Carbon optimization control method and system for dust removal system of steel plant based on three-flow integration

Abstract

The application discloses a steel plant dust removal system carbon optimization control method and system based on three-flow integration, comprising: constructing carbon source flow sequence, energy consumption flow sequence and pollution flow sequence; time sequence slicing is performed on the sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences and pollution flow subsequences; feature aggregation is respectively performed on the subsequences until carbon source intensity coefficients, carbon efficiency coefficients and carbon risk coefficients are obtained; carbon source flow subsequences, energy consumption flow subsequences and pollution flow subsequences corresponding to the same time sequence number are anchored; sequence coding is performed on the subsequences until K three-flow coupling vectors are generated; a three-flow coupling model for predicting carbon state evolution trajectories of the dust removal system within a future rated time length is constructed according to the K three-flow coupling vectors; and carbon optimization control of the steel plant is performed based on the three-flow coupling model; the application can deduce carbon state evolution trajectories within the future rated time length in real time, thereby realizing control transformation from passive response to active prediction.

Description

Technical Field

[0001] This invention relates to carbon optimization control of dust removal systems, specifically a carbon optimization control method and system for dust removal systems in steel plants based on the integration of three flows (air, water, and fuel). Background Technology

[0002] In the existing control logic of dust removal systems in steel plants, the instantaneous carbon source parameters (such as iron tapping frequency and feeding cycle time) and instantaneous energy consumption parameters are difficult to strictly match on the time axis. This makes it difficult to identify low-load production periods, often resulting in a phenomenon where fans continue to operate at high power while the workload is light, leading to a large amount of ineffective carbon source input. At the same time, traditional methods only focus on whether instantaneous pollution parameters exceed the limits, ignoring the cumulative effect of the exceedance magnitude and duration, and cannot calculate the carbon risk coefficient that characterizes the comprehensive environmental risk. More importantly, due to the lack of exploration of the coupling patterns between historical carbon source flow sequences, energy consumption flow sequences, and pollution flow sequences, existing technologies are unable to predict the future carbon state evolution trajectory. The above factors have led to the common defect of "three-flow separation" in dust removal systems in steel plants, that is, the carbon source flow (production cycle time), energy consumption flow (fan energy consumption), and pollution flow (emission status) are independent of each other during operation and cannot form a closed loop for synergistic optimization. This kind of lagging and fragmented control mode means that the system cannot generate feedforward compensation commands before emissions exceed the standard, nor can it dynamically adjust the fan frequency when inefficient operating conditions are predicted. As a result, the dust removal system is in a passive operating state with high carbon consumption and high risk for a long time, which makes it difficult to meet the needs of the steel industry for refined carbon management. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a carbon optimization control method and system for a steel plant dust removal system based on the integration of three flows, which solves the technical problems mentioned in the background art by introducing a three-flow coupling vector.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] In a first aspect, the present invention provides a carbon optimization control method for a dust removal system in a steel plant based on the integration of three flows, comprising the following steps:

[0006] S1. Construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods;

[0007] S2. Perform time-series slicing on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow sequences, respectively;

[0008] Among them, the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence of the same slice share the same time sequence number;

[0009] S3. Perform feature aggregation on carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number until the carbon source intensity coefficient, carbon efficiency coefficient and carbon risk coefficient corresponding to K times with the same time number are obtained.

[0010] The subsequence includes instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector within J consecutive sampling windows;

[0011] S4. Among the K carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number, anchor the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence corresponding to the same time number;

[0012] S5. Sequence encoding is performed on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time series number until K three-flow coupling vectors are generated; wherein each three-flow coupling vector inherits its corresponding time series number;

[0013] S6. Based on K three-flow coupling vectors, construct a three-flow coupling model to predict the carbon state evolution trajectory of the dust removal system within the rated time in the future;

[0014] S7. Carbon optimization control in steel plants based on a three-flow coupling model.

[0015] In some specific embodiments, the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods are constructed, including:

[0016] S1-1. Obtain the instantaneous state parameters within M sampling windows; where the instantaneous state parameters include: instantaneous carbon source parameters, instantaneous energy consumption parameters, and instantaneous pollution parameters;

[0017] S1-2. Characterize the instantaneous state parameters to generate M instantaneous state features corresponding to each sampling window; wherein, the instantaneous state features include: M instantaneous carbon source features, M instantaneous energy consumption features, and instantaneous pollution features;

[0018] S1-3. Perform feature concatenation on the instantaneous state features of each sampling window until the instantaneous carbon source vector, instantaneous energy consumption vector and instantaneous pollution vector corresponding to M sampling windows are obtained.

[0019] S1-4. Arrange the instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector in sequence according to the time order of their sampling windows to construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during the historical observation period.

[0020] In some specific embodiments, the instantaneous state parameters within M sampling windows are obtained, including:

[0021] S1-1-1. Establish a sliding sampling window on the operating timeline of the dust removal system in the steel plant;

[0022] S1-1-2. Based on preset parameter labels, obtain instantaneous state parameters within the sampling window by category, including: N instantaneous carbon source parameters, N instantaneous energy consumption parameters, and N instantaneous pollution parameters;

[0023] S1-1-3. Slide the sampling window with the standard sampling step size until the instantaneous state parameters within M sampling windows are continuously acquired.

[0024] In some specific embodiments, the instantaneous state parameters are characterized to generate M instantaneous state features corresponding to each sampling window, including:

[0025] S1-2-1. Determine the target category parameter from the instantaneous carbon source parameter, instantaneous energy consumption parameter, and instantaneous pollution parameter;

[0026] S1-2-2. Among the M instantaneous state parameters of the target category parameters, anchor the maximum state parameter and the minimum state parameter;

[0027] S1-2-3. Calculate the difference between the maximum and minimum state parameters to obtain the total state deviation.

[0028] S1-2-4. Select the current state parameter from the M instantaneous state parameters;

[0029] S1-2-5. Calculate the difference between the current state parameter and the minimum state parameter to obtain the current state deviation;

[0030] S1-2-6. Calculate the ratio between the total state deviation and the current state deviation to generate the instantaneous state characteristics corresponding to the current state parameters.

[0031] S1-2-7. Traverse the M instantaneous state parameters until the instantaneous state features corresponding to the M sampling windows are obtained.

[0032] In some specific embodiments, the characteristic polymerization step of the carbon source intensity coefficient includes:

[0033] A1. Among the K carbon source subsequences with the same time series number, select the target carbon source subsequence corresponding to the current time series number;

[0034] A2. Calculate the vector magnitude of each component in the J consecutive instantaneous carbon source vectors contained in the target carbon source subsequence.

[0035] A3. Accumulate the vector magnitudes of all instantaneous carbon source vectors within the target carbon source subsequence to obtain the total cumulative carbon source amount corresponding to the subsequence;

[0036] A4. Multiply the total accumulated carbon source with the preset carbon emission factor per unit material to generate the carbon source intensity coefficient corresponding to the current time series number.

[0037] A5. Traverse the K carbon source subsequences with the same time sequence number, and repeatedly generate the carbon source intensity coefficient until the carbon source intensity coefficients corresponding to the K time sequence numbers are obtained.

[0038] In some specific embodiments, the characteristic polymerization step of the carbon efficiency coefficient includes:

[0039] B1. Among K energy consumption flow subsequences and pollution flow subsequences with the same time sequence number, anchor the target energy consumption flow subsequence and target pollution flow subsequence corresponding to the same time sequence number;

[0040] B2. Extract the fan power component from the J consecutive instantaneous energy consumption vectors contained in the target energy consumption subsequence and perform time integration to obtain the total energy consumption value corresponding to the subsequence.

[0041] B3. Extract the emission concentration component and the pipeline air volume component from the J consecutive instantaneous pollution vectors contained in the target pollution flow subsequence, calculate the product of the two and perform time integration to obtain the total dust removal amount corresponding to the subsequence.

[0042] B4. Using the total energy consumption as the dividend and the total dust removal amount as the divisor, perform a ratio calculation to generate the carbon efficiency coefficient corresponding to the current time sequence number.

[0043] B5. Traverse the K subsequence pairs with the same time sequence number, and repeatedly generate carbon efficiency coefficients until the carbon efficiency coefficients corresponding to the K time sequence numbers are obtained.

[0044] In some specific embodiments, the characteristic aggregation step of the carbon risk coefficient includes:

[0045] C1. Among the K contaminated flow subsequences with the same time sequence number, select the target contaminated flow subsequence corresponding to the current time sequence number;

[0046] C2. Preset environmental emission thresholds, and anchor the emission concentration component component one by one in the J consecutive instantaneous pollution vectors contained in the target pollution flow sequence;

[0047] C3. Calculate the difference between the emission concentration component and the environmental emission threshold, select the difference greater than zero as the instantaneous exceedance range, and set the difference less than or equal to zero to zero;

[0048] C4. Accumulate all instantaneous exceedances within the target contamination flow subsequence and multiply by the sampling window time length to obtain the total exceedance area corresponding to the subsequence;

[0049] C5. Map the total area exceeding the standard to the carbon risk coefficient corresponding to the current time series number;

[0050] C6. Traverse the K pollution flow subsequences with the same time sequence number, and repeatedly generate carbon risk coefficients until the carbon risk coefficients corresponding to the K time sequence numbers are obtained.

[0051] In some specific embodiments, the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time sequence number are sequence encoded until K three-flow coupling vectors are generated, including:

[0052] S5-1. In a set of K subsequences with the same time series number, anchor the target carbon source flow subsequence, target energy consumption flow subsequence, and target pollution flow subsequence corresponding to the same time series number k;

[0053] S5-2. Align the sampling windows for the target carbon source flow sequence, the target energy consumption flow sequence, and the target pollution flow sequence, respectively.

[0054] S5-3. Concatenate the three sub-sequences aligned with the sampling window to construct the three-stream fusion sequence under this time series number;

[0055] S5-4. Perform average pooling on each channel of the three-stream fusion sequence to extract a fixed-dimensional vector that represents the integrated state of the three streams under that time series number, and define it as the three-stream coupling vector of that time series number k.

[0056] S5-5. Mark the time series number k on the generated three-stream coupling vector to inherit its time series representation;

[0057] S5-6. Traverse K identical timing numbers and repeatedly generate three-stream coupling vectors until K three-stream coupling vectors with independent timing numbers are generated.

[0058] In some specific embodiments, a three-flow coupling model is constructed based on K three-flow coupling vectors to predict the carbon state evolution trajectory of the dust removal system over a future rated duration, including:

[0059] S6-1. Anchor the target input vector and the supervision label one by one along the time sequence numbering direction of the three-stream coupling vector;

[0060] The supervision label is characterized as: the carbon source intensity coefficient, carbon efficiency coefficient, and carbon risk coefficient associated with the three-flow coupling vector along the time sequence numbering direction and at a distance P from the target input vector from the rated prediction time.

[0061] S6-2. Pair the target input vector with the supervision label to construct carbon trajectory samples for carbon state evolution trajectory prediction until G carbon trajectory samples are obtained; where G=KP and G is less than K.

[0062] S6-3. Input G supervised samples into the time series prediction model for iterative supervised training to generate a three-flow coupled model for predicting the carbon state evolution trajectory of the dust removal system within the rated time period in the future.

[0063] This invention provides a carbon optimization control method and system for a steel plant dust removal system based on the integration of three flows, with the following beneficial effects:

[0064] This invention achieves strict temporal alignment of instantaneous carbon source parameters, instantaneous energy consumption parameters, and instantaneous pollution parameters by constructing a sliding sampling window. Based on temporal slicing and feature aggregation, it unifies heterogeneous three-stream data into carbon source intensity coefficient, carbon efficiency coefficient, and carbon risk coefficient, solving the problem of multi-source data evaluation bias in traditional technologies. Furthermore, this invention compresses variable-length three-stream subsequences into fixed-length three-stream coupling vectors carrying temporal numbers based on sequence encoding. Based on this, a three-stream coupling model is trained to predict the carbon state evolution trajectory within the rated time period in real time. This realizes the control transformation from passive response to active prediction. It can generate feedforward compensation commands for fan frequency conversion and valve opening in advance based on the predicted expected carbon risk coefficient and expected carbon efficiency coefficient. While avoiding the risk of exceeding emission standards, it accurately matches the production load to eliminate ineffective energy consumption, significantly improving the carbon efficiency optimization control of the dust removal system.

[0065] Secondly, the present invention provides a carbon optimization control system for a steel plant dust removal system based on the integration of three flows, characterized in that it includes:

[0066] A flow sequence construction unit is used to construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods;

[0067] The subsequence slicing unit is used to perform time-series slicing on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow sequences, respectively.

[0068] Among them, the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence of the same slice share the same time sequence number;

[0069] The feature aggregation unit is used to perform feature aggregation on carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number, until the carbon source intensity coefficient, carbon efficiency coefficient and carbon risk coefficient corresponding to K times with the same time number are obtained.

[0070] The subsequence includes instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector within J consecutive sampling windows;

[0071] The subsequence anchoring unit is used to anchor the carbon source flow subsequence, energy consumption flow subsequence, and pollution flow subsequence corresponding to the same time number among K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow subsequences with the same time number;

[0072] The vector encoding unit is used to sequence-encode the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time sequence number until K three-flow coupling vectors are generated; each three-flow coupling vector inherits its corresponding time sequence number;

[0073] The coupling model building unit is used to construct a three-flow coupling model for predicting the carbon state evolution trajectory of the dust removal system within the rated time period based on K three-flow coupling vectors.

[0074] Carbon optimization control unit, used for carbon optimization control in steel plants based on a three-flow coupling model.

[0075] Compared with the prior art, the beneficial effects of the carbon optimization control system for steel plant dust removal system based on the integration of three flows in this invention are the same as the beneficial effects of the carbon optimization method for steel plant dust removal system based on the integration of three flows described above, so they will not be repeated here. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of the carbon optimization control method for a steel plant dust removal system based on the integration of three flows according to the present invention.

[0077] Figure 2 This is a schematic diagram of the process for generating the carbon source intensity coefficient according to the present invention;

[0078] Figure 3 This is a schematic diagram of the process for generating the carbon efficiency coefficient described in this invention;

[0079] Figure 4 This is a schematic diagram of the process for generating the carbon risk coefficient described in this invention;

[0080] Figure 5 This is a structural block diagram of the carbon optimization control system for a steel plant dust removal system based on the integration of three flows (air, water, and fuel). Detailed Implementation

[0081] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0082] Example 1: Please refer to Figures 1 to 4This invention provides a carbon optimization control method for a steel plant dust removal system based on the integration of three flows, including the following steps:

[0083] S1. Construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods;

[0084] S2. Perform time-series slicing on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow sequences, respectively;

[0085] Among them, the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence of the same slice share the same time sequence number;

[0086] Specifically, in this embodiment, the time-series slicing refers to: sequentially extracting continuous data segments on the time axis according to a preset subsequence length J and slice sliding step s, and assigning a unique incremental time sequence number k (k=1,2,...,K) to each extracted segment, so that adjacent subsequences form overlapping or seamless time coverage.

[0087] S3. Perform feature aggregation on carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number until the carbon source intensity coefficient, carbon efficiency coefficient and carbon risk coefficient corresponding to K times with the same time number are obtained.

[0088] The subsequence includes instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector within J consecutive sampling windows;

[0089] Specifically, the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence are uniformly transformed into carbon semantic quantitative indicators, including carbon source intensity coefficient (carbon input corresponding to material consumption), carbon efficiency coefficient (the ratio of energy consumption to dust removal effect), and carbon risk coefficient (duration of transient emission exceedance × amplitude).

[0090] S4. Among the K carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number, anchor the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence corresponding to the same time number;

[0091] S5. Sequence encoding is performed on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time series number until K three-flow coupling vectors are generated; wherein each three-flow coupling vector inherits its corresponding time series number;

[0092] S6. Based on K three-flow coupling vectors, construct a three-flow coupling model to predict the carbon state evolution trajectory of the dust removal system within the rated time in the future;

[0093] S7. Carbon optimization control in steel plants based on a three-flow coupling model.

[0094] Specifically, in this embodiment, the carbon optimization control includes the following execution steps:

[0095] S7-1. During the real-time operation of the dust removal system, update the instantaneous state parameters in the sampling window and construct them as a real-time three-flow coupling vector.

[0096] S7-2. Using the real-time three-flow coupling vector as the input of the three-flow coupling model, the carbon state evolution trajectory of the steel plant dust removal system within the future rated prediction time is simulated in real time.

[0097] S7-3. Carbon optimization control is performed based on the predicted carbon state evolution trajectory of the dust removal system in the steel plant over a future rated period.

[0098] Furthermore, carbon optimization control is implemented based on a predefined carbon optimization control strategy, which includes:

[0099] If the expected carbon risk coefficient is higher than the safety threshold, a feedforward compensation instruction is generated to increase the frequency of the wind turbines or open the backup branch valves in advance to avoid future emissions exceeding the standard.

[0100] If the expected carbon efficiency coefficient is lower than the baseline energy efficiency line (indicating high energy consumption and low output), an energy-saving adjustment command will be generated to dynamically reduce the fan frequency or close the opening of the branch valve in the non-operating area to match the future low production load.

[0101] Finally, the generated fan frequency conversion command and valve opening command are sent to the underlying PLC actuator to complete the closed-loop carbon optimal control of the dust removal system.

[0102] In this embodiment, S1 specifically includes:

[0103] S1-1. Obtain the instantaneous state parameters within M sampling windows; where the instantaneous state parameters include: instantaneous carbon source parameters, instantaneous energy consumption parameters, and instantaneous pollution parameters;

[0104] S1-2. Characterize the instantaneous state parameters to generate M instantaneous state features corresponding to each sampling window; wherein, the instantaneous state features include: M instantaneous carbon source features, M instantaneous energy consumption features, and instantaneous pollution features;

[0105] S1-3. Perform feature concatenation on the instantaneous state features of each sampling window until the instantaneous carbon source vector, instantaneous energy consumption vector and instantaneous pollution vector corresponding to M sampling windows are obtained.

[0106] S1-4. Arrange the instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector in sequence according to the time order of their sampling windows to construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during the historical observation period.

[0107] In this embodiment, by transforming discrete instantaneous multidimensional parameters into a continuous time series structure, the temporal dependence of material consumption, energy fluctuations, and pollution emissions in the steel production process is fully preserved.

[0108] Furthermore, step S1-1 also includes:

[0109] S1-1-1. Establish a sliding sampling window on the operating timeline of the dust removal system in the steel plant;

[0110] Specifically, the setup process includes pre-setting and initializing the length of the sampling window and the sliding step size.

[0111] S1-1-2. Based on preset parameter labels, obtain instantaneous state parameters within the sampling window by category, including: N instantaneous carbon source parameters, N instantaneous energy consumption parameters, and N instantaneous pollution parameters;

[0112] Specifically, the instantaneous carbon source parameters are used to reflect the production cycle time of the production operation, including components such as raw material feeding cycle time, blast furnace tapping frequency, and material consumption; the instantaneous energy consumption parameters are used to reflect the energy consumption status of the dust removal equipment, including components such as fan power, frequency conversion parameters, and pipeline pressure; the instantaneous pollution parameters are used to reflect the emission status of environmental emissions, including components such as emission concentration, transient peak dust concentration, and dust removal efficiency; and the instantaneous carbon source parameters, instantaneous energy consumption parameters, and instantaneous pollution parameters all have the same number of dimensions N to ensure the alignment of the three streams of data in the feature space for parallel processing.

[0113] S1-1-3. Slide the sampling window with the standard sampling step size until the instantaneous state parameters in M ​​sampling windows are continuously acquired;

[0114] The standard sampling step size is set by a sliding strategy and is smaller than the length of the sampling window to ensure data overlap between adjacent sampling windows, thereby capturing continuous dynamic change features.

[0115] In this embodiment, by setting an overlapping sliding sampling mechanism, the problem of interruption and loss of production condition change points (such as the moment when iron is tapped) caused by fixed slices is effectively avoided, ensuring the continuity of the three streams data on the time axis.

[0116] Furthermore, step S1-2 also includes:

[0117] S1-2-1. Determine the target category parameter from the instantaneous carbon source parameter, instantaneous energy consumption parameter, and instantaneous pollution parameter;

[0118] S1-2-2. Among the M instantaneous state parameters of the target category parameters, anchor the maximum state parameter and the minimum state parameter;

[0119] S1-2-3. Calculate the difference between the maximum and minimum state parameters to obtain the total state deviation.

[0120] S1-2-4. Select the current state parameter from the M instantaneous state parameters;

[0121] S1-2-5. Calculate the difference between the current state parameter and the minimum state parameter to obtain the current state deviation;

[0122] S1-2-6. Calculate the ratio between the total state deviation and the current state deviation to generate the instantaneous state characteristics corresponding to the current state parameters.

[0123] S1-2-7. Traverse the M instantaneous state parameters until the instantaneous state features corresponding to the M sampling windows are obtained.

[0124] In this embodiment, by constructing a dynamic normalization interval based on the maximum and minimum values ​​within the current observation window, physical parameters of different dimensions (such as kilowatt-level power and milligram-level concentration) are uniformly mapped to the standard feature space of [0,1], ensuring the numerical stability of the three-stream data during feature fusion and model training.

[0125] Specifically, in this embodiment, the characteristic polymerization step of the carbon source intensity coefficient includes:

[0126] A1. Among the K carbon source subsequences with the same time series number, select the target carbon source subsequence corresponding to the current time series number;

[0127] A2. Calculate the vector magnitude of each component in the J consecutive instantaneous carbon source vectors contained in the target carbon source subsequence.

[0128] The formula for calculating the vector magnitude is:

[0129] ;

[0130] in, This represents the weighted vector magnitude of the instantaneous carbon source vector within the current sampling window;

[0131] N represents the number of dimensions of the instantaneous carbon source vector;

[0132] This represents the value of the i-th component in the instantaneous carbon source vector (i.e., the normalized value of the raw material charging cycle, blast furnace tapping frequency, or material consumption).

[0133] The carbon contribution weight coefficient of the i-th component is used to characterize the degree of difference in the impact of different production parameters on carbon source intensity, and the sum of the weight coefficients of all components is 1.

[0134] A3. Accumulate the vector magnitudes of all instantaneous carbon source vectors within the target carbon source subsequence to obtain the total cumulative carbon source amount corresponding to the subsequence;

[0135] A4. Multiply the total accumulated carbon source with the preset carbon emission factor per unit material to generate the carbon source intensity coefficient corresponding to the current time series number.

[0136] Among them, the carbon emission factor per unit material represents the theoretical carbon dioxide emissions generated per unit mass or unit volume of steel production materials during processing; it is obtained by referring to the default value in the latest greenhouse gas emission accounting guidelines published by the industry, or by regression analysis based on historical measured data.

[0137] A5. Traverse the K carbon source subsequences with the same time sequence number, and repeatedly generate the carbon source intensity coefficient until the carbon source intensity coefficients corresponding to the K time sequence numbers are obtained.

[0138] In this embodiment, different production parameters are weighted differently by introducing a carbon contribution weighting coefficient, and the total theoretical carbon input under different production rhythms is quantified by combining vector magnitude accumulation calculation.

[0139] Specifically, in this embodiment, the characteristic polymerization step of the carbon efficiency coefficient includes:

[0140] B1. Among K energy consumption flow subsequences and pollution flow subsequences with the same time sequence number, anchor the target energy consumption flow subsequence and target pollution flow subsequence corresponding to the same time sequence number;

[0141] B2. Extract the fan power component from the J consecutive instantaneous energy consumption vectors contained in the target energy consumption subsequence and perform time integration to obtain the total energy consumption value corresponding to the subsequence.

[0142] B3. Extract the emission concentration component and the pipeline air volume component from the J consecutive instantaneous pollution vectors contained in the target pollution flow subsequence, calculate the product of the two and perform time integration to obtain the total dust removal amount corresponding to the subsequence.

[0143] B4. Using the total energy consumption as the dividend and the total dust removal amount as the divisor, perform a ratio calculation to generate the carbon efficiency coefficient corresponding to the current time sequence number.

[0144] B5. Traverse the K subsequence pairs with the same time sequence number, and repeatedly generate carbon efficiency coefficients until the carbon efficiency coefficients corresponding to the K time sequence numbers are obtained.

[0145] In this embodiment, by simultaneously calculating the energy consumption integral and the total dust removal integral within the same time window and obtaining the ratio, a leap from instantaneous power matching to time-period energy efficiency assessment is achieved. This truly reflects the unit carbon consumption efficiency of the dust removal system within a specific production cycle and provides a quantitative basis for identifying inefficient operating intervals such as "strong winds accompanying small tasks".

[0146] Specifically, in this embodiment, the feature aggregation step of the carbon risk coefficient includes:

[0147] C1. Among the K contaminated flow subsequences with the same time sequence number, select the target contaminated flow subsequence corresponding to the current time sequence number;

[0148] C2. Preset environmental emission thresholds, and anchor the emission concentration component component one by one in the J consecutive instantaneous pollution vectors contained in the target pollution flow sequence;

[0149] C3. Calculate the difference between the emission concentration component and the environmental emission threshold, select the difference greater than zero as the instantaneous exceedance range, and set the difference less than or equal to zero to zero;

[0150] C4. Accumulate all instantaneous exceedances within the target contamination flow subsequence and multiply by the sampling window time length to obtain the total exceedance area corresponding to the subsequence;

[0151] C5. Map the total area exceeding the standard to the carbon risk coefficient corresponding to the current time series number;

[0152] Specifically, the total area of ​​exceedance for each subsequence can be normalized to obtain the carbon risk coefficient.

[0153] C6. Traverse the K pollution flow subsequences with the same time sequence number, and repeatedly generate carbon risk coefficients until the carbon risk coefficients corresponding to the K time sequence numbers are obtained.

[0154] In this embodiment, by calculating the product of the exceedance magnitude and duration (total exceedance area) rather than just judging whether the exceedance occurs instantaneously, it is possible to capture two different forms of environmental risks: short-term high-frequency fluctuations and long-term slight exceedances, thus transforming implicit emission fluctuations into explicit quantitative indicators of carbon compliance risks.

[0155] In this embodiment, S5 specifically includes:

[0156] S5-1. In a set of K subsequences with the same time series number, anchor the target carbon source flow subsequence, target energy consumption flow subsequence, and target pollution flow subsequence corresponding to the same time series number k;

[0157] S5-2. Align the sampling windows for the target carbon source flow sequence, the target energy consumption flow sequence, and the target pollution flow sequence, respectively.

[0158] S5-3. Concatenate the three sub-sequences aligned with the sampling window to construct the three-stream fusion sequence under this time series number;

[0159] Specifically, sampling window alignment refers to verifying whether the timestamps of corresponding position vectors in the three subsequences are consistent, ensuring strict synchronization of the three streams of data on the time axis.

[0160] S5-4. Perform average pooling on each channel of the three-stream fusion sequence to extract a fixed-dimensional vector that represents the integrated state of the three streams under that time series number, and define it as the three-stream coupling vector of that time series number k.

[0161] Specifically, average pooling refers to calculating the arithmetic mean of the three-stream fused sequence in the time dimension (i.e., the J sampling window directions), compressing the sequence of length J into a feature vector of length 1. Therefore, the three-stream coupling vector contains comprehensive statistical features in the same dimension as the number of channels in the three-stream fused sequence.

[0162] S5-5. Mark the time series number k on the generated three-stream coupling vector to inherit its time series representation;

[0163] S5-6. Traverse K identical timing numbers and repeatedly generate three-stream coupling vectors until K three-stream coupling vectors with independent timing numbers are generated.

[0164] In this embodiment, through a combination of encoding operations of time alignment, channel concatenation and average pooling, the originally heterogeneous and variable-length three-stream subsequences are compressed into fixed-length feature vectors, thus fully preserving the comprehensive statistical features and temporal correlation features of the three-stream data under this time series number.

[0165] In this embodiment, S6 specifically includes:

[0166] S6-1. Anchor the target input vector and the supervision label one by one along the time sequence numbering direction of the three-stream coupling vector;

[0167] The supervision label is characterized as: the carbon source intensity coefficient, carbon efficiency coefficient, and carbon risk coefficient associated with the three-flow coupling vector along the time sequence numbering direction and at a distance P from the target input vector from the rated prediction time.

[0168] S6-2. Pair the target input vector with the supervision label to construct carbon trajectory samples for carbon state evolution trajectory prediction until G carbon trajectory samples are obtained; where G=KP and G is less than K.

[0169] S6-3. Input G supervised samples into the time series prediction model for iterative supervised training to generate a three-flow coupling model for predicting the carbon state evolution trajectory of the dust removal system within the rated time period in the future.

[0170] Specifically, the carbon state evolution trajectory represents the set of evolution paths of the expected carbon source intensity coefficient, carbon efficiency coefficient, and carbon risk coefficient in the time series at future moments.

[0171] In this embodiment, the time series prediction model preferably adopts a time series prediction network or a long short-term memory network (LSTM) based on the Transformer architecture. Its network architecture parameters include: input layer dimension (the dimension of the matching three-stream coupling vector), number of hidden layer nodes (e.g., 128 or 256), number of attention heads (e.g., 4 or 8), dropout ratio (e.g., 0.1), and output layer dimension (corresponding to the predicted values ​​of the three carbon coefficients).

[0172] Example 2: This Example 2 differs from Example 1 in that it also provides a carbon optimization control system for a steel plant dust removal system based on the integration of three flows (flow, water, and air). This system is used to implement the above-described method embodiments. Details already described will not be repeated. The terms "module," "unit," and "subunit" used below refer to combinations of software and / or hardware that achieve a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0173] like Figure 5 As shown, Figure 5 This is a structural block diagram of the carbon optimization control system for a steel plant dust removal system based on the integration of three flows (air, water, and fuel). The system includes:

[0174] A flow sequence construction unit is used to construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods;

[0175] The subsequence slicing unit is used to perform time-series slicing on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow sequences, respectively.

[0176] Among them, the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence of the same slice share the same time sequence number;

[0177] The feature aggregation unit is used to perform feature aggregation on carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number, until the carbon source intensity coefficient, carbon efficiency coefficient and carbon risk coefficient corresponding to K times with the same time number are obtained.

[0178] The subsequence includes instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector within J consecutive sampling windows;

[0179] The subsequence anchoring unit is used to anchor the carbon source flow subsequence, energy consumption flow subsequence, and pollution flow subsequence corresponding to the same time number among K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow subsequences with the same time number;

[0180] The vector encoding unit is used to sequence-encode the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time sequence number until K three-flow coupling vectors are generated; each three-flow coupling vector inherits its corresponding time sequence number;

[0181] The coupling model building unit is used to construct a three-flow coupling model for predicting the carbon state evolution trajectory of the dust removal system within the rated time period based on K three-flow coupling vectors.

[0182] Carbon optimization control unit, used for carbon optimization control in steel plants based on a three-flow coupling model.

[0183] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A carbon optimization control method for a steel plant dust removal system based on the integration of three flows, characterized in that, include: S1. Construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods; S2. Perform time-series slicing on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow sequences, respectively; Among them, the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence of the same slice share the same time sequence number; S3. Perform feature aggregation on carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number until the carbon source intensity coefficient, carbon efficiency coefficient and carbon risk coefficient corresponding to K times with the same time number are obtained. The subsequence includes instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector within J consecutive sampling windows; S4. Among the K carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number, anchor the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence corresponding to the same time number; S5. Sequence encoding is performed on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time series number until K three-flow coupling vectors are generated; wherein each three-flow coupling vector inherits its corresponding time series number; S6. Based on K three-flow coupling vectors, construct a three-flow coupling model to predict the carbon state evolution trajectory of the dust removal system within the rated time in the future; S7. Carbon optimization control in steel plants based on a three-flow coupling model.

2. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 1, characterized in that, Constructing the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods includes: S1-1. Obtain the instantaneous state parameters within M sampling windows; where the instantaneous state parameters include: instantaneous carbon source parameters, instantaneous energy consumption parameters, and instantaneous pollution parameters; S1-2. Characterize the instantaneous state parameters to generate M instantaneous state features corresponding to each sampling window; wherein, the instantaneous state features include: M instantaneous carbon source features, M instantaneous energy consumption features, and M instantaneous pollution features; S1-3. Perform feature concatenation on the instantaneous state features of each sampling window until the instantaneous carbon source vector, instantaneous energy consumption vector and instantaneous pollution vector corresponding to M sampling windows are obtained. S1-4. Arrange the instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector in sequence according to the time order of their sampling windows to construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during the historical observation period.

3. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 2, characterized in that, Obtain the instantaneous state parameters within M sampling windows, including: S1-1-1. Establish a sliding sampling window on the operating timeline of the dust removal system in the steel plant; S1-1-2. Based on preset parameter labels, obtain instantaneous state parameters within the sampling window by category, including: N instantaneous carbon source parameters, N instantaneous energy consumption parameters, and N instantaneous pollution parameters; S1-1-3. Slide the sampling window with the standard sampling step size until the instantaneous state parameters within M sampling windows are continuously acquired.

4. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 2, characterized in that, The instantaneous state parameters are characterized to generate M instantaneous state features corresponding to each sampling window, including: S1-2-1. Determine the target category parameter from the instantaneous carbon source parameter, instantaneous energy consumption parameter, and instantaneous pollution parameter; S1-2-2. Among the M instantaneous state parameters of the target category parameters, anchor the maximum state parameter and the minimum state parameter; S1-2-3. Calculate the difference between the maximum and minimum state parameters to obtain the total state deviation. S1-2-4. Select the current state parameter from the M instantaneous state parameters; S1-2-5. Calculate the difference between the current state parameter and the minimum state parameter to obtain the current state deviation; S1-2-6. Calculate the ratio between the total state deviation and the current state deviation to generate the instantaneous state characteristics corresponding to the current state parameters. S1-2-7. Traverse the M instantaneous state parameters until the instantaneous state features corresponding to the M sampling windows are obtained.

5. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 4, characterized in that, The characteristic polymerization steps for the carbon source intensity coefficient include: A1. Among the K carbon source subsequences with the same time series number, select the target carbon source subsequence corresponding to the current time series number; A2. Calculate the vector magnitude of each component in the J consecutive instantaneous carbon source vectors contained in the target carbon source subsequence. A3. Accumulate the vector magnitudes of all instantaneous carbon source vectors within the target carbon source subsequence to obtain the total cumulative carbon source amount corresponding to the subsequence; A4. Multiply the total accumulated carbon source with the preset carbon emission factor per unit material to generate the carbon source intensity coefficient corresponding to the current time series number. A5. Traverse the K carbon source subsequences with the same time sequence number, and repeatedly generate the carbon source intensity coefficient until the carbon source intensity coefficients corresponding to the K time sequence numbers are obtained.

6. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 4, characterized in that, The characteristic polymerization steps for the carbon efficiency coefficient include: B1. Among K energy consumption flow subsequences and pollution flow subsequences with the same time sequence number, anchor the target energy consumption flow subsequence and target pollution flow subsequence corresponding to the same time sequence number; B2. Extract the fan power component from the J consecutive instantaneous energy consumption vectors contained in the target energy consumption subsequence and perform time integration to obtain the total energy consumption value corresponding to the subsequence. B3. Extract the emission concentration component and the pipeline air volume component from the J consecutive instantaneous pollution vectors contained in the target pollution flow subsequence, calculate the product of the two and perform time integration to obtain the total dust removal amount corresponding to the subsequence. B4. Using the total energy consumption as the dividend and the total dust removal amount as the divisor, perform a ratio calculation to generate the carbon efficiency coefficient corresponding to the current time sequence number. B5. Traverse the K subsequence pairs with the same time sequence number, and repeatedly generate carbon efficiency coefficients until the carbon efficiency coefficients corresponding to the K time sequence numbers are obtained.

7. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 4, characterized in that, The characteristic aggregation step of the carbon risk coefficient includes: C1. Among the K contaminated flow subsequences with the same time sequence number, select the target contaminated flow subsequence corresponding to the current time sequence number; C2. Preset environmental emission thresholds, and anchor the emission concentration component component one by one in the J consecutive instantaneous pollution vectors contained in the target pollution flow sequence; C3. Calculate the difference between the emission concentration component and the environmental emission threshold, select the difference greater than zero as the instantaneous exceedance range, and set the difference less than or equal to zero to zero; C4. Accumulate all instantaneous exceedances within the target contamination flow subsequence and multiply by the sampling window time length to obtain the total exceedance area corresponding to the subsequence; C5. Map the total area exceeding the standard to the carbon risk coefficient corresponding to the current time series number; C6. Traverse the K pollution flow subsequences with the same time sequence number, and repeatedly generate carbon risk coefficients until the carbon risk coefficients corresponding to the K time sequence numbers are obtained.

8. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 7, characterized in that, Sequence encoding is performed on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time sequence number until K three-flow coupling vectors are generated, including: S5-1. In a set of K subsequences with the same time series number, anchor the target carbon source flow subsequence, target energy consumption flow subsequence, and target pollution flow subsequence corresponding to the same time series number k; S5-2. Align the sampling windows for the target carbon source flow sequence, the target energy consumption flow sequence, and the target pollution flow sequence, respectively. S5-3. Concatenate the three sub-sequences aligned with the sampling window to construct the three-stream fusion sequence under this time series number; S5-4. Perform average pooling on each channel of the three-stream fusion sequence to extract a fixed-dimensional vector that represents the integrated state of the three streams under that time series number, and define it as the three-stream coupling vector of that time series number k. S5-5. Mark the time series number k on the generated three-stream coupling vector to inherit its time series representation; S5-6. Traverse K identical timing numbers and repeatedly generate three-stream coupling vectors until K three-stream coupling vectors with independent timing numbers are generated.

9. The carbon optimization control method for a steel plant dust removal system based on the integration of three flows as described in claim 8, characterized in that, Based on K three-flow coupling vectors, a three-flow coupling model is constructed to predict the carbon state evolution trajectory of the dust removal system within the rated time period, including: S6-1. Anchor the target input vector and the supervision label one by one along the time sequence numbering direction of the three-stream coupling vector; The supervision label is characterized as: the carbon source intensity coefficient, carbon efficiency coefficient, and carbon risk coefficient associated with the three-flow coupling vector along the time sequence numbering direction and at a distance P from the target input vector from the rated prediction time. S6-2. Pair the target input vector with the supervision label to construct carbon trajectory samples for carbon state evolution trajectory prediction until G carbon trajectory samples are obtained; where G=KP and G is less than K. S6-3. Input G supervised samples into the time series prediction model for iterative supervised training to generate a three-flow coupled model for predicting the carbon state evolution trajectory of the dust removal system within the rated time period in the future.

10. A carbon optimization control system for a steel plant dust removal system based on the integration of three flows, characterized in that, include: A flow sequence construction unit is used to construct the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence of the dust removal system during historical observation periods; The subsequence slicing unit is used to perform time-series slicing on the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence to obtain K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow sequences, respectively. Among them, the carbon source flow sequence, energy consumption flow sequence and pollution flow sequence of the same slice share the same time sequence number; The feature aggregation unit is used to perform feature aggregation on carbon source flow sequence, energy consumption flow sequence and pollution flow sequence with the same time number, until the carbon source intensity coefficient, carbon efficiency coefficient and carbon risk coefficient corresponding to K times with the same time number are obtained. The subsequence includes instantaneous carbon source vector, instantaneous energy consumption vector, and instantaneous pollution vector within J consecutive sampling windows; The subsequence anchoring unit is used to anchor the carbon source flow subsequence, energy consumption flow subsequence, and pollution flow subsequence corresponding to the same time number among K carbon source flow subsequences, energy consumption flow subsequences, and pollution flow subsequences with the same time number; The vector encoding unit is used to sequence-encode the carbon source flow sequence, energy consumption flow sequence, and pollution flow sequence corresponding to the same time sequence number until K three-flow coupling vectors are generated; each three-flow coupling vector inherits its corresponding time sequence number; The coupling model building unit is used to construct a three-flow coupling model for predicting the carbon state evolution trajectory of the dust removal system within the rated time period based on K three-flow coupling vectors. Carbon optimization control unit, used for carbon optimization control in steel plants based on a three-flow coupling model.

Images