Method and system for controlling smelting combustion process based on air-fuel ratio self-adaptation

Abstract

The present application relates to smelting combustion process control technical field, disclose a kind of smelting combustion process control method and system based on air-fuel ratio self-adaption, this method specifically includes: collecting smelting combustion process data, based on smelting combustion process data, data time sequence alignment and disturbance response feature extraction are carried out, obtain combustion process time series dataset;Based on combustion process time series dataset, event anchoring factor graph is constructed, and interference factor contribution deduction and flue gas feedback credibility assessment are carried out, obtain real air-fuel ratio deviation sequence and flue gas feedback credibility sequence;Based on real air-fuel ratio deviation sequence and flue gas feedback credibility sequence, generate fuel valve opening degree regulating quantity and combustion air valve opening degree regulating quantity, obtain smelting combustion process control result;The present application enhances the stability of smelting combustion process control, improves the accuracy of furnace temperature regulation.

Description

Technical Field

[0001] This invention relates to the field of smelting combustion process control, and more specifically to a smelting combustion process control method and system based on adaptive air-fuel ratio. Background Technology

[0002] In the combustion process control of aluminum alloy smelting furnaces, it is necessary to dynamically adjust the air-fuel ratio according to the combustion state and smelting load changes in the furnace, and maintain the furnace temperature within a suitable range for smelting by coordinating the fuel supply and combustion air supply. The aluminum alloy smelting process typically includes operations such as charging, melting, slag removal, heat preservation, and heat storage reversal. The furnace door is frequently opened and closed, cold material absorbs heat significantly, and the exhaust negative pressure also changes with operating conditions, which can easily cause turbulence in the furnace airflow and fluctuations in flue gas composition, making it difficult to accurately judge the combustion state.

[0003] In existing technologies, the smelting combustion process control method based on adaptive air-fuel ratio typically collects data on flue gas oxygen content, flue gas carbon monoxide content, furnace temperature, fuel flow rate, and combustion air flow rate. Based on this, it determines combustion deviations and then adjusts the fuel and combustion air supply. However, during aluminum alloy smelting, furnace door opening, cold material addition, slag removal, regenerator reversal, and flue gas negative pressure fluctuations can alter the furnace air leakage state and flue gas flow path. This makes it difficult for the flue gas feedback at the exhaust end to accurately represent the true combustion state of the flame zone. It can easily misjudge air leakage and mixing or sudden changes in smelting heat load as actual air-fuel ratio deviations, leading to unstable air-fuel ratio correction directions and resulting in poor stability of the smelting combustion process control and poor accuracy of furnace temperature regulation. Summary of the Invention

[0004] The purpose of this invention is to provide a smelting combustion process control method and system based on adaptive air-fuel ratio. By constructing an event anchoring factor diagram and decoupling and suppressing interference factors such as air leakage and mixing, sudden changes in heat load, and deviations in air supply execution, a true air-fuel ratio deviation sequence and a flue gas feedback credibility sequence are generated. This enables dynamic correction control of the air-fuel ratio based on credibility weighting, enhancing the stability of the smelting combustion process control and improving the accuracy of furnace temperature regulation.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] On the one hand, the present invention provides a smelting combustion process control method based on adaptive air-fuel ratio, comprising:

[0007] Collect data on the smelting and combustion process;

[0008] Based on the smelting and combustion process data, data time series alignment and disturbance response feature extraction are performed to obtain a combustion process time series dataset;

[0009] Based on the combustion process time series dataset, an event anchoring factor map is constructed, and the contribution of interference factors is deducted and the credibility of flue gas feedback is evaluated to obtain the true air-fuel ratio deviation sequence and the flue gas feedback credibility sequence.

[0010] Based on the true air-fuel ratio deviation sequence and the flue gas feedback reliability sequence, the fuel valve opening adjustment amount and the combustion air valve opening adjustment amount are generated to obtain the control results of the smelting combustion process.

[0011] Specifically, based on the smelting and combustion process data, data time-series alignment and disturbance response feature extraction are performed to obtain a combustion process time-series dataset, including:

[0012] The data in the smelting and combustion process are divided into continuous data and state data. The continuous data is time-aligned and the state data is event-preserving aligned to obtain the time-aligned process dataset.

[0013] Based on the time-aligned process dataset, the data change amount of each data in the smelting and combustion process data is extracted to obtain the process change amount dataset;

[0014] Based on the process change dataset, the data change of the combustion air supply data is compared for response consistency and the deviation is extracted to obtain the combustion air control response.

[0015] Based on the process change dataset, the suction disturbance intensity is calculated by the data changes of furnace airflow state data and furnace door opening to obtain the suction mixed gas volume.

[0016] Based on the combustion air control response and the amount of gas extracted, the time-series augmentation of the process change dataset is performed to obtain the combustion process time-series dataset.

[0017] Specifically, based on the combustion process time-series dataset, an event anchoring factor map is constructed, and interference factor contributions are deducted and flue gas feedback credibility is evaluated to obtain the true air-fuel ratio deviation sequence and the flue gas feedback credibility sequence, including:

[0018] Based on the combustion process time series dataset, an event node set and a response node set are constructed, and the window response difference of the response nodes is calculated to generate candidate association edges;

[0019] Based on candidate association edges, event node sets, and response node sets, association strength is filtered, association type is classified, and factor node is mapped to construct an event anchoring factor graph.

[0020] Based on the event anchoring factor graph, the initial contribution of each factor node is calculated along the effective association edge, and the initial contribution of nodes sharing the same response node is cross-suppressed to obtain the interference factor contribution set.

[0021] Based on the combustion process time series dataset, the original air-fuel ratio deviation evidence is generated, and combined with the interference factor contribution set, the original air-fuel ratio deviation evidence is subtracted in layers to obtain the true air-fuel ratio deviation factor.

[0022] Based on the true air-fuel ratio deviation factor, time-series deviation organization and flue gas feedback reliability assessment are performed to obtain the true air-fuel ratio deviation sequence and the flue gas feedback reliability sequence.

[0023] Specifically, based on the combustion process time-series dataset, an event node set and a response node set are constructed, and the window response differences of the response nodes are calculated to generate candidate association edges, including:

[0024] Based on the combustion process time series dataset, the data change of smelting operation event data, the data change of furnace airflow state data, the combustion air control response and the amount of gas extracted are extracted to construct an event node set;

[0025] Based on the combustion process time series dataset, the data changes of combustion feedback data, fuel supply data and combustion air supply data are extracted to construct a response node set;

[0026] Based on the event node set, the event occurrence time corresponding to each event node is identified, and the event occurrence time is used as the anchor point to divide the pre-event baseline window and the post-event response window.

[0027] Based on the set of response nodes, the baseline response value of each response node in the baseline window before the event and the response change value in the response window after the event are calculated to generate candidate associated edges.

[0028] Specifically, based on candidate association edges, event node sets, and response node sets, association strength is filtered, association type is classified, and factor node mapping is performed to construct an event anchoring factor graph, including:

[0029] Based on the baseline response value and the response change value, the response association strength corresponding to each candidate association edge is calculated, and the validity of the candidate association edges is determined to obtain the valid association edges.

[0030] Based on the effective associated edges, the associated types are divided to generate leakage factor nodes, heat load factor nodes, air-fuel ratio factor nodes and air supply disturbance factor nodes, thus obtaining a set of factor nodes;

[0031] Based on the event node set, response node set, factor node set, and effective association edges, establish the association relationships from event nodes to response nodes and from response nodes to factor nodes, and construct the event anchoring factor graph.

[0032] Specifically, based on the event anchoring factor graph, the initial contribution of each factor node is calculated along the effective association edges, and cross-suppression is applied to the initial contributions of nodes sharing the same response, resulting in a set of interference factor contributions, including:

[0033] Based on the event anchoring factor graph, the effective association edges connecting event nodes, response nodes and factor nodes are extracted, and the corresponding event node, response node, factor node and response association strength are labeled for each effective association edge to obtain the attributed association edge set;

[0034] Based on the attributed association edge set, the data change of the event node, the window response difference of the response node, and the response association strength are passed to the corresponding factor node along the effective association edge to obtain the initial interference contribution set.

[0035] Based on the attributed association edge set, valid association edges connected to different factor nodes and having the same response node are identified, and factor node response overlap relationships are generated.

[0036] Based on the initial interference contribution set and the overlapping relationship of factor node responses, cross-suppression is performed on the factor contributions that share the same response node to obtain the interference factor contribution set.

[0037] Specifically, based on the attributed association edge set, the data change of the event node, the window response difference of the response node, and the response association strength are passed along the effective association edges to the corresponding factor nodes to obtain the initial interference contribution set, including:

[0038] Based on the attributed association edge set, the data change amount of the event node, the window response difference of the response node, the response association strength and factor node corresponding to each effective association edge are extracted to obtain the edge contribution parameter set;

[0039] Based on the edge contribution parameter set, the edge contribution of each effective associated edge is calculated, and the edge contributions are accumulated to obtain the initial contribution of air leakage and mixing, the initial contribution of heat load change, and the initial contribution of air supply disturbance.

[0040] Based on the effective associated edges corresponding to the air supply disturbance factor nodes, the follow-up deviation of the combustion air supply data is identified to obtain the air supply execution deviation.

[0041] The initial contribution of air supply disturbance is corrected based on the air supply execution deviation, and the corrected initial contribution of air supply disturbance, the initial contribution of air leakage and mixing, and the initial contribution of heat load change are aggregated to the corresponding factor nodes to obtain the initial disturbance contribution set.

[0042] Specifically, based on the initial interference contribution set and the overlap relationship of factor node responses, cross-suppression is performed on the factor contributions sharing the same response node to obtain the interference factor contribution set, including:

[0043] Based on the overlapping relationship of factor node responses, cross-suppression is performed on the factor contributions that share the same response node in the initial interference contribution set to obtain the cross-suppression contribution set;

[0044] Based on the set of associated edges with attributes, the event nodes, response nodes and factor nodes corresponding to each valid associated edge are extracted, and the response start time and response decay time corresponding to the response node are determined.

[0045] Based on the event occurrence time corresponding to the event node, the response start time and response decay time corresponding to the response node, a perturbation holding segment is generated and mapped to the corresponding factor node;

[0046] Based on the disturbance holding segment, the factor contributions in the cross-suppression contribution set are time-based weighted to generate the leakage air mixing contribution, heat load mutation contribution, and air supply disturbance contribution, thus obtaining the disturbance factor contribution set.

[0047] Specifically, based on the combustion process time-series dataset, original air-fuel ratio deviation evidence is generated. Combined with the interference factor contribution set, the original air-fuel ratio deviation evidence is stratified and subtracted to obtain the true air-fuel ratio deviation factor, including:

[0048] Based on the process change dataset, the data changes of flue gas oxygen content, flue gas carbon monoxide content and combustion air supply are extracted to generate original air-fuel ratio deviation evidence.

[0049] Based on the contributions of air leakage and mixing, heat load mutation, and air supply disturbance, the abnormal deduction amounts for flue gas oxygen content, flue gas carbon monoxide content, and combustion air supply were determined respectively.

[0050] The total deduction is generated based on the abnormal deductions for flue gas oxygen content, flue gas carbon monoxide content, and combustion air supply.

[0051] Based on the total deduction, the original air-fuel ratio deviation evidence is reduced by a limit to obtain the true air-fuel ratio deviation factor.

[0052] On the other hand, the present invention provides a smelting combustion process control system based on air-fuel ratio adaptation, comprising:

[0053] The data acquisition module is used to collect data on the smelting and combustion process.

[0054] The time series construction module performs data time series alignment and disturbance response feature extraction based on smelting and combustion process data to obtain a combustion process time series dataset;

[0055] The deviation assessment module constructs an event anchoring factor map based on the combustion process time series dataset, and performs interference factor contribution deduction and flue gas feedback credibility assessment to obtain the true air-fuel ratio deviation sequence and flue gas feedback credibility sequence.

[0056] The predictive control module generates the adjustment amounts of the fuel valve opening and the combustion air valve opening based on the actual air-fuel ratio deviation sequence and the flue gas feedback reliability sequence, thereby obtaining the control results of the smelting combustion process.

[0057] Compared with existing technologies, the beneficial effects of this invention include: by constructing an event anchoring factor graph, establishing a temporal correspondence between furnace door opening, furnace airflow state, combustion air response, and flue gas feedback changes, the impact of smelting operation disturbances and furnace airflow disturbances on flue gas feedback is identified separately, reducing the interference of flue gas feedback distortion on combustion deviation judgment; calculating the initial contribution of factor nodes along effective correlation edges, and cross-suppressing the initial contributions of nodes sharing the same response, weakening the overlapping effects of air leakage and mixing, sudden changes in heat load, and air supply disturbances in flue gas feedback, reducing the probability of smelting disturbances being misjudged as true air-fuel ratio deviations; and by using the interference factor contribution set to perform layered subtraction on the original air-fuel ratio deviation evidence, generating a true air-fuel ratio deviation sequence and a flue gas feedback credibility sequence, making the adjustment basis of fuel valves and combustion air valves closer to the true combustion deviation, enhancing the stability of smelting combustion process control, and improving the accuracy of furnace temperature control. Attached Figure Description

[0058] Figure 1 A flowchart of the smelting combustion process control method based on adaptive air-fuel ratio provided by the present invention;

[0059] Figure 2 A schematic diagram of interference factor contribution transmission and cross-suppression provided by the present invention;

[0060] Figure 3 The diagram shows the structure of the smelting combustion process control system based on adaptive air-fuel ratio provided by this invention. Detailed Implementation

[0061] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0062] Example 1

[0063] See Figures 1-2 This embodiment provides a smelting combustion process control method based on adaptive air-fuel ratio, including the following specific steps:

[0064] Step S1: Collect smelting combustion process data, which includes combustion feedback data, fuel supply data, combustion air supply data, smelting operation event data, and furnace airflow status data. The combustion feedback data includes flue gas oxygen content and flue gas carbon monoxide content, and the smelting operation event data includes furnace door opening.

[0065] In this embodiment, a unified acquisition period is determined comprehensively based on the time constant of the smelting furnace combustion process, the response time of the fuel valve and the combustion air valve, the response time of the flue gas analyzer, and the control accuracy requirements of the control system. Among them, the time constant of the smelting furnace combustion process is the lag time of the change in flue gas composition caused by the change in fuel flow; the response time of the fuel valve and the combustion air valve is the time required from the issuance of the valve control command to the actual valve opening reaching 90% of the target value; the response time of the flue gas analyzer is the time required from the change in flue gas composition to 90% of the stable output value of the analyzer; and the unified acquisition period is taken as one-tenth to one-fifth of the minimum value of the above three times.

[0066] Combustion feedback data is collected, including the oxygen content and carbon monoxide content in the flue gas. The oxygen content and carbon monoxide content in the flue gas are continuously collected by a flue gas analyzer installed in the flue gas duct of the smelting furnace, located between the outlet of the regenerator and the inlet of the induced draft fan. The data collected at this location represents the average composition of the flue gas discharged from the furnace.

[0067] Fuel supply data is collected, including fuel flow rate and actual opening degree of fuel valve. Fuel flow rate is collected by a vortex flow meter installed on the fuel delivery pipeline downstream of the fuel valve. This position can measure the actual fuel flow rate entering the furnace. The actual opening degree of fuel valve is collected by a potentiometer-type position feedback device installed on the fuel valve actuator.

[0068] The combustion air supply data includes combustion air flow rate, actual opening degree of combustion air valve, combustion air fan speed, and combustion air pipeline pressure. The combustion air flow rate is collected by an orifice plate flow meter installed on the combustion air delivery pipeline downstream of the combustion air valve. The actual opening degree of the combustion air valve is collected by a potentiometer-type position feedback device installed on the combustion air valve actuator. The combustion air fan speed is collected by a Hall-effect speed sensor installed on the combustion air fan motor shaft. The combustion air pipeline pressure is collected by a pressure transmitter installed on the combustion air delivery pipeline between the combustion air valve and the combustion air flow rate measuring device.

[0069] Data on smelting operation events is collected, including furnace door opening, charging status, slag removal status, and heat storage reversal status. The furnace door opening is collected by a pull-wire displacement sensor installed on the furnace door lifting mechanism. The charging status, slag removal status, and heat storage reversal status are collected by the operating status signals of the corresponding operating actuators. The charging status, slag removal status, and heat storage reversal status are represented by a binary method, with one value taken when the corresponding operating actuator is in the running state and zero value taken when the corresponding operating actuator is in the stopped state.

[0070] The furnace airflow status data includes flue gas negative pressure and furnace pressure. Flue gas negative pressure is collected by a pressure transmitter installed in the flue gas duct and located upstream of the flue gas analyzer. Furnace pressure is collected by a pressure transmitter installed on the side wall of the smelting furnace and located in the upper middle part of the furnace. The data collected at this location represents the average pressure inside the furnace.

[0071] According to the set unified collection cycle, all the above types of data are collected synchronously to obtain smelting and combustion process data.

[0072] For example, the combustion process time constant of the aluminum alloy melting furnace is 2 seconds, the response time of the fuel valve and the combustion air valve is 1 second, and the response time of the flue gas analyzer is 1.5 seconds. Taking one-tenth of the minimum value of 1 second among the above three times, the unified acquisition period is set to 100 milliseconds.

[0073] At the 1000th data collection point, the following data were collected: flue gas oxygen content 8.2%, flue gas carbon monoxide content 20 parts per million, fuel flow rate 120 standard cubic meters per hour, actual fuel valve opening 35%, combustion air flow rate 3000 standard cubic meters per hour, actual combustion air valve opening 49%, combustion air fan speed 1500 rpm, combustion air pipeline pressure 120 Pa, furnace door opening 0%, feeding status 0, slag removal status 0, heat storage reversal status 0, flue gas negative pressure -20 Pa, and furnace pressure -5 Pa.

[0074] At the 1001st data collection point, the following data were collected: flue gas oxygen content 8.5%, flue gas carbon monoxide content 18 parts per million, fuel flow rate 122 standard cubic meters per hour, actual fuel valve opening 36%, combustion air flow rate 3150 standard cubic meters per hour, actual combustion air valve opening 50.8%, combustion air fan speed 1530 rpm, combustion air pipeline pressure 125 Pa, furnace door opening 0%, feeding status 1, slag removal status 0, heat storage reversal status 0, flue gas negative pressure -25 Pa, and furnace pressure -7 Pa.

[0075] All data were collected continuously at 1000 time points in the same manner to obtain smelting and combustion process data containing 1000 sets of data records.

[0076] Step S2: Based on the smelting and combustion process data, perform data time-series alignment and disturbance response feature extraction to obtain the combustion process time-series dataset.

[0077] The specific steps of step S2 are as follows:

[0078] Step S201: Divide the data in the smelting and combustion process into continuous data and state data, perform time alignment on the continuous data, and perform event-preserving alignment on the state data to obtain the time-aligned process dataset.

[0079] In this embodiment, the oxygen content of flue gas, carbon monoxide content of flue gas, fuel flow rate, actual opening degree of fuel valve, combustion air flow rate, actual opening degree of combustion air valve, combustion air fan speed, combustion air pipeline pressure, furnace door opening, flue gas negative pressure and furnace pressure are classified as continuous data; the feeding state, slag removal state and heat storage reversal state are classified as state data.

[0080] For continuous data, linear interpolation is used for time alignment. For any continuous quantity, if the unified acquisition time is between two original sampling times, the value of the unified acquisition time is calculated as follows: First, calculate the difference between the values ​​of the two original sampling times. Then, calculate the proportion of the time difference between the unified acquisition time and the previous original sampling time to the total time difference between the two original sampling times. Multiply the value difference by this proportion, and finally add the value of the previous original sampling time to obtain the continuous quantity value of the unified acquisition time. If the unified acquisition time is earlier than the first original sampling time, the value of the first original sampling time is directly taken. If the unified acquisition time is later than the last original sampling time, the value of the last original sampling time is directly taken.

[0081] The event-based time-alignment method is used for state variable data. For any state variable, if the unified acquisition time is between two original sampling times, the value of the previous original sampling time is kept unchanged until the value of the next original sampling time changes.

[0082] After completing the time alignment of all data, the time alignment process dataset is obtained.

[0083] For example, the original sampled value of a certain continuous quantity is 10 at 1 second and 20 at 2 seconds. The unified sampling time is 1.3 seconds, which is between 1 second and 2 seconds. The difference between the values ​​at the two original sampling times is 10, and the time difference ratio is 0.3. The calculated value of the continuous quantity at 1.3 seconds at the unified sampling time is 10 plus 10 multiplied by 0.3, which is 13.

[0084] The initial sampled value of a certain state variable is 0 at 1 second, and becomes 1 at 3 seconds. The unified sampling time is 2 seconds, which is between 1 second and 3 seconds. The value of this state variable remains 0 at 2 seconds.

[0085] Step S202: Based on the time-aligned process dataset, extract the data change amount of each data in the smelting and combustion process data to obtain the process change amount dataset.

[0086] In this embodiment, for each data item in the time alignment process dataset, the change at each acquisition time is calculated sequentially. The change at the current acquisition time is equal to the value at the current time minus the value at the previous acquisition time.

[0087] For the first data collection moment, since there is no data from the previous moment, the change is uniformly set to zero.

[0088] The changes in all data at all collection points are organized in chronological order to obtain the process change dataset.

[0089] For example, the data has a value of 8 at the first acquisition time, a value of 10 at the second acquisition time, a value of 10 at the 1000th acquisition time, and a value of 12 at the 1001st acquisition time.

[0090] The change at the first data acquisition time is set to zero.

[0091] The change at the second acquisition time was 10 minus 8, resulting in 2.

[0092] The change at the 1001st acquisition time is 12 minus 10, resulting in 2.

[0093] Step S203: Based on the process change dataset, perform response consistency comparison and deviation extraction on the data change of the combustion air supply data to obtain the combustion air control response.

[0094] In this embodiment, the changes in the control command of the combustion air valve, the actual opening degree of the combustion air valve, the speed of the combustion air fan, and the flow rate of the combustion air are extracted from the process change dataset.

[0095] The four variables mentioned above are normalized by dividing each variable by the sum of the standard deviation of the variable in the historical stable combustion segment and a small constant to prevent the denominator from being zero. The small constant is 0.001. The parameters of the historical stable combustion segment are consistent with those in step S1.

[0096] Calculate the absolute value of the difference between the change in the control command of the combustion air valve and the actual change in the opening degree of the combustion air valve to obtain the valve command opening degree deviation.

[0097] Calculate the absolute value of the difference between the actual change in the opening degree of the combustion air valve and the change in the combustion air flow rate to obtain the opening-flow deviation.

[0098] Calculate the absolute value of the difference between the change in the combustion fan speed and the change in the combustion air flow rate to obtain the speed-flow deviation.

[0099] The valve command opening deviation, opening flow deviation, and speed flow deviation are added together to obtain the combustion air execution deviation. The combustion air execution deviation is used to characterize the degree of deviation between the combustion air control command and the actual execution effect. The larger the value of the combustion air execution deviation, the more the combustion air execution state deviates from the control command.

[0100] Based on the changes in the control command of the combustion air valve, the actual changes in the opening degree of the combustion air valve, and the changes in the combustion air flow rate, the effective response of the combustion air is calculated. The effective response of the combustion air is the difference obtained by subtracting the execution deviation of the combustion air from the absolute value of the change in the control command of the combustion air valve.

[0101] If the effective response of the combustion air is less than zero, the control response of the combustion air is corrected to zero, and the execution deviation of the combustion air is retained as the air supply disturbance characteristic; if the effective response of the combustion air is greater than or equal to zero, the effective response of the combustion air is used as the control response of the combustion air.

[0102] The combustion air control response and combustion air execution deviation are simultaneously collected into the data records at the corresponding acquisition time, so that the effective response of the combustion air and the air supply execution deviation can be used to construct the subsequent event anchoring factor map.

[0103] For example, at the 1001st data acquisition time, the original change in the control command of the combustion air valve was 2% opening, the original change in the actual opening of the combustion air valve was 1.8% opening, the original change in the rotational speed of the combustion air fan was 30 revolutions per minute, and the original change in the flow rate of the combustion air was 150 standard cubic meters per hour.

[0104] The corresponding standard deviations of the historical stable combustion segments are 2% opening, 2% opening, 150 revolutions per minute, and 150 standard cubic meters per hour, respectively.

[0105] After normalization, the change in the control command of the combustion air valve is approximately 1.00, the change in the actual opening degree of the combustion air valve is approximately 0.90, the change in the speed of the combustion air blower is approximately 0.20, and the change in the flow rate of the combustion air is approximately 1.00.

[0106] The absolute value of the valve command opening deviation was calculated to be 1.00 minus 0.90, which is 0.10.

[0107] The deviation of the flow rate at the opening is the absolute value of 0.90 minus 1.00, which results in 0.10.

[0108] The deviation between rotational speed and flow rate is the absolute value of 0.20 minus 1.00, resulting in 0.80.

[0109] The deviation of the combustion air output is 0.10 plus 0.10 plus 0.80, and the result is 1.00.

[0110] The effective response of the combustion air is the absolute value of 1.00 minus 1.00, which results in 0.00. Therefore, the control response of the combustion air is taken as 0.00, and the execution deviation of the combustion air of 1.00 is retained as the characteristic of air supply disturbance.

[0111] If the normalized change of the combustion air valve control command is 1.10 and the combustion air execution deviation is still 1.00, then the effective response of the combustion air is 1.10 minus 1.00, which is 0.10. The control response of the combustion air is taken as 0.10, and the execution deviation of the combustion air of 1.00 is still retained as the air supply disturbance characteristic.

[0112] Step S204: Based on the process change dataset, calculate the suction disturbance intensity of the furnace airflow state data and furnace door opening data changes to obtain the suction mixed gas volume.

[0113] In this embodiment, the changes in flue gas negative pressure, furnace pressure, and furnace door opening are extracted from the process change dataset.

[0114] The three variables mentioned above are normalized, and the normalization method is consistent with that in step S203.

[0115] The larger value between the change in negative pressure of the exhaust smoke and zero is taken to obtain the contribution value of negative pressure of the exhaust smoke.

[0116] The larger of the negative value and zero of the furnace pressure change is taken to obtain the furnace pressure contribution value.

[0117] The larger of the change in furnace door opening and zero is taken to obtain the contribution value of furnace door opening.

[0118] The contribution values ​​of flue gas negative pressure, furnace pressure, and furnace door opening are added together to obtain the amount of gas to be drawn in.

[0119] For example, at the 1001st data collection time, the initial change in flue gas negative pressure was -5 Pa, the initial change in furnace pressure was -2 Pa, and the initial change in furnace door opening was 0%.

[0120] The corresponding standard deviations for the historical stable combustion segments were 2.5 Pa, 1 Pa, and 10%, respectively. After normalization, the changes in flue gas negative pressure, furnace pressure, and furnace door opening were approximately -2.00 and 0.00, respectively.

[0121] The calculated contribution values ​​are: flue gas negative pressure 0, furnace pressure 2.00, and furnace door opening 0.00.

[0122] The gas mixing volume is 0 + 2.00 + 0.00, and the result is 2.00.

[0123] Step S205: Based on the combustion air control response and the amount of gas extracted, perform time-series enhancement on the process change dataset to obtain the combustion process time-series dataset.

[0124] In this embodiment, the combustion wind control response and the amount of gas mixture extracted calculated at each acquisition moment are used as two independent feature dimensions and added to the end of the record at the corresponding acquisition moment in the normalized process change dataset.

[0125] After adding features to all acquisition moments, a time-series dataset of the combustion process is obtained.

[0126] For example, the normalized process change record at the 1001st acquisition time includes: flue gas oxygen content change 1.50, flue gas carbon monoxide content change -0.40, fuel flow rate change 1.00, actual fuel valve opening change 1.00, combustion air valve control command change 1.00, combustion air valve actual opening change 0.90, combustion air fan speed change 0.20, combustion air flow rate change 1.00, combustion air pipeline pressure change 1.00, furnace door opening change 0.00, feeding status change 1.00, slag removal status change 0.00, heat storage reversal status change 0.00, flue gas negative pressure change -2.00, and furnace pressure change -2.00, for a total of 15 original change features.

[0127] After adding the combustion air control response of 0.10 and the intake air mixing amount of 2.00, the record at this moment contains a total of 17 features, constituting a record in the combustion process time series dataset.

[0128] Step S3: Based on the combustion process time series dataset, construct the event anchoring factor map, and perform interference factor contribution deduction and flue gas feedback credibility assessment to obtain the true air-fuel ratio deviation sequence and flue gas feedback credibility sequence.

[0129] The specific steps of step S3 are as follows:

[0130] Step S301: Based on the combustion process time series dataset, construct an event node set and a response node set, calculate the window response difference of the response nodes, and generate candidate association edges.

[0131] The specific steps of step S301 are as follows:

[0132] Step S3011: Based on the combustion process time series dataset, extract the data change amount of smelting operation event data, the data change amount of furnace airflow state data, the combustion air control response amount, and the gas extraction mixing amount to construct an event node set.

[0133] In this embodiment, the following data are extracted from the combustion process time-series dataset: changes in furnace door opening, changes in feeding status, changes in slag removal status, changes in heat storage reversal status, changes in flue gas negative pressure, changes in furnace pressure, combustion air control response, and combustion air execution deviation. These data cover all external operational events and internal airflow disturbance events that may cause flue gas feedback distortion during aluminum alloy smelting. Among them, smelting operation events directly change the furnace's heat load and air leakage status, furnace airflow status events change the flue gas flow path and mixing ratio, and combustion air-related events reflect the execution status of the control system itself.

[0134] Each of the above data types is treated as an independent event node, and an event node set is constructed.

[0135] For example, the event node data at the 1001st acquisition time is as follows: furnace door opening change 0.00, feeding status change 1.00, slag removal status change 0.00, heat storage reversal status change 0.00, flue gas negative pressure change -2.00, furnace pressure change -2.00, combustion air control response 0.10, combustion air execution deviation 1.00.

[0136] Step S3012: Based on the combustion process time series dataset, extract the data changes of combustion feedback data, fuel supply data and combustion air supply data, and construct a response node set.

[0137] In this embodiment, the changes in flue gas oxygen content, flue gas carbon monoxide content, furnace temperature, fuel flow rate, and combustion air flow rate are extracted from the combustion process time series data. These data are the core feedback quantities of the combustion state. Among them, the flue gas oxygen content and carbon monoxide content directly reflect the degree of combustion completeness, the furnace temperature reflects the heat output of combustion, and the fuel and combustion air flow rates reflect the input conditions of combustion.

[0138] Treat each of the above data types as an independent response node to construct a set of response nodes.

[0139] For example, the response node data at the 1001st acquisition time is as follows: flue gas oxygen content change 1.50, flue gas carbon monoxide content change -0.40, furnace temperature change -0.80, fuel flow rate change 1.00, and combustion air flow rate change 1.00.

[0140] Step S3013: Based on the event node set, identify the event occurrence time corresponding to each event node, and use the event occurrence time as the anchor point to divide the pre-event baseline window and the post-event response window.

[0141] In this embodiment, each event node in the event node set is traversed, and the moment when the data change of each event node is not zero is identified, and that moment is determined as the event occurrence time.

[0142] For events whose duration exceeds one acquisition cycle, the first moment of the event is determined as the event occurrence time.

[0143] Using the moment each event occurs as an anchor point, divide the window into a baseline window before the event and a response window after the event.

[0144] The pre-event baseline window consists of m data acquisition moments before the event occurs, used to calculate the stable baseline value of the combustion state before the event. The post-event response window consists of r data acquisition moments after the event occurs, used to calculate the response amplitude of the combustion state after the event. m and r are determined based on the combustion response time of the smelting furnace. m is taken as 0.6 times the number of data acquisition moments corresponding to the combustion response time to ensure that the baseline window is long enough to represent the stable combustion state before the event. r is taken as 1.2 times the number of data acquisition moments corresponding to the combustion response time to ensure that the response window can cover the entire combustion response process caused by the event.

[0145] For example, the combustion response time of the aluminum alloy melting furnace is 10 seconds, the uniform acquisition period is 100 milliseconds, corresponding to 100 acquisition moments.

[0146] The calculations show that m represents 60 data acquisition times and r represents 120 data acquisition times.

[0147] The feeding status changes at the 1001st data collection point, which is taken as the time of the event occurrence.

[0148] The baseline window before the event is from the 941st acquisition time to the 1000th acquisition time, a total of 60 times.

[0149] The post-event response window spans from the 1001st to the 1120th data acquisition time, a total of 120 time points.

[0150] Step S3014: Based on the set of response nodes, calculate the baseline response value of each response node in the baseline window before the event and the response change value in the response window after the event, and generate candidate associated edges.

[0151] In this embodiment, for each response node, the arithmetic mean of its values ​​at all acquisition times within the baseline window before the event is calculated to obtain the baseline response value, which represents the stability level of the response node before the event occurs.

[0152] Calculate the arithmetic mean of the values ​​of the response node at all acquisition times within the response window after the event to obtain the response mean. The response mean represents the average level of the response node after the event occurs.

[0153] The absolute value of the difference between the mean response value and the baseline response value is calculated and divided by the sum of the standard deviation of the response node in the historical stable combustion segment and the small constant to prevent the denominator from being zero. The small constant is taken as 0.001 to obtain the window response difference. The window response difference represents the normalized deviation of the steady-state change amplitude of the response node before and after the event disturbance from the historical normal fluctuation, and is used to describe the strength of the event's disturbance to the combustion state.

[0154] A candidate association edge is generated between each event node and each response node, and the weight of the candidate association edge is the corresponding window response difference.

[0155] For example, the average value of the flue gas oxygen content response node is 0.05 in the pre-event baseline window 941 to 1000 and 1.20 in the post-event response window 1001 to 1120.

[0156] The historical standard deviation of oxygen content in flue gas is 0.2%.

[0157] Calculate the absolute value of 1.20 minus 0.05, divide by the sum of 0.2 and 0.001, and you get a window response difference of approximately 5.75.

[0158] Candidate association edges are generated between the feeding status event node and the flue gas oxygen content response node, with an edge weight of 5.75.

[0159] Generate 40 candidate association edges between 8 event nodes and 5 response nodes in the same way.

[0160] Step S302: Based on the candidate association edges, event node set and response node set, perform association strength screening, association type classification and factor node mapping to construct the event anchoring factor graph.

[0161] The specific steps of step S302 are as follows:

[0162] Step S3021: Based on the baseline response value and the response change value, calculate the response association strength corresponding to each candidate association edge, and determine the validity of the candidate association edges to obtain the valid association edges.

[0163] In this embodiment, the weight of the candidate associated edge is used as the corresponding response association strength, and the response association strength is the window response difference value corresponding to the candidate associated edge, which is used to describe the strength of the causal association between the event node and the response node.

[0164] Based on the response fluctuation distribution under historical normal combustion conditions, a response correlation strength threshold is set. Specifically, the differences in response of all windows under 100 hours of historical normal combustion conditions are statistically analyzed to obtain a sample set of window response differences, and the 95th percentile of the sample set of window response differences is calculated.

[0165] The candidate threshold range was set at 1.2 to 1.8 times the 95th percentile, and each candidate threshold was used to effectively filter historical disturbance samples for association edges. Based on manually labeled records of furnace door opening, feeding, slag removal, heat storage reversal, and abnormal air supply, the disturbance identification accuracy and false association rate under different candidate thresholds were calculated. The candidate threshold with the highest disturbance identification accuracy and a false association rate lower than a preset upper limit was selected as the response association strength threshold. The preset upper limit was determined based on the statistical distribution of the false association rate in the historical disturbance samples. Specifically, the historical disturbance samples were manually labeled, the distribution of the false association rate under each candidate threshold was statistically analyzed, the mean and standard deviation of the false association rate distribution were calculated, and the mean plus twice the standard deviation was used as the preset upper limit.

[0166] When there are insufficient historical perturbation samples, 1.5 times the 95th percentile is used as the threshold for response association strength.

[0167] Candidate association edges with a response association strength greater than the response association strength threshold are determined as valid association edges, and the remaining candidate association edges are eliminated.

[0168] For example, the 95th percentile of the differences in response across all windows under historical normal combustion conditions is 0.2, and the threshold for response correlation strength is set at 0.3.

[0169] The candidate edge weight for the relationship between the feeding status and the oxygen content of the flue gas is 5.75, which is greater than 0.3, and is therefore determined to be a valid edge.

[0170] The candidate associated edge weight for furnace door opening and fuel flow rate is 0.15, which is less than 0.3, so it is discarded.

[0171] After filtering 40 candidate edges, 12 valid edges are obtained, forming a set of valid edges.

[0172] Step S3022: Based on the effective associated edges, perform association type classification to generate leakage factor nodes, heat load factor nodes, air-fuel ratio factor nodes and air supply disturbance factor nodes, and obtain factor node set.

[0173] In this embodiment, the effective associated edges connecting the event nodes of changes in furnace door opening, changes in flue gas negative pressure, and changes in furnace pressure are classified as air leakage associated types. These events cause outside air to enter the furnace, dilute the flue gas composition, increase the oxygen content of the flue gas, and cause a misjudgment of an excessively high air-fuel ratio.

[0174] The valid associated edges connecting the event nodes of the feeding state and the slag removal state are classified into the heat load associated type. The event will cause a sudden increase in the furnace heat load, incomplete fuel combustion, and an increase in the carbon monoxide content in the flue gas, resulting in a misjudgment of an excessively low air-fuel ratio.

[0175] The valid associated edges connecting the combustion air execution deviation event nodes are classified into air supply disturbance associated types. These events reflect the execution deviation of the combustion air control system, which can lead to inconsistencies between the actual air supply and the control command.

[0176] For valid correlation edges that are not classified into the leakage correlation type, heat load correlation type, and air supply disturbance correlation type, if the response node corresponding to the valid correlation edge includes flue gas oxygen content, flue gas carbon monoxide content, fuel flow rate, or combustion air flow rate, and the valid correlation edge reflects the matching deviation between changes in fuel supply and changes in combustion air supply, then it is classified into the air-fuel ratio correlation type.

[0177] When the same valid associated edge satisfies multiple associated types at the same time, it is prioritized for classification according to the following order: air leakage associated type, heat load associated type, air supply disturbance associated type, and air-fuel ratio associated type.

[0178] When the same event node connects to multiple response nodes, the types are classified based on the response node and response association strength corresponding to each valid association edge, and multiple valid association edges are not forcibly merged into the same association type.

[0179] Based on the association type classification results, factor node mapping and instantiation are performed on the effective association edges corresponding to different association types to generate leakage factor nodes, heat load factor nodes, air-fuel ratio factor nodes and air supply disturbance factor nodes, and construct factor node sets.

[0180] For example, the effective correlation edge between the feeding status and the oxygen content of the flue gas belongs to the heat load correlation type and is mapped to the heat load factor node.

[0181] The effective correlation edge between the change in flue gas negative pressure and the oxygen content of flue gas belongs to the air leakage correlation type and is mapped to the air leakage factor node.

[0182] The effective correlation edge between the combustion air execution deviation and the combustion air flow rate belongs to the air supply disturbance correlation type and is mapped to the air supply disturbance factor node.

[0183] The effective correlation edge between flue gas oxygen content and fuel flow rate belongs to the air-fuel ratio correlation type and is mapped to the air-fuel ratio factor node.

[0184] Step S3023: Based on the event node set, response node set, factor node set, and effective association edges, establish the association relationships from event nodes to response nodes and from response nodes to factor nodes, and construct the event anchoring factor graph.

[0185] In this embodiment, an event node is connected to the corresponding response node through a valid associated edge, establishing a first-level association relationship.

[0186] Establish a second-level association between the response node and the corresponding factor node through the corresponding association type.

[0187] By integrating all event nodes, response nodes, factor nodes, and valid associated edges, a three-layer event anchoring factor graph is constructed.

[0188] For example, the event anchoring factor graph contains 8 event nodes, 5 response nodes, 4 factor nodes, and 12 valid association edges; among them, the feeding status event node is connected to the flue gas oxygen content response node through a valid association edge, the flue gas oxygen content response node is connected to the heat load factor node through a heat load association type, the exhaust negative pressure change event node is connected to the flue gas oxygen content response node through a valid association edge, and the flue gas oxygen content response node is connected to the air leakage factor node through an air leakage association type.

[0189] like Figure 2 As shown, step S303: Based on the event anchoring factor graph, calculate the initial contribution of each factor node along the effective association edge, and perform cross-suppression on the initial contributions of nodes sharing the same response to obtain the interference factor contribution set.

[0190] The specific steps of step S303 are as follows:

[0191] Step S3031: Based on the event anchoring factor graph, extract the valid association edges connecting event nodes, response nodes, and factor nodes, and label each valid association edge with the corresponding event node, response node, factor node, and response association strength to obtain an attributed association edge set.

[0192] In this embodiment, all valid association edges connecting event nodes, response nodes, and factor nodes are extracted from the event anchoring factor graph.

[0193] For each valid associated edge, label the corresponding unique identifier of the event node, the unique identifier of the response node, the unique identifier of the factor node, and the response association strength.

[0194] All labeled and valid associated edges are organized into a set to obtain the associated edge set with attributes.

[0195] For example, one attributed edge has the following attributes: event node identifier E2, indicating feeding status; response node identifier R1, indicating flue gas oxygen content; factor node identifier F2, indicating heat load factor; and response association strength 5.75. Another attributed edge has the following attributes: event node identifier E5, indicating flue gas negative pressure change; response node identifier R1, indicating flue gas oxygen content; factor node identifier F1, indicating air leakage factor; and response association strength 3.20.

[0196] Step S3032: Based on the attributed association edge set, the data change of the event node, the window response difference of the response node, and the response association strength are transmitted to the corresponding factor node along the effective association edge to obtain the initial interference contribution set.

[0197] The specific steps of step S3032 are as follows:

[0198] Step S30321: Based on the attribute-based association edge set, extract the data change amount of the event node corresponding to each valid association edge, the window response difference of the response node, the response association strength and factor node, to obtain the edge contribution parameter set.

[0199] In this embodiment, each valid associated edge in the set of associated edges with attributes is traversed, and the normalized data change of the event node corresponding to the edge at the time of the event occurrence, the window response difference of the response node, the response association strength, and the factor node identifier are extracted.

[0200] The above four parameters are organized into edge contribution parameters for each valid associated edge, and the edge contribution parameters of all valid associated edges constitute the edge contribution parameter set.

[0201] For example, the edge contribution parameters corresponding to the effective correlation between the feeding status and the oxygen content of the flue gas are: event change 1.00, window response difference 5.75, response correlation strength 5.75, and factor node identifier F2.

[0202] Step S30322: Based on the edge contribution parameter set, calculate the edge contribution of each effective associated edge, and accumulate the edge contributions to obtain the initial contribution of air leakage and mixing, the initial contribution of heat load change, and the initial contribution of air supply disturbance.

[0203] In this embodiment, the data change of the event node, the window response difference of the response node, and the response correlation strength are all normalized quantities. The data change of the event node represents the intensity of the disturbance input, the window response difference of the response node represents the magnitude of the combustion feedback change, and the response correlation strength represents the reliability of the correspondence between the event node and the response node.

[0204] For each valid association edge, the edge contribution is obtained by multiplying the absolute value of the data change of the event node, the window response difference of the response node, and the response association strength. The above three quantities correspond to the disturbance input, response output, and association reliability, respectively. Therefore, the edge contribution is used to represent the degree of contribution of the valid association edge to the corresponding factor node.

[0205] To avoid excessive edge contribution caused by a single abnormal sampling, the edge contribution distribution of similar effective associated edges in historical stable combustion segments is statistically analyzed. The edge contribution sample set of similar effective associated edges in historical stable combustion segments is statistically analyzed. The edge contribution sample set consists of the edge contribution values ​​of effective associated edges corresponding to each time point in the historical stable combustion segments. The 99th percentile is calculated based on its statistical distribution. The limit value is set to 1.2 times the 99th percentile of the edge contribution distribution. If the calculated edge contribution exceeds the limit value, the edge contribution is corrected to the limit value.

[0206] The initial contribution of a factor node is obtained by summing the edge contributions of all valid associated edges mapped to the same factor node.

[0207] The initial contributions of the air leakage factor node, the heat load factor node, and the air supply disturbance factor node are calculated separately to obtain the initial contributions of air leakage and air mixing, the initial contributions of heat load mutation, and the initial contributions of air supply disturbance.

[0208] For example, calculating 1.00 multiplied by 5.75 multiplied by 5.75 yields an effective edge contribution of approximately 33.06 for the correlation between the feeding state and the oxygen content of the flue gas.

[0209] The heat load factor node is only connected to this valid associated edge, so the initial contribution of the heat load mutation is 33.06.

[0210] The air leakage factor node is connected to two valid associated edges, with edge contributions of 12.80 and 8.50 respectively. The sum of these contributions yields an initial contribution of 21.30 for the air leakage mixture.

[0211] The air supply disturbance factor node is connected to a valid associated edge, and the edge contribution is 15.50. Therefore, the initial contribution of the air supply disturbance is 15.50.

[0212] Step S30323: Based on the effective associated edges corresponding to the air supply disturbance factor nodes, identify the following deviation of the combustion air supply data to obtain the air supply execution deviation.

[0213] In this embodiment, the changes in the control command of the combustion air valve, the actual opening degree of the combustion air valve, the speed of the combustion air fan, and the flow rate of the combustion air are extracted from the combustion process time series data.

[0214] Calculate the absolute value of the difference between the change in the control command of the combustion air valve and the actual change in the opening degree of the combustion air valve to obtain the valve command opening degree deviation.

[0215] Calculate the absolute value of the difference between the actual change in the opening degree of the combustion air valve and the change in the combustion air flow rate to obtain the opening-flow deviation.

[0216] Calculate the absolute value of the difference between the change in the combustion fan speed and the change in the combustion air flow rate to obtain the speed-flow deviation.

[0217] Adding the three deviations together gives the air supply execution deviation. The air supply execution deviation ranges from 0 to positive infinity. The larger the value, the greater the execution deviation of the combustion air control system and the greater the difference between the actual air supply and the control command.

[0218] For example, at the 1001st data acquisition time, the change in the control command of the combustion air valve is 1.00, the change in the actual opening degree of the combustion air valve is 0.90, the change in the speed of the combustion air fan is 0.20, and the change in the flow rate of the combustion air is 1.00.

[0219] The calculated valve command opening deviation is 0.10, the opening flow deviation is 0.10, and the speed flow deviation is 0.80.

[0220] Calculate 0.10 plus 0.10 plus 0.80, and the air supply execution deviation is 1.00.

[0221] Step S30324: Correct the initial contribution of air supply disturbance based on the air supply execution deviation, and aggregate the corrected initial contribution of air supply disturbance, the initial contribution of air leakage and mixing, and the initial contribution of heat load change to the corresponding factor nodes to obtain the initial disturbance contribution set.

[0222] In this embodiment, the normalized value of the air supply execution deviation is calculated. The normalization method is to divide the current air supply execution deviation by the sum of the historical maximum value of the air supply execution deviation and 0.001. The historical maximum value of the air supply execution deviation is obtained by statistically analyzing the operating data of the past 100 hours.

[0223] The corrected initial contribution of air supply disturbance is obtained by multiplying the initial contribution of air supply disturbance by the normalized value of air supply execution deviation.

[0224] The corrected initial contributions of air supply disturbance, air leakage and mixing, and heat load mutation are respectively aggregated to the corresponding factor nodes to obtain the initial disturbance contribution set.

[0225] For example, the historical maximum value of the air supply execution deviation is 5.00. Calculating 1.00 divided by the sum of 5.00 and 0.001, we get a normalized value of approximately 0.20.

[0226] The initial contribution of the air supply disturbance is 15.50. Multiplying 15.50 by 0.20, we get the corrected initial contribution of the air supply disturbance as 3.10.

[0227] The initial disturbance contribution set includes an initial contribution of 21.30 for air leakage and mixing, an initial contribution of 33.06 for sudden changes in heat load, and an initial contribution of 3.10 for air supply disturbance.

[0228] Step S3033: Based on the attribute-associated edge set, identify valid associated edges that connect to different factor nodes and have the same response node, and generate factor node response overlap relationships.

[0229] In this embodiment, the set of attributed associated edges is traversed to identify all valid associated edges that connect to different factor nodes but have the same response node.

[0230] The factor nodes corresponding to these effective associated edges are marked as having response overlap. Response overlap indicates that multiple different interference factors will simultaneously affect the same combustion response node, resulting in overlapping causes of flue gas feedback distortion.

[0231] Generate all factor node pairs with overlapping responses to form factor node response overlap relationships.

[0232] For example, the flue gas oxygen content response node is connected to both the air leakage factor node and the heat load factor node, so the air leakage factor node and the heat load factor node have an overlapping response relationship.

[0233] The combustion air flow response node is connected to both the air supply disturbance factor node and the air-fuel ratio factor node, so the air supply disturbance factor node and the air-fuel ratio factor node have an overlapping response relationship.

[0234] Step S3034: Based on the initial interference contribution set and the overlapping relationship of factor node responses, cross-suppression is performed on the factor contributions that share the same response node to obtain the interference factor contribution set.

[0235] The specific steps of step S3034 are as follows:

[0236] Step S30341: Based on the overlapping relationship of factor node responses, perform cross-suppression on the factor contributions that share the same response node in the initial interference contribution set to obtain the cross-suppression contribution set.

[0237] In this embodiment, for each group of factor nodes with overlapping responses, the response node shared by the group of factor nodes is first determined, and then the initial contribution obtained by each factor node through the response node is extracted.

[0238] Calculate the shared response contribution, which represents the contribution of multiple factor nodes that are repeatedly interpreted through the same response node. The shared response contribution is the minimum of the initial contributions obtained by the group of factor nodes through the same response node, or the sum of the initial contributions obtained by the group of factor nodes through the same response node multiplied by the overlap ratio.

[0239] The overlap ratio is determined based on historical labeled samples. Specifically, the contribution ratio of the same response node explained by multiple factor nodes is statistically analyzed in the historical labeled samples, and the average value of the contribution ratio is used as the overlap ratio. The overlap ratio is preferably between 0.2 and 0.5.

[0240] The overlapping deduction share corresponding to each factor node is determined based on the proportion of the initial contribution obtained by each factor node through the response node to the total initial contribution of the group of factor nodes.

[0241] The cross-suppression factor contribution is obtained by subtracting the corresponding overlap reduction from the initial contribution of each factor node.

[0242] If the factor contribution after cross-suppression is less than zero, then the factor contribution after cross-suppression will be corrected to zero.

[0243] After performing cross-suppression on all factor node groups with overlapping response relationships, the cross-suppression contribution set is obtained.

[0244] For example, the initial contribution of the air leakage factor node is 21.30, and the initial contribution of the heat load factor node is 33.06, with a total contribution of 54.36.

[0245] The contribution ratio of the air leakage factor node is approximately 0.39, and the contribution ratio of the heat load factor node is approximately 0.61.

[0246] After cross-suppression, multiplying 21.30 by the difference of 1 minus 0.61 yields a contribution of approximately 8.31 to the leakage factor node.

[0247] Multiplying 33.06 by the difference of 1 minus 0.39, we get the contribution of the heat load factor node as approximately 20.17.

[0248] The total contribution after cross-suppression is 28.48, which is less than the original total contribution of 54.36, thus eliminating the overlap in contributions.

[0249] Step S30342: Based on the set of attribute-related edges, extract the event nodes, response nodes, and factor nodes corresponding to each valid related edge, and determine the response start time and response decay time corresponding to the response node.

[0250] In this embodiment, for each response node, its instantaneous response deviation value at each acquisition time is calculated. The instantaneous response deviation value is equal to the value of the response node at the current time minus the absolute value of its baseline response value in the baseline window before the event, divided by the sum of the historical standard deviation of the response node and the minimum positive smoothing term, wherein the minimum positive smoothing term is preferably 0.001.

[0251] Set a response start threshold and a response decay threshold. The response start threshold is two-thirds of the response correlation strength threshold, and the response decay threshold is one-third of the response correlation strength threshold. When the instantaneous response deviation is greater than the response start threshold and the condition is met for two consecutive sampling times, the corresponding sampling time is determined as the response start time. This is to avoid misjudgment caused by a single random fluctuation.

[0252] When the instantaneous response deviation is less than the response decay threshold and the condition is met for three consecutive acquisition times, the corresponding acquisition time is determined as the response decay time to ensure that the response has completely decayed to the normal level.

[0253] For example, the response association strength threshold is 0.3, therefore the response initiation threshold is 0.2, and the response decay threshold is 0.1.

[0254] The instantaneous response deviation of the flue gas oxygen content response node was 0.22 at the 1002nd sampling time and 0.25 at the 1003rd sampling time. Since the deviation was greater than 0.2 at two consecutive sampling times, the response start time was determined to be the 1002nd sampling time.

[0255] The instantaneous response deviation value at the 1080th acquisition time was 0.09, at the 1081st it was 0.08, and at the 1082nd it was 0.07. Since the deviation was less than 0.1 for three consecutive times, the response decay time was determined to be the 1082nd acquisition time.

[0256] Step S30343: Based on the event occurrence time corresponding to the event node, the response start time corresponding to the response node, and the response decay time, generate a perturbation holding segment and map the perturbation holding segment to the corresponding factor node.

[0257] In this embodiment, the response start time is taken as the start time of the disturbance holding segment, and the response decay time is taken as the end time of the disturbance holding segment, thereby generating the disturbance holding segment corresponding to the response node. The disturbance holding segment represents the time period during which the disturbance caused by the event corresponding to the event node has a continuous impact on the combustion state.

[0258] Map the perturbation-preserving segment to all factor nodes connected to the response node.

[0259] For example, the disturbance holding period corresponding to the flue gas oxygen content response node is from the 1002nd acquisition time to the 1082nd acquisition time, a total of 81 times.

[0260] The response node is connected to both the air leakage factor node and the heat load factor node, so the disturbance holding segment is mapped to both the air leakage factor node and the heat load factor node.

[0261] Step S30344: Based on the disturbance holding segment, perform time-based weighting on the factor contributions in the cross-suppression contribution set to generate the leakage air mixing contribution, heat load mutation contribution, and air supply disturbance contribution, thus obtaining the disturbance factor contribution set.

[0262] In this embodiment, for each factor node, all its corresponding perturbation-preserving segments are traversed.

[0263] For cases where the current acquisition time is within a certain disturbance holding period, the disturbance holding coefficient is calculated. The disturbance holding coefficient is equal to the difference between the end time of the disturbance holding period and the current acquisition time, divided by the sum of the end time of the disturbance holding period and the start time of the disturbance holding period plus 1. The value range of the disturbance holding coefficient is from 0 to 1, and it gradually decreases over time, representing the degree of residual influence of the disturbance at the current time.

[0264] The factor contribution at the current acquisition time is obtained by multiplying the cross-suppression contribution of the factor node by the perturbation preservation coefficient.

[0265] If the current acquisition time is not within any perturbation hold period, the factor contribution is set to zero.

[0266] After performing time-based weighting on all data collection times, the leakage and mixing contribution sequences, heat load mutation contribution sequences, and air supply disturbance contribution sequences are generated and integrated to obtain the interference factor contribution set.

[0267] For example, the current acquisition time is the 1042nd, which is within the disturbance holding period from 1002 to 1082.

[0268] Dividing the difference between 1082 and 1042 by the sum of 1082 minus 1002 plus 1, we get a disturbance preservation coefficient of approximately 0.49.

[0269] The cross-suppression contribution of the heat load factor node is 20.17. Multiplying 20.17 by 0.49, we get the heat load mutation contribution at the current acquisition time as approximately 9.88.

[0270] The current acquisition time is the 1083rd, which is not within any disturbance holding period, and the contribution of heat load mutation is zero.

[0271] exist Figure 2 In the diagram, the entire schematic diagram includes, from left to right, an event node region, a response node region, and a factor node region. This is used to represent the process in which event nodes are connected to response nodes through effective associated edges, and then the response nodes are mapped to factor nodes, with contribution cross-suppression occurring between factor nodes.

[0272] Figure 2 The solid vertical rounded frame on the left represents the event node area. From top to bottom, the event node area contains the first event node V1, the second event node V2, and the last event node V3. The three circular nodes are used to represent a portion of the event nodes in the event node set, and do not mean that the event node set includes only three event nodes.

[0273] The topmost circle within the vertical rounded frame on the left solid line represents the furnace door opening change event node, i.e., V1, which characterizes the change in furnace air leakage caused by the opening or closing of the furnace door; the middle circle represents the charging status change event node, i.e., V2, which characterizes the change in furnace heat load caused by the charging operation; and the bottommost circle represents the combustion air execution deviation event node, i.e., V3, which characterizes the deviation between the combustion air control command and the actual air supply execution status.

[0274] The vertical ellipsis within the solid-lined, rounded-corner box on the left, located between the second event node V2 and the last event node V3, indicates... Figure 2Other event nodes are omitted from the display. These other event nodes include slag removal status change event nodes, heat storage reversal status change event nodes, flue gas negative pressure change event nodes, furnace pressure change event nodes, combustion air control response event nodes, and gas extraction mixing volume event nodes.

[0275] Figure 2 The vertical rounded frame in the middle represents the response node area. From top to bottom, the response node area contains the first response node V4, the second response node V5, and the last response node V6. The three circular nodes are used to represent a portion of the response nodes in the response node set.

[0276] The topmost circle within the vertical rounded frame of the dashed line represents the response node for changes in flue gas oxygen content, i.e., V4. The waveform inside the circle, which first increases and then decreases, indicates that the flue gas oxygen content increases and gradually recovers after the event. The middle circle represents the response node for changes in flue gas carbon monoxide content, i.e., V5. The step-shaped waveform inside the circle indicates that the flue gas carbon monoxide content undergoes a continuous change after the event. The bottommost circle represents the response node for changes in combustion air flow, i.e., V6. The waveform inside the circle, which first decreases and then recovers, indicates that the combustion air flow decreases and gradually recovers after the event.

[0277] The waveforms inside the circles of each response node are only used to distinguish different response nodes and represent different forms of timing response trends, and do not imply that each response node must have [certain characteristics]. Figure 2 The fixed waveform shown.

[0278] The vertical ellipsis within the dashed vertical rounded frame between the second response node V5 and the last response node V6 indicates... Figure 2 Other response nodes are omitted from the display, including furnace temperature change response nodes and fuel flow rate change response nodes.

[0279] The solid arrow between the event node area on the left and the response node area in the middle represents the valid association edges retained after filtering by response association strength. The starting point of the solid arrow corresponds to the event node, and the ending point of the solid arrow corresponds to the response node. The direction of the solid arrow indicates that event disturbance and response association information are transmitted from the event node to the response node.

[0280] The horizontal solid arrow pointing from the furnace door opening change event node V1 to the flue gas oxygen content change response node V4, i.e., SH1, indicates that there is an effective correlation between the furnace door opening change and the flue gas oxygen content change. The entry of outside air into the furnace caused by the opening of the furnace door may cause a change in the flue gas oxygen content.

[0281] The slanted solid arrow pointing from the event node V1 of the change in furnace door opening to the response node V5 of the change in flue gas carbon monoxide content, i.e., SH2, indicates that there is an effective correlation between the change in furnace door opening and the change in flue gas carbon monoxide content. The furnace airflow disturbance caused by the change in furnace door opening may cause a change in the flue gas carbon monoxide content.

[0282] The slanted solid arrow pointing from the feeding status change event node V2 to the flue gas oxygen content change response node V4, i.e., SH3, indicates that there is an effective correlation between the feeding status change and the flue gas oxygen content change. The change in heat load caused by the feeding operation may change the flue gas oxygen content.

[0283] The near-horizontal solid arrow pointing from the feed status change event node V2 to the flue gas carbon monoxide content change response node V5, i.e., SH4, indicates that there is an effective correlation between the feed status change and the flue gas carbon monoxide content change. The sudden change in heat load caused by the addition of cold feed may cause incomplete combustion and change the flue gas carbon monoxide content.

[0284] The diagonal solid arrow pointing from the fuel feeding status change event node V2 to the combustion air flow change response node V6, i.e., SH5, indicates that there is a valid correlation between the fuel feeding status change and the combustion air flow change, and the combustion control process after fuel feeding may adjust the combustion air supply accordingly.

[0285] The slanted solid arrow pointing from combustion air execution deviation event node V3 to flue gas carbon monoxide content change response node V5, i.e., SH6, indicates that there is a valid correlation between combustion air execution deviation and flue gas carbon monoxide content change. Insufficient actual air supply or delayed response may cause flue gas carbon monoxide content to increase.

[0286] The horizontal solid arrow pointing from the combustion air execution deviation event node V3 to the combustion air flow change response node V6, i.e., SH7, indicates that there is a direct and effective correlation between the combustion air execution deviation and the combustion air flow change, which is used to reflect the actual flow response of the combustion air actuator.

[0287] A node combination that does not have a solid arrow drawn between the event node and the response node indicates that the response association strength of the corresponding candidate association edge has not reached the response association strength threshold, or that the corresponding association relationship has not been retained as a valid association edge.

[0288] Figure 2 The solid vertical rounded frame on the right represents the factor node region. From top to bottom, the factor node region contains the first factor node V7, the second factor node V8, and the last factor node V9. The three circular nodes are used to represent a portion of the factor nodes in the factor node set.

[0289] The topmost circle within the solid vertical rounded frame on the right represents the air leakage and mixing factor node, namely V7, which is used to collect the contributions of changes in furnace door opening, flue gas negative pressure, and furnace pressure to the air leakage and mixing generated by flue gas feedback. The bar chart inside the circle, which gradually decreases from left to right, indicates that the contribution of different effective associated edges to the air leakage and mixing factor node varies.

[0290] The circle in the middle of the vertical rounded frame on the right represents the heat load mutation factor node, namely V8, which is used to collect the contribution of the heat load mutation caused by the changes in feeding state and slag removal state to the flue gas feedback. The bar chart inside the circle, with the middle being higher and the sides being lower, indicates that the various contributions received by the heat load mutation factor node have different relative magnitudes.

[0291] The bottom circle within the solid vertical rounded frame on the right represents the air supply disturbance factor node, namely V9, which is used to collect the contribution of combustion air execution deviation to the air supply disturbance caused by flue gas feedback and combustion air flow. The bar chart inside the circle, which gradually increases from left to right, shows that the different contributions received by the air supply disturbance factor node have an increasing schematic distribution.

[0292] The three factor nodes use different types of bar charts to distinguish the contribution structure corresponding to different factor nodes. The height of the bars only represents the relative difference between different contribution amounts, and does not mean that the actual contribution value must change according to the proportion in the chart.

[0293] The vertical ellipsis within the solid-lined, rounded-corner box on the right, located between the second factor node V8 and the last factor node V9, indicates... Figure 2 Other factor nodes, including the air-fuel ratio factor node, are omitted from the display.

[0294] The dashed arrow between the middle response node region and the right factor node region represents the mapping relationship from the response node to the factor node. The starting point of the dashed arrow corresponds to the response node, and the ending point of the dashed arrow corresponds to the factor node. This indicates that the window response difference of the response node and the response association strength of the corresponding effective associated edge are passed to the corresponding factor node.

[0295] The horizontal dashed arrow pointing from the flue gas oxygen content change response node V4 to the air leakage mixing factor node V7, i.e., XU1, indicates that the flue gas oxygen content change can be used as a response basis for identifying the contribution of air leakage mixing.

[0296] The downward-sloping dashed arrow pointing from the flue gas oxygen content change response node V4 to the heat load mutation factor node V8, i.e., XU2, indicates that the flue gas oxygen content change may also be caused by heat load mutations due to feeding or slag removal. Therefore, the flue gas oxygen content change response can be mapped to the heat load mutation factor node.

[0297] The downward-sloping dashed arrow pointing from the flue gas oxygen content change response node V4 to the air supply disturbance factor node V9, i.e., XU3, indicates that abnormal combustion air supply may also cause changes in flue gas oxygen content. Therefore, the flue gas oxygen content change response can be mapped to the air supply disturbance factor node.

[0298] The upward-sloping dashed arrow pointing from the flue gas carbon monoxide content change response node V5 to the air leakage mixing factor node V7, i.e., XU4, indicates that furnace air leakage and changes in airflow path may alter the flue gas carbon monoxide content. Therefore, the flue gas carbon monoxide content change response can be mapped to the air leakage mixing factor node.

[0299] The horizontal dashed arrow pointing from the flue gas carbon monoxide content change response node V5 to the heat load mutation factor node V8, i.e., XU5, indicates that the change in flue gas carbon monoxide content can be used as a response basis for identifying the contribution of heat load mutation.

[0300] The downward-sloping dashed arrow pointing from the flue gas carbon monoxide content change response node V5 to the air supply disturbance factor node V9, i.e., XU6, indicates that abnormal combustion air execution may cause incomplete combustion and change the flue gas carbon monoxide content. Therefore, the flue gas carbon monoxide content change response can be mapped to the air supply disturbance factor node.

[0301] The upward-sloping dashed arrow pointing from the combustion air flow change response node V6 to the heat load mutation factor node V8, i.e., XU7, indicates that heat load changes may cause the combustion control system to adjust the combustion air flow. Therefore, the combustion air flow change response can be mapped to the heat load mutation factor node.

[0302] The horizontal dashed arrow pointing from the combustion air flow change response node V6 to the air supply disturbance factor node V9, i.e., XU8, indicates that the combustion air flow change can be directly used as the response basis for identifying the contribution of air supply disturbance.

[0303] A node combination for which no straight dashed arrow is drawn between the response node and the factor node indicates that the response node is not mapped to the corresponding factor node, or that the corresponding mapping relationship does not meet the association type classification conditions.

[0304] The right-side factor nodes are connected by arc-shaped dashed arrows labeled W1, W2, W3, and W4, which are used to indicate the response overlap and cross-inhibition relationships formed by different factor nodes due to sharing response nodes. The arrowheads of the arc-shaped dashed arrows point to the factor nodes whose contributions have been deducted or corrected.

[0305] The inner bidirectional arc-shaped dashed arrow marked W1 connects the air leakage mixing factor node V7 and the heat load mutation factor node V8, indicating that V7 and V8 form a response overlap relationship because they receive the mapping contribution of the same response node. The arrow pointing from V7 to V8 indicates the part of the heat load mutation contribution that is suppressed from overlapping with the air leakage mixing contribution, and the arrow pointing from V8 to V7 indicates the part of the air leakage mixing contribution that is suppressed from overlapping with the heat load mutation contribution.

[0306] The double-curved dashed arrow on the outer side of W2 connects the air leakage mixing factor node V7 and the air supply disturbance factor node V9, indicating that V7 and V9 form a response overlap relationship because they receive the mapping contribution of the same response node. The arrow pointing from V7 to V9 indicates the part of the air supply disturbance contribution that overlaps with the air leakage mixing contribution, and the arrow pointing from V9 to V7 indicates the part of the air leakage mixing contribution that overlaps with the air supply disturbance contribution.

[0307] The inner bidirectional arc-shaped dashed arrow marked W3 connects the heat load mutation factor node V8 and the air supply disturbance factor node V9, indicating that V8 and V9 form a first response overlap relationship; the arrow pointing from V8 to V9 indicates the part of the air supply disturbance contribution that is suppressed from overlapping with the heat load mutation contribution, and the arrow pointing from V9 to V8 indicates the part of the heat load mutation contribution that is suppressed from overlapping with the air supply disturbance contribution.

[0308] The double-curved dashed arrow marked W4 connects the heat load mutation factor node V8 and the air supply disturbance factor node V9, indicating that another response overlap relationship is formed between V8 and V9; the arrow pointing from V8 to V9 indicates the suppression of the corresponding repeated contribution from the air supply disturbance contribution, and the arrow pointing from V9 to V8 indicates the suppression of the corresponding repeated contribution from the heat load mutation contribution.

[0309] W3 and W4 represent two sets of response overlap relationships formed by heat load mutation factor node V8 and air supply disturbance factor node V9 based on different shared response nodes, respectively, and are used to distinguish the repeated contributions generated by different shared response sources.

[0310] The cross suppression shown by the arc-shaped dashed arrows W1 to W4 is not simply a one-way deduction of one factor node from another, but rather the overlapping deduction share of each factor node is determined based on the initial contribution ratio obtained by each factor node through the corresponding shared response node.

[0311] Figure 2 When solid or dashed lines intersect in a blank area, no circular node or connection marker is set at the intersection. Therefore, the intersection only indicates that the drawing paths intersect on the plane, and does not indicate that two related edges form a new connection relationship at the intersection.

[0312] The solid rounded corner boxes on the left and right sides, and the dashed rounded corner box in the middle, are only used to divide different types of nodes into regions and do not belong to the nodes or associated edges in the event anchoring factor diagram; the middle rounded corner box is drawn with a dashed line and is only used to visually distinguish the response node region, and does not indicate that the response node or response data has uncertainty.

[0313] Figure 2 The arrowheads indicate the direction of information, response differences, or factor contributions. Solid arrows SH1 to SH7 represent the effective correlation between event nodes and response nodes. Straight dashed arrows XU1 to XU8 represent the mapping relationship between response nodes and factor nodes. Arc-shaped dashed arrows W1 to W4 represent the response overlap and cross-inhibition relationship between factor nodes. Step S304: Based on the combustion process time-series dataset, generate original air-fuel ratio deviation evidence. Combined with the interference factor contribution set, perform stratified subtraction on the original air-fuel ratio deviation evidence to obtain the true air-fuel ratio deviation factor.

[0314] The specific steps of step S304 are as follows:

[0315] Step S3041: Based on the combustion process time series dataset, extract the data changes in flue gas oxygen content, flue gas carbon monoxide content, and combustion air supply data to generate original air-fuel ratio deviation evidence.

[0316] In this embodiment, the normalized change in oxygen content of flue gas, the normalized change in carbon monoxide content of flue gas, the normalized change in combustion air flow rate of combustion air supply data, and the normalized change in fuel flow rate of fuel supply data are extracted from the combustion process time series dataset.

[0317] The larger of the change in carbon monoxide content in flue gas and zero is calculated to obtain the carbon monoxide anomalous evidence component, which reflects the degree of incomplete combustion.

[0318] Calculate the absolute value of the change in oxygen content in flue gas to obtain the evidence component of abnormal oxygen content, which reflects the degree of deviation in oxygen content in flue gas caused by excessive air or air leakage.

[0319] The absolute value of the difference between the normalized change in combustion air flow rate and the normalized change in fuel flow rate is calculated to obtain the air supply matching evidence component, which reflects the degree of matching between fuel and combustion air inputs.

[0320] The original air-fuel ratio deviation evidence is obtained by multiplying the carbon monoxide abnormality evidence component, the oxygen content abnormality evidence component, and the air supply matching evidence component by their respective weighting coefficients and then adding them together.

[0321] The weighting coefficients were determined through orthogonal experiments. The factors in the orthogonal experiments included the weighting of evidence components for abnormal carbon monoxide, abnormal oxygen content, and air supply matching. Each factor was set with three levels: 0.2, 0.3, and 0.4. The sum of the three weighting coefficients was required to be 1.

[0322] Using the manually confirmed air-fuel ratio deviation direction and furnace temperature recovery time from historical operating data as evaluation criteria, the deviation direction identification accuracy and furnace temperature recovery time under different weight combinations were calculated respectively. The weight combination with the highest deviation direction identification accuracy and the shortest furnace temperature recovery time was selected as the weight coefficient of the original air-fuel ratio deviation evidence.

[0323] For example, the normalized change in carbon monoxide in the flue gas at the 1001st acquisition time was -0.40, and the carbon monoxide anomalous evidence component was 0.

[0324] The normalized change in oxygen content in flue gas was 1.50, and the evidence component for abnormal oxygen content was 1.50.

[0325] The normalized change in combustion air flow rate is 1.00, the normalized change in fuel flow rate is 1.00, and the evidence component for air supply matching is the absolute value of 1.00 minus 1.00, which results in 0.

[0326] After comparing different weight combinations using orthogonal experiments, the accuracy of identifying the deviation direction in historical samples was highest when the weight of the abnormal evidence component of carbon monoxide was 0.4, the weight of the abnormal evidence component of oxygen content was 0.3, and the weight of the air supply matching evidence component was 0.3, and the accuracy of the furnace temperature recovery time was shorter. Therefore, the above three weight coefficients were set to 0.4, 0.3, and 0.3, respectively.

[0327] Calculate 0.4 multiplied by 0, add 0.3 multiplied by 1.50, add 0.3 multiplied by 0, and the original air-fuel ratio deviation evidence is 0.45.

[0328] Step S3042: Based on the contributions of air leakage and mixing, heat load mutation, and air supply disturbance, determine the abnormal deduction amount for flue gas oxygen content, abnormal deduction amount for flue gas carbon monoxide content, and abnormal deduction amount for combustion air supply, respectively.

[0329] In this embodiment, the normalized value of the air leakage contribution is calculated by dividing the air leakage contribution at the current moment by the sum of the maximum air leakage contribution at all times and 0.001.

[0330] Air leakage and mixing can affect the oxygen content of flue gas, causing it to rise and leading to a misjudgment of an excessively high air-fuel ratio.

[0331] The normalized value of the contribution of air leakage and mixing is multiplied by the weighted evidence corresponding to the abnormal oxygen content evidence component in the original air-fuel ratio deviation evidence to obtain the abnormal oxygen content deduction in the flue gas.

[0332] For example, the contribution of air leakage and gas mixing at the 1001st acquisition time is 12.99, the maximum contribution of air leakage and gas mixing at all times is 25.00, and the normalized value is approximately 0.52.

[0333] Multiplying 0.3 by 1.50 yields a weighted evidence of 0.45 for the oxygen content abnormality evidence component in the original air-fuel ratio deviation evidence.

[0334] Multiplying 0.45 by 0.52, we get an abnormal reduction of approximately 0.23 in the oxygen content of the flue gas.

[0335] Step S3043: Generate the total deduction amount based on the abnormal deduction amount of flue gas oxygen content, the abnormal deduction amount of flue gas carbon monoxide content, and the abnormal deduction amount of combustion air supply.

[0336] In this embodiment, the normalized value of the heat load mutation contribution is calculated by dividing the heat load mutation contribution at the current moment by the maximum value of the heat load mutation contribution at all times plus 0.001.

[0337] Sudden changes in heat load can affect the carbon monoxide content in flue gas, causing it to rise and leading to a misjudgment of an excessively low air-fuel ratio.

[0338] The normalized value of the contribution of the heat load mutation is multiplied by the weighted evidence corresponding to the carbon monoxide abnormality evidence component in the original air-fuel ratio deviation evidence to obtain the abnormal deduction of flue gas carbon monoxide content.

[0339] For example, the heat load mutation contribution at the 1001st acquisition time is 20.07, the maximum heat load mutation contribution at all times is 40.00, and the normalized value is approximately 0.50.

[0340] Multiplying 0.4 by 0 yields a weighted evidence of 0 corresponding to the carbon monoxide anomaly evidence component in the original air-fuel ratio deviation evidence.

[0341] The abnormal deduction for carbon monoxide content in flue gas is 0 multiplied by 0.50, resulting in 0.

[0342] Step S3044: Based on the total deduction, the original air-fuel ratio deviation evidence is reduced by a limit to obtain the true air-fuel ratio deviation factor.

[0343] In this embodiment, the normalized value of the air supply disturbance contribution is calculated by dividing the current air supply disturbance contribution by the maximum value of the air supply disturbance contribution at all times plus 0.001.

[0344] Disturbances in the air supply can affect the combustion air flow, causing discrepancies between the actual air supply and the control command, and resulting in misjudgments of the air-fuel ratio deviation.

[0345] The normalized value of the contribution of air supply disturbance is multiplied by the weighted evidence corresponding to the air supply matching evidence component in the original air-fuel ratio deviation evidence to obtain the abnormal reduction in combustion air supply.

[0346] The total deduction is obtained by adding the deductions for abnormal oxygen content in flue gas, abnormal carbon monoxide content in flue gas, and abnormal combustion air supply.

[0347] When the total deduction exceeds the upper limit of the original air-fuel ratio deviation evidence, the total deduction is corrected to the upper limit. The upper limit is the original air-fuel ratio deviation evidence multiplied by the upper limit deduction ratio. The upper limit deduction ratio is set based on the accuracy of air-fuel ratio deviation direction identification and furnace temperature recovery time under historical disturbance conditions. Specifically, multiple candidate upper limit deduction ratios are set in the range of 0.8 to 0.95 at intervals of 0.05. Based on historical operating data, each candidate upper limit deduction ratio is used for limited deduction, and the corresponding accuracy of air-fuel ratio deviation direction identification and furnace temperature recovery time are calculated. The candidate upper limit deduction ratio with the highest accuracy of air-fuel ratio deviation direction identification is selected as the final upper limit deduction ratio. When there are multiple candidate upper limit deduction ratios with the same accuracy of air-fuel ratio deviation direction identification, the candidate upper limit deduction ratio with the shortest furnace temperature recovery time is selected as the final upper limit deduction ratio. The upper limit deduction ratio is used to retain the true air-fuel ratio deviation evidence that has not been explained by interference factors.

[0348] The true air-fuel ratio deviation factor is obtained by subtracting the corrected total deduction from the original air-fuel ratio deviation evidence.

[0349] For example, the contribution of air supply disturbance at the 1001st acquisition time is 3.10, the maximum contribution of air supply disturbance at all times is 10.00, and the normalized value is about 0.31.

[0350] Multiplying 0.3 by 0, the weighted evidence corresponding to the air supply matching evidence component in the original air-fuel ratio deviation evidence becomes 0.

[0351] The abnormal reduction in combustion air supply is calculated by multiplying 0 by 0.31, resulting in 0.

[0352] The abnormal deduction for flue gas oxygen content is 0.23, the abnormal deduction for flue gas carbon monoxide content is 0, the abnormal deduction for combustion air supply is 0, and the total deduction is 0.23.

[0353] The original air-fuel ratio deviation evidence is 0.45. The upper limit of the deduction ratio is 0.9. Multiplying 0.45 by 0.9, we get the upper limit of the deduction as 0.405.

[0354] Since the total deduction of 0.23 is less than the deduction limit of 0.405, the total deduction remains at 0.23.

[0355] Calculate the difference between 0.45 and 0.23 to obtain the true air-fuel ratio deviation factor as 0.22.

[0356] If the total deduction is 0.50 at a certain moment, the original air-fuel ratio deviation evidence is 0.45, and the upper limit of the deduction is 0.405, then the total deduction is corrected to 0.405. Subtracting 0.405 from 0.45 yields a true air-fuel ratio deviation factor of 0.045.

[0357] Step S305: Based on the true air-fuel ratio deviation factor, perform time-series deviation organization and flue gas feedback reliability assessment to obtain the true air-fuel ratio deviation sequence and the flue gas feedback reliability sequence.

[0358] In this embodiment, the true air-fuel ratio deviation factor at each acquisition time is arranged in chronological order to obtain a preliminary true air-fuel ratio deviation sequence.

[0359] Determine the direction of air-fuel ratio deviation at each moment, and set thresholds for carbon monoxide increase, oxygen content decrease, oxygen content increase, and carbon monoxide decrease.

[0360] The thresholds are all determined based on statistical data under historical stable combustion conditions. Specifically, the standard deviations of the normalized change in carbon monoxide and the normalized change in oxygen content in flue gas under normal combustion conditions over the past 100 hours are calculated, and one time the standard deviation is taken as the corresponding threshold. The absolute values ​​of the carbon monoxide increase threshold and the carbon monoxide decrease threshold are equal, and the absolute values ​​of the oxygen content increase threshold and the oxygen content decrease threshold are equal.

[0361] When the normalized change in carbon monoxide in flue gas is greater than the carbon monoxide increase threshold, and the normalized change in oxygen content in flue gas is less than the oxygen decrease threshold, the flue gas composition is determined to be insufficient air.

[0362] When the normalized change in oxygen content in flue gas is greater than the oxygen content increase threshold and the normalized change in carbon monoxide content in flue gas is less than the carbon monoxide decrease threshold, the flue gas composition is determined to be excessive air.

[0363] Based on the normalized change in fuel flow rate and the normalized change in combustion air flow rate, the direction of change in the air-fuel ratio on the supply side is calculated. If the normalized change in combustion air flow rate decreases relative to the normalized change in fuel flow rate, the supply side determines that there is insufficient air. If the normalized change in combustion air flow rate increases relative to the normalized change in fuel flow rate, the supply side determines that there is excess air.

[0364] When the flue gas composition determination result is consistent with the supply-side determination result, the consistent determination result is taken as the deviation direction at the current sampling time. Insufficient air corresponds to a deviation direction of +1, and excessive air corresponds to a deviation direction of -1.

[0365] When the flue gas composition determination result is inconsistent with the supply-side determination result, the flue gas feedback reliability at the current collection time is compared with the reliability threshold. If the flue gas feedback reliability is greater than the reliability threshold, the flue gas composition determination result is adopted. If the flue gas feedback reliability is less than or equal to the reliability threshold, the supply-side determination result is adopted. If neither the flue gas composition determination result nor the supply-side determination result meets the direction determination condition, the deviation direction is set to 0.

[0366] The credibility threshold is determined based on historical operational data, with a preferred value of 0.5 to 0.7.

[0367] Multiply the true air-fuel ratio deviation factor by the deviation direction to obtain the true air-fuel ratio deviation amount.

[0368] The reliability of flue gas feedback equals the actual air-fuel ratio deviation factor, plus the stable feedback retention coefficient, multiplied by 1 minus the normalized value of the contribution of air leakage and mixing, multiplied by 1 minus the normalized value of the contribution of heat load mutation, and multiplied by 1 minus the normalized value of the contribution of air supply disturbance.

[0369] The stability feedback retention coefficient was determined through verification using historical disturbance samples. Specifically, multiple candidate values ​​were selected within the range of 0.5 to 0.7. Each candidate value was substituted into the historical disturbance samples for replay calculation. The number of samples where the valve adjustment direction was inconsistent with the manually confirmed correct adjustment direction was counted and divided by the total number of historical disturbance samples to obtain the valve adjustment direction error rate. The time elapsed from the moment the disturbance occurred until the furnace temperature recovered and remained continuously within the allowable deviation range of the target temperature was counted to obtain the furnace temperature recovery time. The candidate value with the lowest valve adjustment direction error rate and the shortest furnace temperature recovery time was selected as the stability feedback retention coefficient.

[0370] The calculated flue gas feedback reliability is limited to between 0 and 1.

[0371] The reliability of flue gas feedback is subjected to first-order exponential smoothing. The smoothing coefficient is determined through historical operating data. Specifically, multiple candidate values ​​are selected within the range of 0.6 to 0.8, and each candidate value is used to perform first-order exponential smoothing on the historical flue gas feedback reliability. The average absolute value of the difference in reliability of smoothed flue gas feedback between adjacent acquisition times is calculated to obtain the reliability fluctuation range. The times when the reliability of unsmoothed flue gas feedback first exceeds the reliability threshold and the times when the reliability of smoothed flue gas feedback first exceeds the reliability threshold in the same direction are collected. The time difference between the times when the reliability of unsmoothed flue gas feedback first exceeds the reliability threshold and the times when the reliability of smoothed flue gas feedback first exceeds the reliability threshold in the same direction is taken as the control response lag time. Candidate values ​​with small reliability fluctuation range and control response lag time not exceeding the allowable range are selected as smoothing coefficients. The allowable range is set according to the response time of fuel valve and combustion air valve. Specifically, the smaller value between the response time of fuel valve and the response time of combustion air valve is taken as the allowable upper limit of control response lag time, and zero to the allowable upper limit is determined as the allowable range.

[0372] Arrange the smoothed flue gas feedback credibility in chronological order to obtain the flue gas feedback credibility sequence.

[0373] Arrange the actual air-fuel ratio deviations in chronological order to obtain the final actual air-fuel ratio deviation sequence.

[0374] For example, the standard deviation of the normalized change in carbon monoxide in flue gas under historically stable combustion conditions is 0.2, and the standard deviation of the normalized change in oxygen content in flue gas is 0.2.

[0375] The threshold for increasing carbon monoxide is set to 0.2, the threshold for decreasing carbon monoxide is set to -0.2, the threshold for increasing oxygen content is set to 0.2, and the threshold for decreasing oxygen content is set to -0.2.

[0376] The credibility threshold is set to 0.6, the stable feedback retention coefficient is set to 0.6, and the smoothing coefficient is set to 0.7.

[0377] At the 1001st sampling time, the normalized change of carbon monoxide in the flue gas was -0.40, which is less than -0.2, and the normalized change of oxygen content in the flue gas was 1.50, which is greater than 0.2. Therefore, the flue gas composition was determined to be excessive air.

[0378] The normalized change in combustion air flow rate at the 1001st data collection time was 1.00, and the normalized change in fuel flow rate was 1.00. The difference between the two was 0. Therefore, the supply-side judgment result did not meet the judgment conditions for insufficient or excessive air.

[0379] Since the flue gas composition determination result indicates an excess of air, and the supply-side determination result does not meet the direction determination condition, the flue gas composition determination result is adopted, and the deviation direction is taken as negative 1.

[0380] The true air-fuel ratio deviation factor is 0.22. Multiplying 0.22 by -1 gives a true air-fuel ratio deviation of -0.22.

[0381] If the flue gas composition determination result at another sampling time is insufficient air, the supply side determination result is excessive air, and the current flue gas feedback confidence is 0.75, which is greater than the confidence threshold of 0.6, then the flue gas composition determination result is adopted, and the deviation direction is positive 1.

[0382] If the current flue gas feedback confidence level is 0.40, which is less than the confidence level threshold of 0.6, then the supply-side judgment result is adopted, and the deviation direction is negative 1.

[0383] When calculating the reliability of flue gas feedback, the normalized value of the contribution of air leakage and mixing at the 1001st acquisition time was 0.52, the normalized value of the contribution of heat load change was 0.50, and the normalized value of the contribution of air supply disturbance was 0.31.

[0384] The reliability of the flue gas feedback is 0.22 plus 0.6 multiplied by 1 minus 0.52 multiplied by 1 minus 0.50 multiplied by 1 minus 0.31, which is 0.22 plus 0.6 multiplied by 0.48 multiplied by 0.50 multiplied by 0.69. Adding 0.10 to 0.22, we get the final result of the flue gas feedback reliability as 0.32.

[0385] Even after limiting the range to between 0 and 1, the reliability of the flue gas feedback remains at 0.32.

[0386] The smoothing confidence level of the previous time step was 0.30. Multiplying 0.7 by 0.30 and adding 0.3 by 0.32, we get a smoothed confidence level of 0.31.

[0387] The true air-fuel ratio deviation and flue gas feedback reliability were calculated in the same way for all 1000 data collection times, resulting in a true air-fuel ratio deviation sequence and a flue gas feedback reliability sequence containing data from 1000 times.

[0388] Step S4: Based on the true air-fuel ratio deviation sequence and the flue gas feedback reliability sequence, generate the fuel valve opening adjustment amount and the combustion air valve opening adjustment amount to obtain the smelting combustion process control results.

[0389] In this embodiment, the valve opening adjustment amount is generated by a rolling predictive control method based on the time series dataset of the smelting and combustion process, the flue gas feedback reliability sequence, and the actual air-fuel ratio deviation sequence.

[0390] Obtain the opening degree of the fuel valve and the combustion air valve in the current control cycle.

[0391] Construct a set of candidate fuel valve opening adjustment values ​​and a set of candidate combustion air valve opening adjustment values.

[0392] The range of the candidate adjustment set is determined based on the maximum adjustment speed of the valve and the stability requirements of the combustion process. If the range is too large, it will cause the adjustment to be too abrupt and cause combustion fluctuations. If the range is too small, the adjustment speed will be too slow and the deviation cannot be corrected in time. The step size is determined based on the control accuracy requirements. If the step size is too large, it will lead to insufficient control accuracy. If the step size is too small, it will increase the amount of calculation and have limited improvement on the control effect.

[0393] Each element in the candidate fuel valve opening adjustment set is paired with each element in the candidate combustion air valve opening adjustment set to obtain a candidate control quantity set. Each candidate control quantity consists of one fuel valve opening adjustment quantity and one combustion air valve opening adjustment quantity.

[0394] For each candidate control variable in the candidate control variable set, predict the furnace temperature, flue gas oxygen content, and flue gas carbon monoxide content for the next control cycle.

[0395] For each candidate control variable in the candidate control variable set, extract the current furnace temperature, candidate fuel valve opening adjustment, candidate combustion air valve opening adjustment, current furnace temperature change, current true air-fuel ratio deviation, and current heat load mutation contribution. Multiply the candidate fuel valve opening adjustment, candidate combustion air valve opening adjustment, current furnace temperature change, current true air-fuel ratio deviation, and current heat load mutation contribution by their respective furnace temperature prediction coefficients and then sum them to obtain the furnace temperature prediction change for the next control cycle. Add the current furnace temperature to the furnace temperature prediction change to obtain the furnace temperature prediction value.

[0396] Extract the current flue gas oxygen content, the adjustment amount of the candidate combustion air valve opening, the adjustment amount of the candidate fuel valve opening, the change in current flue gas oxygen content, the current actual air-fuel ratio deviation, and the current contribution of air leakage. Multiply the adjustment amount of the candidate combustion air valve opening, the adjustment amount of the candidate fuel valve opening, the change in current flue gas oxygen content, the current actual air-fuel ratio deviation, and the current contribution of air leakage by their respective flue gas oxygen content prediction coefficients, and then add them together to obtain the predicted change in flue gas oxygen content for the next control cycle. Add the current flue gas oxygen content to the predicted change in flue gas oxygen content to obtain the predicted value of flue gas oxygen content.

[0397] Extract the current flue gas carbon monoxide content, candidate fuel valve opening adjustment amount, candidate combustion air valve opening adjustment amount, current flue gas carbon monoxide content change amount, current true air-fuel ratio deviation amount, and current heat load mutation contribution amount. Multiply the candidate fuel valve opening adjustment amount, candidate combustion air valve opening adjustment amount, current flue gas carbon monoxide content change amount, current true air-fuel ratio deviation amount, and current heat load mutation contribution amount by the corresponding flue gas carbon monoxide content prediction coefficient, and then add them to obtain the predicted flue gas carbon monoxide content change amount for the next control cycle. Add the current flue gas carbon monoxide content to the predicted flue gas carbon monoxide content change amount to obtain the predicted flue gas carbon monoxide content value.

[0398] The predicted coefficients for furnace temperature, flue gas oxygen content, and flue gas carbon monoxide content are determined based on historical operating data. Specifically, the corresponding input quantity for each control cycle in the historical operating data is used as the independent variable, and the actual change of the corresponding data for the next control cycle relative to the current control cycle is used as the dependent variable. The prediction coefficients corresponding to each input quantity are solved using the least squares method.

[0399] Obtain the current aluminum alloy grade and smelting stage. Based on the current aluminum alloy grade and smelting stage, obtain the allowable range of furnace temperature, allowable range of flue gas oxygen content, and allowable upper limit of flue gas carbon monoxide content specified in the corresponding process specification. Use the median value of the allowable furnace temperature range as the target furnace temperature value. Screen historical stable combustion segments with the same aluminum alloy grade and smelting stage where the furnace temperature is within the allowable range, the flue gas carbon monoxide content does not exceed the allowable upper limit, and the smelting quality is qualified. Calculate the median of the flue gas oxygen content in the historical stable combustion segments to obtain the target value of flue gas oxygen content. Calculate the median of the flue gas carbon monoxide content in the historical stable combustion segments to obtain the target value of flue gas carbon monoxide content. When the target value of flue gas oxygen content exceeds the allowable range, correct it to the corresponding allowable range boundary. When the target value of flue gas carbon monoxide content exceeds the allowable upper limit of flue gas carbon monoxide content, correct it to the allowable upper limit of flue gas carbon monoxide content.

[0400] The flue gas deviation is corrected based on the reliability of flue gas feedback. The correction deviation for flue gas oxygen content is equal to the reliability of flue gas feedback multiplied by the difference between the predicted value and the target value of flue gas oxygen content. The correction deviation for flue gas carbon monoxide content is equal to the reliability of flue gas feedback multiplied by the difference between the predicted value and the target value of flue gas carbon monoxide content.

[0401] Calculate the furnace temperature deviation, which is equal to the difference between the predicted furnace temperature and the target furnace temperature.

[0402] For each candidate control variable, an evaluation function value is calculated. This evaluation function value equals the absolute value of the furnace temperature deviation multiplied by the furnace temperature weighting coefficient, plus the absolute value of the flue gas oxygen content correction deviation multiplied by the flue gas oxygen content weighting coefficient, plus the absolute value of the flue gas carbon monoxide content correction deviation multiplied by the flue gas carbon monoxide content weighting coefficient, plus the absolute value of the current true air-fuel ratio deviation multiplied by the air-fuel ratio weighting coefficient, plus the sum of the absolute values ​​of the fuel valve opening adjustment and the combustion air valve opening adjustment multiplied by the adjustment amplitude weighting coefficient. Each weighting coefficient is determined through orthogonal experiments based on process requirements and is used to balance the importance of different control objectives. The furnace temperature weighting coefficient is usually the largest because furnace temperature is the most critical factor affecting smelting quality; the adjustment amplitude weighting coefficient is used to limit frequent valve adjustments and extend the service life of the actuator.

[0403] The candidate control variable with the smallest evaluation function value is selected as the optimal control variable. The smaller the evaluation function value, the better the overall effect of the control scheme in satisfying all control objectives.

[0404] The fuel valve opening adjustment in the optimal control quantity is used as the fuel valve opening adjustment in the current control cycle, and the combustion air valve opening adjustment in the optimal control quantity is used as the combustion air valve opening adjustment in the current control cycle.

[0405] Update the fuel valve opening and combustion air valve opening. The updated fuel valve opening is equal to the current fuel valve opening plus the fuel valve opening adjustment amount, and is limited to the minimum and maximum opening of the fuel valve.

[0406] The updated combustion air valve opening is equal to the current combustion air valve opening plus the combustion air valve opening adjustment amount, and is limited to the minimum and maximum opening of the combustion air valve.

[0407] The process controller sends the fuel valve opening adjustment amount to the fuel valve actuator and the combustion air valve opening adjustment amount to the combustion air valve actuator.

[0408] The fuel valve actuator and the combustion air valve actuator adjust the valve opening according to the received adjustment amount to control the smelting and combustion process and obtain the smelting and combustion process control result.

[0409] Example 2

[0410] See Figure 3 This embodiment provides a smelting combustion process control system based on adaptive air-fuel ratio, including: a data acquisition module, a timing construction module, a deviation evaluation module, and a predictive control module.

[0411] The data acquisition module is used to collect data on the smelting and combustion process.

[0412] The timing construction module performs data timing alignment and disturbance response feature extraction based on the smelting and combustion process data to obtain a combustion process timing dataset.

[0413] The deviation assessment module constructs an event anchoring factor graph based on the combustion process time series dataset, and performs interference factor contribution deduction and flue gas feedback credibility assessment to obtain the true air-fuel ratio deviation sequence and flue gas feedback credibility sequence.

[0414] The predictive control module generates the adjustment amount of the fuel valve opening and the adjustment amount of the combustion air valve opening based on the actual air-fuel ratio deviation sequence and the flue gas feedback reliability sequence, thereby obtaining the control results of the smelting combustion process.

[0415] The specific functions of each module described above are explained in the relevant content of the method in Embodiment 1, and will not be repeated here.

[0416] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0417] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smelting combustion process control method based on adaptive air-fuel ratio, characterized in that, include: Collect smelting and combustion process data, which includes combustion feedback data, fuel supply data, combustion air supply data, smelting operation event data, and furnace airflow status data. The combustion feedback data includes flue gas oxygen content and flue gas carbon monoxide content. The smelting operation event data includes furnace door opening. Based on the smelting and combustion process data, data time series alignment and disturbance response feature extraction are performed to obtain a combustion process time series dataset; Based on the combustion process time series dataset, an event anchoring factor map is constructed, and the contribution of interference factors is deducted and the credibility of flue gas feedback is evaluated to obtain the true air-fuel ratio deviation sequence and the flue gas feedback credibility sequence. Based on the true air-fuel ratio deviation sequence and the flue gas feedback reliability sequence, the fuel valve opening adjustment amount and the combustion air valve opening adjustment amount are generated to obtain the control results of the smelting combustion process.

2. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 1, characterized in that, The combustion process data, based on smelting and combustion process data, undergoes time-series alignment and perturbation response feature extraction to obtain a combustion process time-series dataset, including: The data in the smelting and combustion process are divided into continuous data and state data. The continuous data is time-aligned and the state data is event-preserving aligned to obtain the time-aligned process dataset. Based on the time-aligned process dataset, the data change amount of each data in the smelting and combustion process data is extracted to obtain the process change amount dataset; Based on the process change dataset, the data change of the combustion air supply data is compared for response consistency and the deviation is extracted to obtain the combustion air control response. Based on the process change dataset, the suction disturbance intensity is calculated by the data changes of furnace airflow state data and furnace door opening to obtain the suction mixed gas volume. Based on the combustion air control response and the amount of gas extracted, the time-series augmentation of the process change dataset is performed to obtain the combustion process time-series dataset.

3. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 1, characterized in that, Based on the combustion process time-series dataset, an event anchoring factor map is constructed, and interference factor contributions are deducted and flue gas feedback credibility is evaluated to obtain the true air-fuel ratio deviation sequence and the flue gas feedback credibility sequence, including: Based on the combustion process time series dataset, an event node set and a response node set are constructed, and the window response difference of the response nodes is calculated to generate candidate association edges; Based on candidate association edges, event node sets, and response node sets, association strength is filtered, association type is classified, and factor node is mapped to construct an event anchoring factor graph. Based on the event anchoring factor graph, the initial contribution of each factor node is calculated along the effective association edge, and the initial contribution of nodes sharing the same response node is cross-suppressed to obtain the interference factor contribution set. Based on the combustion process time series dataset, the original air-fuel ratio deviation evidence is generated, and combined with the interference factor contribution set, the original air-fuel ratio deviation evidence is subtracted in layers to obtain the true air-fuel ratio deviation factor. Based on the true air-fuel ratio deviation factor, time-series deviation organization and flue gas feedback reliability assessment are performed to obtain the true air-fuel ratio deviation sequence and the flue gas feedback reliability sequence.

4. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 3, characterized in that, The process involves constructing an event node set and a response node set based on a combustion process time-series dataset, calculating the window response differences of the response nodes, and generating candidate association edges, including: Based on the combustion process time series dataset, the data change of smelting operation event data, the data change of furnace airflow state data, the combustion air control response and the amount of gas extracted are extracted to construct an event node set; Based on the combustion process time series dataset, the data changes of combustion feedback data, fuel supply data and combustion air supply data are extracted to construct a response node set; Based on the event node set, the event occurrence time corresponding to each event node is identified, and the event occurrence time is used as the anchor point to divide the pre-event baseline window and the post-event response window. Based on the set of response nodes, the baseline response value of each response node in the baseline window before the event and the response change value in the response window after the event are calculated to generate candidate associated edges.

5. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 4, characterized in that, The process of constructing an event anchoring factor graph based on candidate association edges, event node sets, and response node sets includes: filtering association strength, classifying association types, and mapping factor nodes. Based on the baseline response value and the response change value, the response association strength corresponding to each candidate association edge is calculated, and the validity of the candidate association edges is determined to obtain the valid association edges. Based on the effective associated edges, the associated types are divided to generate leakage factor nodes, heat load factor nodes, air-fuel ratio factor nodes and air supply disturbance factor nodes, thus obtaining a set of factor nodes; Based on the event node set, response node set, factor node set, and effective association edges, establish the association relationships from event nodes to response nodes and from response nodes to factor nodes, and construct the event anchoring factor graph.

6. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 3, characterized in that, The event-anchored factor graph is used to calculate the initial contribution of each factor node along the effective association edges, and cross-suppression is applied to the initial contributions of nodes sharing the same response to obtain the interference factor contribution set, including: Based on the event anchoring factor graph, the effective association edges connecting event nodes, response nodes and factor nodes are extracted, and the corresponding event node, response node, factor node and response association strength are labeled for each effective association edge to obtain the attributed association edge set; Based on the attributed association edge set, the data change of the event node, the window response difference of the response node, and the response association strength are passed to the corresponding factor node along the effective association edge to obtain the initial interference contribution set. Based on the attributed association edge set, valid association edges connected to different factor nodes and having the same response node are identified, and factor node response overlap relationships are generated. Based on the initial interference contribution set and the overlapping relationship of factor node responses, cross-suppression is performed on the factor contributions that share the same response node to obtain the interference factor contribution set.

7. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 6, characterized in that, The set of attribute-based association edges is used to propagate the data changes of event nodes, the window response differences of response nodes, and the response association strength along the effective association edges to the corresponding factor nodes, thus obtaining an initial interference contribution set, including: Based on the attributed association edge set, the data change amount of the event node, the window response difference of the response node, the response association strength and factor node corresponding to each effective association edge are extracted to obtain the edge contribution parameter set; Based on the edge contribution parameter set, the edge contribution of each effective associated edge is calculated, and the edge contributions are accumulated to obtain the initial contribution of air leakage and mixing, the initial contribution of heat load change, and the initial contribution of air supply disturbance. Based on the effective associated edges corresponding to the air supply disturbance factor nodes, the follow-up deviation of the combustion air supply data is identified to obtain the air supply execution deviation. The initial contribution of air supply disturbance is corrected based on the air supply execution deviation, and the corrected initial contribution of air supply disturbance, the initial contribution of air leakage and mixing, and the initial contribution of heat load change are aggregated to the corresponding factor nodes to obtain the initial disturbance contribution set.

8. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 6, characterized in that, The method involves cross-suppression of factor contributions sharing the same response node based on the initial interference contribution set and the overlap relationship of factor node responses, resulting in an interference factor contribution set, including: Based on the overlapping relationship of factor node responses, cross-suppression is performed on the factor contributions that share the same response node in the initial interference contribution set to obtain the cross-suppression contribution set; Based on the set of associated edges with attributes, the event nodes, response nodes and factor nodes corresponding to each valid associated edge are extracted, and the response start time and response decay time corresponding to the response node are determined. Based on the event occurrence time corresponding to the event node, the response start time and response decay time corresponding to the response node, a perturbation holding segment is generated and mapped to the corresponding factor node; Based on the disturbance holding segment, the factor contributions in the cross-suppression contribution set are time-based weighted to generate the leakage air mixing contribution, heat load mutation contribution, and air supply disturbance contribution, thus obtaining the disturbance factor contribution set.

9. The smelting combustion process control method based on adaptive air-fuel ratio according to claim 8, characterized in that, The original air-fuel ratio deviation evidence is generated based on the combustion process time-series dataset. Combined with the interference factor contribution set, the original air-fuel ratio deviation evidence is stratified and subtracted to obtain the true air-fuel ratio deviation factor, including: Based on the combustion process time series dataset, the data changes of flue gas oxygen content, flue gas carbon monoxide content and combustion air supply data are extracted to generate original air-fuel ratio deviation evidence. Based on the contributions of air leakage and mixing, heat load mutation, and air supply disturbance, the abnormal deduction amounts for flue gas oxygen content, flue gas carbon monoxide content, and combustion air supply were determined respectively. The total deduction is generated based on the abnormal deductions for flue gas oxygen content, flue gas carbon monoxide content, and combustion air supply. Based on the total deduction, the original air-fuel ratio deviation evidence is reduced by a limit to obtain the true air-fuel ratio deviation factor.

10. A smelting combustion process control system based on adaptive air-fuel ratio, used to implement the smelting combustion process control method based on adaptive air-fuel ratio as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect data on the smelting and combustion process. The time series construction module performs data time series alignment and disturbance response feature extraction based on smelting and combustion process data to obtain a combustion process time series dataset; The deviation assessment module constructs an event anchoring factor map based on the combustion process time series dataset, and performs interference factor contribution deduction and flue gas feedback credibility assessment to obtain the true air-fuel ratio deviation sequence and flue gas feedback credibility sequence. The predictive control module generates the adjustment amounts of the fuel valve opening and the combustion air valve opening based on the actual air-fuel ratio deviation sequence and the flue gas feedback reliability sequence, thereby obtaining the control results of the smelting combustion process.

Images