A method for carbon emission flow tracking and optimized scheduling in a multi-product co-production process in fluorochemicals

Abstract

This invention relates to the field of industrial process management and optimization scheduling technology, and particularly to a method for tracking and optimizing carbon emission flows in a multi-product co-production process in the fluorochemical industry. First, a directed graph model of the product chain with process units as vertices is constructed. A multi-dimensional carbon emission flow calculation model is established, including direct carbon flows from fuel combustion, indirect carbon flows from purchased electricity and heat, direct carbon flows from the process itself, and carbon flows from the emission of fluorinated greenhouse gases. A dynamic evolution algorithm for node carbon potential and an iterative algorithm for reverse tracing of branch carbon flows are employed to achieve accurate tracking of spatiotemporally coupled carbon flows throughout the entire process. Using the time-varying carbon intensity coefficient as the core parameter, a dual-objective optimization scheduling model is established to minimize both total system carbon emissions and overall operating costs. An adaptive penalty mechanism with carbon constraints is introduced to solve for the optimal scheduling scheme, constructing a rolling optimization closed-loop control link for the production process. This invention effectively solves the problems of calculating carbon flows from the emission of fluorinated gases and allocating carbon flows among multiple products in the fluorochemical industry, achieving a synergistic improvement in carbon emission reduction and economic benefits.

Description

Technical Field

[0001] This invention relates to the field of industrial process management and optimization scheduling technology, specifically a method for carbon emission flow tracking and optimization scheduling in a multi-product co-production process of fluorochemicals. Background Technology

[0002] Current carbon emission management technologies in the fluorochemical industry still have many shortcomings. Traditional carbon emission accounting methods mainly rely on static calculations based on annual statistical data, which can only obtain the total carbon emissions for the entire plant or process, but cannot accurately allocate carbon flows to each process node and each final product. Most accounting methods oversimplify the treatment of carbon emissions from the escape of fluorinated greenhouse gases, using only fixed empirical coefficients for estimation, without considering the impact of equipment sealing levels, operating pressure and temperature, and fluctuations in operating conditions on the escape rate. The calculation results have large errors and cannot meet the needs of refined carbon management.

[0003] In the field of production scheduling, traditional scheduling methods have long focused on maximizing output and minimizing operating costs as the sole optimization objectives, failing to incorporate carbon emission factors into the scheduling decision-making system. Most scheduling methods that introduce carbon constraints employ a fixed average carbon intensity coefficient, ignoring the dynamic characteristics of carbon flow across time and space. They are unable to dynamically adjust production plans and energy allocation schemes based on real-time carbon flow distribution, making it difficult to achieve synergistic optimization of carbon emission reduction and economic benefits.

[0004] Most existing carbon flow tracing technologies employ simple mass proportioning methods, which cannot accurately resolve ambiguities in carbon flow allocation for recirculating backflow branches and collinear separation networks present in multi-product co-production processes. This leads to distorted life-cycle carbon intensity calculations for each product, failing to provide reliable data support for product carbon footprint accounting and carbon asset management. Furthermore, most existing carbon flow tracing technologies are offline analysis tools, unable to track the dynamic evolution of carbon flows during production in real time, and thus cannot provide timely and accurate feedback parameters for online optimization and scheduling.

[0005] Furthermore, most existing technologies fail to achieve deep integration between carbon emission management and production scheduling systems. Carbon emission data and production operation data are fragmented, failing to form a complete technological chain from carbon flow quantification and carbon flow tracking to scheduling optimization and closed-loop control. This makes it difficult to dynamically adjust production operation parameters based on real-time carbon emission status and to effectively address various uncertainties during the production process, such as raw material fluctuations, changes in equipment operating conditions, and adjustments to carbon market policies.

[0006] Therefore, there is an urgent need to develop a carbon emission flow tracking and optimization scheduling method that can adapt to the characteristics of multi-product co-production processes in the fluorochemical industry. This method should achieve accurate quantification and dynamic tracking of carbon emissions throughout the entire process and at all times, establish a dual-objective optimization scheduling system that deeply integrates carbon flow and production scheduling, and construct a closed-loop control link for continuous rolling optimization of the production process. This will help fluorochemical companies improve the precision of their carbon management, maximizing economic benefits while meeting carbon emission reduction requirements. Summary of the Invention

[0007] The purpose of this invention is to provide a method for tracking and optimizing the scheduling of carbon emission flows in a multi-product co-production process in the fluorochemical industry, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A method for carbon emission flow tracking and optimized scheduling in a multi-product co-production process of fluorochemicals includes the following steps:

[0010] S1: Obtain process topology data, time-series material flow data, and time-series energy flow data for the co-production of multiple products in fluorochemicals. After data cleaning and feature mapping, construct a directed graph model of the product chain with process units as vertices and material branches and energy branches as directed edges. Vertices include raw material input nodes, reaction and conversion nodes, separation and purification nodes, product output nodes, utility nodes, and intermediate buffer nodes.

[0011] S2: Based on the product chain directed graph model, the input and output material balance, energy consumption, process reaction carbon emission factor and fluorine-containing greenhouse gas emission kinetic parameters of each process node are extracted to construct a multi-dimensional carbon emission flow calculation model that includes direct carbon flow from fuel combustion, indirect carbon flow from purchased electricity / heat, direct carbon flow from the process and carbon flow from fluorine-containing greenhouse gas emission.

[0012] S3: The dynamic evolution equation of node carbon potential based on mass-energy-carbon conservation and the reverse source tracing iterative algorithm of branch carbon flow are used to perform spatiotemporal coupling carbon flow tracing on the directed graph model of the product chain, and calculate the dynamic carbon potential distribution of each process node at different scheduling periods and the carbon emission intensity distribution of each material branch and energy branch.

[0013] S4: The dynamic carbon potential distribution and the branch carbon emission intensity distribution are tensor-mapped according to the time series and product flow direction to construct a dynamic carbon flow intensity matrix. The time-varying carbon intensity coefficient output by the dynamic carbon flow intensity matrix is ​​used as the core parameter to establish a multi-objective optimization scheduling model with the dual objectives of minimizing the total carbon emissions of the system and minimizing the comprehensive operating cost. The multi-objective optimization scheduling model is coupled with product planned output constraints, process node capacity and load ramping constraints, energy-carbon supply and demand coupling constraints, and dynamic carbon quota constraints.

[0014] S5: The multi-objective optimization solution algorithm with carbon constraint adaptive penalty mechanism and fuzzy preference projection decision is used to iteratively solve the multi-objective optimization scheduling model to obtain the Pareto optimal scheduling solution set, and the optimal scheduling instruction sequence is extracted from the Pareto optimal scheduling solution set based on the carbon-economic preference weight preset by the decision-maker.

[0015] S6: Send the optimal scheduling instruction sequence to the production management system to perform rolling optimization and dynamic adjustment of production task allocation, key process node operating parameter settings, and energy medium configuration schemes for multi-product joint production lines.

[0016] As can be seen from the technical solution provided by the present invention above, the carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals provided by the present invention has the following beneficial effects:

[0017] This invention constructs a multidimensional carbon emission flow calculation model that includes direct carbon flow from fuel combustion, indirect carbon flow from purchased electricity and heat, direct carbon flow from the process, and carbon flow from the escape of fluorinated greenhouse gases. For the first time, it incorporates the pressure difference-driven and concentration diffusion coupling mechanism of fluorinated greenhouse gas escape into the carbon flow quantification system, solving the industry problem of accurately calculating carbon emissions from the escape of high-GWP fluorinated gases in the fluorochemical industry. By adopting an iterative algorithm that combines forward evolution of node carbon potential with reverse tracing of branch carbon flow, it effectively solves the ambiguity problem of carbon flow allocation caused by the circulation return branch and collinear separation network in multi-product co-production processes, and realizes accurate tracking of carbon flow at each node and branch throughout the entire process from raw material input to product output.

[0018] This invention uses the time-varying carbon intensity coefficient output by the dynamic carbon flow intensity matrix as the core parameter for optimized scheduling, breaking the limitation of traditional production scheduling that only focuses on the single objective of operating cost. It couples dynamic carbon quota constraints with energy carbon supply and demand constraints, introduces an adaptive penalty mechanism for carbon constraints to guide the population to evolve towards the low-carbon feasible domain, and generates a Pareto optimal scheduling solution set that represents the boundary between carbon emission reduction and economic trade-off. Combined with the fuzzy preference projection decision method, it transforms the decision-maker's subjective preferences into objective decision-making basis, and achieves the optimal balance between minimizing the total carbon emissions of the system and minimizing the overall operating cost.

[0019] This invention overcomes the bottleneck of static scheduling schemes being unable to adapt to dynamic fluctuations in production. By collecting multi-dimensional operational data in real time, it establishes a production deviation assessment and online dynamic compensation mechanism. When the actual operating state deviates from the scheduling target, it calls the lightweight node carbon potential extrapolation module to quickly update the carbon flow allocation weight, generate a dynamic adjustment instruction set, and inject it into the optimization model of the next rolling window. This forms a complete closed-loop control link for instruction issuance, status acquisition, deviation assessment, carbon potential extrapolation parameter compensation, and window rolling, ensuring that the scheduling scheme remains effective in actual operation and can cope with various random disturbances such as raw material fluctuations and changes in equipment operating conditions.

[0020] This invention accurately depicts the complex material and energy flow coupling topology of the multi-product co-production process in fluorochemicals based on a product chain directed graph model, supporting dynamic tracking and optimized scheduling of carbon flow under different product output allocation ratios. The method fully considers the strong coupling characteristics of reaction conversion and separation purification units in fluorochemical processes, as well as the linkage between utility systems and production systems, and can provide customized carbon emission management and optimized scheduling solutions for various multi-product co-production units in fluorochemicals.

[0021] This invention can accurately locate the carbon emission contribution of each process node and the carbon intensity of each final product throughout its entire life cycle, providing reliable data support for enterprise carbon accounting, carbon auditing, and carbon asset management; by optimizing the allocation of production tasks and the configuration of energy media, it can effectively reduce the total carbon emissions of the system while ensuring product output and quality, and reduce the cost of purchasing carbon allowances; under the carbon market trading mechanism, it helps enterprises achieve synergistic improvement in economic and environmental benefits, and enhance their sustainable development capabilities and market competitiveness.

[0022] This invention is developed based on industrial general data interfaces and standard communication protocols, and can be seamlessly integrated with existing production execution systems, energy management systems and distributed control systems without the need for large-scale transformation of existing production facilities. The overall technical framework of the method is universal and can be extended to carbon emission management and optimization scheduling in other high-energy-consuming multi-product co-production industries such as petrochemicals, chemicals, and metallurgy by adjusting the industry carbon emission factor library and process topology model. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the steps of a carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0025] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0026] like Figure 1 As shown in the figure, this invention provides a method for carbon emission flow tracking and optimized scheduling in a multi-product co-production process in fluorochemicals, comprising the following steps:

[0027] S1: Obtain process topology data, time-series material flow data, and time-series energy flow data for the co-production of multiple products in fluorochemicals. After data cleaning and feature mapping, construct a directed graph model of the product chain with process units as vertices and material branches and energy branches as directed edges. Vertices include raw material input nodes, reaction and conversion nodes, separation and purification nodes, product output nodes, utility nodes, and intermediate buffer nodes.

[0028] S2: Based on the product chain directed graph model, the input and output material balance, energy consumption, process reaction carbon emission factor and fluorine-containing greenhouse gas emission kinetic parameters of each process node are extracted to construct a multi-dimensional carbon emission flow calculation model that includes direct carbon flow from fuel combustion, indirect carbon flow from purchased electricity / heat, direct carbon flow from the process and carbon flow from fluorine-containing greenhouse gas emission.

[0029] S3: The dynamic evolution equation of node carbon potential based on mass-energy-carbon conservation and the reverse source tracing iterative algorithm of branch carbon flow are used to perform spatiotemporal coupling carbon flow tracing on the directed graph model of the product chain, and calculate the dynamic carbon potential distribution of each process node at different scheduling periods and the carbon emission intensity distribution of each material branch and energy branch.

[0030] S4: The dynamic carbon potential distribution and the branch carbon emission intensity distribution are tensor-mapped according to the time series and product flow direction to construct a dynamic carbon flow intensity matrix. The time-varying carbon intensity coefficient output by the dynamic carbon flow intensity matrix is ​​used as the core parameter to establish a multi-objective optimization scheduling model with the dual objectives of minimizing the total carbon emissions of the system and minimizing the comprehensive operating cost. The multi-objective optimization scheduling model is coupled with product planned output constraints, process node capacity and load ramping constraints, energy-carbon supply and demand coupling constraints, and dynamic carbon quota constraints.

[0031] S5: The multi-objective optimization solution algorithm with carbon constraint adaptive penalty mechanism and fuzzy preference projection decision is used to iteratively solve the multi-objective optimization scheduling model to obtain the Pareto optimal scheduling solution set, and the optimal scheduling instruction sequence is extracted from the Pareto optimal scheduling solution set based on the carbon-economic preference weight preset by the decision-maker.

[0032] S6: Send the optimal scheduling instruction sequence to the production management system to perform rolling optimization and dynamic adjustment of production task allocation, key process node operating parameter settings, and energy medium configuration schemes for multi-product joint production lines.

[0033] In this embodiment, the core function of step S1 is to comprehensively collect process and operational data of the multi-product co-production process in fluorochemicals, complete data standardization preprocessing and physical entity abstraction mapping, and construct a directed graph model of the product chain that can accurately reflect the coupling relationship between the material flow and energy flow of the entire process, providing a topological foundation and data support for subsequent carbon emission flow calculation and carbon flow tracking; the detailed steps are as follows:

[0034] Step S1-1: Extraction and Standardization Preprocessing of Multi-Source Production Operation Data

[0035] Three types of core data are extracted from the production execution system, energy management system, and real-time database of the multi-product co-production process in fluorochemicals. The first type is the process topology data associated with the production process piping and instrumentation flow diagram, which includes a list of physical equipment, equipment connection relationships, and pipeline attribute information. The second type is the time-series records of mass flow rate and component concentration of each material pipeline, covering the entire process of material transfer data from raw materials, intermediate materials, and final products. The third type is the time-series records of the consumption of various utility media, including real-time consumption data of energy media such as steam, electricity, circulating water, and compressed air.

[0036] Three standardized preprocessing operations are performed on the extracted raw data. The first is missing value imputation, which uses linear interpolation to handle single-point missing values ​​in continuous time series data and the historical mean method under the same operating conditions to handle multiple consecutive missing values, ensuring the continuity of the data sequence. The second is abnormal operating condition segment removal, which identifies abnormal data segments that deviate from the normal operating range based on the 3σ criterion and calculates the mean and standard deviation of a certain time series data sequence using the following formula: ,in, The mean of the time series data. The length of the time-series data sequence. For the first One data point; ,in, The standard deviation of the time series data sequence;

[0037] The value falls within the interval Data points outside of these categories are identified as outliers. Consecutive outliers constitute an abnormal operating condition segment, which is then removed after confirmation in conjunction with the process log. The third item is multi-source data timestamp alignment. Using the unified clock of the scheduling system as a reference, data with different sampling frequencies are resampled to a unified scheduling time step to eliminate time deviations from different data sources.

[0038] After preprocessing, a standardized set of process topology data, a standardized set of time-series material flow data, and a standardized set of time-series energy flow data are generated, providing standardized data input for subsequent feature mapping.

[0039] Step S1-2: Abstract feature mapping of physical devices and pipelines:

[0040] The standardized process topology data set is used to perform process function clustering and identification of each physical device. Based on the core role of the equipment in the production process, the physical devices are mapped into six types of abstract nodes: raw material input nodes correspond to raw material storage tanks and feed pump groups, responsible for receiving and transporting various raw materials required for production; reaction and conversion nodes correspond to various reactors and supporting heat exchange equipment, responsible for completing the chemical reaction and conversion process of fluorochemical raw materials; separation and purification nodes correspond to distillation columns, extraction columns, and filtration equipment, responsible for separating and purifying reaction products into qualified intermediate materials or final products; product output nodes correspond to product storage tanks and loading systems, responsible for storing and outputting final products; utility nodes correspond to boilers, steam turbines, and circulating water stations, responsible for providing various energy media for the production process; intermediate buffer nodes correspond to intermediate material storage tanks and buffer tanks, responsible for buffering intermediate materials to balance production load fluctuations.

[0041] The material conveying pipelines connecting various physical devices are mapped as directed edges of material branches, and the energy supply pipelines connecting energy supply equipment and energy consumption equipment are mapped as directed edges of energy branches; a bidirectional mapping index table is established between physical device identifiers, node category identifiers and abstract node identifiers to realize fast bidirectional retrieval and association between physical entities and abstract model elements.

[0042] Step S1-3: Dynamic construction and verification of the directed graph model of the product chain:

[0043] The standardized time-series material flow data set drives the dynamic assignment of flow and direction attributes of directed edges of each material branch, and updates the transmission status of each material branch in real time; the standardized time-series energy flow data set drives the dynamic assignment of energy consumption medium type and consumption intensity attributes of directed edges of each energy branch, and reflects the consumption status of each energy medium in real time.

[0044] Based on the connection relationship of physical equipment in the process topology data, abstract nodes with flow direction association are cascaded in sequence according to the material flow direction and energy transfer direction to construct a product chain directed graph model that reflects the coupling topology of material flow and energy flow in the entire process of multi-product production line.

[0045] Two topology verification rules are applied to the constructed product chain directed graph model. The first verification rule is that the starting and ending vertices of any material branch directed edge belong to the abstract node set, and the direction of the material branch directed edge corresponds to the actual flow of materials in the fluorochemical multi-product co-production process. The second verification rule is that the ending vertex of any energy branch directed edge is a utility node or reaction conversion node that consumes the energy, and the starting vertex is a purchased energy input node or energy output node.

[0046] After successful verification, the topology file and dynamic attribute mapping table of the product chain directed graph model are saved, completing the construction of the product chain directed graph model and providing a foundation for the subsequent establishment of a multidimensional carbon emission flow calculation model.

[0047] In this embodiment, the core function of step S2 is to extract the operating parameters and carbon emission-related attributes of each process node based on the constructed directed graph model of the product chain, call the industry standard parameter library, and construct four types of carbon flow calculation sub-modules respectively. Through multi-dimensional orthogonal superposition, a multi-dimensional carbon emission flow calculation model with spatiotemporal mechanism three-dimensional coupling is formed, realizing the accurate quantification of carbon emissions throughout the entire process of multi-product co-production in fluorochemicals, and providing underlying data support for subsequent carbon flow tracking; the detailed steps are as follows:

[0048] Step S2-1: Extracting and binding process node operating parameters and attributes:

[0049] Based on the directed graph model of the product chain, all vertices and their associated directed edges of material branches and energy branches are traversed to establish the association mapping matrix between nodes and branches; from the standardized time-series material flow data set and the standardized time-series energy flow data set, sliced ​​according to the scheduling time window, the total mass of input materials, the total mass of output materials, the mass ratio of each component material, the consumption of various energy media, and the energy media supply pressure and temperature parameters of each vertex in the corresponding time period are extracted;

[0050] Calculate the input and output material balance at each vertex using the following formula: ,in, For node material balance, For the first The material quality of each input material branch. For the first The material quality of each output material branch. Enter the total number of material branches for each node. Output the total number of material branches for each node;

[0051] By combining the bidirectional mapping index table, the process reaction type, equipment sealing level and operating condition attributes corresponding to the physical equipment identifier are bound to the corresponding abstract node; the operating condition attributes include the node's real-time operating pressure, real-time operating temperature, equipment operating load rate and continuous running time; a dynamic update mechanism for node attributes is established, and the attribute information of all nodes is automatically refreshed after each scheduling time window to ensure that the attribute information of each abstract node is completely synchronized with the operating status of the actual physical equipment.

[0052] Step S2-2: Industry carbon emission parameter library access and parameter mapping:

[0053] The system calls upon a pre-built knowledge base of carbon emission factors for the fluorochemical industry and a database of kinetic parameters for the emission of fluorinated greenhouse gases. The knowledge base of carbon emission factors for the fluorochemical industry includes carbon emission factors for various fuel combustion processes, carbon emission factors for various fluorochemical process reactions, and carbon content coefficients for various carbon sequestration byproducts. The database of kinetic parameters for the emission of fluorinated greenhouse gases includes equivalent resistance coefficients for leakage paths of equipment with different sealing levels, pressure differential-driven emission rate index, concentration diffusion coefficient, and weighted averages of global warming potential values ​​for various fluorinated greenhouse gases.

[0054] The process reaction types bound to nodes are mapped to the corresponding process reaction carbon emission factors; for composite reaction nodes containing multiple reaction types, the weighted average process reaction carbon emission factor is calculated according to the proportion of material conversion of each reaction; the equipment sealing level, real-time operating pressure and real-time operating temperature attributes bound to nodes are mapped to the corresponding fluorine-containing greenhouse gas emission kinetic parameters; for the same equipment under different operating conditions, the emission kinetic parameters corresponding to the real-time operating conditions are calculated using linear interpolation, realizing the dynamic correlation between process operating status and carbon emission calculation parameters;

[0055] Step S2-3: Independent construction of four types of carbon flow calculation submodules:

[0056] Using the directed edges of the material branch and the directed edges of the energy branch as carbon flow transmission channels, four independent carbon flow calculation sub-modules are constructed respectively. Each sub-module includes a data input layer, a parameter calculation layer and a result output layer.

[0057] The first submodule is the direct carbon flow calculation submodule for fuel combustion. This submodule first identifies the fuel medium type in the directed edges of the energy branch, extracts the corresponding fuel consumption and lower heating value, and calculates the direct carbon flow of combustion at each vertex based on the fuel carbon oxidation rate. The calculation formula is as follows: ,in, For the direct carbon flow of fuel combustion, This refers to fuel medium consumption. For the lower heating value of fuel, The carbon emission factor corresponding to the calorific value of fuel. Fuel carbon oxidation rate;

[0058] For nodes that consume multiple fuels simultaneously, the combustion carbon flow rate of each fuel is calculated separately and then summed to obtain the total direct carbon flow rate of fuel combustion at that node.

[0059] The second submodule is the indirect carbon flow calculation submodule for purchased electricity and heat. This submodule first identifies the purchased energy access identifiers in the directed edges of the energy branches, distinguishing between purchased electricity, purchased steam, and purchased hot water; it then couples the time-series boundary curve of the purchased energy carbon intensity to calculate the indirect carbon flow at each vertex. The calculation formula is as follows: ,in, Carbon flow indirect from purchased electric heating For the consumption of purchased electric heating, for The time-series value of carbon intensity for externally purchased energy;

[0060] For purchased energy consumption at different times, the carbon intensity time series value of the corresponding time period is used for calculation to accurately reflect the time-varying characteristics of carbon intensity of purchased energy.

[0061] The third submodule is the direct carbon flow calculation submodule for the process. This submodule first calculates the material conversion amount participating in the reaction based on the input and output material balance and component concentration of the directed edges of the material branches; then, it calculates the direct carbon flow of the reaction process at each vertex by combining the process reaction carbon emission factor and the material conversion rate. The calculation formula is as follows: ,in, The direct carbon flow rate in the process. The amount of material converted to participate in the reaction. The carbon emission factor corresponding to the process reaction. For material conversion rate, For the first The quality of carbon fixation by-products For the first Carbon content coefficient of carbon fixation byproducts This represents the total number of types of carbon sequestration byproducts.

[0062] The amount of carbon carried away by carbon fixation byproducts is deducted to ensure the accuracy of carbon flow rate calculation in the process.

[0063] The fourth submodule is the carbon flow calculation submodule for fluorinated greenhouse gases; this submodule comprehensively considers both pressure difference-driven emission and concentration diffusion emission mechanisms to calculate the carbon equivalent of fluorinated greenhouse gases at each vertex. The calculation formula is as follows: ,in, The carbon equivalent of fluorine-containing greenhouse gases emitted. The equivalent resistance coefficient of the leakage path. The pressure difference between the inside and outside of the equipment. The differential pressure-driven dissipation rate exponent, The concentration diffusion coefficient is... The difference in concentration of fluorinated greenhouse gases inside and outside the equipment, The equivalent area of ​​the leakage path. The duration of the scheduling time window, Weights are assigned to the global warming potential values ​​corresponding to fluorinated greenhouse gases.

[0064] For nodes containing multiple fluorinated greenhouse gases, the total fluorinated greenhouse gas carbon equivalent of the node is obtained by summing the efflux carbon equivalent of each gas after calculating the efflux carbon equivalent of each gas.

[0065] Step S2-4: Orthogonal superposition, integration, and verification of multidimensional carbon emission flow calculation models:

[0066] The output carbon flows of the four carbon flow calculation submodules are orthogonally superimposed in three dimensions: the first dimension is the vertex space topological position dimension, which corresponds to all abstract nodes of the product chain directed graph model; the second dimension is the scheduling time series window dimension, which corresponds to all preset scheduling periods; and the third dimension is the carbon flow generation mechanism dimension, which corresponds to the above four types of carbon flows.

[0067] After superposition, a multidimensional carbon emission flow calculation model is generated, which takes the set of vertices of the product chain directed graph model as the spatial basis, the scheduling period as the time basis, and the four types of carbon flow as the attribute dimensions. The output of this model forms a dynamic mapping relationship with the vertices and directed edges of the product chain directed graph model, forming a carbon emission flow structure coupled in three dimensions of spatiotemporal mechanism.

[0068] A global carbon balance check is performed on the completed multidimensional carbon emission flow calculation model. The total carbon input and total carbon output of the system are calculated. The total carbon input is the sum of the carbon brought in by raw materials and the carbon brought in by purchased energy. The total carbon output is the sum of the carbon brought out by products, the carbon brought out by carbon sequestration by-products, and the total carbon emissions. When the relative deviation between the total carbon input and total carbon output is less than one percent, the model is considered to have passed the check. If the deviation exceeds the allowable range, the parameter settings and calculation logic of each carbon flow calculation submodule are checked back, the errors are corrected, and the check is re-executed.

[0069] After completing model integration and verification, the parameter configuration file and calculation interface of the multidimensional carbon emission flow calculation model are saved to provide underlying carbon flow input for subsequent dynamic evolution of node carbon potential and carbon flow tracking.

[0070] In this embodiment, the core function of step S3 is to use the spatiotemporal mechanism three-dimensional coupled carbon emission flow quantification structure output by the multidimensional carbon emission flow calculation model. It employs an iterative algorithm combining forward evolution of node carbon potential and reverse tracing of branch carbon flow to resolve the ambiguity in carbon flow allocation caused by circulating backflow and collinear separation networks during multi-product co-production. This accurately calculates the dynamic carbon potential distribution of each process node at different scheduling periods and the carbon emission intensity distribution of each material and energy branch, providing core carbon intensity parameters for subsequent optimized scheduling. The detailed steps are as follows:

[0071] Step S3-1: Initial state construction and parameter setting for carbon flow tracking:

[0072] Using the spatiotemporal mechanism three-dimensional coupled carbon emission flow structure output by the multidimensional carbon emission flow calculation model as the tracking benchmark, the node operation parameters and carbon flow data at the start time of each time period are extracted sequentially according to the scheduling time window. The extracted content includes the input and output material balance of each node, the consumption of various energy media, the direct carbon flow of fuel combustion, the indirect carbon flow of purchased electricity and heat, the direct carbon flow of the process, and the carbon flow of fluorine-containing greenhouse gases.

[0073] Construct an initial carbon potential state matrix for nodes. The rows of the matrix correspond to all abstract nodes in the directed graph model of the product chain, and the columns of the matrix correspond to four types of carbon flow. The matrix elements are the initial carbon flow values ​​of the corresponding nodes and their corresponding carbon flow types. Construct an initial carbon flow allocation vector for branches. The vector dimension is equal to the total number of all material branches and energy branches in the directed graph model of the product chain. The vector elements are the initial carbon flow allocation values ​​of the corresponding branches. The initial allocation values ​​are evenly distributed according to the branch flow ratio.

[0074] Set the carbon potential evolution convergence tolerance and the maximum iteration period; set the carbon potential evolution convergence tolerance to 0.001 and the maximum iteration period to 100 times; establish an iteration process state monitoring mechanism to record the node carbon potential value and branch carbon flow intensity value in real time for each iteration, and provide data support for subsequent convergence determination.

[0075] Step S3-2: Forward evolution calculation of nodal carbon potential:

[0076] Based on the initial state matrix of node carbon potential and the initial allocation vector of branch carbon flow, the forward evolution calculation of node carbon potential is performed sequentially along the forward path of material flow and energy transfer in the directed graph of the product chain, in topological order.

[0077] Based on the principle of mass, energy, and carbon conservation, the total carbon flow input from upstream related material and energy branches is weighted and aggregated according to the corresponding branch flow ratio. This is then combined with the direct carbon flow from the process at this node and the carbon flow from the emission of fluorine-containing greenhouse gases. After deducting the carbon carried out by carbon fixation byproducts within the node, the local carbon potential baseline value for each node in the current time period is calculated. The formula for calculating the node carbon potential is: ,in, For the first Time period The carbon potential of each node, For the first The set of all input branches of a node For the first Time period The carbon flux intensity of each branch, For the first Time period Material or energy flow of the branch, For the first Time period The carbon flow rate is directly related to the process flow at each node. For the first Time period Fluorine-containing greenhouse gas carbon emission flux at each node For the first Time period Total output material flow of each node For the first Scheduling period The node generates the first The quality of carbon fixation by-products For the first Carbon content coefficient of carbon fixation byproducts;

[0078] After calculating the carbon potential of all nodes in topological order, a forward evolution sequence of node carbon potential is generated; this sequence contains the carbon potential values ​​of all nodes in the current iteration round, providing input for the reverse tracing of carbon flow in subsequent branches;

[0079] Step S3-3: Iterative calculation of branch carbon flow in reverse tracing:

[0080] Using the forward evolution sequence of node carbon potential as the input for reverse tracing, the reverse tracing iterative calculation of branch carbon flow is initiated for the circulating backflow branch and collinear separation network in the multi-product co-production process topology.

[0081] Using the planned carbon intensity of the product output node as the inversion anchor point, the process traces backward along the material branch to the upstream intermediate buffer node and utility node; based on the material allocation ratio, reaction selectivity coefficient, and energy cascade utilization rate, the accumulated carbon potential of the downstream node is proportionally distributed to each input branch; the branch carbon flow allocation calculation formula is as follows: ,in, The first obtained by reverse tracing Time period The carbon flux intensity of each branch, For the first Carbon potential at downstream nodes during the time period The set of all input branches of the downstream node;

[0082] The difference between the branch carbon flow intensity obtained from reverse tracing and the branch carbon flow intensity obtained from forward evolution is calculated to obtain the branch carbon flow allocation deviation. The carbon flow allocation deviation of all branches is summarized to generate a global carbon flow allocation deviation vector.

[0083] Step S3-4: Convergence determination of carbon flow tracing and spatiotemporal result splicing:

[0084] The branch carbon flow distribution deviation is fed back and superimposed onto the node carbon potential forward evolution sequence for closed-loop correction. The node carbon potential forward evolution calculation and branch carbon flow reverse source tracing and allocation steps are repeated.

[0085] Within each iteration cycle, the changes in the carbon potential values ​​of all nodes and the carbon flow intensity of branches obtained from two adjacent iterations are compared. When the maximum relative deviation between the carbon potential values ​​of all nodes and the carbon flow intensity of all branches is less than the convergence tolerance, or when the number of iterations reaches the maximum iteration cycle, the carbon flow tracking is determined to be converged. The steady-state solutions of the node carbon potential and the steady-state solutions of the branch carbon flow distribution that satisfy the global mass-energy carbon conservation constraints are output.

[0086] The node carbon potential steady-state solution and the branch carbon flow distribution steady-state solution after convergence in each scheduling period are dynamically spliced ​​and topologically aligned along the time series axis to construct a spatiotemporal coupled carbon flow tracking result set covering the entire process unit; the dynamic carbon potential distribution sequence of each process node in different scheduling periods, as well as the carbon emission intensity distribution sequence of each material branch and energy branch in different scheduling periods are extracted from the result set.

[0087] After completing carbon flow tracing, save the spatiotemporally coupled carbon flow tracing result set file to provide core data input for subsequent construction of dynamic carbon flow intensity matrix and establishment of multi-objective optimization scheduling model.

[0088] In this embodiment, the core function of step S4 is to perform multi-dimensional tensor mapping between the dynamic carbon potential distribution obtained from spatiotemporal coupled carbon flow tracking and the branch carbon emission intensity distribution, constructing a dynamic carbon flow intensity matrix that reflects the spatiotemporal evolution of carbon flow and its correlation with product flow direction, extracting time-varying carbon intensity coefficients as core feedback parameters, and establishing a dual-objective optimization scheduling model that minimizes the total carbon emissions and overall operating costs of the system. This model couples four types of key operational constraints, providing a rigorous mathematical model foundation for subsequently solving the optimal scheduling scheme. The detailed steps are as follows:

[0089] Step S4-1: Dynamic carbon flux intensity tensor mapping and time-varying carbon intensity coefficient extraction:

[0090] The dynamic carbon potential distribution and branch carbon emission intensity distribution in the spatiotemporally coupled carbon flow tracking result set are subjected to tensor mapping operation in three orthogonal dimensions: the first dimension is the time series dimension, corresponding to all scheduling periods; the second dimension is the spatial dimension, corresponding to all process unit nodes and material and energy branches in the directed graph model of the product chain; the third dimension is the product flow dimension, corresponding to all independent production paths from raw material input to each final product output in the multi-product co-production topology.

[0091] The third-order carbon flux intensity tensor is constructed using the following formula: ,in, For the third-order carbon flux intensity tensor, This represents the total number of time periods within the scheduling cycle. This represents the total number of process unit nodes. The total number of material and energy branches. This refers to the number of final product types.

[0092] The third-order carbon flow intensity tensor is subjected to dimensionality compression and matrix transformation to generate a dynamic carbon flow intensity matrix. The matrix rows correspond to each scheduling period, the matrix columns correspond to each final product, and the matrix elements are the cumulative carbon intensity values ​​of the corresponding products for the corresponding period. The time-varying carbon intensity coefficient sequence corresponding to each scheduling window is obtained by extracting the time-varying carbon intensity coefficients row by row from the dynamic carbon flow intensity matrix according to product category. A dynamic update mechanism for the time-varying carbon intensity coefficients is established, and the dynamic carbon flow intensity matrix and the time-varying carbon intensity coefficient sequence are automatically refreshed after each carbon flow tracking calculation is completed.

[0093] Step S4-2: Construction of the bi-objective optimization function and definition of scheduling decision variables:

[0094] The node production load allocation rate, branch material diversion ratio, and utility media call sequence are defined as scheduling decision variables; the node production load allocation rate represents the ratio of the actual operating load of each reaction conversion node and separation purification node to the design rated load; the branch material diversion ratio represents the ratio of the flow of each branch material branch to the total upstream output flow; the utility media call sequence represents the call volume and allocation scheme of various utility media in each time period.

[0095] Using the time-varying carbon intensity coefficient as a dynamic weighting parameter, the first optimization objective function is constructed, namely, the objective function for minimizing the total carbon emissions of the system. The calculation formula is as follows: ,in, This represents the total carbon emissions of the system during the scheduling period. For the first Time period Time-varying carbon intensity coefficient of the product For the first Time period The output of this product;

[0096] Construct a second optimization objective function, namely the objective function for minimizing overall operating cost. This function is obtained by weighted aggregation of four cost factors, and the calculation formula is as follows: ,in, The overall operating cost of the system within the scheduling period, The total cost of purchased energy, The total cost of raw material consumption. Total losses for equipment load regulation. The total cost of dynamic settlement of carbon allowances;

[0097] The specific calculation methods for each cost are as follows: The total cost of purchased energy is the sum of the products of the consumption of each type of purchased energy in each time period and the unit price of energy in the corresponding time period; the total cost of raw material consumption is the sum of the products of the total consumption of each type of raw material and the unit price of the corresponding raw material; the total loss of equipment load regulation is the sum of the products of the load change of each node in adjacent time periods and the unit load regulation loss coefficient; the total cost of dynamic carbon quota settlement is the difference between the total carbon emissions of the system and the total amount of free carbon quotas multiplied by the real-time trading price of the regional carbon market.

[0098] Step S4-3: Mathematical modeling and boundary setting of four types of operational constraints:

[0099] The scheduling decision variables are logically bound to the bi-objective optimization function, and four types of operational constraints are sequentially implanted to clarify the feasible domain boundaries of each decision variable.

[0100] The first type of constraint is the planned output constraint for products; this constraint achieves boundary control by limiting the deviation tolerance between the cumulative output of each product output node at the end of the scheduling cycle and the preset planned target. The calculation formula is as follows: ,in, For the first Minimum planned output of this product For the first The maximum planned output of this product;

[0101] The second type of constraint is the process node capacity and load ramping constraint; the capacity constraint limits the operating load of each node to not be lower than the minimum allowable load and not higher than the maximum allowable load; the load ramping constraint limits the change in node load between adjacent scheduling periods to not exceed the maximum allowable rate of change, and the calculation formula is: ,in, For the first Time period The operating load rate of each node, For the first The minimum allowable load rate for each node. For the first The maximum allowable load rate of each node For the first The maximum allowable load change rate for each node;

[0102] The third type of constraint is the energy carbon supply and demand coupling constraint. This constraint is based on the medium quality balance relationship of the directed edge of the energy branch, and links and matches the time series boundary curve of the carbon intensity of purchased energy with the energy cascade utilization network in the plant to ensure that the energy dispatch sequence and the carbon emission source intensity respond synchronously in the spatiotemporal dimension. The constraint requires that the total supply of various energy media in each period is equal to the total consumption, and the amount of purchased energy dispatch shall not exceed the maximum transmission capacity of the energy supply system.

[0103] The fourth type of constraint is the dynamic carbon quota constraint; this constraint combines the real-time quota allocation curve of the regional carbon market with the historical carbon footprint rollover mechanism to set a dynamic threshold range for allowable carbon emissions within each scheduling period. The calculation formula is as follows: ,in, For the first The minimum carbon flow allowed for emissions during a given period. For the first The maximum carbon flow allowed for emissions during a given period;

[0104] Step S4-4: Structured encapsulation and verification of the multi-objective optimization scheduling model:

[0105] The bi-objective optimization function and four types of operational constraints are encapsulated in a structured manner to form a multi-objective optimization scheduling model with time-varying carbon intensity coefficient as the pre-feedback parameter. The solution input of this model receives a standardized sequence of scheduling decision variables, and the output generates a set of candidate scheduling schemes that satisfy global carbon flow conservation and the economic feasible region.

[0106] Three checks are performed on the completed multi-objective optimization scheduling model: the first is constraint consistency check, which checks whether there are any contradictions between the constraints and ensures that the feasible region is not empty; the second is carbon flow conservation check, which verifies that the relative deviation between the total carbon emissions of the system calculated by the model and the total carbon emissions output by the multi-dimensional carbon emission flow calculation model is less than one percent; the third is economic feasibility check, which ensures that the overall operating cost of the candidate scheduling schemes output by the model is within a reasonable range.

[0107] If the verification fails, backtrack to check the parameter settings of the objective function and the boundary settings of the constraints, correct the errors, and re-execute the verification; if the verification passes, save the configuration file and solution interface of the multi-objective optimization scheduling model to provide a model foundation for the execution of subsequent multi-objective optimization solution algorithms.

[0108] In this embodiment, the core function of step S5 is to iteratively solve the dual-objective optimization scheduling model using a multi-objective co-evolutionary algorithm with an adaptive penalty mechanism that incorporates carbon constraints, generating a Pareto optimal scheduling solution set that represents the boundary between carbon emission reduction and economic efficiency. Then, a fuzzy preference projection decision method is used to transform the decision-maker's subjective preferences into objective decision-making criteria. The optimal scheduling scheme is then selected from the solution set and decoded into an executable instruction sequence, providing a clear control basis for subsequent production scheduling execution. The detailed steps are as follows:

[0109] Step S5-1: Initial scheduling population generation and algorithm parameter setting:

[0110] The sequence of scheduling decision variables in the multi-objective optimization scheduling model is encoded with real numbers to generate an initial scheduling population. Each individual in the population corresponds to a complete scheduling scheme, and its encoding dimension is equal to the sum of the number of node production load allocation rates, the number of branch material diversion ratios, and the length of the utility media call sequence. The initial population is generated using the Latin hypercube sampling method to ensure that the initial individuals are evenly distributed within the feasible domain of the decision variables and to avoid premature convergence of the population.

[0111] The baseline values ​​of the bi-objective fitness of each scheduling scheme within the population are initialized based on the time-varying carbon intensity coefficient; the total carbon emissions and comprehensive operating costs of the system corresponding to each individual are calculated and used as the individual's bi-objective fitness value; the algorithm iteration termination condition and constraint violation tolerance threshold are set; the maximum number of iterations is set to 200, and the algorithm is considered to have converged when the hypervolume change rate of the Pareto front is less than 0.001 for 30 consecutive generations; the constraint violation tolerance threshold is set to 0.01 to distinguish between feasible and infeasible solutions.

[0112] Step S5-2: Carbon Constraint Adaptive Penalty Mechanism and Feasible Domain Guidance:

[0113] Based on the initial scheduling population, an adaptive penalty mechanism for carbon constraints is introduced to perform hierarchical mapping processing on four types of operational constraints; the product planned output constraint and the process node capacity load ramping constraint are used as hard constraint boundaries to directly filter out all individuals that violate the hard constraints, ensuring that the remaining individuals meet the basic physical constraints of the production process; the energy carbon supply and demand coupling constraint and the dynamic carbon quota constraint are transformed into soft constraint penalty terms and added to the fitness function of the individuals.

[0114] Construct an adaptive penalty function, calculated as follows: ,in, For individuals The fitness value after punishment. For individuals Total constraint violations For the first An adaptive penalty factor for iteration;

[0115] The total constraint violation is the weighted sum of the energy carbon supply and demand coupling constraint violation and the dynamic carbon quota constraint violation; the adaptive penalty factor is dynamically updated with the number of iterations and the population carbon quota deviation, and the calculation formula is: ,in, As the initial penalty factor, The factor representing the influence of the number of iterations. This is the coefficient representing the influence of carbon quota deviation. For the first Cumulative deviation of average carbon quota from generation population;

[0116] The cumulative deviation of the current iteration population from the dynamic carbon quota threshold is statistically analyzed in real time. Based on this, the growth slope of the penalty factor is adaptively adjusted so that the fitness value of the high deviation scheme decays non-linearly, guiding the population to evolve towards the feasible region that satisfies the carbon constraint, and outputting a set of candidate scheduling schemes that meet the feasible region requirements.

[0117] Step S5-3: Multi-objective co-evolutionary iteration and Pareto optimal solution set generation:

[0118] Multi-objective co-evolutionary iterations are performed based on a set of candidate scheduling schemes. In each iteration, the population is first sorted using a fast non-dominated ordering method, dividing the individuals into different non-dominated levels. Then, the crowding degree of individuals within each non-dominated level is calculated using the following formula: ,in, For the first The degree of crowding of individuals For the first The objective function value, For the first The maximum value of each objective function. For the first The minimum value of an objective function. In the same non-dominated hierarchy, according to the first After sorting the nth objective function value, the th... The first adjacent individual after the first individual The objective function value, In the same non-dominated hierarchy, according to the first After sorting the nth objective function value, the th... The first adjacent individual of the previous individual One objective function value;

[0119] Individuals are selected based on non-dominated level and crowding, with priority given to individuals with low non-dominated level and high crowding to enter the next generation of the population; adaptive crossover and recombination and local mutation perturbation operations are performed on the selected individuals; the crossover probability and mutation probability are adaptively adjusted with the number of iterations, with higher crossover probability and mutation probability used in the early stage to enhance the global search capability, and lower crossover probability and mutation probability used in the later stage to enhance the local search capability.

[0120] Update the population state and recalculate the bi-objective fitness value and the penalized fitness value of all individuals; iteratively perform non-dominated sorting, individual selection, crossover and recombination and mutation operations until the population front distribution is stable or the preset maximum number of iterations is reached; after convergence, generate a Pareto optimal scheduling solution set that represents the boundary of the trade-off between carbon emission reduction efficiency and operational economy.

[0121] Step S5-4: Fuzzy Preference Projection Decision and Optimal Scheduling Instruction Sequence Decoding:

[0122] A fuzzy preference projection decision space is constructed for the Pareto optimal scheduling solution set; the carbon economy preference weights preset by the decision-maker are converted into fuzzy membership mapping rules for the dual-objective space; a preference projection reference axis is set with the ideal zero carbon emission extreme value and the lower limit of economic cost as anchor points, and the coordinates of the ideal point are: ,in, For an ideal point in the dual-objective space, This represents the minimum total carbon emissions of the Pareto centralized system. This represents the minimum overall operating cost of the Pareto solution set;

[0123] The fuzzy projection membership values ​​of each scheme in the Pareto optimal scheduling solution set to the preferred projection reference axis are calculated using the following formula: ,in, For the first The fuzzy projection membership values ​​of a Pareto solution. For the first The coordinate vectors of the Pareto solutions in the dual-objective space;

[0124] Based on the membership maximization criterion, the solution set is optimized by dimensionality reduction to lock the single scheduling scheme with the highest comprehensive score of carbon emission reduction efficiency and operation economy; the locked single scheduling scheme is decoded and reverse mapped according to the scheduling time window slice and the process topology node sequence to restore the load allocation rate, material diversion ratio and energy medium call sequence into a time-series control parameter set;

[0125] Assemble and generate the optimal scheduling instruction sequence, which includes time stamp identifiers, node setting adjustment instructions, and utility switching instructions; complete the closed-loop solution and decision output of the multi-objective optimization scheduling model, and save the Pareto optimal scheduling solution set file and the optimal scheduling instruction sequence file to provide data support for subsequent production scheduling execution.

[0126] In this embodiment, the core function of step S6 is to transform the optimal scheduling instruction sequence into an executable control task on-site. Through real-time status feedback and dynamic compensation mechanisms, continuous rolling optimization of production task allocation, process parameter setting, and energy configuration for multi-product joint production lines is achieved, ensuring that the scheduling scheme is accurately implemented in actual operation. At the same time, it can cope with random disturbances and operating condition fluctuations in the production process, maintaining the system at the optimal balance point between carbon emission reduction and economic efficiency. The detailed steps are as follows:

[0127] Step S6-1: Optimal scheduling instruction parsing and underlying control mapping:

[0128] The optimal scheduling instruction sequence is parsed and mapped at the underlying level. The instructions are decomposed into independent control units according to the scheduling time window slice. Each control unit contains three core contents: time stamp identifier, node setpoint adjustment instruction, and utility switching instruction. The time stamp identifier is used to clarify the execution time window of the instruction. The node setpoint adjustment instruction includes the temperature and pressure setpoints of each reaction conversion node, the reflux ratio setpoints of each separation and purification node, and the liquid level setpoints of each intermediate buffer node. The utility switching instruction includes the allocation and switching status of steam, electricity, circulating water, and compressed air in each time period.

[0129] The system performs format conversion and address binding through the communication protocol interface between the production management system and the field distributed control system; supported communication protocols include Industrial Ethernet, Modbus, and OPC UA; it maps the parsed control parameters one-to-one with the input addresses of the field process unit controllers, and maps the utility switching commands one-to-one with the control addresses of the energy dispatch valve groups; it generates an initial execution task list that corresponds one-to-one with the process unit controllers and energy dispatch valve groups, specifying the execution time, execution object, and execution parameters of each control task;

[0130] Step S6-2: Time-sequential rolling execution and multi-dimensional operation status collection:

[0131] The sequential rolling execution process is initiated based on the initial task list; at the start of the current scheduling period, the corresponding control sub-sequence is sent to the field process unit controller and energy dispatch valve group; the controller automatically adjusts the equipment operating status according to the preset parameters, and the energy dispatch valve group adjusts the flow rate and direction of each energy medium according to the instructions;

[0132] Simultaneously, multi-dimensional actual operation data is collected through the production management system, with the collection frequency consistent with the scheduling time window. The collected data includes the actual flow rate and component concentration of each material branch, the actual consumption of each energy medium, the actual operating pressure and temperature of each key process node, and the actual product output and quality indicators of each product output node. All collected data is timestamped to ensure that the time base of all data is consistent. An actual operation status feedback sequence is constructed to provide data support for subsequent deviation assessment.

[0133] Step S6-3: Triggering of operational deviation assessment and online dynamic compensation mechanism:

[0134] The actual operational status feedback sequence is spatiotemporally aligned and compared with the expected target value in the optimal scheduling instruction sequence to calculate three types of core operational deviations; the first type is the production task allocation deviation, calculated using the following formula: ,in, For the first Time period Deviation in production task allocation for this product For the first Time period The actual output of this product For the first Time period Target output for this product;

[0135] The second category is the deviation of process node parameters, calculated using the following formula: ,in, For the first Time period Deviation of process parameters at each node, For the first Time period The actual operating parameter values ​​of each node For the first Time period Target operating parameter values ​​for each node;

[0136] The third category is the energy carbon supply-demand coupling deviation, calculated using the following formula: ,in, For the first Energy carbon supply and demand coupling deviation during a certain period For the first Actual carbon emissions during the period For the first Target carbon emissions for the specified period;

[0137] Three types of deviations are preset with dynamic tolerance thresholds. When the amplitude of any type of deviation exceeds the corresponding tolerance threshold, an online dynamic compensation mechanism is triggered. The lightweight real-time deduction module of the node carbon potential dynamic evolution equation is called to update the branch carbon flow allocation weight in combination with the current actual deviation amount. Based on the updated carbon flow allocation weight, the optimal load allocation of each node and the optimal material diversion ratio of each branch are recalculated to generate a dynamic adjustment instruction set that includes the load redistribution ratio and parameter compensation correction amount.

[0138] Step S6-4: Scrolling window optimization and closed-loop control link construction:

[0139] The dynamic adjustment instruction set is injected as a pre-feedback parameter into the multi-objective optimization scheduling model of the next rolling time window; the time-varying carbon intensity coefficient in the dynamic carbon flow intensity matrix is ​​updated in real time, and the dynamic carbon quota constraint boundary is adjusted according to the actual carbon emissions and the remaining carbon quota; the multi-objective optimization solution process is re-executed to generate an updated rolling scheduling instruction subsequence.

[0140] The updated rolling scheduling instruction subsequence is seamlessly switched to the production management system for continued execution, replacing the original instruction for the next scheduled period. The complete process of instruction issuance, status acquisition, deviation assessment, carbon potential simulation, parameter compensation, and window rolling is repeated, forming a closed-loop control link for instruction issuance, status acquisition, deviation assessment, carbon potential simulation, parameter compensation, and window rolling.

[0141] Establish an emergency response mechanism for abnormal operating conditions. When the system experiences extreme situations such as major equipment failure or raw material supply interruption, the automatic scheduling process will be automatically suspended and switched to manual intervention mode. After the abnormal operating conditions are resolved, the system status will be reinitialized and the automatic rolling optimization scheduling will be restored.

[0142] After the entire scheduling cycle is completed, a scheduling execution effect evaluation report is generated. The actual operation data is compared with the scheduling target data to analyze the carbon emission reduction effect and cost saving effect. The operation data and optimization results of this scheduling are stored in the historical database to provide data support for the parameter optimization and iterative upgrade of the subsequent scheduling model.

[0143] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for carbon emission flow tracking and optimized scheduling in a multi-product co-production process of fluorochemicals, characterized in that: Includes the following steps: S1: Obtain process topology data, time-series material flow data, and time-series energy flow data for the co-production of multiple products in fluorochemicals. After data cleaning and feature mapping, construct a directed graph model of the product chain with process units as vertices and material branches and energy branches as directed edges. Vertices include raw material input nodes, reaction and conversion nodes, separation and purification nodes, product output nodes, utility nodes, and intermediate buffer nodes. S2: Based on the product chain directed graph model, the input and output material balance, energy consumption, process reaction carbon emission factor and fluorine-containing greenhouse gas emission kinetic parameters of each process node are extracted to construct a multi-dimensional carbon emission flow calculation model that includes direct carbon flow from fuel combustion, indirect carbon flow from purchased electricity / heat, direct carbon flow from the process and carbon flow from fluorine-containing greenhouse gas emission. S3: The dynamic evolution equation of node carbon potential based on mass-energy-carbon conservation and the reverse source tracing iterative algorithm of branch carbon flow are used to perform spatiotemporal coupling carbon flow tracing on the directed graph model of the product chain, and calculate the dynamic carbon potential distribution of each process node at different scheduling periods and the carbon emission intensity distribution of each material branch and energy branch. S4: The dynamic carbon potential distribution and the branch carbon emission intensity distribution are tensor-mapped according to the time series and product flow direction to construct a dynamic carbon flow intensity matrix. The time-varying carbon intensity coefficient output by the dynamic carbon flow intensity matrix is ​​used as the core parameter to establish a multi-objective optimization scheduling model with the dual objectives of minimizing the total carbon emissions of the system and minimizing the comprehensive operating cost. The multi-objective optimization scheduling model is coupled with product planned output constraints, process node capacity and load ramping constraints, energy-carbon supply and demand coupling constraints, and dynamic carbon quota constraints. S5: The multi-objective optimization solution algorithm with carbon constraint adaptive penalty mechanism and fuzzy preference projection decision is used to iteratively solve the multi-objective optimization scheduling model to obtain the Pareto optimal scheduling solution set, and the optimal scheduling instruction sequence is extracted from the Pareto optimal scheduling solution set based on the carbon-economic preference weight preset by the decision-maker. S6: Send the optimal scheduling instruction sequence to the production management system to perform rolling optimization and dynamic adjustment of production task allocation, key process node operating parameter settings, and energy medium configuration schemes for multi-product joint production lines.

2. The carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to claim 1, characterized in that: The process topology data, time-series material flow data, and time-series energy flow data of the multi-product co-production process in the fluorochemical industry are acquired. After data cleaning and feature mapping, a directed graph model of the product chain is constructed, with process units as vertices and material and energy branches as directed edges. Specifically, this includes: The process topology data associated with the production process piping instrumentation diagram, the time-series records of mass flow rate and component concentration of each material pipeline, and the time-series records of medium consumption of each utility are extracted from the production execution system, energy management system and real-time database of the multi-product co-production process in fluorochemicals. The extracted data is then filled with missing values, abnormal operating conditions are removed and multi-source data timestamps are aligned to obtain a standardized set of process topology data, a standardized set of time-series material flow data and a standardized set of time-series energy flow data. The process function clustering and identification of each physical device in the standardized process topology data set is performed. The physical devices are mapped to six types of abstract nodes: raw material input nodes, reaction and conversion nodes, separation and purification nodes, product output nodes, utility nodes and intermediate buffer nodes. The pipelines connecting each physical device are mapped to directed edges of material branches, and the energy supply pipelines are mapped to directed edges of energy branches. A bidirectional mapping index table between physical device identifiers, node category identifiers and abstract node identifiers is established. The standardized time-series material flow data set drives the dynamic assignment of flow and direction attributes of directed edges of each material branch, and the standardized time-series energy flow data set drives the dynamic assignment of energy consumption medium type and consumption intensity attributes of directed edges of each energy branch. At the same time, based on the connection relationship of physical equipment in the process topology data, the abstract nodes with flow direction association are cascaded in sequence according to the material flow direction and energy transfer direction to construct a product chain directed graph model that reflects the coupling topology of material flow and energy flow in the entire process of multi-product production line. In the product chain directed graph model, the starting and ending vertices of the directed edge of any material branch belong to the set of abstract nodes, and the direction of the directed edge of the material branch corresponds to the actual direction of the material in the process of multi-product co-production in fluorochemicals; the ending vertex of the directed edge of any energy branch is the utility node or reaction conversion node that consumes the energy, and the starting vertex is the purchased energy input node or energy output node.

3. The carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to claim 1, characterized in that: Based on the product chain directed graph model, the input and output material balance, energy consumption, process reaction carbon emission factor, and fluorinated greenhouse gas emission kinetic parameters of each process node are extracted. A multi-dimensional carbon emission flow calculation model is constructed, including direct carbon flow from fuel combustion, indirect carbon flow from purchased electricity / heat, direct carbon flow from the process, and carbon flow from fluorinated greenhouse gases. Specifically, it includes: Based on the product chain directed graph model, each vertex and its associated material branch directed edges and energy branch directed edges are traversed. The input and output material balance and energy consumption of each vertex are extracted from the standardized time-series material flow data set and the standardized time-series energy flow data set according to the scheduling time window slice. The process reaction type, equipment sealing level and operating condition attribute corresponding to the physical equipment identifier are bound to the corresponding vertex in combination with the bidirectional mapping index table. The system calls upon the pre-set knowledge base of carbon emission factors in the fluorochemical industry and the library of fluorocarbon greenhouse gas emission kinetic parameters to map the process reaction type to the process reaction carbon emission factor, and to map the equipment sealing level, peak real-time operating pressure and temperature attributes to the fluorocarbon greenhouse gas emission kinetic parameters. The fluorocarbon greenhouse gas emission kinetic parameters include the equivalent resistance coefficient of the leakage path, the pressure difference-driven emission rate index and the weight of the global warming potential value. Using the directed edges of the material branch and the directed edges of the energy branch as carbon flow transport channels, submodules for calculating direct carbon flow from fuel combustion, indirect carbon flow from purchased electricity / heat, direct carbon flow from the process, and carbon flow from fluorinated greenhouse gas emissions are constructed respectively. The direct carbon flow calculation submodule from fuel combustion calculates the direct carbon flow of combustion at each vertex based on the fuel medium type and consumption intensity attributes in the directed edges of the energy branch, combined with the carbon oxidation rate. The indirect carbon flow calculation submodule from purchased electricity / heat calculates the indirect carbon flow at each vertex based on the purchased energy access identifier and energy consumption medium type in the directed edges of the energy branch, coupled with the time-series boundary curve of the purchased energy carbon intensity. The direct carbon flow calculation submodule from the process calculates the direct carbon flow of the reaction process at each vertex based on the input and output material balance and component concentration in the directed edges of the material branch, combined with the carbon emission factor of the process reaction and the material conversion rate. The carbon flow calculation submodule from fluorinated greenhouse gas emissions calculates the carbon equivalent of fluorinated greenhouse gas emissions at each vertex based on the real-time operating pressure, temperature attributes, and fluorinated greenhouse gas emission kinetic parameters of the vertex, through a pressure difference-driven and concentration diffusion coupling mechanism. The output carbon flows of the four carbon flow calculation submodules are orthogonally superimposed according to the vertex spatial topological position, scheduling time series window, and carbon flow generation mechanism dimension to generate a multidimensional carbon emission flow calculation model with the vertex set of the product chain directed graph model as the spatial basis, the scheduling period as the time basis, and the four types of carbon flows as the attribute dimensions. The output of the multidimensional carbon emission flow calculation model forms a dynamic mapping relationship with the vertices and directed edges of the product chain directed graph model, forming a carbon emission flow structure coupled in three dimensions of spatiotemporal-mechanism, providing the underlying carbon flow input for the dynamic evolution of carbon potential of subsequent nodes.

4. The carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to claim 1, characterized in that: A dynamic evolution equation for node carbon potential based on mass-energy-carbon conservation and an iterative algorithm for reverse tracing of branch carbon flow are used to perform spatiotemporally coupled carbon flow tracking on a directed graph model of the product chain. This calculates the dynamic carbon potential distribution of each process node during different scheduling periods and the carbon emission intensity distribution of each material and energy branch, specifically including: Using the spatiotemporal-mechanistic three-dimensional coupled carbon emission flow structure output by the multidimensional carbon emission flow calculation model as the tracking benchmark, the node material balance, energy consumption and four types of carbon flow at the beginning of each scheduling period are extracted to construct the node carbon potential initial state matrix and the branch carbon flow initial allocation vector, and the carbon potential evolution convergence tolerance and maximum iteration period are set. Based on the initial state matrix of node carbon potential and the initial allocation vector of branch carbon flow, the forward evolution calculation of node carbon potential is performed along the forward path of material flow and energy transfer in the directed graph of the product chain. According to the principle of mass-energy-carbon conservation, the total amount of carbon flow input from upstream associated material branches and energy branches is weighted and aggregated according to the flow ratio of the corresponding branches. The direct carbon flow from the process reaction at this node and the carbon flow from the emission of fluorine-containing greenhouse gases are superimposed, and the amount of carbon carried out by carbon fixation by-products within the node is deducted. The local carbon potential baseline value of each node in the current time period is calculated, and the forward evolution sequence of node carbon potential is generated. Using the forward evolution sequence of node carbon potential as the input for reverse tracing, the reverse tracing iterative calculation of branch carbon flow is initiated for the circulating return branch and collinear separation network in the multi-product co-production process topology. Taking the planned carbon intensity of the product output node as the inversion anchor point, the calculation is carried out by tracing back along the material branch to the upstream intermediate buffer node and utility node. According to the material allocation ratio, reaction selectivity coefficient and energy cascade utilization rate, the cumulative carbon potential of the downstream node is proportionally and reversely allocated to each input branch, and the branch carbon flow allocation deviation is calculated. The branch carbon flow distribution deviation is fed back and superimposed onto the forward evolution sequence of the node carbon potential for closed-loop correction. The forward evolution calculation and reverse source tracing and allocation steps are repeated. In each iteration cycle, the changes in the node carbon potential value and the branch carbon flow intensity obtained from the two adjacent iterations are compared. When the maximum relative deviation is less than the convergence tolerance or the maximum iteration cycle is reached, the carbon flow tracking is determined to be converged. The steady-state solutions of the node carbon potential and the branch carbon flow distribution that satisfy the global mass-energy-carbon conservation constraints are output. The steady-state solutions of node carbon potential and branch carbon flow distribution after convergence in each scheduling period are dynamically spliced ​​and topologically aligned along the time series axis to construct a spatiotemporally coupled carbon flow tracking result set covering the entire process unit. From this set, the dynamic carbon potential distribution sequence of each process node in different scheduling periods, as well as the carbon emission intensity distribution sequence of each material branch and energy branch in different scheduling periods, are extracted to complete the spatiotemporally coupled carbon flow tracking of the directed graph model of the product chain.

5. The carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to claim 1, characterized in that: The dynamic carbon potential distribution and branch carbon emission intensity distribution are tensor-mapped with product flow direction according to time series to construct a dynamic carbon flow intensity matrix. Using the time-varying carbon intensity coefficient output from the dynamic carbon flow intensity matrix as the core parameter, a multi-objective optimization scheduling model is established with the dual objectives of minimizing total system carbon emissions and minimizing overall operating costs. This multi-objective optimization scheduling model couples product planned output constraints, process node capacity and load ramping constraints, energy-carbon supply and demand coupling constraints, and dynamic carbon quota constraints. Specifically, it includes: The dynamic carbon potential distribution and branch carbon emission intensity distribution are subjected to tensor mapping operation with time series as the time dimension, process unit node and material and energy branch as the spatial dimension, and multi-product co-production topology path as the product flow dimension to generate a dynamic carbon flow intensity matrix. The time-varying carbon intensity coefficients corresponding to each scheduling window are extracted from the dynamic carbon flow intensity matrix according to product category. Using the time-varying carbon intensity coefficient as a dynamic weight parameter, the node production load allocation rate, branch material diversion ratio, and utility media call sequence are defined as scheduling decision variables. A dual-objective optimization function is constructed. The objective of minimizing the total carbon emissions of the system is quantified by the time-varying carbon intensity coefficient and the corresponding branch carbon flow during the scheduling cycle. The objective of minimizing the overall operating cost is quantified by the weighted aggregation of external energy costs, raw material consumption costs, equipment load adjustment losses, and dynamic carbon quota settlement costs. The scheduling decision variables are logically bound to the bi-objective optimization function, and four types of operational constraints are sequentially incorporated: product planned output constraints achieve boundary control by limiting the deviation tolerance between the cumulative output of each product output node at the end of the scheduling cycle and the preset planned target; process node capacity and load ramping constraints physically limit the transient adjustment range of scheduling decision variables based on the design capacity extreme values ​​of each reaction conversion node and separation and purification node and the load change rate threshold between adjacent scheduling periods; energy-carbon supply and demand coupling constraints are based on the medium quality balance relationship of the directed edge of the energy branch, and link the time series boundary curve of carbon intensity of purchased energy with the energy cascade utilization network in the plant to ensure that the energy call sequence and carbon emission source intensity respond synchronously in the spatiotemporal dimension; dynamic carbon quota constraints combine the real-time quota allocation curve of the regional carbon market and the historical carbon footprint rolling carry-over mechanism to set the dynamic threshold range of carbon flow allowed to be emitted in each scheduling period. The bi-objective optimization function and four types of operational constraints are structurally encapsulated to form a multi-objective optimization scheduling model with time-varying carbon intensity coefficient as the pre-feedback parameter. The solution input of this model receives a standardized sequence of scheduling decision variables, and the output generates a set of candidate scheduling schemes that satisfy global carbon flow conservation and the economic feasible region, thus completing the construction of the multi-objective optimization scheduling model.

6. The carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to claim 1, characterized in that: A multi-objective optimization algorithm incorporating a carbon-constrained adaptive penalty mechanism and fuzzy preference projection decision is used to iteratively solve the multi-objective optimization scheduling model, obtaining a Pareto optimal scheduling solution set. Based on the decision-maker's pre-defined carbon-economic preference weights, the optimal scheduling instruction sequence is extracted from the Pareto optimal scheduling solution set, specifically including: The scheduling decision variable sequence in the multi-objective optimization scheduling model is encoded with real numbers to generate an initial scheduling population. The bi-objective fitness baseline value of each scheduling scheme in the population is initialized according to the time-varying carbon intensity coefficient. The algorithm iteration termination condition and constraint violation tolerance threshold are set. Based on the initial scheduling population, an adaptive penalty mechanism for carbon constraints is introduced to perform hierarchical mapping of four types of operational constraints. Product planned output constraints and process node capacity load ramping constraints are used as hard constraint boundaries to directly filter out out-of-bounds schemes. Energy-carbon supply and demand coupling constraints and dynamic carbon quota constraints are transformed into soft constraint penalty terms. The cumulative deviation of the current iteration population from the dynamic carbon quota threshold is counted in real time, and the growth slope of the penalty factor is adaptively adjusted accordingly to make the fitness value of high deviation schemes decay non-linearly, and output a set of candidate scheduling schemes that meet the feasible region guidance. Based on the candidate scheduling scheme set, multi-objective cooperative evolution iteration is performed. In each iteration, non-dominated relationship is determined and the spatial distribution of congestion is evaluated according to the objectives of minimizing the total carbon emissions of the system and minimizing the comprehensive operating cost. Non-dominated front individuals are selected for adaptive crossover and local mutation perturbation, the population state is updated and the fitness of the two objectives is recalculated. The evolution operation is performed cyclically until the distribution of the front of the population is stable or the iteration termination condition is reached, and the Pareto optimal scheduling solution set representing the carbon-economic trade-off boundary is generated. A fuzzy preference projection decision space is constructed for the Pareto optimal scheduling solution set. The carbon-economic preference weights preset by the decision-maker are converted into fuzzy membership degree mapping rules for the dual-objective space. A preference projection reference axis is set with the ideal zero carbon emission extreme value and the lower limit of economic cost as anchor points. The fuzzy projection membership degree values ​​of each scheme in the Pareto optimal scheduling solution set to the preference projection reference axis are calculated. The solution set is dimensionality reduced and optimized according to the membership degree maximization criterion to lock the single scheduling scheme with the highest comprehensive score of carbon emission reduction efficiency and operational economy. The locked single scheduling scheme is decoded and reverse-mapped according to the scheduling time window slice and the process topology node sequence. The load allocation rate, material diversion ratio and energy medium call sequence are restored into a time-sequential control parameter set. The optimal scheduling instruction sequence containing time stamp identifiers, node setpoint adjustment instructions and utility switching instructions is assembled to complete the closed-loop solution and decision output of the multi-objective optimization scheduling model.

7. The carbon emission flow tracking and optimized scheduling method for a multi-product co-production process in fluorochemicals according to claim 1, characterized in that: The optimal scheduling instruction sequence is sent to the production management system to continuously optimize and dynamically adjust the production task allocation, key process node operating parameter settings, and energy medium configuration schemes for multi-product joint production lines. Specifically, this includes: The optimal scheduling instruction sequence is parsed and mapped at the underlying level. The time stamp identifier, node set value adjustment instruction and utility switching instruction are decoupled into executable control sub-sequences according to the scheduling time window slice. The format is converted and the address is bound through the communication protocol interface between the production management system and the field distributed control system, and an initial execution task list corresponding to the process unit controller and energy scheduling valve group is generated. Based on the initial task list, the sequential rolling execution and real-time status acquisition are initiated. During the current scheduling period, the control subsequence is sent to the corresponding process topology node. Simultaneously, the actual material branch flow, actual energy medium consumption, operating pressure and temperature of key process nodes, and actual product output components are collected through the production management system to construct the actual operating status feedback sequence. The actual operating status feedback sequence is spatiotemporally aligned and compared with the expected target value in the optimal scheduling instruction sequence. The deviation of production task allocation, deviation of process node parameters and deviation of energy-carbon supply and demand coupling are calculated. When the deviation amplitude exceeds the preset dynamic tolerance threshold, the online dynamic compensation mechanism is triggered. The lightweight real-time deduction module of the node carbon potential dynamic evolution equation is called. The branch carbon flow allocation weight is updated in reverse by combining the current actual deviation amount, and a dynamic adjustment instruction set including load redistribution ratio and parameter compensation correction amount is generated. The multi-objective optimization scheduling model injects the dynamically adjusted instruction set as a pre-feedback parameter into the next rolling time window, updates the time-varying carbon intensity coefficient and dynamic carbon quota constraint boundary in real time, re-executes the multi-objective optimization solution process to generate an updated rolling scheduling instruction subsequence, and seamlessly switches this rolling scheduling instruction subsequence to the production management system for continued execution, forming a closed-loop control link of instruction issuance - status acquisition - deviation assessment - carbon potential deduction - parameter compensation - window rolling, completing the continuous rolling optimization and dynamic adjustment of production task allocation for multi-product joint production lines, setting of key process node operating parameters, and energy medium configuration scheme.

Images