Carbon neutralization target-oriented electricity-carbon-green certificate multi-target scheduling optimization method and system

Abstract

The invention discloses a carbon neutralization target-oriented electricity-carbon-green certificate multi-target scheduling optimization method and system, and relates to the technical field of carbon neutralization control, and the method comprises the following steps: obtaining first data of a target energy storage unit, constructing a constraint generation network, and dynamically updating an energy storage operation constraint domain based on a dynamic time warping algorithm; based on the updated operation constraint domain, establishing an electricity-carbon coupling model, extracting a nonlinear mapping relation between an electric energy output behavior and carbon emission intensity, and outputting a dynamic incidence matrix through a multi-dimensional time sequence predictor; constructing a carbon flow distribution topological graph according to the dynamic incidence matrix, and obtaining an optimal carbon flow path through a Dijkstra algorithm; based on the operation constraint domain and the optimal carbon flow path, establishing a multi-subject collaborative decision-making mechanism, solving a multi-target unit equilibrium solution by adopting a distributed alternating direction multiplier method, and generating a global scheduling instruction set; according to the method, a carbon flow topology and dynamic constraint domain cooperation mechanism is constructed, so that efficient solving of a multi-main-body carbon response optimization scheduling problem is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon neutrality control, and more specifically, to a multi-objective scheduling optimization method and system for electricity-carbon-green certificate oriented to the carbon neutrality goal. Background Art

[0002] With the increasingly severe global climate change problem, carbon neutrality has become an important development strategic goal for countries around the world. To achieve this goal, the power system, as a key area of carbon emissions, is facing a transformation from a power balance-centered model to a new scheduling mechanism that coordinates "electricity-carbon-green certificate" control. In this transformation process, how to comprehensively consider multi-dimensional elements such as power transmission, carbon emission control, and green power consumption certificate (i.e., green certificate) trading, and achieve low-carbon, efficient, and flexible multi-objective optimal scheduling has become the core challenge in the intelligent and green development of the power system.

[0003] With the large-scale access of new energy, the distributed development of power sources, and the enhancement of load response capabilities, the power system presents complex characteristics of "multi-agent collaboration, multi-variable coupling, and multi-constraint dynamic evolution". This background poses higher requirements for scheduling strategies: not only considering the spatio-temporal uncertainties on the power generation side and load side, but also taking into account the unified optimization of various heterogeneous elements such as energy storage systems, carbon emission paths, carbon intensity indicators, and green certificate weights. In this process, how to dynamically model the regulation capabilities of different agents, identify carbon flow paths, construct a constraint domain with real-time adaptability, and on this basis, achieve differential decomposition of regulation tasks and multi-objective fusion optimization has become the key technical bottleneck for achieving the carbon neutrality scheduling goal.

[0004] For example, a multi-energy coupling system optimal scheduling method considering grid connection safety disclosed in the invention patent with the publication number of CN120016607A includes: constructing a multi-energy coupling system architecture model; the multi-energy coupling system architecture model includes a renewable energy power generation unit, a traditional energy power generation unit, an energy storage unit, an energy conversion unit, and a combined cooling, heating, and power unit; constructing a distributed energy storage optimal operation model based on the multi-energy coupling system architecture model; the constraint conditions include: charge and discharge constraints of battery energy storage, state of charge constraints, and peak shaving constraints; output constraints of energy conversion equipment; operation constraints of electric vehicle charging stations; new energy power generation constraints; power balance constraints; transmission power constraints of public network lines; transmission power constraints of internal distribution network lines; and using an optimization algorithm to solve the distributed energy storage optimal operation model to obtain an optimal scheduling plan. The advantage of the present invention is to provide scientific and reliable technical support for the efficient operation of the multi-energy coupling system. For example, a method and device for constructing a robust operation domain of flexible resources in a new energy power system disclosed in a patent for invention with the publication number of CN120016585A, which relates to the field of power system operation control. The method includes: obtaining the parameters of a generator set, an energy storage system, and an HVDC transmission system, and constructing a flexible resource operation domain model; obtaining the uncertainty sets of the power generation output of a new energy power station and the substation bus load; obtaining the power grid topology information, and constructing the power balance and line power flow security constraints for multiple time periods; constructing the power value or energy value as an uncertain quantity related to the operation domain boundary variable, and calculating the robust operation domain of the flexible resources by means of decision-related uncertainty and robust optimization methods; and performing the operation control of the generator set, the energy storage system, and the HVDC transmission system according to the robust operation domain. The present invention can meet the unexpected requirements of the dynamic regulation of the new energy power system under the action of a stochastic process, and provide a simple and easy-to-execute strategy for the regulation of flexible resources and the consumption of new energy.

[0005] In the above-disclosed technical solution, there are at least the following technical problems: In the prior art, there are generally problems in power dispatching methods such as weak carbon emission modeling ability, lack of carbon flow path tracking mechanism, insufficient coordinated regulation of main bodies, static constraint domain not adapting to new scenarios, and fragmentation of electricity-carbon-green certificate dispatching. Specifically, the carbon emission impact is only embedded in the model as an additional constraint, and the carbon behavior in the energy storage and load response cannot be dynamically reflected; the conduction path of the carbon factor lacks modeling and optimization; the dispatching process is mainly centralized, and it is difficult to achieve the coordinated response between multiple devices; the constraint boundary is solidified and is prone to failure in the face of disturbances; electricity, carbon, and green certificates do not form a unified dispatching optimization system and cannot effectively support the refined regulation of the carbon neutrality goal. In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0006] In order to overcome the above defects of the prior art, an embodiment of the present invention provides an electricity-carbon-green certificate multi-objective dispatching optimization method and system for a carbon neutrality goal. By adopting a constraint generation network combining a feature migration and an adversarial enhancement mechanism, the method effectively addresses the uncertainty and constraint drift problems of the energy storage operation conditions in different scenarios, and significantly improves the response accuracy and coordinated dispatching ability of the multi-energy system under the carbon neutrality goal.

[0007] To achieve the above object, the present invention provides the following technical solutions: A multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal, comprising the following steps: obtaining first data of a target energy storage unit, constructing a constraint generation network, and dynamically updating the energy storage operation constraint domain based on the dynamic time warping algorithm; based on the updated operation constraint domain, establishing an electricity-carbon coupling model, extracting the non-linear mapping relationship between the electric energy output behavior and the carbon emission intensity, and outputting a dynamic correlation matrix through a multi-dimensional time series predictor; constructing a carbon flow distribution topology map with spatio-temporal labels according to the dynamic correlation matrix, and calculating the optimal carbon flow path with the minimum carbon response path cost through an improved Dijkstra algorithm; based on the operation constraint domain and the optimal carbon flow path, establishing a multi-agent collaborative decision-making mechanism, and using the distributed alternating direction multiplier method to solve the equilibrium solutions of the energy storage unit, the load unit and the carbon source unit, and generating a global scheduling instruction set.

[0008] In a preferred embodiment, the obtaining of the first data of the target energy storage unit and the construction of the constraint generation network are specifically as follows: obtaining the real-time operation data of the target energy storage unit and performing data preprocessing on the operation data, where the operation data includes real-time power, state of charge and historical carbon response records; constructing an adversarial constraint enhancement network including a perturbation generator and a discriminator, where the perturbation generator is used to generate perturbation scenario samples, and the discriminator is used to evaluate the recognizability of the original scenario and the perturbation scenario; alternately optimizing and training the discriminator and the perturbation generator, and iteratively generating multiple groups of virtual sample sets that meet the preset constraint conditions; inputting the virtual samples into the constraint output layer, performing multi-dimensional feature comparison and analysis with the historical actual regulation data, extracting the comparison data, and constructing the constraint generation network.

[0009] In a preferred embodiment, the dynamically updating the energy storage operation constraint domain based on the dynamic time warping algorithm is specifically as follows: calculating the matching degree score between the comparison data of each group of virtual samples and the actual operation behavior sequence based on the dynamic time warping algorithm, and mapping it to a scenario fitness value; according to the scenario fitness value, assigning dynamic weights to the virtual samples through a non-linear weight allocation model to generate a weighted virtual sample population; based on the weighted virtual sample population, using a multi-objective optimization algorithm to dynamically update the boundary parameters of the operation constraint domain of the energy storage system.

[0010] In a preferred embodiment, based on the weighted virtual sample population, a multi-objective optimization algorithm is used to dynamically update the boundary parameters of the operation constraint domain of the energy storage system, specifically as follows: statistically analyze the confidence interval of the power parameters in the weighted virtual sample population. If the current operation data exceeds the confidence interval, the power constraint boundary of the energy storage system is dynamically expanded; based on the SOC data of the virtual samples, a state transition path set and a transition matrix are constructed, and low-frequency transition paths that satisfy the law of energy conservation are identified, incorporated into the constraint domain, and a directed state transition graph with a probability threshold is constructed to obtain state transition paths; according to the carbon emission intensity corresponding to the time period of the energy storage operation in the virtual samples, a carbon response intensity mapping relationship is established. If the carbon response intensity is lower than the preset threshold, the carbon emission weight in the constraint domain is dynamically weighted and amplified; based on the updated power constraint boundary, state transition paths, and carbon emission weights, the boundary parameters of the operation constraint domain of the energy storage system are dynamically updated through a multi-objective optimization algorithm.

[0011] In a preferred embodiment, based on the updated operation constraint domain, an electricity-carbon coupling model is established to extract the non-linear mapping relationship between the electricity output behavior and the carbon emission intensity, and a dynamic correlation matrix is output through a multi-dimensional time series predictor, specifically as follows: multi-dimensional operation characteristics are extracted based on the operation constraint domain of the energy storage system, a function mapping relationship between the electricity output behavior and the carbon emission impact is established, and an electricity-carbon coupling model is constructed to output an electricity-carbon interaction dynamic sequence; according to the electricity-carbon interaction dynamic sequence and the operation characteristics, a multi-dimensional time series predictor is constructed through an attention mechanism, and the predictor includes a time dependence analysis module and a feature cross-correlation module; an input feature set is extracted from the multi-dimensional time series predictor, and a three-dimensional input feature tensor is constructed. The input feature set includes power fluctuation characteristics, SOC trajectory characteristics, and carbon response intensity characteristics; the input feature tensor is subjected to vector space mapping through a self-attention network to generate a query vector matrix and a key vector matrix, and a dynamic correlation matrix is calculated.

[0012] In a preferred embodiment, a carbon flow distribution topology graph with spatio-temporal tags is constructed according to the dynamic correlation matrix, specifically as follows: electricity-carbon coupling variables are extracted from the dynamic correlation matrix output by the multi-dimensional time series predictor, and a multi-dimensional feature vector set is generated by encoding according to the weight coefficients of the coupling variables; a node set and an edge set of a graph structure are constructed. The node set includes an energy storage node, a load node, and a regional carbon source node, and the edge set includes an electricity flow path edge and a carbon response path edge; spatio-temporal tag information is attached to each edge to construct a weighted spatio-temporal carbon flow distribution topology graph, and the spatio-temporal tags include a timestamp and a geographic coordinate vector.

[0013] In a preferred embodiment, the optimal carbon flow path with the minimum cost of the carbon response path is calculated by improving the Dijkstra algorithm as follows: Based on the pre-constructed weighted spatio-temporal carbon flow distribution topology graph, the starting node and the ending node of the path are selected; the shortest path from the starting node to the ending node is calculated by improving the Dijkstra algorithm, and with the minimization of the total path cost as the optimization goal, an optimal carbon flow path sequence is generated.

[0014] In a preferred embodiment, the multi-agent collaborative decision-making mechanism is established based on the operation constraint domain and the optimal carbon flow path as follows: The collaborative data in the system operation constraint domain and the carbon flow path are extracted, and after data preprocessing, a unified input data set is formed; according to the functions of the devices in the carbon flow path, the system is divided into power generation source nodes, transmission path nodes and power consumption response nodes: based on the global regulation goal and the carbon flow topology graph, the overall task is decomposed into regional sub-tasks, and an initial scheduling instruction is generated by combining the constraint capabilities of each node and the carbon emission weight; the initial instruction is adapted according to the local regulation conditions to generate an executable node instruction; during the execution of the node, the change of the carbon flow density is monitored in real time, and the local regulation parameters are dynamically feedback-regulated.

[0015] In a preferred embodiment, the distributed alternating direction method of multipliers is used to solve the equilibrium solutions of the energy storage unit, the load unit and the carbon source unit, and a global scheduling instruction set is generated as follows: According to the evaluation results of task response consistency and path feasibility and the division results of each control node, the overall regulation problem is decomposed into multiple local optimization sub-problems, the shared variables between the local problems are extracted, and the initial consistency conditions are set; corresponding distributed optimization agents are deployed at each control unit, and the agents report the local solution results and the status of the relevant shared variables to the central scheduling coordinator; after receiving the solutions of each sub-problem, the central scheduling coordinator evaluates whether the consistency between the shared variables meets the set conditions and judges whether the system reaches global equilibrium; if the equilibrium conditions are met, the optimal solutions of each sub-node are integrated to generate a global scheduling instruction set under consistency constraints.

[0016] System for an electro-carbon-green certificate multi-objective scheduling optimization method towards carbon neutrality target, characterized by comprising a constraint domain module, an association matrix module, a carbon flow path screening module, and an instruction generation module, with connections among the modules; the constraint domain module is used to obtain first data of a target energy storage unit, construct a constraint generation network, and dynamically update the energy storage operation constraint domain based on the dynamic time warping algorithm; the association matrix module is used to establish an electro-carbon coupling model based on the updated operation constraint domain, extract the non-linear mapping relationship between the electric energy output behavior and the carbon emission intensity, and output a dynamic association matrix through a multi-dimensional time series predictor; the carbon flow path screening module is used to construct a carbon flow distribution topology map with spatio-temporal labels according to the dynamic association matrix, and calculate the optimal carbon flow path with the minimum carbon response path cost through an improved Dijkstra algorithm; the instruction generation module is used to establish a multi-agent collaborative decision-making mechanism based on the operation constraint domain and the optimal carbon flow path, solve the equilibrium solutions of the energy storage unit, the load unit, and the carbon source unit by using the distributed alternating direction method of multipliers, and generate a global scheduling instruction set.

[0017] Technical effects and advantages of an electro-carbon-green certificate multi-objective scheduling optimization method and system of the present invention towards carbon neutrality target: 1. By constructing a dynamic energy storage constraint generation network, combining with the dynamic time warping algorithm, and introducing a matching mechanism between virtual disturbance scenarios and actual operation behaviors, the present invention forms an operation boundary with high scenario adaptability and dynamic update ability. This mechanism can significantly improve the robustness and response ability of the energy storage system in an uncertain environment, and solves the problem that the constraint domain in traditional methods is static and cannot adapt to changing working conditions.

[0018] 2. Through electro-carbon coupling modeling and multi-dimensional time series prediction mechanism, the present invention can accurately depict the influence relationship of electric energy output behavior on carbon emissions, and build a dynamic association matrix with the help of the attention mechanism to ensure that the model has the ability to distinguish different operating conditions at different times. Compared with the previous method of only performing coarse-grained modeling with an average carbon emission coefficient, the present invention has stronger refined dynamic modeling ability and can better reflect the real carbon emission distribution characteristics. Description of the Drawings

[0019] Figure 1 It is a schematic flow chart of an electro-carbon-green certificate multi-objective scheduling optimization method of the present invention towards carbon neutrality target.

[0020] Figure 2 It is a schematic structural diagram of a system of an electro-carbon-green certificate multi-objective scheduling optimization method of the present invention towards carbon neutrality target. Detailed Embodiments

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] Embodiment 1 Figure 1 A multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal of the present invention is given, including the following steps: S1. Obtain the first data of the target energy storage unit, construct a constraint generation network, and dynamically update the energy storage operation constraint domain based on the dynamic time warping algorithm.

[0023] In this embodiment, obtaining the first data of the target energy storage unit and constructing a constraint generation network are as follows: Collect the operation data of the energy storage unit. The collected operation data includes: charge and discharge power time series data (recording the charge and discharge power curves of the energy storage unit during actual operation, with a sampling period of 5 minutes to 15 minutes, reflecting its response ability to different scheduling instructions), energy state migration paths (tracking the energy state transfer process between different operation stages of the energy storage unit, including the transfer from the charging state to the discharging state, standby state, etc.), and the power grid carbon emission intensity map (obtaining the carbon emission intensity data per unit electricity of the power grid during the energy storage operation period, divided by region and time period to form a time series map reflecting the change of carbon intensity); Input the first data into the feature transfer learning model, use the historical operation data as the source domain and the current scenario data as the target domain to construct a constraint generation network. The structure of the constraint generation network includes a feature extraction layer, a transfer mapping layer, and a constraint output layer.

[0024] The structure of the constraint generation network includes a feature extraction layer, a transfer mapping layer, and a constraint output layer, where: The feature extraction layer is responsible for extracting the SOC (state of charge) change rate, power volatility, and carbon intensity correlation index from the collected data; The transfer mapping layer realizes the cross-time domain alignment of the feature space based on the maximum mean discrepancy (MMD) or adversarial domain adaptation (such as DANN); The constraint output layer outputs the operation boundary conditions of the energy storage unit adapted to the current system state, including the maximum / minimum charge and discharge power, the allowed SOC change path, and the continuous operation duration.

[0025] In this embodiment, dynamically updating the energy storage operation constraint domain based on the dynamic time warping algorithm is as follows: Based on the dynamic time warping algorithm, a perturbation generator is constructed to impose small - amplitude perturbations on the input features, and the discriminator is used to evaluate the recognizability of the original scenario and the perturbed scenario. By alternately optimizing the discriminator and the perturbation generator, multiple groups of virtual sample sets covering boundary cases are generated; The virtual samples are input into the constrained output layer and compared with the historical actual regulation data to obtain comparison data, which includes the power time - series deviation rate, the difference in state - transfer paths, and the matching degree of the corresponding carbon intensity intervals; The comparison data of each group of virtual samples is matched and scored with the actual operation behavior to output the scenario fitness; Dynamic weights are assigned to each virtual sample according to the scenario fitness to obtain a weighted virtual sample population, and the weight - assignment function meets the following requirements: samples with high fitness account for a larger proportion in subsequent constraint correction, a weight - boosting factor is set for low - frequency extreme scenarios to enhance system resilience, and the weighting method uses the soft - label method, and overfitting is prevented through regularization terms; According to the weighted virtual sample population, the energy - storage operation constraint domain is dynamically updated.

[0026] In this embodiment, according to the weighted virtual sample population, the energy - storage operation constraint domain is dynamically updated as follows: The confidence intervals of the power upper and lower limits in the virtual sample population are statistically analyzed. If out - of - sample extreme value behaviors occur during the current operation, the power boundary is expanded; Based on the SOC data in the virtual sample population, a state - transfer path set is constructed, and a state - transfer matrix is constructed. Transfer paths with low frequencies but conforming to the law of energy conservation in the state - transfer matrix are identified, and it is judged whether the single - step energy change of the newly added SOC path is within the allowable charge - discharge rate of the device. If the condition is met, it is included in the constraint domain, and each state - transfer path is recorded by a directed graph and a transition probability threshold is set; The carbon emission intensity corresponding to the energy - storage actions (such as "high - frequency discharging") in the virtual samples is statistically analyzed, and a carbon response intensity mapping relationship is established. If a certain type of energy - storage scheduling action frequently appears in high - carbon sections (such as the high overlap between the discharging peak and the carbon - intensity peak), it is determined that its carbon response intensity is low, and its carbon response intensity weight is weighted and amplified.

[0027] S2. Based on the updated operation constraint domain, an electricity - carbon coupling model is established, the non - linear mapping relationship between the electric - energy output behavior and the carbon emission intensity is extracted, and a dynamic correlation matrix is output through a multi - dimensional time - series predictor.

[0028] In this embodiment, based on the updated operation constraint domain, an electricity - carbon coupling model is established, the non - linear mapping relationship between the electric - energy output behavior and the carbon emission intensity is extracted, and a dynamic correlation matrix is output through a multi - dimensional time - series predictor, as follows: Extract operation characteristics according to the constraint domain, where the operation characteristics include power behavior, state transition path set, carbon response intensity, and scenario fitness; Establish a functional mapping relationship between the power output behavior and the carbon emission impact according to the operation characteristics, construct an electricity-carbon coupling model, and output a dynamic sequence. Quantify the impact of different control actions on the system carbon emissions in the current environment through the electricity-carbon coupling model, so as to support the optimization of dispatching decisions under carbon constraint conditions; Construct a multi-dimensional time series predictor based on the attention mechanism according to the dynamic sequence, power behavior, and state transition path; Extract input features in the system according to the multi-dimensional time series predictor, where the input features include the historical time window length and operation characteristics; Construct an input feature tensor according to the input features , where is the input feature tensor, is the historical time window length, is the number of input features, ; Perform vector mapping on the input feature tensor through the self-attention network structure to obtain a query vector and a key vector, and output a dynamic correlation matrix.

[0029] The electricity-carbon coupling model is specifically as follows:

[0030] In the formula: represents the carbon emission estimation value at time t, is the power behavior of the energy storage unit, is the energy storage state transition path, is the carbon response intensity, is the scenario fitness.

[0031] The dynamic correlation matrix is specifically as follows:

[0032] In the formula: is the dynamic correlation matrix, representing the dynamic coupling strength between input variables at the t-th moment, query vector, is the key vector, represents the transpose of the key vector, is the vector dimension, The operation is used for normalization to ensure that the matrix is normalized to the interval [0,1].

[0033] The structure of the multi-dimensional time series predictor includes: Time series encoder: Use a bidirectional GRU or Transformer structure to extract the time correlation between variables; Self-attention module: Extract the degree of mutual influence between variables at each time step; Mapping output layer: Output the carbon emission prediction value for the next moment and the influence matrix between variables.

[0034] S3. Construct a carbon flow distribution topology map with spatio-temporal labels based on the dynamic correlation matrix, and calculate the optimal carbon flow path with the minimum cost of the carbon response path through an improved Dijkstra algorithm.

[0035] In this embodiment, a carbon flow distribution topology map with spatio-temporal labels is constructed based on the dynamic correlation matrix, and the optimal carbon flow path with the minimum cost of the carbon response path is calculated through an improved Dijkstra algorithm, specifically as follows: Extract the coupled variables from the dynamic correlation matrix output by the multi-dimensional time series predictor, and encode them into a set of feature vectors according to the coupling weights of the coupled variables. The coupled variables include state transition trend, charge and discharge rate, carbon intensity, and regional load intensity; Take the energy storage nodes, load nodes, and regional carbon source nodes as the graph node set, and construct an edge set based on the power flow path and carbon response path. The initial weight of the edges in the graph is initialized by the coupling matrix output by the predictor and is dynamically updated during model training; Attach spatio-temporal label information to each edge to construct a spatio-temporal carbon flow distribution topology map with time series update ability. The spatio-temporal label information includes the trigger timestamp of the edge, the region to which the node belongs, and the load level; According to the carbon flow distribution topology map, combined with the actual operating state and carbon response requirements of the energy storage unit, select some nodes as the start and end points of the path. Take the energy storage behavior trigger node as the starting node and the load absorption node as the ending node, and perform the shortest path calculation through the Dijkstra algorithm. With the minimum weighted carbon response path cost as the optimization goal, obtain the optimal carbon flow path in the current scenario; The energy storage behavior trigger node refers to the node corresponding to the energy storage device with regulation behavior (such as charging or discharging action) within the current scheduling period; The load absorption node refers to the terminal node with the ability to absorb carbon emissions or associated with the green certificate registration system and low-carbon load units.

[0036] The specific Dijkstra algorithm is as follows:

[0037] In the formula: is the set of feasible paths from the starting node to the ending node, is the starting node, is the ending node, represents the carbon response weight of the edge connecting node i and node j at time t; S4. Based on the operation constraint domain and the optimal carbon flow path, establish a multi-agent collaborative decision-making mechanism, and use the distributed alternating direction method of multipliers to solve the equilibrium solutions of the energy storage unit, the load unit, and the carbon source unit, and generate a global scheduling instruction set.

[0038] In this embodiment, based on the operation constraint domain and the optimal carbon flow path, establish a multi-agent collaborative decision-making mechanism as follows: Extract the collaborative data in the constraint domain and the carbon flow path, and perform fusion and sorting to form a unified data input set for collaborative optimization. The collaborative data includes power boundaries, state paths, action frequency constraints, path nodes, channel selection priorities, and key control points; According to the device functions and the positions of the carbon flow paths, respectively demarcate the power generation source nodes (including energy storage power sources and renewable power sources), the transmission path nodes (such as regional interconnection interfaces), and the electricity consumption response nodes (including load adjustable devices, carbon response units, etc.). Each node participates in the regulation task as an independent collaborative control unit; Inject the constraint domain into the corresponding energy storage unit according to the functions of each node, and allocate the boundary conditions with higher carbon response sensitivity in the path to the key load nodes and transmission channel nodes to construct a set of local regulation conditions with multi-node differences; According to the global regulation objectives (such as peak shaving and valley filling, utilization of carbon reduction priority paths), combined with the carbon flow distribution topology map, decompose the overall regulation task into several regional task units, and generate initial scheduling instructions according to the operation constraint domain and carbon emission weights of each node; Each node localizes and adapts the initial scheduling instructions according to the set of local regulation conditions, and generates node execution data on the premise of meeting the local constraint conditions. The node execution data includes adjustable ranges, response priorities, path switching tolerances, etc., to form a targeted node execution plan; While each node executes the initial scheduling instructions, it monitors its own state response, and adjusts the local regulation conditions based on the changes in the carbon flow density and control information transmitted in the path; After each control unit completes the task response, the system integrates all node feedback information, and evaluates the task response consistency and path feasibility based on the ADMM residual evaluation method, providing an input basis for subsequent global optimal scheduling.

[0039] In this embodiment, use the distributed alternating direction method of multipliers to solve the equilibrium solutions of the energy storage unit, the load unit, and the carbon source unit, and generate a global scheduling instruction set as follows: According to the evaluation results of task response consistency and path feasibility and the division results of each control node, decompose the overall regulation problem into multiple local optimization sub-problems. Each sub-problem only involves the constraint parameters, regulation variables, and response objectives of this node and its directly adjacent nodes, thereby reducing the computational coupling degree and facilitating parallel solution; Extract shared variables among local problems (such as transmission power between nodes, carbon flow distribution ratio, control boundary crossing intervals, etc.), and set initial consistency conditions. The shared variables serve as a collaborative bridge among multiple sub-problems for subsequent consistency convergence judgment and feedback correction; Deploy corresponding distributed optimization agents at each control unit. The agents run the optimization process in parallel through local modeling and edge computing resources, and report their local calculation results to the central coordinator. This architecture can achieve system-level parallel optimization and iterative scheduling with the support of edge computing resources; Based on the solutions of each sub-problem and the status of shared variables collected, the scheduling coordinator determines whether the overall consistency conditions of the system are met (such as whether the carbon emission estimation values at each node converge to the target values and whether the power of the connected paths is balanced). If not, it adjusts the estimated values of shared variables or local target weights according to the adjustment rules, and feeds back the correction information to each agent to start the next round of calculation; When the system meets the equilibrium condition, that is, the shared variables among the output solutions of all agents reach the convergence state, the system considers that the multi-agent collaborative equilibrium solution has been obtained, and the central coordinator integrates the final optimization outputs of each sub-node to generate a set of global scheduling instruction sets covering the entire path, all nodes, and the entire cycle.

[0040] Embodiment 2 Figure 2 A system of the multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal of the present invention is provided, which is characterized by including a constraint domain module, an association matrix module, a carbon flow path screening module, and an instruction generation module, and there are connections among the modules; The constraint domain module is used to obtain the first data of the target energy storage unit, construct a constraint generation network, and dynamically update the energy storage operation constraint domain based on the dynamic time warping algorithm; The association matrix module is used to establish an electricity-carbon coupling model based on the updated operation constraint domain, extract the non-linear mapping relationship between the electric energy output behavior and the carbon emission intensity, and output a dynamic association matrix through a multi-dimensional time series predictor; The carbon flow path screening module is used to construct a carbon flow distribution topology graph with spatio-temporal tags according to the dynamic association matrix, and calculate the optimal carbon flow path with the minimum carbon response path cost through an improved Dijkstra algorithm; The instruction generation module is used to establish a multi-agent collaborative decision-making mechanism based on the operation constraint domain and the optimal carbon flow path, solve the equilibrium solutions of the energy storage unit, the load unit, and the carbon source unit by using the distributed alternating direction multiplier method, and generate a global scheduling instruction set.

[0041] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0042] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0043] Those of ordinary skill in the art will realize that the modules and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented with electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0044] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0045] As described above, this is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0046] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal, characterized in that Including: Obtain the first data of the target energy storage unit, construct a constraint generation network, and dynamically update the energy storage operation constraint domain based on the dynamic time warping algorithm; Based on the updated operation constraint domain, establish an electricity-carbon coupling model, extract the non-linear mapping relationship between the electric energy output behavior and the carbon emission intensity, and output a dynamic correlation matrix through a multi-dimensional time series predictor; Construct a carbon flow distribution topology map with spatio-temporal labels according to the dynamic correlation matrix, and calculate the optimal carbon flow path with the minimum carbon response path cost through an improved Dijkstra algorithm; Based on the operation constraint domain and the optimal carbon flow path, establish a multi-agent collaborative decision-making mechanism, and use the distributed alternating direction multiplier method to solve the equilibrium solutions of the energy storage unit, the load unit and the carbon source unit, and generate a global scheduling instruction set.

2. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 1, wherein The obtaining of the first data of the target energy storage unit and the construction of the constraint generation network are specifically as follows: Obtain the real-time operation data of the target energy storage unit and perform data preprocessing on the operation data. The operation data includes real-time power, state of charge and historical carbon response records; Construct an adversarial constraint enhancement network including a perturbation generator and a discriminator, where the perturbation generator is used to generate perturbation scenario samples, and the discriminator is used to evaluate the recognizability of the original scenario and the perturbation scenario; By alternately optimizing and training the discriminator and the perturbation generator, iteratively generate multiple sets of virtual sample sets that meet the preset constraint conditions; Input the virtual samples into the constraint output layer, perform multi-dimensional feature comparison and analysis with the historical actual regulation data, extract the comparison data, and construct a constraint generation network.

3. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 2, characterized in that The dynamic update of the energy storage operation constraint domain based on the dynamic time warping algorithm is specifically as follows: Calculate the matching degree score between the comparison data of each group of virtual samples and the actual operation behavior sequence based on the dynamic time warping algorithm, and map it to a scenario fitness value; According to the scenario fitness value, assign dynamic weights to the virtual samples through a non-linear weight allocation model to generate a weighted virtual sample population; Based on the weighted virtual sample population, use a multi-objective optimization algorithm to dynamically update the boundary parameters of the energy storage system's operation constraint domain.

4. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 3, characterized in that The dynamic update of the boundary parameters of the energy storage system's operation constraint domain based on the weighted virtual sample population is specifically as follows: Statistically analyze the confidence interval of the power parameters in the weighted virtual sample population. If the current operation data exceeds the confidence interval, dynamically expand the power constraint boundary of the energy storage system; Construct a state transition path set and a transition matrix based on the SOC data of the virtual samples, identify the low-frequency transition paths that satisfy the law of energy conservation, include them in the constraint domain and construct a directed state transition graph with a probability threshold to obtain the state transition paths; Establish a carbon response intensity mapping relationship according to the carbon emission intensity corresponding to the energy storage action time period in the virtual samples. If the carbon response intensity is lower than the preset threshold, dynamically weight and amplify the carbon emission weight in the constraint domain; Based on the updated power constraint boundary, state transition path and carbon emission weight, use a multi-objective optimization algorithm to dynamically update the boundary parameters of the energy storage system's operation constraint domain.

5. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 4, characterized in that, Based on the updated operation constraint domain, an electricity-carbon coupling model is established to extract the non-linear mapping relationship between electricity output behavior and carbon emission intensity, and a dynamic correlation matrix is output through a multi-dimensional time series predictor, which is specifically as follows: Extract multi-dimensional operation characteristics based on the operation constraint domain of the energy storage system, establish a functional mapping relationship between electricity output behavior and carbon emission impact, construct an electricity-carbon coupling model, and output an electricity-carbon interaction dynamic sequence; According to the electricity-carbon interaction dynamic sequence and operation characteristics, construct a multi-dimensional time series predictor through an attention mechanism. The predictor includes a time dependence analysis module and a feature cross-correlation module; Extract the input feature set from the multi-dimensional time series predictor and construct a three-dimensional input feature tensor. The input feature set includes power fluctuation characteristics, SOC trajectory characteristics, and carbon response intensity characteristics; Perform vector space mapping on the input feature tensor through a self-attention network to generate a query vector matrix and a key vector matrix, and calculate the dynamic correlation matrix.

6. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 5, wherein The carbon flow distribution topology map with spatio-temporal labels is constructed according to the dynamic correlation matrix, which is specifically as follows: Extract the electricity-carbon coupling variables from the dynamic correlation matrix output by the multi-dimensional time series predictor, and generate a multi-dimensional feature vector set according to the weight coefficients of the coupling variables; Construct a node set and an edge set of the graph structure. The node set includes energy storage nodes, load nodes, and regional carbon source nodes. The edge set includes electricity flow path edges and carbon response path edges; Attach spatio-temporal label information to each edge to construct a weighted spatio-temporal carbon flow distribution topology map. The spatio-temporal label includes a timestamp and a geographical coordinate vector.

7. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 6, wherein The optimal carbon flow path with the minimum cost of the carbon response path is calculated through an improved Dijkstra algorithm, which is specifically as follows: Based on the pre-constructed weighted spatio-temporal carbon flow distribution topology map, select the starting node and the ending node of the path; Calculate the shortest path from the starting node to the ending node through an improved Dijkstra algorithm, and generate an optimal carbon flow path sequence with the minimum total path cost as the optimization goal.

8. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 7, characterized in that Based on the operation constraint domain and the optimal carbon flow path, a multi-agent collaborative decision-making mechanism is established, which is specifically as follows: Extract the collaborative data in the system operation constraint domain and the carbon flow path, and form a unified input data set after data preprocessing; According to the functions of the devices in the carbon flow path, the system is divided into power generation source nodes, transmission path nodes, and electricity consumption response nodes: Based on the global regulation goal and the carbon flow topology map, decompose the overall task into regional sub-tasks, and generate an initial scheduling instruction by combining the constraint capabilities of each node and the carbon emission weight; Adapt the initial instruction according to the local regulation conditions to generate an executable node instruction; During the execution of the node, monitor the change of carbon flow density in real time, and dynamically feedback and adjust the local regulation parameters.

9. The multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal according to claim 8, wherein The distributed alternating direction multiplier method is used to solve the equilibrium solutions of the energy storage unit, the load unit, and the carbon source unit, and generate a global scheduling instruction set, which is specifically as follows: According to the task response consistency and path feasibility evaluation results and the division results of each control node, decompose the overall regulation problem into multiple local optimization sub-problems, extract the shared variables between each local problem, and set the initial consistency condition; Deploy corresponding distributed optimization agents at each control unit, and the agents report the local solution results and the status of relevant shared variables to the central scheduling coordinator; After receiving the solutions of each sub-problem, the central scheduling coordinator evaluates whether the consistency among the shared variables meets the set conditions to determine whether the system reaches global equilibrium; If the equilibrium condition is satisfied, integrate the optimal solutions of each sub-node to generate a global scheduling instruction set under consistency constraints.

10. A system using the multi-objective scheduling optimization method for electricity-carbon-green certificate oriented to the carbon neutrality goal as described in any one of claims 1-9, characterized in that, It includes a constraint domain module, an incidence matrix module, a carbon flow path screening module, and an instruction generation module, and there are connections between the modules; The constraint domain module is used to obtain the first data of the target energy storage unit, construct a constraint generation network, and dynamically update the energy storage operation constraint domain based on the dynamic time warping algorithm; The incidence matrix module is used to establish an electricity-carbon coupling model based on the updated operation constraint domain, extract the non-linear mapping relationship between the electricity output behavior and the carbon emission intensity, and output a dynamic incidence matrix through a multi-dimensional time series predictor; The carbon flow path screening module is used to construct a carbon flow distribution topology map with spatio-temporal labels according to the dynamic incidence matrix, and calculate the optimal carbon flow path with the minimum carbon response path cost through an improved Dijkstra algorithm; The instruction generation module is used to establish a multi-agent collaborative decision-making mechanism based on the operation constraint domain and the optimal carbon flow path, solve the equilibrium solutions of the energy storage unit, the load unit, and the carbon source unit by using the distributed alternating direction method of multipliers, and generate a global scheduling instruction set.

Images