t)≤Q_max,f_fan^min≤f_fan,cmd(t)≤f_fan^max; Where μ_k is the enterprise's internal control safety margin coefficient, preferably less than or equal to 1.

[0207] 2. Execution Feedback and Dynamic Model Correction

[0208] After each system is executed, it will transmit back the actual control parameters, actual emissions data, actual energy consumption data, quality results, and production cycle time. The system calculates control deviation and performance deviation. e_u,m=(u_m^exec-u_m^ ) / Range(u_m); e_y,n=(y_n^actual- _n(u_i^ )) / B_n; Where e_u,m represents the execution deviation of the m-th control variable, e_y,n represents the performance deviation of the n-th target indicator, Range(u_m) represents the adjustable range of the control variable, and B_n represents the normalized benchmark of the n-th target indicator.

[0209] If the performance deviation exceeds the preset threshold, the system will recursively correct the sensitivity factor matrix or prediction model. s_mn^new=(1-α_s)·s_mn^old+α_s·[(y_n^actual-y_n^base) / B_n] / [(u_m^exec-u_m^base) / Range(u_m)]; θ_model^new=θ_model^old+K·[y^actual-F(x_i,u_i^exec;θ_model^old)]; Where α_s is the sensitivity factor update coefficient, θ_model is the prediction model parameter, and K is the recursive correction gain.

[0210] For the next control cycle, the adjusted control variables are generated based on the deviation correction results: u_i^adj=Proj_Ω[u_i^ -K_u·e_y]; 3. Automatic distribution and manual confirmation / judgment To ensure safety at industrial sites, the output method is determined based on path risk, control scope, and data reliability: Mode_i=Auto, if Conf_i≥Conf_min and Risk_i≤Risk_max and max_m|Δu_m|≤Δu_m^safe; If Mode_i=Review, and none of the above conditions are met; Among them, Auto means that it can be automatically sent to MES, EMS, DCS or environmental control system; Review means that it only outputs operation suggestions or scheduling suggestions, which are then confirmed and executed by process personnel.

[0211] The system output mapping table is shown in Table 8.

[0212] Table 8

[0213] 4. Output path

[0214] For the current furnace cycle i, let the control variable before optimization be u_i^0, and the control variable after optimization be u_i^ The optimized path output module encapsulates the difference between the two, the execution period, and the target improvement amount into a path package Path_i, and distributes it to the production, energy, environmental protection, and equipment management systems according to different execution objects.

[0215] u_i^ ={u_scrap^ ,P_arc^ (t),Q_O2^ (t),Q_NG^ (t),f_dust^ (t),z_green^ (t),P_ch^ (t),P_dis^ (t),a_tech^ }(1)

[0216] Path_i={Δu_scrap,ΔP_arc(t),ΔQ_O2(t),ΔQ_NG(t),Δf_dust(t),T_green^ ,S_bat^ ,PRI_tech,ΔY_i}(2)

[0217] Among them, u_scrap^ This indicates the optimized scrap steel ratio; P_arc^ (t) represents the optimized electric furnace power delivery curve; Q_O2^ (t) and Q_NG^ (t) represent the oxygen injection rate and the natural gas injection rate, respectively; f_dust^ (t) represents the frequency of the dust collector fan; z_green^ (t) represents the variable for selecting green electricity consumption periods; P_ch^ (t), P_dis^ (t) represent the energy storage charging power and discharging power, respectively; a_tech^ This indicates the result of the technological upgrading measures selection; ΔY_i represents the improvement in carbon emissions, pollutant emissions, cost, quality risk, and cycle time disturbance corresponding to the optimized path. The output is shown in Table 9. Table 9

[0218] (1) Suggested output formula for scrap steel ratio

[0219] The system represents scrap steel structures as a proportion vector. The sum of the proportions of various types of scrap steel is 1, and is subject to constraints such as inventory, residual element control of steel grades, basket loading capacity, and quality risk boundary.

[0220] u_scrap^ =[r_1^ ,r_2^ ,...,r_R^ ],Σ_{r=1}^{R}r_r^ =1(3)

[0221] M_{scrap,r}^ =r_r^ ·M_scrap^ (4)

[0222] Δr_r=r_r^ -r_r^0,ΔM_{scrap,r}=M_{scrap,r}^ -M_{scrap,r}^0(5)

[0223] Among them, r_r^ M_scrap^ represents the optimal proportion of type r scrap steel. M_{scrap,r}^ represents the total amount of scrap steel fed into the current heat after optimization. This indicates the recommended feed amount for type r scrap. When output to MES or WMS, it can generate a batching recommendation of "increase / decrease of type r scrap by ΔM_{scrap,r} tons".

[0224] 0≤M_{scrap,r}^ ≤I_r(6)

[0225] r_r^{min}≤r_r^ ≤r_r^{max}(7)

[0226] C_e^{min}≤Σ_{r=1}^{R}r_r^ ·c_{r,e}+c_{add,e}≤C_e^{max}(8)

[0227] Q_slag^ =F_slag(u_scrap^ Flux^ )≤Q_slag^{max}(9)

[0228] Where I_r represents the inventory of scrap steel of type r; c_{r,e} represents the content of element e in scrap steel of type r; C_e^{min} and C_e^{max} represent the control range of element e for the target steel grade; Q_slag^ This indicates the predicted slag quantity or inclusion load. Equation (8) is used to ensure steel grade quality constraints, and Equation (9) is used to limit the increase in slag quantity and pollutant load caused by high inclusion scrap steel.

[0229] (2) Formula for adjusting the power supply curve

[0230] The system parameterizes the electric furnace power supply curve according to the smelting stage or time segment, and preferably divides it into the arc-starting segment after charging, the early melting stage, the middle melting stage, the adjustment period after melting and clearing, and the heat preservation segment before tapping.

[0231] P_arc^ (t)=P_arc^0(t)+ΔP_s,t∈τ_s (10)

[0232] ΔP_s=P_s^ -P_s^0(11)

[0233] 0≤P_arc^ (t)≤P_tr^{max}(12)

[0234] |P_arc^ (t+Δt)-P_arc^ (t)| / Δt≤R_P^{max}(13)

[0235] E_arc^ = _{T_i}P_arc^ (t)dt(14)

[0236] Where τ_s represents the time window of the s-th smelting stage; P_tr^{max} represents the maximum active power allowed by the transformer or power supply system; and R_P^{max} represents the upper limit of the power ramp-up rate. When output to the DCS, it can generate a power supply curve adjustment instruction or suggestion of "increasing / decreasing the power plateau ΔP_s during the middle of melting for a duration of τ_s".

[0237] (3) Oxygen / natural gas injection parameter output formula

[0238] Oxygen and natural gas injection parameters can be output in stages, including injection flow rate, injection duration, injection intensity, and stage cumulative amount.

[0239] Q_O2^ (t)=Q_O2^0(t)+ΔQ_{O2,s},t∈τ_s (15)

[0240] Q_NG^ (t)=Q_NG^0(t)+ΔQ_{NG,s},t∈τ_s(16)

[0241] V_{O2,s}^ = _{τ_s}Q_O2^ (t)dt,V_{NG,s}^ = _{τ_s}Q_NG^ (t)dt(17)

[0242] Q_{O2}^{min}≤Q_O2^ (t)≤Q_{O2}^{max}(18)

[0243] Q_{NG}^{min}≤Q_NG^ (t)≤Q_{NG}^{max}(19)

[0244] CO^{min}≤CO^ (t)≤CO^{safe}(20)

[0245] NOx^ (t)≤NOx^{limit}(21)

[0246] Wherein, V_{O2,s}^ and V_{NG,s}^ These represent the cumulative injection volumes of oxygen and natural gas in stage s, respectively; CO^{safe} represents the CO safety boundary; NOx^{limit} represents the nitrogen oxide emission limit or the enterprise's internal control limit. This output can be used for DCS injection control and can be linked with the environmental control system to verify NOx risk.

[0247] (4) Formula for the frequency output of dust collector fan

[0248] The frequency output of the dust collector fan aims to meet particulate matter control and reduce fan power consumption. The system determines the staged frequency setpoint based on furnace door status, flue gas temperature, particulate matter concentration, flue gas flow rate, and bag filter pressure difference.

[0249] f_dust^ (t)=f_dust^0(t)+Δf_s,t∈τ_s (22)

[0250] Q_dust^ (t)=a_f·f_dust^ (t)+b_f(23)

[0251] E_fan^ = _{T_i}k_f·[f_dust^ (t)]^3dt(24)

[0252] C_PM^ (t)≤L_PM,C_SO2^ (t)≤L_SO2,C_NOx^ (t)≤L_NOx(25)

[0253] DP_bag^{min}≤DP_bag^ (t)≤DP_bag^{max}(26)

[0254] Among them, Q_dust^ (t) represents the dust removal air volume; E_fan^ This represents the power consumption of the fan; C_PM^ (t), C_SO2^ (t), C_NOx^ (t) represents the predicted emission concentrations of particulate matter, SO2, and NOx, respectively; DP_bag^ (t) represents the pressure difference of the filter bag. Equation (24) reflects the nonlinear relationship between the fan frequency and power consumption, and can be used to identify the risk of "pollution reduction but carbon increase".

[0255] (5) Green electricity consumption period output formula

[0256] The system outputs the low-carbon energy consumption periods for adjustable furnace cycles based on the electric furnace production plan, time-of-use electricity price, time-varying carbon factor, available green electricity, and continuous casting matching constraints.

[0257] z_{i,t}∈{0,1},Σ_{t∈T}z_{i,t}=1 (27)

[0258] T_green^ =argmin_{t∈T}{E_i·EF_grid(t)+Cost_i(t)+Penalty_i(t)}(28)

[0259] E_i·z_{i,t}≤G_t^{avail}(29)

[0260] t_i^{start,min}≤Σ_{t∈T}t·z_{i,t}≤t_i^{start,max}(30)

[0261] Match_CC(i,t)=1(31)

[0262] Where z_{i,t} indicates whether furnace i is scheduled for time period t; EF_grid(t) represents the time-varying carbon factor; G_t^{avail} represents the available green electricity for time period t; Penalty_i(t) represents the production scheduling, waiting, or continuous casting matching penalty caused by adjusting the time period; Match_CC(i,t) indicates whether the furnace meets the continuous casting connection requirements in time period t. When output to EMS or MES, a suggestion to "schedule furnace i to use energy in time period t" can be generated.

[0263] (6) Output formula of energy storage charging and discharging strategy

[0264] When enterprises configure energy storage systems, the system can output energy storage charging and discharging strategies based on electric furnace load, green electricity fluctuations, electricity prices, and carbon factors.

[0265] SOC(t+1)=SOC(t)+η_ch·P_ch(t)Δt-P_dis(t)Δt / η_dis(32)

[0266] SOC^{min}≤SOC(t)≤SOC^{max}(33)

[0267] 0≤P_ch(t)≤P_ch^{max},0≤P_dis(t)≤P_dis^{max}(34)

[0268] P_grid(t)=P_load(t)+P_ch(t)-P_dis(t)-P_pv(t)(35)

[0269] S_bat^ ={P_ch^ (t),P_dis^ (t),SOC^ (t)}(36)

[0270] P_ch(t)·P_dis(t)=0(37)

[0271] Where SOC represents the state of charge of the energy storage; P_ch(t) and P_dis(t) represent the charging power and discharging power, respectively; η_ch and η_dis represent the charging efficiency and discharging efficiency, respectively; and P_grid(t) represents the actual purchased power. Equation (37) is used to avoid simultaneous charging and discharging during the same period. When output to the EMS, a strategy of "charging during low-carbon and low-price periods and discharging during high-carbon and high-price periods or high-load periods of electric furnaces" can be formed.

[0272] (7) Technological Upgrade Priority Output Formula

[0273] For technological upgrades such as scrap steel preheating, waste heat recovery, dust removal system renovation, green electricity expansion, energy storage configuration, and intelligent power transmission control, the system generates priorities based on marginal emission reduction benefits, pollutant reduction benefits, costs, implementation cycle, and feasibility.

[0274] a_l∈{0,1},l=1,2,...,L (38)

[0275] Benefit_l=w_C·ΔCI_l^- / CI_base+w_P·ΔPI_l^- / PI_base+w_Q·ΔRisk_l^-+w_T·ΔTime_l^-(39)

[0276] Cost_l=CAPEX_l+Σ_{y=1}^{Y}OPEX_{l,y} / (1+d)^y(40)

[0277] PRI_l=Benefit_l·Fea_l / Cost_l(41)

[0278] Rank_tech=sort_desc({PRI_l|l=1,...,L})(42)

[0279] Where, a_l indicates whether the l-th technical improvement measure is selected; ΔCI_l^- represents the reduction in carbon emissions per ton of steel; ΔPI_l^- represents the reduction in pollutant emission intensity; ΔRisk_l^- represents the reduction in quality risk; ΔTime_l^- represents the reduction in cycle time disturbance; CAPEX_l and OPEX_l represent investment cost and operating cost, respectively; d represents the discount rate; and Fea_l represents the feasibility coefficient. The system outputs the technical improvement priority from high to low according to PRI_l.

[0280] (8) Path effect quantification and output verification

[0281] The system calculates the path effect vector for each candidate optimization path p and outputs only the paths that satisfy the constraints and have a positive overall benefit.

[0282] ΔY_p=[ΔCI_p,ΔPI_p,ΔCost_p,ΔRisk_p,ΔTime_p] (43)

[0283] ΔCI_p=CI(u_p)-CI(u^0)(44)

[0284] ΔPI_p=PI(u_p)-PI(u^0)(45)

[0285] ΔCost_p=Cost(u_p)-Cost(u^0)(46)

[0286] Score_p=-w_C·ΔCI_p / CI_base-w_P·ΔPI_p / PI_base-w_K·ΔCost_p / Cost_base-w_R·ΔRisk_p-w_T·ΔTime_p (47)

[0287] Output_p=1,ifScore_p≥Score_minandViolation_p=0 (48)

[0288] Here, Violation_p indicates whether the candidate path violates constraints such as steel grade quality, tapping temperature, equipment capacity, emission limits, production cycle time, electricity price, green electricity availability, or safety boundaries. If Violation_p=1, the candidate path will not enter the closed-loop control output.

[0289] In a specific implementation, path optimization can be expressed as finding a combination U of control variables within the constraint set Ω that optimizes the objective function F: minF(U)={C_total(U),P_total(U),Cost(U),QualityRisk(U),TimeDisturbance(U)}; The constraint set Ω includes steel grade composition, tapping temperature, scrap inventory, metal recovery rate, transformer capacity, dust removal fan capacity, emission concentration limits, continuous casting matching, order delivery, electricity price window, green electricity availability, and equipment safety boundaries.

[0290] The path optimization module searches for a combination of control variables U within the constraint set Ω to optimize the objective function F(U). The objective function F(U) includes the carbon emission per ton of steel objective function C_total(U), the pollutant emission objective function P_total(U), the comprehensive cost objective function Cost(U), ​​the quality risk objective function QualityRisk(U), and the cycle disturbance objective function TimeDisturbance(U). Specifically, C_total(U) is calculated based on carbon emissions from electricity, fuel, material processes, electrodes, transportation, and waste heat recovery; P_total(U) is calculated based on particulate matter, SO2, NOx, and fugitive emission intensity; Cost(U) is calculated based on material costs, energy costs, environmental costs, quality costs, delay costs, and technological upgrade costs; QualityRisk(U) is calculated based on steel composition deviation, tapping temperature deviation, inclusion risk, and process quality risk; and TimeDisturbance(U) is calculated based on electric arc furnace smelting cycle deviation, continuous casting waiting time, order delays, and scheduling switching penalties.

[0291] (1) C_total(U): Objective function for carbon emissions per ton of steel

[0292] C_total(U) represents the carbon emission intensity per ton of steel under the control variable combination U, preferably expressed as tCO2 / t steel or kgCO2 / t steel. This objective function is related to factors such as the green electricity usage period, power transmission curve, fuel injection, scrap steel ratio, metal recovery rate, and waste heat recovery.

[0293] C_total(U) = E_CO2(U) / M_q(U); E_CO2(U)=E_elec(U)+E_fuel(U)+E_mat(U)+E_electrode(U)+E_transport(U)-E_recovery(U); E_elec(U)=Σ_tP_grid(t;U)·Δt·[(1-ρ_green(t;U))EF_grid(t)+ρ_green(t;U)EF_green(t)]; E_fuel(U)=Σ_fAD_f(U)·NCV_f·CC_f·OF_f·44 / 12; E_mat(U) = Σ_mAD_m(U)·EF_m; M_q(U)=M_tap(U)·η_yield(U)·η_quality(U); Wherein, E_CO2(U) represents the total carbon emissions per furnace under the control variable combination U; M_q(U) represents the qualified steel output; E_elec(U) represents the indirect carbon emissions corresponding to purchased electricity or in-plant electricity consumption; E_fuel(U) represents the carbon emissions from the combustion of fuels such as natural gas, coal gas, and diesel; E_mat(U) represents the carbon emissions from the material processing of lime, dolomite, carbon powder, alloys, etc.; E_electrode(U) represents the carbon emissions corresponding to electrode consumption; E_transport(U) represents the carbon emissions from the transportation process; and E_recovery(U) represents the emission reductions calculated from waste heat recovery, green electricity substitution, or energy recovery.

[0294] In the formula, P_grid(t;U) represents the purchased power in time period t, Δt is the sampling time interval, ρ_green(t;U) is the green electricity ratio, EF_grid(t) is the time-varying grid emission factor, and EF_green(t) is the green electricity emission factor or certified substitution factor. M_tap(U) is the steel output, η_yield(U) is the metal yield, and η_quality(U) is the quality pass rate.

[0295] C_total(U) mainly corresponds to constraints in the constraint set Ω, such as scrap inventory, metal recovery rate, transformer capacity, electricity price window, green electricity availability, equipment safety boundary, and order delivery. If U violates the above constraints, then even if C_total(U) is low, it cannot be used as an executable optimization path output.

[0296] (2) P_total(U): Objective function for pollutant emissions

[0297] P_total(U) represents the overall pollutant emission intensity or pollutant emission risk target under the combination of control variables U. This objective function can be calculated for at least one of particulate matter, SO2, NOx, CO, and fugitive emissions.

[0298] P_total(U)=Σ_kθ_k·PI_k(U) / L_k; PI_k(U) = E_poll,k(U) / M_q(U); E_poll,k(U)=E_org,k(U)+E_unorg,k(U)+E_trans,k(U); E_org,k(U)=Σ_pΣ_tC_k,p(t;U)·Q_p(t;U)·κ_p(t)·Δt / 10^6; E_unorg,k(U)=F_unorg,k(u_scrap,f_dust(t),Door(t),Yard(t),Transport(t)); Wherein, θ_k represents the weight of pollutant category k; PI_k(U) represents the emission intensity of pollutant category k per ton of steel; L_k represents the emission limit, ultra-low emission limit, or internal control limit of pollutant category k; E_poll,k(U) represents the total emission of pollutant category k; and E_org,k(U), E_unorg,k(U), and E_trans,k(U) represent organized emissions, unorganized emissions, and transport emissions, respectively.

[0299] In the formula, C_k,p(t;U) represents the concentration of pollutant of type k at outlet p in time period t; Q_p(t;U) represents the standard dry flow rate of flue gas; κ_p(t) represents the temperature, pressure, humidity, oxygen content, and standard condition conversion correction coefficient. P_total(U) is related to control variables such as dust removal fan frequency, dust removal air volume, scrap steel inclusions, furnace door opening and closing, oxygen / natural gas injection, and transportation mode.

[0300] P_total(U) mainly corresponds to the constraints in the constraint set Ω, including the dust collector fan capacity, emission concentration limits, equipment safety boundaries, scrap steel inventory, and scrap steel inclusion control constraints. The emission concentration limits can be expressed as C_k,p(t;U)≤L_k,p, and the dust collector fan capacity can be expressed as f_dust(t)≤f_dust^max or Q_dust(t)≤Q_dust^max.

[0301] (3) Cost(U): Overall cost objective function

[0302] Cost(U) represents the comprehensive cost per heat or the comprehensive cost per unit of qualified steel under the combination of control variables U, preferably in yuan / t of steel. This objective function is used to prevent an abnormal increase in production costs caused by simply reducing carbon emissions or pollutant emissions.

[0303] Cost(U)=[Cost_mat(U)+Cost_energy(U)+Cost_env(U)+Cost_quality(U)+Cost_delay(U)+Cost_tech(U)] / M_q(U); Cost_mat(U)=Σ_rM_scrap,r(U)·price_scrap,r+Σ_aM_add,a(U)·price_add,a; Cost_energy(U)=Σ_tP_grid(t;U)·price_elec(t)·Δt+Σ_fAD_f(U)·price_f+Cost_gas(U); Cost_env(U)=Cost_fan(U)+Cost_reagent(U)+Cost_waste(U)+Cost_emission(U); Cost_delay(U)=c_time·max(0,τ_total(U)-τ_plan); Cost_tech(U)=Σ_la_l·[CAPEX_l / Y_l+OPEX_l]; Among them, Cost_mat(U) represents the cost of scrap steel, auxiliary materials and alloys; Cost_energy(U) represents the cost of energy media such as electricity, gas, oxygen, natural gas, etc.; Cost_env(U) represents the cost of dust removal fan power consumption, environmental protection agents, solid waste disposal and emission management; Cost_quality(U) represents the quality cost caused by quality deviation, rework, downgrading or scrapping; Cost_delay(U) represents the cost of production cycle delay; and Cost_tech(U) represents the converted cost of technical improvement measures.

[0304] Cost(U) mainly corresponds to the following in the constraint set Ω: scrap inventory, electricity price window, green electricity availability, order delivery, continuous casting matching, production cycle time, and equipment capacity. For example, the electricity price window affects price_elec(t), green electricity availability affects ρ_green(t;U) and the electricity purchase structure, and order delivery and continuous casting matching affect Cost_delay(U).

[0305] (4) QualityRisk(U): Quality risk objective function

[0306] QualityRisk(U) represents the risk that the combination of control variables U will cause the current heat to deviate from the target steel quality requirements, preferably a dimensionless risk value. This objective function is used to ensure that the pollution reduction and carbon reduction path does not sacrifice the stability of steel composition, tapping temperature, and product quality.

[0307] QualityRisk(U)=R_comp(U)+R_temp(U)+R_incl(U)+R_process(U); R_comp(U)=Σ_eα_e·[max(0,C_e(U)-C_e^max) / Tol_e]^2+Σ_eα_e·[max(0,C_e^min-C_e(U)) / Tol_e]^2; C_e(U)=[Σ_rM_scrap,r(U)c_r,e+Σ_aM_add,a(U)c_a,e-Loss_e(U)] / M_liquid(U); R_temp(U)=α_T·[max(0,|T_tap(U)-T_target|-Tol_T) / Tol_T]^2; R_incl(U)=α_I·F_I(I_scrap(U),Q_slag(U),O2_total(U),Stir(U)); Wherein, R_comp(U) represents the risk of deviation in the chemical composition of the steel grade; R_temp(U) represents the risk of deviation in the tapping temperature; R_incl(U) represents the risk of inclusions, slag quantity, or cleanliness; R_process(U) represents the process quality risk caused by abnormalities in the smelting process. C_e(U) represents the predicted content of element e in the molten steel; C_e^min and C_e^max represent the control range of element e for the target steel grade; Tol_e represents the element control tolerance; T_tap(U) represents the predicted tapping temperature; T_target represents the target tapping temperature; and Tol_T represents the allowable deviation in tapping temperature.

[0308] QualityRisk(U) mainly corresponds to constraints in the constraint set Ω, such as steel composition, tapping temperature, scrap inventory, metal yield, equipment safety boundary, and continuous casting matching. In particular, scrap ratio, oxygen injection rate, power supply curve, and refining time will jointly affect Ce(U), Ttap(U), and inclusion risk.

[0309] (5) TimeDisturbance(U): Objective function of clock disturbance

[0310] TimeDisturbance(U) represents the disturbance caused by the combination of control variables U to the smelting cycle, continuous casting matching, order delivery, and production rhythm, preferably a dimensionless index or minutes. This objective function is used to avoid production organization becoming unenforceable due to pollution reduction and carbon reduction pathways.

[0311] TimeDisturbance(U)=β_1·|τ_EAF(U)-τ_EAF^plan| / τ_EAF^plan+β_2·Wait_CC(U) / τ_CC^plan+β_3·Delay_order(U) / D_order+β_4·Switch_penalty(U); τ_EAF(U)=τ_charge(U)+τ_melt(U)+τ_oxid(U)+τ_reduce(U)+τ_tap(U); Wait_CC(U)=max(0,t_ladle_arrive(U)-t_CC_window); Delay_order(U)=max(0,t_finish(U)-t_due); Wherein, τ_EAF(U) represents the total smelting cycle of the electric arc furnace; τ_charge, τ_melt, τ_oxid, τ_reduce, and τ_tap represent the durations of the charging period, melting period, oxidation period, reduction period, and tapping period, respectively; Wait_CC(U) represents the continuous casting waiting time or mismatch time; Delay_order(U) represents the order delivery delay time; Switch_penalty(U) represents the scheduling penalty caused by steel grade switching, waiting during low-carbon periods, or equipment switching; β_1 to β_4 are weighting coefficients.

[0312] TimeDisturbance(U) primarily corresponds to production cycle time, continuous casting matching, order delivery, electricity price window, green electricity availability, and equipment safety boundaries in the constraint set Ω. For example, scheduling the electric furnace to a low-carbon green electricity period may reduce C_total(U), but if it causes continuous casting interruptions or order delays, it will increase TimeDisturbance(U).

[0313] (6) Formulation of the constraint set Ω

[0314] The constraint set Ω can be represented as the set of all combinations of control variables that satisfy the conditions of production, quality, environmental protection, energy, and safety.

[0315] Ω={U|g_q(U)≤0,g_T(U)≤0,g_inv(U)≤0,g_yield(U)≤0,g_tr(U)≤0,g_fan(U)≤0,g_emis(U)≤0,g_CC(U)≤0,g_order(U)≤0,g_price(U)≤0,g_green(U)≤0,g_safe(U)≤0}; The meanings of each constraint function are shown in Table 10 below.

[0316] Table 10

[0317] (7) The correspondence between the objective function and the constraint set Ω

[0318] A correspondence is established between the five objective functions and each constraint term in the constraint set Ω. This correspondence indicates that the objective functions are used to evaluate the merits of candidate paths, and the constraint set Ω is used to limit whether a candidate path is executable. As shown in Table 11.

[0319] Table 11

[0320] (8) The path output types are shown in Table 12.

[0321] Table 12

[0322] Example 1: Construction of a Data Chain for Secondary Carbon Pollution in Furnaces

[0323] A short-process electric arc furnace steel plant is equipped with an electric arc furnace, an LF refining furnace, a continuous casting machine, a rolling mill, a primary dust removal system, a secondary dust removal system, an energy management system, and an access control and transportation system. The system uses Heat-2026-001 as the heat number ID to retrieve information such as steel grade, scrap steel batching, power supply curve, oxygen flow rate, flue gas flow rate, CEMS concentration, power consumption, fuel gas consumption, steel output, and product batch for that heat.

[0324] The system determines the furnace boundary based on the power supply start time and tapping time, and allocates continuous CEMS data and dust removal fan data according to the furnace flue gas flow rate and fan load to form a carbon pollution profile for that furnace. This profile includes stage power consumption, stage carbon emissions, stage particulate matter emissions, stage NOx emissions, dust removal power consumption, and quality results.

[0325] Example 2: Normalized Evaluation under Similar Working Conditions

[0326] The system detected that the power consumption per ton of steel in Heat-2026-001 was higher than the plant-wide average. If evaluated using the traditional plant-wide average, this heat would be classified as low-performance. However, the system further identified that this heat was for high-alloy steel, requiring high tapping temperatures, had insufficient high-quality scrap steel inventory, and had high inclusion content in the scrap steel. The system then retrieved historical heats with similar steel grades, scrap steel quality, and furnace ages from a benchmark database for comparison.

[0327] After normalization, it was found that the carbon emission intensity of this furnace was still 12% higher than the similar benchmark and the particulate matter emission intensity was 8% higher than the similar benchmark. The system determined it to be a furnace with high risk in both aspects and further called the sensitivity factor matrix to locate the main cause.

[0328] Example 3: Carbon Pollution Conflict Identification and Dust Control

[0329] The particulate matter emission concentration of a certain furnace was below the limit and better than the benchmark for similar operating conditions, but the power consumption of the dust removal fan was significantly higher than that of similar furnaces, resulting in increased indirect carbon emissions. The system determined that this furnace was in the pollution reduction and carbon increase zone.

[0330] The system generates a control path based on furnace door status, flue gas flow rate, particulate matter concentration, fan frequency, and bag filter pressure differential: a high secondary dust removal airflow is maintained during the charging and furnace door opening phases; after melting and cleaning, the fan frequency is gradually reduced, and a particulate matter concentration feedback threshold is set. When the particulate matter concentration approaches the warning value, the high airflow is automatically restored. This path reduces fan power consumption while ensuring emissions meet standards.

[0331] Example 4: Coordinated Optimization of Scrap Steel Proportioning, Power Supply, and Pulse Jetting

[0332] The system identified an excessively high proportion of light and thin materials in a certain batch of furnaces, leading to a prolonged melting cycle, increased power supply time, enhanced flue gas temperature fluctuations, and higher particulate matter load. The sensitivity factor matrix showed that increasing the proportion of high-quality heavy waste and optimizing the power supply ramp-up rate had a synergistic effect on improving both carbon emissions and particulate matter.

[0333] System output path: Increase the proportion of high-quality heavy scrap in the next furnace by 5%, reduce the proportion of high-inclusion scrap by 2%, increase the power ramp-up rate in the early stage of melting, maintain stable power in the middle stage of melting, and adjust the oxygen injection volume based on CO concentration feedback. Expected results include a shorter smelting cycle, lower power consumption per ton of steel, reduced flue gas fluctuations, and lower particulate matter load.

[0334] Example 5: Coordinated Scheduling of Green Electricity Periods and Production Cycle

[0335] The system acquires the next day's production plan, continuous casting matching plan, time-of-use electricity price, available green electricity, and time-varying grid carbon factor. For dispatchable furnaces, the system schedules some high-energy-consuming furnaces to low-carbon electricity periods or green electricity availability periods without affecting order delivery and continuous casting continuity, and generates energy storage charging and discharging suggestions.

[0336] When low-carbon power periods overlap with high-emission-risk operating conditions, the system comprehensively considers carbon emission benefits, pollutant risks, production cycle time, and quality risks to select the scheduling scheme with the best overall performance.

[0337] In a specific implementation, an online evaluation system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production includes: The multi-source data acquisition module is used to acquire multi-source operational data during the electric arc furnace short-process steel production process; The boiler secondary data chain construction module is used to perform time synchronization, boiler binding and data trust verification of multi-source operation data with boiler ID as the core index, forming a boiler secondary carbon pollution data chain. The operating condition stage identification module is used to identify the operating condition stage of a furnace based on the electric furnace power curve, oxygen flow rate, flue gas parameters, and equipment status signals. The Similar Heat Trial Benchmark Library module is used to construct a similar heat trial benchmark library based on steel grade, scrap steel quality, furnace age, tapping temperature and production load, and to calculate normalized carbon emission intensity and normalized pollutant emission intensity. The carbon pollution conflict identification module is used to identify the synergistic or conflicting states of carbon pollution based on the normalization results. The sensitivity factor matrix generation module is used to generate a carbon pollution sensitivity factor matrix based on the marginal impact of adjustable control variables on carbon emissions and pollutant emissions. The multi-constraint path optimization module is used to generate pollution reduction and carbon reduction optimization paths under constraints such as steel grade quality, production cycle time, equipment capacity, emission limits and energy costs. The closed-loop control feedback module is used to output the optimized path to the production scheduling, energy management, electric furnace control or environmental control system, and dynamically correct the model based on the execution feedback.

[0338] In a specific embodiment, an electronic device includes a processor, a memory, and a communication interface. The memory stores a computer program, which the processor executes to implement the above-described method steps. The electronic device can be deployed in enterprise data centers, industrial edge computing devices, energy management platform servers, or cloud-edge collaboration platforms.

[0339] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the above-described online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production.

[0340] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0341] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production, characterized in that, include: S1. Acquire multi-source operational data during the electric arc furnace short-process steel production process; S2. Using the furnace ID as the core index, perform time synchronization, furnace binding, and data trust verification on the multi-source operation data to construct a furnace-level carbon pollution data chain; S3. Based on the electric furnace power curve, oxygen flow rate, flue gas parameters and equipment status signals, identify and classify the working conditions within the furnace. S4. Construct a benchmark library of similar heats based on steel grade, scrap steel quality, furnace age, tapping temperature and production load, and calculate the normalized carbon emission intensity and normalized pollutant emission intensity of the current heat based on the benchmark library of similar heats. S5. Identify the carbon-pollution synergy or conflict state of the current furnace based on the normalized carbon emission intensity and normalized pollutant emission intensity. S6. Based on the marginal impact of adjustable control variables on carbon emissions and pollutant emissions, generate a carbon pollution sensitivity factor matrix; S7. Under the constraints of steel quality, production cycle time, equipment capacity, emission limits and energy costs, and combining the carbon pollution synergy or conflict state with the carbon pollution sensitivity factor matrix, generate an optimized path for pollution reduction and carbon reduction. S8. Output the pollution reduction and carbon reduction optimization path to the production scheduling system, energy management system, electric furnace control system or environmental protection control system to form a closed-loop control.

2. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, The multi-source operational data includes: production data, electric furnace operation data, energy data, environmental protection data, material data, quality data, and transportation data; the furnace-level carbon pollution data chain uses the furnace ID as the primary key to establish an association model of furnace-operating stage-equipment-discharge port-energy flow-material flow-transportation flow-product batch, so that each furnace of steel corresponds to a set of carbon pollution profiles.

3. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, The data reliability verification includes consistency verification of power consumption and power supply time, consistency verification of flue gas flow rate and dust removal fan frequency, consistency verification of material input and steel output, identification of abnormal constant values ​​of pollutant concentration, identification of missing data intervals, and consistency verification of furnace number across systems.

4. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, The S3 operating phase includes at least one of the following: charging phase, melting phase, oxidation phase, reduction phase, tapping phase, and refining phase; the identification rules for each operating phase are as follows: When the scrap steel addition signal is valid, the furnace cover or furnace door is open, and the active power of the electric furnace is lower than the low power standby threshold, the corresponding time window is identified as the charging stage. When the furnace lid is closed, if the active power of the electric furnace is continuously higher than the effective power supply judgment threshold, the power curve enters the arc climbing stage or the high power melting stage, and the flue gas temperature continues to rise, the corresponding time window is identified as the melting stage. When the oxygen injection flow rate is continuously higher than the oxidation blowing flow rate threshold, and the CO volume fraction is higher than the decarbonization reaction judgment threshold or the CO volume fraction change rate changes from positive to negative, the corresponding time window is identified as the oxidation stage. When the oxygen injection flow rate is lower than the oxygen flow rate threshold during the reduction period, the CO volume fraction decreases by more than a preset ratio relative to the peak value during the oxidation period, and the active power of the electric furnace enters the low power heat preservation or intermittent power supply range, the corresponding time window is identified as the reduction stage. When the tapping tilting signal is valid and the electric furnace active power is lower than the tapping low power threshold, the corresponding time window is identified as the tapping stage. When the tapping stage is over, the ladle arrives at the refining station outside the furnace, and the refining start signal is valid or the active power of the refining furnace is higher than the refining power supply judgment threshold, the corresponding time window is identified as the refining stage.

5. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, S4 includes: S41. Construct a feature vector for furnace operation conditions. The feature vector includes steel grade, scrap grade, scrap inclusion index, furnace age, target tapping temperature, production load, smelting cycle, green electricity ratio, and quality constraint features. S42. Select a set of candidate heats from historical heats that meet the requirements of being on the same production line, having comparable steel grades, meeting quality standards, and having reliable data. Calculate the comprehensive operating condition distance between the current heat and the candidate heats and generate a similarity coefficient. S43. Select a preset number of furnaces according to their similarity from high to low to form a set of similar furnaces; S44. Based on the carbon emission intensity and pollutant emission intensity of similar furnace sets, a benchmark carbon emission intensity and a benchmark pollutant emission intensity are generated by weighted calculation. S45. The ratio of the actual carbon emission intensity of the current furnace batch to the benchmark carbon emission intensity is taken as the normalized carbon emission intensity, and the ratio of the actual pollutant emission intensity of the current furnace batch to the benchmark pollutant emission intensity is taken as the normalized pollutant emission intensity.

6. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, The carbon pollution synergy or conflict state includes: synergistic optimization zone, pollution reduction and carbon increase zone, carbon reduction and pollution increase zone, and dual high-risk zone, and the determination rule is as follows: When the normalized carbon emission intensity is ≤1 and the normalized pollutant emission intensity is ≤1, it is identified as a synergistic optimization zone; when the normalized carbon emission intensity is >1 and the normalized pollutant emission intensity is ≤1, it is identified as a pollution reduction and carbon increase zone; when the normalized carbon emission intensity is ≤1 and the normalized pollutant emission intensity is >1, it is identified as a carbon reduction and pollution increase zone; when the normalized carbon emission intensity is >1 and the normalized pollutant emission intensity is >1, it is identified as a dual high-risk zone.

7. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, The adjustable control variables include at least one of the following: scrap steel ratio, power transmission curve, oxygen injection rate, natural gas injection rate, dust removal air volume, green electricity usage period, and technological upgrading measures; the carbon pollution sensitivity factor matrix is ​​used to describe the marginal impact of each adjustable control variable on carbon emission intensity, pollutant emission intensity, production cost, quality risk, and production cycle time; the carbon pollution sensitivity factor matrix is ​​calculated and generated by local weighted regression method, digital twin perturbation method, or historical sample finite difference method.

8. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, In S7, a multi-objective optimization model is established with carbon emission intensity per ton of steel, pollutant emission intensity, comprehensive cost, quality risk, and production cycle disturbance as optimization objectives. The constraints include steel grade quality and tapping temperature constraints, scrap steel inventory and batching constraints, equipment capacity constraints, emission limits and environmental protection facility constraints, production cycle and process connection constraints, electricity price and green electricity availability constraints, and safety boundary and interlocking constraints. The pollution reduction and carbon reduction optimization path includes at least one of the following: low-carbon priority path, pollution reduction priority path, cost balance path, and quality conservative path.

9. The online evaluation method for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production according to claim 1, characterized in that, The S8 includes: decomposing the optimized path into control commands corresponding to different systems and outputting them to the MES, EMS, DCS and environmental control system respectively; collecting the actual execution data and performance results of each system, calculating the control deviation and performance deviation, and dynamically correcting the benchmark library of similar furnaces, the sensitive factor matrix and the optimization model; and selecting the output mode of automatic execution or execution after manual confirmation based on path risk, control amplitude and data reliability.

10. An online evaluation system for the synergistic performance of pollution reduction and carbon reduction in electric arc furnace short-process steel production, characterized in that, include: The multi-source data acquisition module is used to acquire multi-source operational data during the electric arc furnace short-process steel production process; The boiler secondary data chain construction module is used to perform time synchronization, boiler binding and data trust verification of multi-source operation data with boiler ID as the core index, forming a boiler secondary carbon pollution data chain. The operating condition stage identification module is used to identify the operating condition stage of a furnace based on the electric furnace power curve, oxygen flow rate, flue gas parameters, and equipment status signals. The Similar Heat Trial Benchmark Library module is used to construct a similar heat trial benchmark library based on steel grade, scrap steel quality, furnace age, tapping temperature and production load, and to calculate normalized carbon emission intensity and normalized pollutant emission intensity. The carbon pollution conflict identification module is used to identify the synergistic or conflicting states of carbon pollution based on the normalization results. The sensitivity factor matrix generation module is used to generate a carbon pollution sensitivity factor matrix based on the marginal impact of adjustable control variables on carbon emissions and pollutant emissions. The multi-constraint path optimization module is used to generate pollution reduction and carbon reduction optimization paths under constraints such as steel grade quality, production cycle time, equipment capacity, emission limits and energy costs. The closed-loop control feedback module is used to output the optimized path to the production scheduling, energy management, electric furnace control or environmental control system, and dynamically correct the model based on the execution feedback.

Images