Power distribution network carbon flow fine quantification method and system based on dynamic topology self-adaption

Abstract

The invention discloses a power distribution network carbon flow fine quantification method and system based on dynamic topology self-adaption, and the method comprises the steps: firstly evaluating the importance degree of each node of a power distribution network based on a node importance evaluation index, and screening out high-importance nodes in the power distribution network, the node importance evaluation index considering a dynamic topology correlation degree, load sensitivity and a carbon flow influence weight; then, collecting measurement data of each high-importance node in real time, constructing a condition denoising diffusion implicit model taking the measurement data as a condition, and generating complete power distribution network measurement data; and finally, determining the carbon flow of the power distribution network based on the complete measurement data of the power distribution network. According to the method, cost redundancy caused by excessive measurement and information vacancy caused by missing of important measurement objects are fundamentally solved, the problems of missing of measurement data and noise interference are effectively solved, reliable data support is provided for carbon flow calculation, and low-carbon dispatching and operation optimization of a power distribution network are supported.

Description

Technical Field

[0001] This invention belongs to the field of power system carbon emission calculation and deep learning technology, specifically involving a method and system for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation. Background Technology

[0002] Under the "dual carbon" goal, accurate calculation of the carbon emission flow (CEF) of the distribution network is crucial for achieving low-carbon operation of the power system. Carbon flow calculation relies on complete distribution network measurement data, including information such as power generation, load, and energy type of each node. However, the collection of measurement data in the distribution network usually suffers from the following problems: (1) Insufficient coverage of measurement equipment: Many nodes lack real-time measurement devices, resulting in missing data; (2) Data noise and anomalies: Intermittent output of distributed power sources and measurement equipment errors lead to low data reliability.

[0003] Traditional methods for obtaining complete measurement data include physical model interpolation or statistical correction, but these two methods struggle to handle high-dimensional, non-linear data distributions. Deep learning models such as Generative Adversarial Networks (GANs) also exist, but while they can generate data, they suffer from pattern collapse and training instability. Therefore, a new method for processing distribution network data is needed to provide reliable input for carbon flow calculations. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned problems in the existing technology by providing a method and system for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, this invention proposes a method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation, including:

[0007] S1. Based on the node importance assessment index, assess the importance of each node in the distribution network and screen out the high-importance nodes in the distribution network. The node importance assessment index considers dynamic topology correlation, load sensitivity, and carbon flow impact weight.

[0008] S2. Real-time acquisition of measurement data from each high-importance node, construction of a conditional denoising diffusion implicit model based on the measurement data, and generation of complete distribution network measurement data;

[0009] S3. Determine the carbon flow of the distribution network based on complete distribution network measurement data.

[0010] In S1, the importance of each node in the distribution network is calculated using the following formula:

[0011] ;

[0012] ;

[0013] ;

[0014] ;

[0015] In the above formula, For nodes The importance of , , These are the weights for dynamic topological correlation, load sensitivity, and carbon flow influence, respectively. For nodes Dynamic topological correlation degree, For nodes The load sensitivity index For nodes The carbon flow influences the weight. For nodes after topological changes arrive Passing through the node The total number of shortest paths, Nodes after topological change arrive The total number of shortest paths, The topological change coefficient, , They are nodes The maximum and minimum active loads, For nodes The average active load, For nodes Load type weights, This represents the average change in carbon potential across all nodes in the network. This represents the total number of nodes in the distribution network. For nodes The change in active power.

[0016] The The following formula is used to calculate:

[0017] ;

[0018] ;

[0019] ;

[0020] In the above formula, For topological dynamic weights, The topological change coefficient, This is the matrix of topological changes. For the distribution network node-branch correlation matrix, This is the node-branch correlation matrix after topological changes.

[0021] S2 includes:

[0022] S21, based on measurement data Based on the conditional information, a sampling model for the backdiffusion process is constructed:

[0023] ;

[0024] ;

[0025] ;

[0026] In the above formula, In order to be in and Under these conditions, time step The posterior probability of the corresponding data. For time step The corresponding noisy data, For time step The corresponding noisy data, Given the measurement data, For posterior probability The mean, For posterior probability variance For time step The corresponding noise scheduling parameters, For noise;

[0027] S22. Define the conditional denoising function And the mean of the sampling model for the reverse diffusion process. and variance They are respectively represented as and ,in, For the missing data to be generated, Given the known conditional observations, For time step Missing values ​​that need to be generated for The conditional observations known in the data and used to estimate the missing values;

[0028] S23. Train the conditional denoising function by minimizing the loss function, and then... All observations are set as conditional observations. All missing values ​​are set as the generation target. The target is sampled to generate complete power distribution network measurement data;

[0029] The loss function is:

[0030] ;

[0031] The generated target is calculated using the following formula:

[0032] ;

[0033] In the above formula, For loss function, for According to distribution Take the expected value. For conditional denoising functions, For time step The corresponding noise scheduling parameters, To generate the target for interpolation.

[0034] In S3, the carbon flow of the distribution network includes the carbon potential of the distribution network nodes and the distribution of the branch power flow;

[0035] The carbon potential of the distribution network node is calculated using the following formula:

[0036] ;

[0037] ;

[0038] In the above formula, Let be the nodal carbon potential vector. For nodes Carbon emission intensity per unit of active power, For inflow node The set of branch paths, branch road active power, branch road The initial carbon potential, For the first generator injection node power, For the first The carbon emission intensity of a generator;

[0039] The branch power flow distribution matrix is ​​represented as follows: The elements are calculated using the following formula:

[0040] ;

[0041] In the above formula, For the node Flow to Node active power, For nodes voltage, For nodes voltage, , Branch roads The real and imaginary parts of admittance For nodes and nodes The voltage phase angle difference.

[0042] Secondly, this invention proposes a distribution network carbon flow fine quantification system based on dynamic topology adaptation, including a high-importance node screening module, a distribution network measurement data generation module, and a distribution network carbon flow determination module;

[0043] The high importance node screening module is used to evaluate the importance of each node in the distribution network based on the node importance assessment index, and to screen out the high importance nodes in the distribution network. The node importance assessment index considers dynamic topology correlation, load sensitivity, and carbon flow impact weight.

[0044] The distribution network measurement data generation module is used to collect measurement data of each high-importance node in real time, construct a conditional denoising diffusion implicit model based on the measurement data, and generate complete distribution network measurement data.

[0045] The distribution network carbon flow determination module is used to determine the distribution network carbon flow based on complete distribution network measurement data.

[0046] In the high-importance node screening module, the importance of each node in the distribution network is calculated using the following formula:

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] In the above formula, For nodes The importance of , , These are the weights for dynamic topological correlation, load sensitivity, and carbon flow influence, respectively. For nodes Dynamic topological correlation degree, For nodes The load sensitivity index For nodes The carbon flow influences the weight. For nodes after topological changes arrive Passing through the node The total number of shortest paths, Nodes after topological change arrive The total number of shortest paths, The topological change coefficient, , They are nodes The maximum and minimum active loads, For nodes The average active load, For nodes Load type weights, This represents the average change in carbon potential across all nodes in the network. This represents the total number of nodes in the distribution network. For nodes The change in active power.

[0052] The The following formula is used to calculate:

[0053] ;

[0054] ;

[0055] ;

[0056] In the above formula, For topological dynamic weights, The topological change coefficient, This is the matrix of topological changes. For the distribution network node-branch correlation matrix, This is the node-branch correlation matrix after topological changes.

[0057] The power distribution network measurement data generation module includes a sampling model construction unit, a conditional denoising function definition unit, and a training and sampling unit.

[0058] The sampling model construction unit is used to measure data. Based on the conditional information, a sampling model for the backdiffusion process is constructed:

[0059] ;

[0060] ;

[0061] ;

[0062] In the above formula, In order to be in and Under these conditions, time step The posterior probability of the corresponding data. For time step The corresponding noisy data, For time step The corresponding noisy data, Given the measurement data, For posterior probability The mean, For posterior probability variance For time step The corresponding noise scheduling parameters, For noise;

[0063] The conditional denoising function definition unit is used to define the conditional denoising function. And the mean of the sampling model for the reverse diffusion process. and variance They are respectively represented as and ,in, For the missing data to be generated, Given the known conditional observations, For time step Missing values ​​that need to be generated for The conditional observations known in the data and used to estimate the missing values;

[0064] The training and sampling unit is used to train the conditional denoising function by minimizing the loss function, and... All observations are set as conditional observations. All missing values ​​are set as the generation target. The target is sampled to generate complete power distribution network measurement data;

[0065] The loss function is:

[0066] ;

[0067] The generated target is calculated using the following formula:

[0068] ;

[0069] In the above formula, For loss function, for According to distribution Take the expected value. For conditional denoising functions, For time step The corresponding noise scheduling parameters, To generate the target for interpolation.

[0070] The distribution network carbon flow determination module includes a distribution network node carbon potential calculation unit and a branch power flow distribution calculation unit.

[0071] The carbon potential calculation unit for the distribution network node is used to calculate the carbon potential of the distribution network node using the following formula:

[0072] ;

[0073] ;

[0074] In the above formula, Let be the nodal carbon potential vector. For nodes Carbon emission intensity per unit of active power, For inflow node The set of branch paths, branch road active power, branch road The initial carbon potential, For the first generator injection node power, For the first The carbon emission intensity of a generator;

[0075] The branch power flow distribution calculation unit is used to represent the branch power flow distribution matrix as follows: The elements are calculated using the following formula:

[0076] ;

[0077] In the above formula, For the node Flow to Node active power, For nodes voltage, For nodes voltage, , Branch roads The real and imaginary parts of admittance For nodes and nodes The voltage phase angle difference.

[0078] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0079] 1. This invention proposes a method and system for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation. The method first assesses the importance of each node in the distribution network based on a node importance evaluation index, identifying high-importance nodes. This index considers dynamic topology correlation, load sensitivity, and carbon flow impact weight. Then, it collects measurement data from each high-importance node in real time, constructs a conditional denoising diffusion implicit model based on the measurement data, and generates complete distribution network measurement data. Finally, based on the complete distribution network measurement data, it determines the carbon flow in the distribution network. On the one hand, this method constructs a dynamic evaluation system for the importance of distribution network nodes, assesses the importance of each node in the distribution network, screens out high-importance nodes in the distribution network, and configures high-precision measurement devices only for nodes that truly affect carbon flow calculation, fundamentally resolving the cost redundancy caused by over-measurement and significantly reducing capital expenditure. On the other hand, this method uses the measurement data of high-importance nodes as conditional information, utilizes the non-Markovian characteristics of the conditional denoising diffusion implicit model to significantly improve the sampling speed and meet the needs of online calculation. At the same time, it combines real-time measurement data and dynamic adjustment of the generation model based on topology changes to achieve real-time updates of carbon flow distribution, generating complete distribution network measurement data that conforms to physical laws, effectively solving the problems of missing measurement data and noise interference, providing reliable data support for carbon flow calculation, improving the accuracy of carbon flow calculation, and supporting low-carbon scheduling and operation optimization of the distribution network.

[0080] 2. This invention proposes a method and system for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation. When a sudden change occurs in the distribution network topology, this method triggers a reassessment of node importance, re-selects highly important nodes, ensures that measurement data can cover the key areas after the topology change, supplements the core data required for carbon flow calculation, solves the information gap caused by the absence of important measurement objects in the distribution network, improves the accuracy of carbon emission distribution estimation, and maximizes the value of measurement data while keeping investment costs under control. Attached Figure Description

[0081] Figure 1 This is an overall flowchart of the method described in this invention.

[0082] Figure 2 This is a structural diagram of the system described in this invention. Detailed Implementation

[0083] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0084] This invention proposes a method and system for fine-grained quantification of carbon flow in distribution networks based on dynamic topology adaptation. It learns the importance of nodes in a topology-variable distribution network and applies data measurement equipment to high-importance nodes to obtain their measurement data. A denoised diffusion implicit model (C-DDIM) is constructed using measurement data from high-importance nodes as conditions. Self-supervised training based on conditional information improves the accuracy of model data generation and carbon flow prediction. The carbon flow of the distribution network is calculated based on the complete data generated by the model. Conditional generation addresses data loss and noise issues, achieving high-precision, real-time carbon emission tracking and providing core technical support for low-carbon optimization of distribution networks.

[0085] Example 1:

[0086] like Figure 1 As shown, the method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation proceeds in the following steps:

[0087] 1. Based on the node-branch correlation matrix of the distribution network, a topology change coefficient is introduced to characterize the topology change of the distribution network;

[0088] Initialize and construct the node-branch correlation matrix of the distribution network Considering the variable characteristics of the distribution network topology, including branch switching (disconnecting a branch during maintenance, or adding a new tie branch, or connecting a branch to the network), distributed generation (DG) access / de-access (DG access creates a new "node-DG branch," while DG de-access causes the branch to disappear), and load transfer (in extreme cases, the effective connection relationship between nodes may change (e.g., reconstructing the power supply path through switching operations), a topology change coefficient is introduced. It characterizes the changes in the distribution network. Among them, For the number of nodes, For a matrix, the number of branches is... If node It is a side road The "starting node" is... If node It is a side road The "terminal node" is... If node and branch roads If there is no correlation, then .

[0089] To quantify the "magnitude of topological change," the core is to compare the differences in the node-branch association matrix before and after the topological change, achieved through matrix norms, such as the F-norm, which is more sensitive to differences in elements. Specifically:

[0090] Based on the topology change operation, generate the changed distribution network node-branch association matrix. And calculate the difference in the matrix before and after the change. Then, using the commonly used F-norm (the square root of the sum of the squares of all elements of the matrix), the magnitude of change is calculated using the following formula:

[0091] ;

[0092] In the above formula, The topological change coefficient, This is the matrix of topological changes. For the distribution network node-branch correlation matrix, The node-branch incidence matrix after topological transformation, numerator The denominator measures the "size" of the changed portion. The "scale" of the original topology is measured and used for normalization, making... It becomes a relative change ratio.

[0093] Correlation Matrix It is a "mathematical representation of static topology". It is a "quantitative indicator of the magnitude of topology change". By comparing the differences in the correlation matrix before and after the change, it can quantitatively describe the "topology dynamics" and provide support for the "real-time and accuracy of carbon flow calculation in distribution networks".

[0094] 2. Based on topology changes, construct node importance assessment indicators to evaluate the importance of each node in the distribution network and screen out high-importance nodes in the distribution network;

[0095] To address the variable characteristics of distribution network topology (such as branch switching, distributed generation access / exit, load transfer, etc.), a node importance assessment index system for dynamic topology adaptation is constructed from three dimensions: dynamic topology correlation, load sensitivity, and carbon flow impact weight, to ensure that the index can respond to topology changes in real time.

[0096] Dynamic topological correlation is calculated using dynamic betweenness centrality. This index reflects the "bridging" role of nodes in power transmission after topological changes.

[0097] ;

[0098] In the above formula, For nodes Dynamic topological correlation degree, For nodes after topological changes arrive Passing through the node The total number of shortest paths, Nodes after topological change arrive The total number of shortest paths, The topological change coefficient;

[0099] By combining historical load data generated by the model with real-time load fluctuations, a load sensitivity index is defined, which reflects the degree of impact of nodal load fluctuations on carbon flow calculations.

[0100] ;

[0101] In the above formula, For nodes The load sensitivity index , They are nodes The maximum and minimum active loads, For nodes The average active load, For nodes The load type weights are set, with the weights for important nodes such as hospitals and industrial loads set to [value missing]. The weight of ordinary residential load is set as follows: ;

[0102] The impact of nodes on the overall carbon flow distribution is analyzed through offline simulation, and the carbon flow impact weight is calculated. This index quantifies the sensitivity of node power fluctuations to the accuracy of carbon flow calculation.

[0103] ;

[0104] In the above formula, For nodes The carbon flow influences the weight. This represents the average change in carbon potential across all nodes in the network. This represents the total number of nodes in the distribution network. For nodes The change in active power, obtained through measurement or simulation, is the direct factor driving the change in carbon flow.

[0105] Considering the real-time nature of topology changes, a dynamic weighting coefficient is used to fuse the above three indicators. The importance of each node in the distribution network is calculated using the following formula:

[0106] ;

[0107] In the above formula, For nodes The importance of , , The weights for dynamic topological correlation, load sensitivity, and carbon flow influence are respectively, satisfying the following conditions: ;

[0108] Among them, weight The topological dynamic weights are coefficients that vary with the topology. The adaptive adjustment is calculated using the following formula:

[0109] ;

[0110] In the above formula, For topological dynamic weights, That is, the topological change exceeds 20%.

[0111] Each node of the power distribution point is arranged according to Nodes are ranked from largest to smallest importance, and the top 20%-30% are selected as high-importance nodes based on both the deployment cost of measurement equipment and data accuracy requirements. For example, in a 100-node distribution network, 25 high-importance nodes are selected. The essence of selecting nodes based on both deployment cost and data accuracy requirements is that the accuracy baseline must not be breached, and cost optimization is carried out above this baseline: First, the core nodes that must be deployed are identified by breaking down accuracy requirements to ensure that carbon flow calculations do not exceed the accuracy threshold; then, through cost quantification and cost-effectiveness ranking, recommended deployment nodes are supplemented to the maximum extent possible within the budget to improve the precision of carbon flow calculations; ultimately, maximum accuracy is achieved within a limited cost, avoiding both excessive investment in accuracy and sacrificing core accuracy for cost, perfectly meeting the actual needs of distribution network projects.

[0112] If a sudden change occurs in the topology, such as the failure and removal of a main branch, it triggers a reassessment of node importance, re-selecting highly important nodes to ensure that measurement data can cover the critical areas after the topology change.

[0113] Existing methods for selecting measurement nodes in distribution networks are often based on experience or a single dimension (such as considering only load size or topology centrality), leading to two major problems: First, "over-measuring" results in wasted costs, such as deploying high-precision measurement equipment at ordinary residential nodes with minimal impact on carbon flow calculations, where the data value does not match the investment cost. Second, "measurement omissions" lead to missing key data, such as ignoring newly formed power transmission "bridge nodes" after topology changes or nodes where high-carbon emission units connect, causing significant deviations in subsequent carbon flow calculations due to insufficient core data, making it difficult to support low-carbon dispatch decisions. This invention achieves a breakthrough through a multi-dimensional dynamic node importance assessment system, forming a virtuous cycle of "precise selection of measurement nodes - maximization of data value - controllable costs."

[0114] 3. Collect measurement data of each high-importance node in real time, construct a conditional denoising diffusion implicit model based on the measurement data, and generate complete distribution network measurement data;

[0115] Equipment selection and deployment: Deploy intelligent measurement devices with topology awareness capabilities, such as synchronous phasor measurement units (PMUs), with a sampling rate ≥50Hz, at high-importance nodes to collect node voltage amplitude, phase angle, and active / reactive power data in real time;

[0116] Smart meters: support edge computing, can locally filter abnormal data (such as measurement values ​​exceeding 3 standard deviations) to reduce invalid data transmission; during deployment, priority should be given to utilizing existing infrastructure such as poles and distribution boxes to reduce installation costs, and the device communication interface must be compatible with the existing communication network of the power distribution network (such as fiber optic, LoRa).

[0117] Data Acquisition and Preprocessing: Measurement equipment transmits real-time data to the data center at a frequency of 1 minute per transmission, forming a measurement dataset for high-importance nodes. Furthermore, outliers were removed using the 3σ criterion, and missing values ​​were filled using interpolation of the mean of five adjacent time steps to ensure... The integrity and accuracy of the data provide reliable conditional information for constructing a Conditional Denoising Diffusion Implicit Model (C-DDIM).

[0118] A conditional denoising diffusion implicit model based on measurement data is constructed. As a non-Markov generative model, the conditional denoising diffusion implicit model has the advantages of high generation quality and high sampling efficiency. It can generate or correct missing data by combining the actual operating conditions of the distribution network (such as historical data and local measurement values), providing reliable input for carbon flow calculation, as detailed below:

[0119] Based on known measurement datasets of high-importance nodes in distribution networks Using conditional information, a sampling model for the backdiffusion process is constructed using Bayes' theorem. Gradually denoise and generate complete distribution network measurement data;

[0120] The forward diffusion process satisfies ,in, For time step The corresponding noise data, For time step The corresponding noise scheduling parameters are the previous ones. Step noise coefficient The cumulative product reflects the... arrive The "total intensity" of the added noise satisfies This is used to control the degree of noise addition. The larger the value, the smaller the noise impact, and the closer the data is to the condition information. Conversely, the greater the impact of noise, For noise;

[0121] Combining the normal distribution characteristics of the backward and forward diffusion processes, the posterior probability of the backward diffusion process satisfies:

[0122] ;

[0123] ;

[0124] ;

[0125] In the above formula, In order to be in and Under these conditions, time step The posterior probability of the corresponding data. For time step The corresponding noisy data, For posterior probability The mean, For posterior probability variance The noise coefficient is a parameter that controls the noise intensity at each step, and the time step... Corresponding to different This reflects the stages of noise addition. For trainable denoising functions, For time step The corresponding noise scheduling parameters;

[0126] The denoising diffusion implicit model improves speed by establishing a non-Markov chain and modifying the backdiffusion sampling method: assuming Satisfies a normal distribution ,in, , , These are parameters introduced by the model to break the Markov chain constraint and optimize the backsampling process. The specific values ​​will be determined based on the model derivation and optimization objectives, with the aim of significantly improving the sampling speed of back diffusion.

[0127] Combining forward diffusion iteration relationship The specific form of the normal distribution can be derived as follows:

[0128] ;

[0129] Since the reverse process does not require strict step-by-step calculations, let Alternative The generator can be obtained as follows:

[0130] ;

[0131] In the above formula, For time step The corresponding noise data, For time step The corresponding noise scheduling parameters, For time step noise, and Multiple iterations can be spaced out, through a fixed process (from...) arrive Samples are generated to form an implicit probability model.

[0132] By minimizing the loss function, self-supervised training based on conditional information is achieved, improving the efficiency and quality of model data generation.

[0133] Define conditional denoising function And the mean of the sampling model for the reverse diffusion process. and variance They are respectively represented as and :

[0134] ;

[0135] ;

[0136] In the above formula, For missing data to be generated (such as the power of unmeasured nodes). These are known observations (such as the power and voltage of already measured nodes). For time step Missing values ​​that need to be generated for The known conditional observations used to estimate missing values. The parameterized form of the mean. This is the parameterized form of variance;

[0137] By minimizing the loss function, the conditional denoising function is trained to learn how to recover true missing values ​​from noisy conditional data, and... All observations are set as conditional observations. All missing values ​​are set as the generation target. The target is sampled to generate complete power distribution network measurement data;

[0138] The loss function is:

[0139] ;

[0140] The generated target is calculated using the following formula:

[0141] ;

[0142] In the above formula, For loss function, for According to distribution Take the expected value. For conditional denoising functions, For time step The corresponding noise scheduling parameters, To generate the target for interpolation.

[0143] 4. Determine the carbon flow of the distribution network based on complete distribution network measurement data;

[0144] Carbon flow in distribution networks includes the carbon potential at distribution network nodes and the distribution of power flow in branches;

[0145] Define the carbon potential vector of the distribution network node The carbon emission intensity per unit of active power is calculated using the following formula:

[0146] ;

[0147] In the above formula, Let be the nodal carbon potential vector. For nodes Carbon emission intensity per unit of active power, For inflow node The set of branch paths, branch road active power, branch road The initial carbon potential, For the first generator injection node power, For the first The carbon emission intensity of a generator;

[0148] Based on the complete measurement data generated by the model, the branch power flow distribution matrix is ​​calculated using a DC power flow algorithm. The elements satisfy:

[0149] ;

[0150] In the above formula, For the node Flow to Node active power, For nodes voltage, For nodes voltage, , Branch roads The real and imaginary parts of admittance For nodes and nodes The voltage phase angle difference.

[0151] The above formula is transformed into a system of linear equations using matrix operations. ,in, Let be the diagonal matrix of active power flux at the nodes, where each element represents the sum of active power flowing into the nodes. Inject the matrix into the unit, Let be the carbon emission intensity vector of the generator. This is obtained by solving the system of linear equations. .

[0152] Example 2:

[0153] like Figure 2 As shown, the distribution network carbon flow fine quantification system based on dynamic topology adaptation includes a high-importance node screening module, a distribution network measurement data generation module, and a distribution network carbon flow determination module.

[0154] The high importance node screening module is used to evaluate the importance of each node in the distribution network based on the node importance assessment index, and to screen out the high importance nodes in the distribution network. The node importance assessment index considers dynamic topology correlation, load sensitivity, and carbon flow impact weight.

[0155] The distribution network measurement data generation module is used to collect measurement data of each high-importance node in real time, construct a conditional denoising diffusion implicit model based on the measurement data, and generate complete distribution network measurement data.

[0156] The distribution network carbon flow determination module is used to determine the distribution network carbon flow based on complete distribution network measurement data.

[0157] In the high-importance node screening module, the importance of each node in the distribution network is calculated using the following formula:

[0158] ;

[0159] ;

[0160] ;

[0161] ;

[0162] In the above formula, For nodes The importance of , , These are the weights for dynamic topological correlation, load sensitivity, and carbon flow influence, respectively. For nodes Dynamic topological correlation degree, For nodes The load sensitivity index For nodes The carbon flow influences the weight. For nodes after topological changes arrive Passing through the node The total number of shortest paths, Nodes after topological change arrive The total number of shortest paths, The topological change coefficient, , They are nodes The maximum and minimum active loads, For nodes The average active load, For nodes Load type weights, This represents the average change in carbon potential across all nodes in the network. This represents the total number of nodes in the distribution network. For nodes The change in active power.

[0163] The The following formula is used to calculate:

[0164] ;

[0165] ;

[0166] ;

[0167] In the above formula, For topological dynamic weights, The topological change coefficient, This is the matrix of topological changes. For the distribution network node-branch correlation matrix, This is the node-branch correlation matrix after topological changes.

[0168] The power distribution network measurement data generation module includes a sampling model construction unit, a conditional denoising function definition unit, and a training and sampling unit.

[0169] The sampling model construction unit is used to measure data. Based on the conditional information, a sampling model for the backdiffusion process is constructed:

[0170] ;

[0171] ;

[0172] ;

[0173] In the above formula, In order to be in and Under these conditions, time step The posterior probability of the corresponding data. For time step The corresponding noisy data, For time step The corresponding noisy data, Given the measurement data, For posterior probability The mean, For posterior probability variance For time step The corresponding noise scheduling parameters, For noise;

[0174] The conditional denoising function definition unit is used to define the conditional denoising function. And the mean of the sampling model for the reverse diffusion process. and variance They are respectively represented as and ,in, For the missing data to be generated, Given the known conditional observations, For time step Missing values ​​that need to be generated for The conditional observations known in the data and used to estimate the missing values;

[0175] The training and sampling unit is used to train the conditional denoising function by minimizing the loss function, and... All observations are set as conditional observations. All missing values ​​are set as the generation target. The target is sampled to generate complete power distribution network measurement data;

[0176] The loss function is:

[0177] ;

[0178] The generated target is calculated using the following formula:

[0179] ;

[0180] In the above formula, For loss function, for According to distribution Take the expected value. For conditional denoising functions, For time step The corresponding noise scheduling parameters, To generate the target for interpolation.

[0181] The distribution network carbon flow determination module includes a distribution network node carbon potential calculation unit and a branch power flow distribution calculation unit.

[0182] The carbon potential calculation unit for the distribution network node is used to calculate the carbon potential of the distribution network node using the following formula:

[0183] ;

[0184] ;

[0185] In the above formula, Let be the nodal carbon potential vector. For nodes Carbon emission intensity per unit of active power, For inflow node The set of branch paths, branch road active power, branch road The initial carbon potential, For the first generator injection node power, For the first The carbon emission intensity of a generator;

[0186] The branch power flow distribution calculation unit is used to represent the branch power flow distribution matrix as follows: The elements are calculated using the following formula:

[0187] ;

[0188] In the above formula, For the node Flow to Node active power, For nodes voltage, For nodes voltage, , Branch roads The real and imaginary parts of admittance For nodes and nodes The voltage phase angle difference.

Claims

1. A method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation, characterized in that, The method includes: S1. Based on the node importance assessment index, assess the importance of each node in the distribution network and screen out the high-importance nodes in the distribution network. The node importance assessment index considers dynamic topology correlation, load sensitivity, and carbon flow impact weight. S2. Real-time acquisition of measurement data from each high-importance node, construction of a conditional denoising diffusion implicit model based on the measurement data, and generation of complete distribution network measurement data; S3. Determine the carbon flow of the distribution network based on complete distribution network measurement data.

2. The method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation according to claim 1, characterized in that, In S1, the importance of each node in the distribution network is calculated using the following formula: ； ； ； ； In the above formula, For nodes The importance of , , These are the weights for dynamic topological correlation, load sensitivity, and carbon flow influence, respectively. For nodes Dynamic topological correlation degree, For nodes The load sensitivity index For nodes The carbon flow influences the weight. For nodes after topological changes arrive Passing through the node The total number of shortest paths, Nodes after topological change arrive The total number of shortest paths, The topological change coefficient, , They are nodes The maximum and minimum active loads, For nodes The average active load, For nodes Load type weights, This represents the average change in carbon potential across all nodes in the network. This represents the total number of nodes in the distribution network. For nodes The change in active power.

3. The method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation according to claim 2, characterized in that, The The following formula is used to calculate: ； ； ； In the above formula, For topological dynamic weights, The topological change coefficient, This is the matrix of topological changes. For the distribution network node-branch correlation matrix, This is the node-branch correlation matrix after topological changes.

4. The method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation according to claim 1, characterized in that, S2 includes: S21, based on measurement data Based on the conditional information, a sampling model for the backdiffusion process is constructed: ； ； ； In the above formula, In order to be in and Under these conditions, time step The posterior probability of the corresponding data. For time step The corresponding noisy data, For time step The corresponding noisy data, Given the measurement data, For posterior probability The mean, For posterior probability variance For time step The corresponding noise scheduling parameters, For noise; S22. Define the conditional denoising function And the mean of the sampling model for the reverse diffusion process. and variance They are respectively represented as and ,in, For the missing data to be generated, Given the known conditional observations, For time step Missing values ​​that need to be generated for The conditional observations known in the data and used to estimate the missing values; S23. Train the conditional denoising function by minimizing the loss function, and then... All observations are set as conditional observations. All missing values ​​are set as the generation target. The target is sampled to generate complete power distribution network measurement data; The loss function is: ； The generated target is calculated using the following formula: ； In the above formula, For loss function, for According to distribution Take the expected value. For conditional denoising functions, For time step The corresponding noise scheduling parameters, To generate the target for interpolation.

5. The method for fine quantification of carbon flow in distribution networks based on dynamic topology adaptation according to claim 1, characterized in that, In S3, the carbon flow of the distribution network includes the carbon potential of the distribution network nodes and the distribution of the branch power flow; The carbon potential of the distribution network node is calculated using the following formula: ； ； In the above formula, Let be the nodal carbon potential vector. For nodes Carbon emission intensity per unit of active power, For inflow node The set of branch paths, branch road active power, branch road The initial carbon potential, For the first generator injection node power, For the first The carbon emission intensity of a generator; The branch power flow distribution matrix is ​​represented as follows: The elements are calculated using the following formula: ； In the above formula, For the node Flow to Node active power, For nodes voltage, For nodes voltage, , Branch roads The real and imaginary parts of admittance For nodes and nodes The voltage phase angle difference.

6. A distribution network carbon flow fine quantification system based on dynamic topology adaptation, characterized in that, The system includes a high-importance node screening module, a distribution network measurement data generation module, and a distribution network carbon flow determination module. The high importance node screening module is used to evaluate the importance of each node in the distribution network based on the node importance assessment index, and to screen out the high importance nodes in the distribution network. The node importance assessment index considers dynamic topology correlation, load sensitivity, and carbon flow impact weight. The distribution network measurement data generation module is used to collect measurement data of each high-importance node in real time, construct a conditional denoising diffusion implicit model based on the measurement data, and generate complete distribution network measurement data. The distribution network carbon flow determination module is used to determine the distribution network carbon flow based on complete distribution network measurement data.

7. The distribution network carbon flow fine quantification system based on dynamic topology adaptation according to claim 6, characterized in that, In the high-importance node screening module, the importance of each node in the distribution network is calculated using the following formula: ； ； ； ； In the above formula, For nodes The importance of , , These are the weights for dynamic topological correlation, load sensitivity, and carbon flow influence, respectively. For nodes Dynamic topological correlation degree, For nodes The load sensitivity index For nodes The carbon flow influences the weight. For nodes after topological changes arrive Passing through the node The total number of shortest paths, Nodes after topological change arrive The total number of shortest paths, The topological change coefficient, , They are nodes The maximum and minimum active loads, For nodes The average active load, For nodes Load type weights, This represents the average change in carbon potential across all nodes in the network. This represents the total number of nodes in the distribution network. For nodes The change in active power.

8. The distribution network carbon flow fine quantification system based on dynamic topology adaptation according to claim 7, characterized in that, The The following formula is used to calculate: ； ； ； In the above formula, For topological dynamic weights, The topological change coefficient, This is the matrix of topological changes. For the distribution network node-branch correlation matrix, This is the node-branch correlation matrix after topological changes.

9. The distribution network carbon flow fine quantification system based on dynamic topology adaptation according to claim 6, characterized in that, The power distribution network measurement data generation module includes a sampling model construction unit, a conditional denoising function definition unit, and a training and sampling unit. The sampling model construction unit is used to measure data. Based on the conditional information, a sampling model for the backdiffusion process is constructed: ； ； ； In the above formula, In order to be in and Under these conditions, time step The posterior probability of the corresponding data. For time step The corresponding noisy data, For time step The corresponding noisy data, Given the measurement data, For posterior probability The mean, For posterior probability variance For time step The corresponding noise scheduling parameters, For noise; The conditional denoising function definition unit is used to define the conditional denoising function. And the mean of the sampling model for the reverse diffusion process. and variance They are respectively represented as and ,in, For the missing data to be generated, Given the known conditional observations, For time step Missing values ​​that need to be generated for The conditional observations known in the data and used to estimate the missing values; The training and sampling unit is used to train the conditional denoising function by minimizing the loss function, and... All observations are set as conditional observations. All missing values ​​are set as the generation target. The target is sampled to generate complete power distribution network measurement data; The loss function is: ； The generated target is calculated using the following formula: ； In the above formula, For loss function, for According to distribution Take the expected value. For conditional denoising functions, For time step The corresponding noise scheduling parameters, To generate the target for interpolation.

10. The distribution network carbon flow fine quantification system based on dynamic topology adaptation according to claim 6, characterized in that, The distribution network carbon flow determination module includes a distribution network node carbon potential calculation unit and a branch power flow distribution calculation unit. The carbon potential calculation unit for the distribution network node is used to calculate the carbon potential of the distribution network node using the following formula: ； ； In the above formula, Let be the nodal carbon potential vector. For nodes Carbon emission intensity per unit of active power, For inflow node The set of branch paths, branch road active power, branch road The initial carbon potential, For the first generator injection node power, For the first The carbon emission intensity of a generator; The branch power flow distribution calculation unit is used to represent the branch power flow distribution matrix as follows: The elements are calculated using the following formula: ； In the above formula, For the node Flow to Node active power, For nodes voltage, For nodes voltage, , Branch roads The real and imaginary parts of admittance For nodes and nodes The voltage phase angle difference.

Images