A ccus source-sink matching network optimization method considering uncertainty

Abstract

The application discloses a CCUS source-sink matching network optimization method considering uncertainty, and belongs to the field of energy system planning and computer optimization algorithm. The application constructs a double-layer optimization architecture which is completely decoupled in discrete-continuous. The application innovatively reconfigures the objective function of the upper layer simulated annealing module, so that it is changed from solving the static cost optimization of the source-sink network to solving the dynamic stability optimization. At the same time, an intermediate uncertainty module is introduced, and optimization is performed based on information gap decision theory (IGDT). The robustness boundary and opportunity threshold of the system are directly quantified by constructing an extreme parameter set. Finally, a feature feedback mechanism based on adaptive weight is established, and the key topological features of high-quality schemes are used to dynamically guide the algorithm to converge efficiently. The application effectively avoids the curse of dimensionality of large-scale planning, and provides adaptive dynamic risk decision support for the transformation of the CCUS project from the start-up period to the mature period.

Description

Technical Field

[0001] This invention belongs to the field of energy system planning and computer optimization algorithms, specifically relating to an optimization method for CCUS source-sink matching networks that considers uncertainty. More particularly, this invention relates to a two-layer coupled optimization technique based on Information Gap Decision Theory (IGDT) and Simulated Annealing (SA) algorithm, used to solve complex decision-making problems involving the coupling of strategic site selection, infrastructure construction, and operational traffic scheduling in long-term, large-scale CCUS cluster planning. Background Technology

[0002] Currently, carbon capture, utilization, and storage (CCUS) technology has become a key supporting technology for the deep low-carbon transformation of the energy system. For centralized high-emission industries such as coal-fired power plants, steel, and cement, CCUS not only provides a crucial emission reduction pathway but also builds a future low-carbon industrial chain through the resource utilization of carbon dioxide (such as EOR enhanced oil recovery). In the entire CCUS chain, source-sink matching is the most strategically significant link. It determines how captured CO2 is rationally allocated spatially to different storage or utilization ports, and its planning results directly determine pipeline layout, investment scale, and long-term operating costs. Reasonable source-sink matching can not only reduce transportation distances and fixed investments but also achieve a balance between economics and capacity among different types of carbon sinks, which is a prerequisite for the large-scale commercial deployment of CCUS.

[0003] However, CCUS projects typically have a lifecycle of 20 to 30 years, during which time the external environment faces severe challenges of "deep uncertainty." Fluctuations in carbon trading prices, the iterative costs of storage technology, the continuity of policy subsidies, and the volatility of EOR returns due to oil price changes are all highly unpredictable. This long-term deep uncertainty makes traditional static programming methods based on deterministic parameters unsuitable for practical engineering needs.

[0004] However, existing CCUS source-sink matching and pipeline planning technologies have the following three main shortcomings, making it difficult to meet the actual needs of large-scale cluster planning:

[0005] First, existing planning models often employ a single deterministic cost objective and a tightly coupled architecture, making it difficult to directly optimize the system's risk boundary. Most existing technologies construct single-layer models based on mixed-integer linear programming (MILP), using "minimizing the expected total system cost" as the sole optimization objective, and tightly coupling discrete variables such as location selection with continuous variables such as traffic allocation. This modeling approach has two limitations: First, the strong coupling between location selection and traffic leads to an exponential increase in computational complexity with node size, resulting in NP-hard problems and low solution efficiency in large-scale network planning. Second, the single deterministic cost objective ignores the inherent risk resistance of the source-sink network topology. The model cannot mathematically represent and optimize the fluctuations in total cost for each source-sink network solution under changing market conditions, leading to solutions that, while having the lowest expected cost in static scenarios, often lack robustness in the face of market fluctuations.

[0006] Secondly, regarding the problem of parameter uncertainty, existing technologies mainly rely on stochastic programming or traditional robust optimization methods. These methods typically assume that the uncertain parameters follow a specific probability distribution (such as a normal distribution) or belong to a known fuzzy set. However, since the CCUS industry is still in its early stages, there is a lack of sufficient historical data to fit an accurate probability distribution function. In an environment of "deep uncertainty," forcibly assuming a probability distribution often leads to serious biases in the optimization results, failing to provide decision-makers with a truly reliable risk quantification boundary.

[0007] Finally, existing heuristic solution algorithms (such as simulated annealing and genetic algorithms) employ a uniform random sampling strategy, lacking a search feedback mechanism based on historical information, thus limiting their convergence efficiency. This strategy lacks memory, meaning the algorithm fails to effectively utilize historical evaluation data to identify key nodes that significantly contribute to system performance during iteration. Due to the lack of a targeted screening mechanism for high-contribution nodes, the algorithm struggles to distinguish between high-quality and low-quality subsets of the solution space during the search process, resulting in a significant waste of computational resources on ineffective topology combinations. This leads to slow convergence and a tendency to get trapped in local optima, making it difficult to obtain a globally optimal topology with high robustness within a limited time. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and address the core technical challenges faced in planning large-scale CCUS networks: First, existing models often employ a single cost objective and a tightly coupled architecture, which not only suffers from the curse of computational dimensionality but also fails to directly optimize the system's risk boundary. Second, existing stochastic programming methods rely on assumptions about parameter probability distributions, leading to distorted risk assessments in environments with deep uncertainty and a lack of data. Third, traditional heuristic algorithms use random sampling and lack historical information feedback mechanisms, resulting in low convergence efficiency in large-scale combinatorial optimization. Therefore, this invention proposes a CCUS source-sink matching network optimization method that considers uncertainty, achieving efficient solutions to complex source-sink matching problems.

[0009] The specific technical solution adopted in this invention is as follows:

[0010] Firstly, the present invention provides a CCUS source-sink matching network optimization method considering uncertainty, as follows:

[0011] S1: Collect basic data of the CCUS source-sink matching network and construct a physical topology including carbon source set, carbon sink set, and pipeline set; define the nominal values ​​of uncertainty parameters of each carbon sink based on the physical topology to construct the IGDT uncertainty set; finally, set the total cost threshold of the source-sink matching network according to the IGDT uncertainty set; based on a greedy heuristic strategy, generate a binary decision vector of the initial pipeline network topology that satisfies the capacity constraint by calculating the marginal cost of candidate nodes in the IGDT uncertainty set, and use the binary decision vector as the starting state of the main loop of the coupled optimization iteration in S2;

[0012] S2: Construct a coupled optimization model comprising a source-sink matching network solution layer, an uncertainty analysis layer, and a flow solution layer; based on the coupled optimization model, and according to a preset decision-making mode, reconstruct the optimization objective function of the source-sink matching network solution layer into an extreme value search for the risk index; set the initial temperature for the simulated annealing algorithm. and termination temperature Then, it enters the main loop of coupled optimization iteration until the current temperature drops to the termination temperature, at which point the algorithm terminates and outputs the optimal solution under the current decision mode.

[0013] Preferably, S1 is as follows:

[0014] S11: Set the nominal values ​​of key economic parameters in the source-sink matching network. The Including the nominal value of unit storage cost and nominal value of unit economic benefits Based on the above Constructing a system based on uncertainty levels or IGDT uncertainty set with radius or ;

[0015] S12: Set the total cost threshold for the source-sink matching network according to the preset decision-making model:

[0016] If the decision-making model adopts a robust approach, then set an upper limit on the acceptable robustness cost of the source-sink matching network. If the decision-making model adopts an opportunity-based approach, then the expected opportunity cost objective of the source-sink matching network should be set. ;in, For the optimal nominal cost in a deterministic scenario, This is the preset deviation coefficient;

[0017] S13: Calculate each candidate sink based on a greedy heuristic strategy. nominal net cost per unit ;in, For the remittance point The nominal unit cost of sealing; For the remittance point The average nominal unit transportation cost to each source point; For nominal unit economic benefits;

[0018] according to The sinks are activated in ascending order of capacity until the total capacity of the activated sinks meets the emission requirements of all sources. The resulting binary decision vector is the initial site selection scheme. .

[0019] Preferably, in step S2, the optimization objective function of the upper planning layer is reconstructed according to a preset decision-making model, as follows:

[0020] If the decision-making model adopts a robust approach, then the cost ceiling will be met. Under the constraints, the objective function is to maximize the uncertainty tolerance. :

[0021] ;

[0022] If the decision-making model adopts an opportunity-based approach, then in order to meet the profit objective... Under the constraints, the optimization objective function is to minimize the chance bias. :

[0023] ;

[0024] in, The decision vector characterizing the source-sink topology. For the operating costs reported by the lower levels, For source-sink matching networks in topology and uncertainty parameters The total cost function under this condition.

[0025] Preferably, in S2, the main loop of the coupling optimization iteration is to repeatedly execute S3~S5, as follows:

[0026] S3: The source-sink matching network solution layer receives the dynamic weight vector fed back from S5. ; using the above Non-uniform probability sampling is performed on both disabled and enabled nodes to generate new candidate source-sink network schemes within the neighborhood of the current solution. and the Pass the inner IGDT optimization evaluation sub-loop;

[0027] The IGDT optimization evaluation sub-loop is to repeatedly execute S4, as follows:

[0028] S41: Based on the current level of uncertainty in the trial , get S3 Substitute the parameters into the uncertainty analysis layer and perform IGDT optimization to construct an extreme parameter set that makes the optimization objective function of the source-sink matching network reach the boundary. ;

[0029] S42: The flow calculation layer will receive... As a physical constraint, As computational parameters, a linear programming model is constructed to solve for the optimal flow allocation matrix; by solving for the optimal carbon dioxide flow allocation at this point, the current candidate source-sink network scheme is further derived. The operating cost is calculated; the operating cost is fed back to S41 and compared with a preset performance threshold. The uncertainty level is then updated using a binary search method. ;

[0030] S43: Repeat S41~S42 until the search converges, at which point the uncertainty level is... As a critical risk indicator for this candidate source-sink network scheme or Output;

[0031] S5: Using the critical risk index solved in S43 as a feedback signal, we find a source-sink matching network topology to optimize the critical risk index, and update each carbon sink node through a dynamic scoring equation. And will update Feedback is sent to the upper planning layer for the next round of sampling; simultaneously, the difference in critical risk indicators between the old and new schemes is calculated, and the Metropolis criterion is used to determine whether to accept the candidate source-sink network scheme. and perform temperature cooling operation. .

[0032] Furthermore, in step S3, new candidate source-sink network schemes are generated within the neighborhood of the current solution. The specific strategies are as follows:

[0033] When attempting to activate a new node, the node... Probability of being selected Its dynamic weight vector Proportional, that is ;

[0034] When attempting to remove an active node, the node... Probability of being selected Its dynamic weight vector Inversely proportional, that is .

[0035] Furthermore, in S41, the uncertainty analysis layer employs an IGDT optimization strategy, constructing an extreme parameter set based on the IGDT uncertainty set described in S1 by extracting the set boundary values. , Includes pessimistic parameter set And optimistic parameter set The details are as follows:

[0036] S41-1: If the decision-making model adopts a robust model, based on the IGDT uncertainty set in S1. Construct a pessimistic boundary parameter set that maximizes the operating cost of the source-sink matching network. ;in accordance with The definition range is determined by taking its upper or lower bound as... The possible values ​​of:

[0037] ;

[0038] By using the current pessimistic boundary parameter set Substituting into S42, we solve for the optimal flow allocation and operating cost, combine this with the pipeline construction cost to form the total expected cost, and compare it with the preset cost ceiling; then, we iteratively adjust the cost using a binary search method. Solve for the maximum value that satisfies the upper limit of cost constraint. value:

[0039] ;

[0040] in, This represents the total expected cost of the current site selection plan;

[0041] S41-2: If the decision-making model adopts the opportunity model, based on the IGDT uncertainty set in S1. Construct an optimistic boundary parameter set that minimizes the operating cost of the source-sink matching network. ;in accordance with The definition range is determined by taking its lower or upper bound as... The possible values ​​of:

[0042] ;

[0043] By using the current optimistic boundary parameter set Substituting into S42, we solve for the optimal flow allocation and operating cost, combine this with the pipeline construction cost to form the total expected cost, and compare it with the opportunity cost objective; then we iteratively adjust the result using a binary search method. Solve for the minimum value that satisfies the profit objective constraint. value:

[0044] .

[0045] Furthermore, in S42, the linear programming model is specifically as follows:

[0046] ;

[0047] in, For source-end capture integration cost, For the cost of sealing, For pipeline transportation costs, For the overall system maintenance cost, For economic gain;

[0048] Meanwhile, the linear programming model satisfies the following linear constraints:

[0049] Node traffic balancing constraints ;

[0050] Pipeline transport capacity constraints ;

[0051] Carbon sink location and capacity constraints ;

[0052] in, , For each node Flow to Node ,node Flow to Node CO2 transport flow; Source point capture quantity; For the remittance point The actual amount sealed; This indicates the sink's enabled status as determined by the upper layer, with a value of 0 or 1. This represents the maximum storage capacity of the remittance point; For candidate carbon aggregation; This is a set of candidate carbon sources.

[0053] Furthermore, in S5, the dynamic weights in the feature feedback mechanism The update rules are as follows:

[0054] ;

[0055] in, This refers to the set of activated carbon sink nodes included in the scheme whose risk indicators are better than the historical best solution during the current iteration. This is a preset topological excitation factor; This is the updated dynamic weight value; This is the dynamic weight value before the update.

[0056] Furthermore, in S5, the Metropolis criterion is used to determine whether to accept the candidate source-sink network scheme. The specific methods are as follows:

[0057] First, calculate the new solution. Compared with the current plan fitness difference The fitness value is taken from the risk index defined in S2; in the robust decision-making mode... In decision-making mode of opportunity mode ;

[0058] Secondly, calculate the probability of acceptance of the proposed solution. And decide whether to update the solution:

[0059] ;

[0060] in, The current temperature parameter; if the generated random number Less than If the new plan is accepted, it will be accepted; otherwise, no update will be made.

[0061] Finally, following the cooling equation Perform temperature updates; where, The coefficient of performance is the cooling factor. The temperature parameter is for the (k+1)th iteration. The temperature parameter is given in the k-th iteration.

[0062] when Reduced to the preset lower limit When the algorithm converges, it outputs the CCUS source-sink matching network topology, total expected cost, and carbon dioxide flow allocation with the optimal risk index under the preset decision-making model.

[0063] Secondly, this invention provides a CCUS source-sink matching network optimization system that considers uncertainty, comprising:

[0064] The data acquisition and initialization module is used to collect basic data of the CCUS source-sink matching network and construct a physical topology including carbon source set, carbon sink set, and pipeline set. Based on the physical topology, it defines the nominal values ​​of uncertainty parameters of each carbon sink to construct an IGDT uncertainty set. Finally, it sets the total cost threshold of the source-sink matching network according to the IGDT uncertainty set. Based on a greedy heuristic strategy, it generates a binary decision vector of the initial pipeline network topology that satisfies the capacity constraint by calculating the marginal cost of candidate nodes in the IGDT uncertainty set, and uses the binary decision vector as the starting state of the main loop of the coupled optimization iteration in the two-layer coupled optimization calculation module.

[0065] A two-layer coupled optimization computation module is used to construct a coupled optimization model comprising a source-sink matching network solution layer, an uncertainty analysis layer, and a flow solution layer. Based on the coupled optimization model, and according to a preset decision-making mode, the optimization objective function of the source-sink matching network solution layer is reconstructed into an extreme value search for a risk indicator. The initial temperature of the simulated annealing algorithm is set. and termination temperature Then, it enters the main loop of coupled optimization iteration until the current temperature drops to the termination temperature, at which point the algorithm terminates and outputs the optimal solution under the current decision mode.

[0066] Compared with the prior art, the present invention has the following advantages:

[0067] This invention firstly separates the complex source-sink topology generation from the flow calculation by constructing a completely decoupled discrete-continuous two-layer optimization architecture, effectively avoiding the "curse of dimensionality" faced by large-scale network planning, and using lower-layer linear programming to ensure the physical connectivity and capacity compliance of all solutions. Secondly, it establishes a feature feedback mechanism based on dynamic scoring, enabling the upper-layer model to prioritize the high-quality topology features solved by the middle-layer module based on the score, improving the optimization efficiency and convergence quality of the algorithm. Thirdly, it introduces the IGDT optimization strategy to overcome the limitation of traditional stochastic programming relying on prior data of probability distribution, and realizes accurate quantification of uncertainty risk under the condition of lack of historical data. Finally, by reconstructing the upper-layer optimization objective function, the source-sink network planning orientation is changed from a single "lowest static cost" to "equal emphasis on cost and risk indicators", so that the characteristics of the obtained optimal solution change from the lowest static cost to the optimal risk indicators in dynamic scenarios. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0069] Figure 1 A system flowchart for the CCUS source-sink matching network optimization method considering uncertainties;

[0070] Figure 2 The diagram shows the convergence speed of the total expected cost of the adaptive dynamic screening mechanism based on IGDT feedback.

[0071] Figure 3 This is a graph showing the convergence speed of the uncertain parameters in the adaptive dynamic screening mechanism based on IGDT feedback. Detailed Implementation

[0072] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. Technical features in various embodiments of the present invention can be combined accordingly without mutual conflict.

[0073] like Figure 1 As shown, this invention provides a CCUS source-sink matching network optimization method that considers uncertainty. Through the following innovative improvements, it effectively overcomes the above-mentioned technical challenges:

[0074] First, addressing the problem that existing planning models have a single objective and cannot directly quantify the stability of candidate solutions, this invention reconstructs the non-probabilistic risk measurement objective function of the upper-level two-layer planning model. Unlike the conventional approach in existing technologies that uses a defined expected total system cost as the single optimization objective, this invention constructs a topology-operation decoupled Max-Min and Min-Min complementary model, and derives two core risk measurement equations as the objective function of the upper-level simulated annealing algorithm. Specifically, for a risk-averse strategy, a robust objective function is constructed, namely, finding a market fluctuation that the system can withstand under the constraint of cost increases. The topology is maximized; for risk-preference strategies, an opportunistic objective function is constructed, which seeks the minimum market advantage deviation required to achieve the objective while satisfying the constraint of cost reduction. Minimize the topology.

[0075] Second, addressing the problem of difficulty in quantifying risk due to the lack of probability distribution data in uncertain environments, this invention embeds an IGDT module that does not rely on prior probability distributions between traditional topology generation and traffic allocation. This module replaces the traditional probability density function by constructing an interval uncertainty set centered on the nominal parameter and with the risk indicator as the radius. During the evaluation process, this module does not perform random sampling but instead constructs deterministic extreme boundary scenarios according to the decision-making mode: in robust mode, the cost parameter is pushed to the upper limit of the interval, and the revenue parameter is pushed to the lower limit of the interval, forcing candidate source-sink solutions to be exposed to the worst market environment, thereby observing the magnitude of change in their total expected cost; in opportunistic mode, the opposite is true. This mechanism can directly quantify the upper limit of parameter fluctuations that the system can tolerate to meet preset performance thresholds in uncertain environments.

[0076] Third, addressing the problem of blind searching and low convergence efficiency caused by uniform random sampling in existing heuristic algorithms, this invention proposes an adaptive dynamic screening mechanism based on IGDT feedback. This invention establishes a closed-loop feedback system based on intermediate computational data. It utilizes the risk index output by the IGDT module during the iteration process (… or As a feedback signal, the system constructs a dynamic score for each carbon sink node. During algorithm iteration, a dynamic scoring equation is used to assign weights to key nodes appearing in schemes with higher risk indicators. Furthermore, when the upper-level model generates new candidate topologies, a derived non-uniform sampling probability density function replaces the traditional uniform distribution, giving it a greater probability of selecting carbon sinks with higher weights. This dynamic evolution mechanism endows the model with the ability to learn historical information, enabling it to proactively identify key topologies that significantly contribute to improving system robustness.

[0077] The method of this invention mainly includes an upper-layer source-sink network generation module, a middle-layer uncertainty analysis module, and a lower-layer linear flow solution module, in order to realize the stability measurement of CCUS source-sink matching schemes and the identification and utilization of source-sink topology features. Specifically, it includes the following steps:

[0078] S1, Data Acquisition, Parameter Definition, and Initial State Generation:

[0079] Basic data of the CCUS source-sink matching network is collected, and a physical topology including carbon source set, carbon sink set, and pipeline set is constructed. Based on the obtained physical topology, the nominal values ​​of each carbon sink uncertainty parameter (unit absorption cost and unit economic benefit) are defined to construct the IGDT uncertainty set. Finally, the total cost threshold of the source-sink matching network is set according to the obtained IGDT uncertainty set. Based on the greedy heuristic strategy, the binary decision vector of the initial pipeline network topology that satisfies the capacity constraint is generated by calculating the marginal cost of candidate nodes in the obtained IGDT uncertainty set, and the obtained binary decision vector is used as the starting state of the main loop of the coupled optimization iteration in S2.

[0080] As a preferred embodiment of the present invention, in order to construct the computing environment required for the source-sink network generation module and the uncertainty analysis module and to generate a high-quality initial solution, the specific operation includes the following three sub-processes:

[0081] S11, Initialize the set of uncertain parameters:

[0082] Set the nominal values ​​of key economic parameters in the source-sink matching network. (Including the nominal value of unit storage cost) and nominal value of unit economic benefits ); based on the results Constructing a system based on uncertainty levels or IGDT uncertainty set with radius or .

[0083] by For example, that is, allowing and With the nominal value as the center, Fluctuation within a radius:

[0084] ;

[0085] in, A vector of actual cost or benefit parameters; for The nominal value; Let represent the level of uncertainty to be solved, and let represent the true value. Relative to nominal value The maximum percentage deviation range. This set correlates market volatility with the level of uncertainty for subsequent solutions.

[0086] S12, Set the system total cost threshold:

[0087] Set the total cost threshold for the source-sink matching network based on the preset decision-making model (i.e., the decision-maker's risk preference):

[0088] If the decision-making model adopts a robust approach, then set an upper limit on the acceptable robustness cost of the source-sink matching network. :

[0089] ;

[0090] If the decision-making model adopts an opportunity-based approach, then the expected opportunity cost objective of the source-sink matching network is set as follows: :

[0091] ;

[0092] in, For the optimal nominal cost in a deterministic scenario, This is the preset deviation coefficient;

[0093] S13, Generate the initial location selection scheme based on a greedy heuristic strategy:

[0094] To improve the convergence speed of subsequent two-layer optimization, a greedy heuristic strategy is used to compute each candidate sink. nominal net cost per unit :

[0095] ;

[0096] in, For the remittance point The nominal unit cost of sealing; For the remittance point The average nominal unit transportation cost to each source point; For nominal unit economic benefits;

[0097] according to The sinks are activated in ascending order of capacity until the total capacity of the activated sinks meets the emission requirements of all sources. The resulting binary decision vector is the initial site selection scheme. .

[0098] S2, Build the optimization model and start the iteration:

[0099] A coupled optimization model is constructed, comprising a source-sink matching network solution layer, an uncertainty analysis layer, and a flow solution layer. Based on the obtained coupled optimization model, and according to a preset decision-making mode (robust mode or opportunistic mode), the optimization objective function of the source-sink matching network solution layer is reconstructed to address the risk indicator (uncertainty tolerance). or chance bias Extreme value search; setting the initial temperature for the simulated annealing algorithm. and termination temperature Then, it enters the main loop of coupled optimization iteration until the current temperature drops to the termination temperature, at which point the algorithm terminates and outputs the optimal solution under the current decision mode.

[0100] In a preferred embodiment of the present invention, in this step, the optimization objective function of the upper planning layer is reconstructed according to a preset decision-making model, aiming to search for the extreme value of the uncertainty level, as follows:

[0101] If the decision-making model adopts a robust approach, then the cost ceiling will be met. Under the constraints, the objective function is to maximize the uncertainty tolerance. Find the parameter that can withstand the largest fluctuation range. topology :

[0102] ;

[0103] If the decision-making model adopts an opportunity-based approach, then in order to meet the profit objective... Under the constraints, the optimization objective function is to minimize the chance bias. To find the minimum external favorable deviation required to achieve this goal. topology :

[0104] ;

[0105] in, The decision vector characterizing the source-sink topology. For the operating costs reported by the lower levels, For the system (i.e., the source-sink matching network) in topology and uncertainty parameters The total cost function consists of the annualized fixed investment cost and the operating costs fed back from the lower level. and These are the sets of uncertainty parameters for the corresponding modes. and The specific value is calculated by calling the uncertainty analysis module in step S4 and is directly used as the objective function of the upper-level module.

[0106] In a preferred embodiment of the present invention, in this step, the main loop of the coupling optimization iteration is to repeatedly execute S3~S5, as follows:

[0107] S3, a source-sink matching network scheme based on weighted neighborhood search:

[0108] The source-sink matching network solver layer receives the dynamic weight vector fed back from S5. ;use Non-uniform probability sampling is performed on both disabled and enabled nodes to generate new candidate source-sink network schemes within the neighborhood of the current solution. and will Pass the inner IGDT optimization evaluation sub-loop.

[0109] The IGDT optimization evaluation sub-loop is to repeatedly execute S4, as follows:

[0110] S4, Uncertainty Analysis:

[0111] Candidate source-sink network solutions Import the uncertainty analysis layer for solution, as follows:

[0112] S41, Construct the extreme parameter set:

[0113] Based on the current level of uncertainty , get S3 Substitute the parameters into the uncertainty analysis layer and perform IGDT optimization to construct the extreme parameter set that makes the optimization objective function of the source-sink matching network reach the boundary. .

[0114] S42, Flow rate calculation and risk index calculation:

[0115] The flow calculation layer will receive As a physical constraint, As computational parameters, a linear programming model is constructed to solve for the optimal flow allocation matrix; by solving for the optimal carbon dioxide flow allocation at this point, the current candidate source-sink network scheme is further derived. The operating costs are calculated; the obtained operating costs are fed back to S41, compared with the preset performance threshold, and the uncertainty level is updated using a binary search method. ;

[0116] S43: Repeat S41~S42 until the search converges, at which point the uncertainty level is... As a critical risk indicator for this candidate source-sink network scheme or Output;

[0117] S5, Feature Feedback Update and Decision Convergence Determination:

[0118] Using the critical risk index solved by S43 as a feedback signal, a source-sink matching network topology is sought to optimize the critical risk index. Each carbon sink node is then updated using a dynamic scoring equation. And will update Feedback is sent to the upper planning layer for the next round of sampling; simultaneously, the difference in critical risk indicators between the old and new schemes is calculated, and the Metropolis criterion is used to determine whether to accept the candidate source-sink network scheme. and perform temperature cooling operation. .

[0119] In a preferred embodiment of the present invention, in step S3 above, a new candidate source-sink network scheme is generated in the neighborhood of the current solution. The specific strategies are as follows:

[0120] When attempting to activate a new node, the node... Probability of being selected Its dynamic weight vector Proportional, that is ;

[0121] When attempting to remove an active node, the node... Probability of being selected Its dynamic weight vector Inversely proportional, that is ;

[0122] Metropolis Criterion Acceptance Probability Determined by the following formula:

[0123] ;

[0124] in, The difference between the risk indicators of the old and new plans. This is the current temperature parameter.

[0125] In a preferred embodiment of the present invention, in step S41 above, the uncertainty analysis layer adopts the IGDT optimization strategy, and constructs the extreme parameter set by extracting the set boundary values ​​based on the IGDT uncertainty set in S1. , Includes pessimistic parameter set And optimistic parameter set The specific construction and calculation logic are as follows:

[0126] S41-1: If the decision-making model adopts a robust model, based on the IGDT uncertainty set in S1. Construct a pessimistic boundary parameter set that maximizes the operating cost of the source-sink matching network. ;in accordance with The definition range is determined by taking its upper or lower bound as... The possible values ​​of:

[0127] ;

[0128] Subsequently, by using the current pessimistic boundary parameter set Substituting into S42, we solve for the optimal flow allocation and operating cost, combine this with the pipeline construction cost to form the total expected cost, and compare it with the preset cost ceiling; then, we iteratively adjust the cost using a binary search method. Solve for the maximum value that satisfies the upper limit of cost constraint. value:

[0129] ;

[0130] in, This represents the total expected cost of the current site selection plan;

[0131] S41-2: If the decision-making model adopts the opportunity model, based on the IGDT uncertainty set in S1. Construct an optimistic boundary parameter set that minimizes the operating cost of the source-sink matching network. ;in accordance with The definition range is determined by taking its lower or upper bound as... The possible values ​​of:

[0132] ;

[0133] Subsequently, by using the current optimistic boundary parameter set Substituting into S42, we solve for the optimal flow allocation and operating cost, combine this with the pipeline construction cost to form the total expected cost, and compare it with the opportunity cost objective; then we iteratively adjust the result using a binary search method. Solve for the minimum value that satisfies the profit objective constraint. value:

[0134] .

[0135] In a preferred embodiment of the present invention, in step S42 above, the lower-level flow calculation module will calculate the flow rate based on the pipeline topology determined in step S3. As fixed physical boundary constraints, a linear programming (LP) model is constructed with the objective of minimizing actual operating costs:

[0136] ;

[0137] in, For source-end capture integration cost, For the cost of sealing, For pipeline transportation costs, For the overall system maintenance cost, The economic benefit; the storage cost and economic benefit are determined by the set of extreme parameters passed in step S41. The rest are taken from the basic data obtained in step S1.

[0138] Meanwhile, to ensure physical feasibility, the following linear constraints must be met:

[0139] 1) Node flow balance constraint (mass conservation):

[0140] ;

[0141] 2) Pipeline transport capacity constraints:

[0142] ;

[0143] 3) Carbon sink location and capacity constraints:

[0144] ;

[0145] in, , For each node Flow to Node ,node Flow to Node CO2 transport flow; Source point capture quantity; For the remittance point The actual amount sealed; This indicates the sink's enabled status as determined by the upper layer, with a value of 0 or 1. This represents the maximum storage capacity of the remittance point; For candidate carbon aggregation; Given a set of candidate carbon sources, the solution yields the minimum operating cost. This will be returned as a feedback signal to the upper-layer source-sink network generation module.

[0146] In a preferred embodiment of the present invention, in step S5 above, the feature feedback mechanism is based on the risk index solved in step S4. or This involves constructing a dynamic scoring equation to weight and incentivize key source-sink topologies appearing in high-quality solutions. The specific implementation logic is as follows:

[0147] First, for each candidate carbon sink node Establish initial dynamic scores (or weights). The initial value is set to 1.0; secondly, during the iteration process, when the upper-level module discovers a new solution, and the risk index of this solution is better than the current historical best solution (i.e., or When this happens, the set of carbon sink nodes that are active in the scheme is locked. Finally, the set is updated using a dynamic scoring equation. Weights of each node:

[0148] ;

[0149] in, This refers to the set of activated carbon sink nodes included in the scheme whose risk indicators are better than the historical best solution during the current iteration. A preset topology incentive factor (ranging from 0.1 to 0.2) is used to reward nodes that significantly contribute to improving system robustness. Dynamic scoring. This will be returned as a feature feedback signal to the upper-layer source-sink network generation module. When generating new neighborhood perturbation solutions in the upper layer, the probability of a node being selected... Its dynamic rating Positive correlation (i.e.) The algorithm is guided to actively preserve highly robust topological features, thereby improving solution efficiency.

[0150] In a preferred embodiment of the present invention, in step S5 above, the simulated annealing mechanism uses the Metropolis criterion to control the iteration and convergence process of the two-layer model, that is, it uses the Metropolis criterion to determine whether to accept candidate source-sink network schemes. The specific methods are as follows:

[0151] First, calculate the new solution. Compared with the current plan fitness difference The fitness value is taken from the risk index defined in S2. or ); when the decision-making model is robust (pursuing) Maximize) In decision-making mode that is opportunity-driven (pursuit of opportunities) Minimize) ;

[0152] Secondly, calculate the probability of acceptance of the proposed solution. And decide whether to update the solution:

[0153] ;

[0154] in, The current temperature parameter; if the generated random number Less than If the new plan is accepted, it will be accepted; otherwise, no update will be made.

[0155] Finally, temperature updates and convergence checks are performed, following the cooling equation. Perform temperature updates; where, The coefficient of performance is the cooling factor. The temperature parameter is for the (k+1)th iteration. Let be the temperature parameter in the k-th iteration; when Reduced to the preset lower limit When the algorithm converges, it outputs the CCUS source-sink matching network topology, total expected cost, and carbon dioxide flow allocation with the optimal risk index under the preset decision-making model.

[0156] This invention constructs a two-layer decoupled optimization architecture based on Information Gap Decision Theory (IGDT) and Simulated Annealing (SA). It replaces the traditional cost objective function with a newly derived risk measurement objective function, achieving efficient collaborative solutions for discrete location decisions and continuous operational traffic allocation. Simultaneously, an additional IGDT module, independent of prior probability distributions, is embedded in the traditional two-layer optimization model to directly measure the robustness boundary and opportunistic threshold of the system under uncertain environments, enhancing the risk resistance of large-scale CCUS cluster planning schemes. Finally, an adaptive dynamic selection mechanism based on IGDT feedback is proposed. This allows the upper-layer model to use the risk indicators output by the IGDT module during iteration as feedback signals to identify source-sink topology characteristics and proactively select a superior source-sink network topology, improving model solution efficiency.

[0157] The following examples will illustrate the method and performance effects of the present invention.

[0158] Example

[0159] This embodiment takes a 6 million-ton-level CCUS (Carbon Capture, Utilization and Storage) cluster planning project in a coastal area as an example and conducts a CCUS source-sink matching network optimization method that takes uncertainty into account.

[0160] This region contains one major coal-fired power plant emission source and 117 carbon sinks for purposes including chemical utilization, oil recovery, and geological storage. It faces a complex market environment characterized by uncertain waste disposal costs, fluctuating chemical product prices, and changes in international oil prices. This embodiment utilizes the method described in this invention to address these uncertainties, following... Figure 1 The process shown illustrates a source-sink matching and pipeline layout structure and CO2 flow distribution scheme that offers optimal stability and maximum revenue potential in complex market environments.

[0161] S1: Data Acquisition, Parameter Definition, and Initial State Generation

[0162] S11: Loading Basic Data

[0163] This step takes a 10 million-ton-level CCUS cluster planning project in a coastal area as an example, completing the construction and initialization of the basic database through the system interface. The system reads the emission source of one major coal-fired power plant in the area, with a total capture capacity of 6 million tons per year, and the source coordinates are used as the starting point for planning. Simultaneously, the system loads 117 candidate carbon sink nodes in and around the area, covering economic carbon sinks, onshore brackish water layers, and seabed storage points, with Sink_78 (synthetic methanol) as an example. As an economic carbon sink, its nominal absorption cost is recorded as 500 yuan per ton, the selling price of the finished product is converted to 400 yuan per ton, and the maximum absorption capacity is 350,000 tons per year. Regarding pipeline network data, the system defines 10 standardized pipeline specifications, and in this initialization phase, Type_4 pipeline is selected as the benchmark, with its unit construction cost set at 7.6293 million yuan per km for subsequent initial screening and estimation. Based on these costs, a database is constructed for all sinks with an uncertainty level. or IGDT uncertainty set with radius or .

[0164] S12: Set total cost threshold

[0165] The system sets parameters and thresholds to address uncertainties throughout the entire lifecycle. The project operating cycle is set at 30 years, and fixed investments are converted into annualized costs based on this. For economic parameters lacking historical data, the system defines that the uncertainty fluctuations of their nominal forecasts follow an interval distribution with a standard deviation factor of 0.15. Based on this, the system sets a total cost threshold according to the decision-making model: for the robust model, a budget coefficient of 1.1 is set, meaning that under the premise of a 10% increase in total cost, the system seeks to minimize the risk indicator (…). The optimal source-exchange matching scheme; for high-yield opportunity models, a target coefficient of 0.9 is set, meaning that under the premise of expecting a 10% reduction in total cost, the search should be conducted to find a risk indicator that ( The smallest solution.

[0166] S13: Generate initial network topology

[0167] Finally, to generate a high-quality initial solution for initiating the two-level iteration, the system executes a greedy selection strategy based on minimum spanning tree (MST) marginal cost. The system sets the number of candidate carbon sinks. In each iteration, all unselected candidate sinks are traversed, and the change in total network investment caused by adding them to the current network is calculated. After 10 iterations, 10 sinks with optimal spatial distribution, including Sink_111 and Sink_113, are selected, generating the initial network topology solution. .

[0168] Step S2: Objective function reconstruction and startup coupling optimization of the main loop

[0169] A discrete-continuous fully decoupled two-level programming model was constructed, and the optimization objective function of the upper-level programming module was reconstructed according to the decision-making mode. The system first considers no uncertainty (i.e.,...) In the nominal scenario, a standard optimization was performed to establish the system's baseline cost. After iterative search, the system output the nominally optimal solution (Normal). The annualized total expected cost of this solution is calculated to be 3.096 billion yuan per year. For the robust mode, the system sets the budget coefficient to 1.1. Robust cost threshold. Under this constraint, the objective function of the upper-level source-sink network generation module is reconstructed to maximize the uncertainty tolerance:

[0170] ;

[0171] That is: to find a topological structure This allows it to withstand the greatest fluctuations in market parameters without a total cost not exceeding 3.251 billion yuan.

[0172] For the opportunity model, the system sets the budget coefficient to 0.9. Robust cost threshold. Under this constraint, the objective function of the upper-level module is refactored to minimize the chance bias:

[0173] ;

[0174] That is: to find a topological structure This allows it to reduce total costs to below 2.786 billion yuan with minimal market benefits (such as product price increases or cost reductions).

[0175] Step S3: Accept the dynamic weight vector and generate new candidate solutions.

[0176] Receive dynamic weight vector fed back by S5 ;use Non-uniform probability sampling is performed on both disabled and enabled nodes to generate new candidate source-sink network schemes within the neighborhood of the current solution. and will Pass the inner IGDT optimization evaluation sub-loop;

[0177] Step S4: Execute IGDT optimization sub-loop

[0178] Step S41: Determine the extreme parameter set

[0179] This step embeds an analysis module based on Information Gap Decision Theory (IGDT) into the middle layer of the two-layer optimization architecture. For each candidate topology scheme generated in the upper layer, an optimization strategy is used to quantify its risk resistance. Based on the uncertainty set defined in step S1, the boundaries of the scheme are explored by constructing an extreme parameter set.

[0180] Taking the robust mode as an example, under the robust cost threshold constraint calculated in step S2, the IGDT module performs optimization to solve the objective function determined in step S2 and find the maximum uncertainty tolerance. The system initializes with a tentative level of uncertainty. And construct the corresponding pessimistic boundary parameter set. Specifically, the system reduces storage costs. Relative to nominal value upward At the same time, it increases economic benefits (such as EOR benefits). Downward adjustment ,Right now:

[0181] ;

[0182] Step S42: Flow calculation and operating cost accounting

[0183] Subsequently, the module calls the traffic calculation layer to calculate the extreme operating cost under this pessimistic scenario. The uncertainty level is continuously adjusted using a binary search method until a critical point is found that brings the operating cost close to but does not exceed 3.251 billion yuan.

[0184] Step S43: Repeat S41~S42 until the search converges.

[0185] After multiple rounds of iterative calculations, the system ultimately determined the maximum risk index of the current robust solution to be 0.2988. This indicates that the current source-sink network can withstand a 29.88% deterioration in market parameters compared to the nominal predicted value, while the total cost only increases by 10%. Similarly, the maximum risk index of the opportunistic solution is 0.0410, representing that the current source-sink network can withstand only a 4.10% advantage in market parameters compared to the nominal predicted value, while the total cost decreases by 10%. Through the above calculations, this step transforms the abstract topology into quantifiable risk indicators for comparison.

[0186] Step S5: Feature Feedback and Convergence Determination

[0187] Establish a feature feedback closed loop based on a dynamic scoring equation. Apply the risk indicators (such as the maximum uncertainty tolerance in robust mode) obtained in step S3. As a feedback signal, a dynamic score is constructed for each candidate carbon sink node. During the algorithm initialization phase, the system assigns uniform initial weights to the 117 candidate carbon sink nodes within the region. This means that all nodes have an equal probability of being selected in the initial state. When a candidate topology scheme generated by the upper-layer module is solved by the middle-layer module and its risk index is better than the current historical best solution, the feature feedback mechanism will be triggered. In the robustness optimization process of this embodiment, the system discovered a new set of topologies with an uncertainty tolerance of The value was increased to 0.0557, significantly better than the previous scheme. The system identified carbon aggregation in this scheme. The set specifically includes 10 nodes: Sink_111, Sink_113, Sink_16, and Sink_7. Comparison revealed that, compared to the nominally optimal solution, this robust solution removed the poorer Sink_28 and Sink_51, and identified the better Sink_7 and Sink_16.

[0188] To solidify this topology, the system applies dynamic assignment equations to... The key nodes in the process are weighted and incentivized. The specific calculation logic is shown in the following formula:

[0189] ;

[0190] in, The preset topology activation factor is set to 0.1 in this embodiment. This means that the selection probability weights of nodes such as Sink_7 and Sink_16 increase by 10%, while the weights of nodes not appearing in this optimal solution remain unchanged or decrease relatively. Updated weight vector This is then used as a feature feedback signal and returned to the upper-layer source-sink network generation module. When generating new neighborhood perturbation solutions in the upper layer, the system no longer flips the node states completely randomly, but instead follows the weight vector. Make a selection. For example... Figure 2 , Figure 3 As shown, the guided algorithm actively converges to the solution space region containing these topologies, improving the solution efficiency and the robustness of the final solution.

[0191] The Metropolis criterion is used to control the acceptance of solutions, and the weights calculated in step S4 are introduced during the neighborhood solution generation stage. In each iteration, the system needs to base its solution on the current scheme. Generate a new candidate solution The system no longer employs a random perturbation strategy, but instead constructs a weight-based... Non-uniform sampling probability distribution:

[0192] When the system decides to use a set of currently inactive sinks. When a node is selected to be activated, the node... Probability of being selected It is proportional to its weight:

[0193] ;

[0194] This means that nodes with increased weights (such as Sink_7 and Sink_16) have a higher probability of being selected to enter the source-sink network. And when the system selects from the currently active set of sink nodes... When a node is removed from the node, the node Probability of being selected It is inversely proportional to its weight:

[0195] ;

[0196] This means that nodes with reduced weights (such as Sink_28, which was removed by the robust solution) have a very high probability of being removed first.

[0197] After generating candidate solutions based on the aforementioned weighting strategy, the system calls the mid-level module to calculate the difference in their risk indicators. (e.g., in robust mode) ).like That is, the new solution has stronger risk resistance, and the system directly accepts the new solution; if Then based on probability Accepting inferior solutions to maintain population diversity. This embodiment sets the initial temperature. Cooling coefficient As the iteration proceeds, the temperature Gradually decrease. When the temperature parameter drops to the preset lower limit... When the algorithm converges, the system determines that the optimal solution has been found and outputs the final robust solution. The network topology of this solution includes 10 core sinks: Sink_81, Sink_82, Sink_90, and Sink_113. Its maximum uncertainty tolerance is... The system ultimately converges to 0.2988, indicating that it can withstand a 29.88% fluctuation in all parameters. The annualized total expected cost of this scheme is 3.124 billion yuan, approximately 0.9% higher than the nominally optimal scheme, but it can withstand greater market volatility. The calculation process for the opportunity-optimal scheme is similar.

[0198] This invention addresses the problems of strong coupling between location selection and scheduling, and the difficulty in quantifying long-term risks in existing technologies, by constructing a completely decoupled discrete-continuous two-layer optimization architecture. The upper layer uses simulated annealing for topology heuristic search, while the lower layer uses linear programming for precise traffic allocation and full cost calculation. This invention innovatively reconstructs the objective function of the upper-layer simulated annealing module, transforming it from solving for static cost optimization of the source-sink network to solving for dynamic stability optimization. Simultaneously, an intermediate uncertainty module is introduced, performing optimization based on gap information theory (IGDT), directly quantifying the robustness boundary and opportunistic threshold of the system by constructing an extreme parameter set. Finally, a feature feedback mechanism based on adaptive weights is established, dynamically guiding the algorithm to converge efficiently using key topological features of high-quality solutions. This invention effectively avoids the curse of dimensionality in large-scale planning, providing suitable dynamic risk decision support for CCUS projects transitioning from the initial to the mature stage.

[0199] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A CCUS source-sink matching network optimization method considering uncertainty, characterized in that, Specifically as follows: S1: Collect basic data of the CCUS source-sink matching network and construct a physical topology including carbon source set, carbon sink set, and pipeline set; define the nominal values ​​of uncertainty parameters of each carbon sink based on the physical topology to construct the IGDT uncertainty set; finally, set the total cost threshold of the source-sink matching network according to the IGDT uncertainty set; based on a greedy heuristic strategy, generate a binary decision vector of the initial pipeline network topology that satisfies the capacity constraint by calculating the marginal cost of candidate nodes in the IGDT uncertainty set, and use the binary decision vector as the starting state of the main loop of the coupled optimization iteration in S2; S2: Construct a coupled optimization model comprising a source-sink matching network solution layer, an uncertainty analysis layer, and a flow solution layer; based on the coupled optimization model, and according to a preset decision-making mode, reconstruct the optimization objective function of the source-sink matching network solution layer into an extreme value search for the risk index; set the initial temperature for the simulated annealing algorithm. and termination temperature Then it enters the main loop of coupled optimization iteration until the current temperature drops to the termination temperature, at which point the algorithm terminates and outputs the optimal solution under the current decision mode. In step S2, the optimization objective function of the upper planning layer is reconstructed according to a preset decision-making model, as follows: If the decision-making model adopts a robust approach, then the cost ceiling will be met. Under the constraints, the objective function is to maximize the uncertainty tolerance. : ； If the decision-making model adopts an opportunity-based approach, then in order to meet the profit objective... Under the constraints, the optimization objective function is to minimize the chance bias. : ； in, The decision vector characterizing the source-sink topology. For the operating costs reported by the lower levels, For source-sink matching networks in topology and uncertainty parameters The total cost function under; In S2, the main loop of the coupling optimization iteration is to repeatedly execute S3~S5, as follows: S3: The source-sink matching network solution layer receives the dynamic weight vector fed back from S5. ; using the above Non-uniform probability sampling is performed on both disabled and enabled nodes to generate new candidate source-sink network schemes within the neighborhood of the current solution. and the Pass the inner IGDT optimization evaluation sub-loop; The IGDT optimization evaluation sub-loop is to repeatedly execute S4, as follows: S41: Based on the current level of uncertainty in the trial , get S3 Substitute the parameters into the uncertainty analysis layer and perform IGDT optimization to construct an extreme parameter set that makes the optimization objective function of the source-sink matching network reach the boundary. ; S42: The flow calculation layer will receive... As a physical constraint, As computational parameters, a linear programming model is constructed to solve for the optimal flow allocation matrix; by solving for the optimal carbon dioxide flow allocation at this point, the current candidate source-sink network scheme is further derived. The operating cost is calculated; the operating cost is fed back to S41 and compared with a preset performance threshold. The uncertainty level is then updated using a binary search method. ; S43: Repeat S41~S42 until the search converges, at which point the uncertainty level is... As a critical risk indicator for this candidate source-sink network scheme or Output; S5: Using the critical risk index solved in S43 as a feedback signal, we find a source-sink matching network topology to optimize the critical risk index, and update each carbon sink node through a dynamic scoring equation. And will update Feedback is sent to the upper planning layer for the next round of sampling; simultaneously, the difference in critical risk indicators between the old and new schemes is calculated, and the Metropolis criterion is used to determine whether to accept the candidate source-sink network scheme. and perform temperature cooling operation. .

2. The CCUS source-sink matching network optimization method considering uncertainty according to claim 1, characterized in that, S1 is specifically as follows: S11: Set the nominal values ​​of key economic parameters in the source-sink matching network. The Including the nominal value of unit storage cost and nominal value of unit economic benefits Based on the above Constructing a system based on uncertainty levels or IGDT uncertainty set with radius or ; S12: Set the total cost threshold for the source-sink matching network according to the preset decision-making model: If the decision-making model adopts a robust approach, then set an upper limit on the acceptable robustness cost of the source-sink matching network. If the decision-making model adopts an opportunity-based approach, then the expected opportunity cost objective of the source-sink matching network should be set. ;in, For the optimal nominal cost in a deterministic scenario, This is the preset deviation coefficient; S13: Calculate each candidate sink based on a greedy heuristic strategy. nominal net cost per unit ;in, For the remittance point The nominal unit cost of sealing; For the remittance point The average nominal unit transportation cost to each source point; For nominal unit economic benefits; according to The sinks are activated in ascending order of capacity until the total capacity of the activated sinks meets the emission requirements of all sources. The resulting binary decision vector is the initial site selection scheme. .

3. The CCUS source-sink matching network optimization method considering uncertainty according to claim 1, characterized in that, In step S3, a new candidate source-sink network scheme is generated within the neighborhood of the current solution. The specific strategies are as follows: When attempting to activate a new node, the node... Probability of being selected Its dynamic weight vector Proportional, that is ; When attempting to remove an active node, the node... Probability of being selected Its dynamic weight vector Inversely proportional, that is .

4. The CCUS source-sink matching network optimization method considering uncertainty according to claim 1, characterized in that, In step S41, the uncertainty analysis layer employs an IGDT optimization strategy. Based on the IGDT uncertainty set described in S1, it constructs an extreme parameter set by extracting the set boundary values. , Includes pessimistic parameter set and optimistic parameter set The details are as follows: S41-1: If the decision-making model adopts a robust model, based on the IGDT uncertainty set in S1. Construct a pessimistic boundary parameter set that maximizes the operating cost of the source-sink matching network. ;in accordance with The definition range is determined by taking its upper or lower bound as... The possible values ​​of: ； By using the current pessimistic boundary parameter set Substituting into S42, we solve for the optimal flow allocation and operating cost, combine this with the pipeline construction cost to form the total expected cost, and compare it with the preset cost ceiling; then, we iteratively adjust the cost using a binary search method. Solve for the maximum value that satisfies the upper limit of cost constraint. value: ； in, This represents the total expected cost of the current site selection plan; S41-2: If the decision-making model adopts the opportunity model, based on the IGDT uncertainty set in S1. Construct an optimistic boundary parameter set that minimizes the operating cost of the source-sink matching network. ;in accordance with The definition range is determined by taking its lower or upper bound as... The possible values ​​of: ； By using the current optimistic boundary parameter set Substituting into S42, we solve for the optimal flow allocation and operating cost, combine this with the pipeline construction cost to form the total expected cost, and compare it with the opportunity cost objective; then we iteratively adjust the result using a binary search method. Solve for the minimum value that satisfies the profit objective constraint. value: 。 5. The CCUS source-sink matching network optimization method considering uncertainty according to claim 1, characterized in that, In S42, the linear programming model is as follows: ； in, For source-end capture integration cost, For the cost of sealing, For pipeline transportation costs, For the overall system maintenance cost, For economic gain; Meanwhile, the linear programming model satisfies the following linear constraints: Node traffic balancing constraints ; Pipeline transport capacity constraints ; Carbon sink location and capacity constraints ; in, , For each node Flow to Node ,node Flow to Node CO2 transport flow; Source point capture quantity; For the remittance point The actual amount sealed; This indicates the sink's enabled status as determined by the upper layer, with a value of 0 or 1. This represents the maximum storage capacity of the remittance point; For candidate carbon aggregation; This is a set of candidate carbon sources.

6. The CCUS source-sink matching network optimization method considering uncertainty according to claim 1, characterized in that, In S5, the dynamic weights in the feature feedback mechanism The update rules are as follows: ； in, This refers to the set of activated carbon sink nodes included in the scheme whose risk indicators are better than the historical best solution during the current iteration. This is a preset topological excitation factor; This is the updated dynamic weight value; This is the dynamic weight value before the update.

7. The CCUS source-sink matching network optimization method considering uncertainty according to claim 1, characterized in that, In step S5, the Metropolis criterion is used to determine whether to accept a candidate source-sink network scheme. The specific methods are as follows: First, calculate the new solution. Compared with the current plan fitness difference The fitness value is taken from the risk index defined in S2; in the robust decision-making mode... In decision-making mode of opportunity mode ; Secondly, calculate the probability of acceptance of the proposed solution. And decide whether to update the solution: ； in, The current temperature parameter; if the generated random number Less than If the new plan is accepted, it will be accepted; otherwise, no update will be made. Finally, following the cooling equation Perform temperature updates; where, The coefficient of performance is the cooling factor. The temperature parameters are for the low k+1 iterations. The temperature parameter is given in the k-th iteration. when Reduced to the preset lower limit When the algorithm converges, it outputs the CCUS source-sink matching network topology, total expected cost, and carbon dioxide flow allocation with the optimal risk index under the preset decision-making model.

8. A CCUS source-sink matching network optimization system considering uncertainty, characterized in that, include: The data acquisition and initialization module is used to collect basic data of the CCUS source-sink matching network and construct a physical topology including carbon source set, carbon sink set, and pipeline set. Based on the physical topology, it defines the nominal values ​​of uncertainty parameters of each carbon sink to construct an IGDT uncertainty set. Finally, it sets the total cost threshold of the source-sink matching network according to the IGDT uncertainty set. Based on a greedy heuristic strategy, it generates a binary decision vector of the initial pipeline network topology that satisfies the capacity constraint by calculating the marginal cost of candidate nodes in the IGDT uncertainty set, and uses the binary decision vector as the starting state of the main loop of the coupled optimization iteration in the two-layer coupled optimization calculation module. A two-layer coupled optimization computation module is used to construct a coupled optimization model comprising a source-sink matching network solution layer, an uncertainty analysis layer, and a flow solution layer. Based on the coupled optimization model, and according to a preset decision-making mode, the optimization objective function of the source-sink matching network solution layer is reconstructed into an extreme value search for a risk indicator. The initial temperature of the simulated annealing algorithm is set. and termination temperature Then it enters the main loop of coupled optimization iteration until the current temperature drops to the termination temperature, at which point the algorithm terminates and outputs the optimal solution under the current decision mode. In the dual-layer coupled optimization calculation module, the optimization objective function of the upper planning layer is reconstructed according to a preset decision-making model, as follows: If the decision-making model adopts a robust approach, then the cost ceiling will be met. Under the constraints, the objective function is to maximize the uncertainty tolerance. : ； If the decision-making model adopts an opportunity-based approach, then in order to meet the profit objective... Under the constraints, the optimization objective function is to minimize the chance bias. : ； in, The decision vector characterizing the source-sink topology. For the operating costs reported by the lower levels, For source-sink matching networks in topology and uncertainty parameters The total cost function is as follows; In the aforementioned dual-layer coupled optimization calculation module, the main loop of the coupled optimization iteration is to repeatedly execute S3~S5, as follows: S3: The source-sink matching network solution layer receives the dynamic weight vector fed back from S5. ; using the above Non-uniform probability sampling is performed on both disabled and enabled nodes to generate new candidate source-sink network schemes within the neighborhood of the current solution. and the Pass the inner IGDT optimization evaluation sub-loop; The IGDT optimization evaluation sub-loop is to repeatedly execute S4, as follows: S41: Based on the current level of uncertainty in the trial , get S3 Substitute the parameters into the uncertainty analysis layer and perform IGDT optimization to construct an extreme parameter set that makes the optimization objective function of the source-sink matching network reach the boundary. ; S42: The flow calculation layer will receive... As a physical constraint, As computational parameters, a linear programming model is constructed to solve for the optimal flow allocation matrix; by solving for the optimal carbon dioxide flow allocation at this point, the current candidate source-sink network scheme is further derived. The operating cost is calculated; the operating cost is fed back to S41 and compared with a preset performance threshold. The uncertainty level is then updated using a binary search method. ; S43: Repeat S41~S42 until the search converges, at which point the uncertainty level is... As a critical risk indicator for this candidate source-sink network scheme or Output; S5: Using the critical risk index solved in S43 as a feedback signal, we find a source-sink matching network topology to optimize the critical risk index, and update each carbon sink node through a dynamic scoring equation. And will update Feedback is sent to the upper planning layer for the next round of sampling; simultaneously, the difference in critical risk indicators between the old and new schemes is calculated, and the Metropolis criterion is used to determine whether to accept the candidate source-sink network scheme. and perform temperature cooling operation. .

Images