A method and system for dynamic synergistic analysis of electrocarbon

Abstract

This invention relates to the field of power-carbon synergy technology, and particularly to a method and system for dynamic power-carbon synergy analysis. The method includes: extracting features from real-time operational data of the power system's generation, transmission, distribution, and consumption links to obtain a full-chain correlation quantification representation; generating a full-chain spatiotemporal correlation matrix based on the full-chain correlation quantification representation; using a multi-branch convolutional neural network to obtain a spatiotemporal mapping prediction result of grid carbon emissions based on the full-chain spatiotemporal correlation matrix and carbon emission parameters of the power system's generation, transmission, distribution, and consumption links, thereby acquiring correlation data for the target region; and introducing an attention mechanism to dynamically optimize the full-chain spatiotemporal correlation matrix to obtain a power-carbon synergy coupling matrix. Based on the power-carbon synergy coupling matrix, a generative adversarial network is used to quantify the spatial distribution of low-carbon effects in power grid planning, obtaining the dynamic power-carbon synergy analysis result. This invention, through spatiotemporal correlation modeling and dynamic optimization mechanisms, can comprehensively and accurately analyze the dynamic power-carbon synergy relationships in the power system's generation, transmission, distribution, and consumption links.

Description

Technical Field

[0001] This invention relates to the field of electro-carbon synergy technology, and in particular to a method and system for dynamic electro-carbon synergy analysis. Background Technology

[0002] With the transformation of the global energy structure, the synergistic analysis of power systems and carbon emissions has become a key research direction in the energy field. As renewable energy is integrated into the grid on a large scale, grid planning is no longer limited to meeting the balance between power supply and demand, but needs to promote the realization of carbon emission reduction while ensuring the safe and reliable operation of the power system.

[0003] However, existing technologies have significant shortcomings in handling the synergy between power grid planning and carbon emissions, making it difficult to meet practical application needs. On the one hand, traditional methods mostly rely on static statistics or simple linear models, making it difficult to capture the dynamic characteristics and nonlinear relationships in power system operation. On the other hand, most existing research focuses only on the local optimization of a single link on the power supply or consumption side, lacking a systematic consideration of the synergistic effects of the entire power chain. This local optimization approach often leads to insufficient synergy or even contradictions in planning results when facing actual operating scenarios, making it difficult to comprehensively quantify the long-term impact of planning schemes and hindering the deep integration of power systems with low-carbon development. In summary, existing technologies have many shortcomings in the synergistic analysis of power grid planning and carbon emissions. Therefore, researching an analytical method that can dynamically quantify the synergistic relationship between power grid planning and carbon emissions has become a key issue that urgently needs to be addressed in this field. Summary of the Invention

[0004] To address the above technical problems, this invention provides a method and system for dynamic synergistic analysis of electrocarbon energy.

[0005] In a first aspect, the present invention provides a method for dynamic synergistic analysis of electrocarbon, the method comprising the following steps:

[0006] Feature extraction is performed on real-time operational data from the power system's generation, transmission, distribution, and consumption links to obtain a quantitative representation of the entire chain of correlations;

[0007] Based on the full-chain correlation quantization representation, a multi-dimensional coupling relationship model is performed on the nonlinear coupling relationship between the power system generation, transmission, distribution and consumption links to generate a full-chain spatiotemporal correlation matrix;

[0008] Based on the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system's generation, transmission, distribution and consumption links, a multi-branch convolutional neural network is used to learn the spatiotemporal mapping relationship between the power grid's operating status and carbon emissions, and the spatiotemporal mapping prediction result of the power grid's carbon emissions is obtained.

[0009] Based on the target area association data where the carbon emission deviation exceeds a preset deviation threshold in the spatiotemporal mapping prediction results of the power grid carbon emissions, an attention mechanism is introduced to dynamically optimize the full-chain spatiotemporal association matrix to obtain the power-carbon synergistic coupling matrix.

[0010] Based on the aforementioned electric-carbon synergistic coupling matrix, a generative adversarial network is used to quantify the spatial distribution of the low-carbon effects of power grid planning, thereby obtaining the results of dynamic synergistic analysis of electric-carbon.

[0011] In a further implementation scheme, the step of extracting features from the real-time operational data of the power system's generation, transmission, distribution, and consumption links to obtain a full-chain correlation quantification representation includes:

[0012] Collect real-time and historical operating data from the power system's generation, transmission, distribution, and consumption stages to form multi-source operating data encompassing these stages.

[0013] The multi-source operational data is standardized to obtain standardized multi-source operational data, and the standardized multi-source operational data is segmented using a sliding window to construct standardized time-series data slices.

[0014] Spatiotemporal feature extraction is performed on the standardized time-series data slices to obtain a hybrid feature vector;

[0015] Based on the hybrid feature vector, feature compression and association encoding are performed using a fully connected layer to obtain a full-chain association quantization representation.

[0016] In a further implementation, the step of performing multi-dimensional coupling relationship modeling on the nonlinear coupling relationship between the power system's generation, transmission, distribution, and consumption links based on the full-chain correlation quantization representation, and generating a full-chain spatiotemporal correlation matrix, includes:

[0017] A dynamic topology diagram is constructed based on the topological connections of the power system's generation, transmission, distribution, and consumption links;

[0018] The dynamic topology graph is input into a multi-layer graph convolutional network to extract cross-link coupling features layer by layer, generating a global-local interaction correlation matrix.

[0019] The global-local interaction correlation matrix is ​​spatiotemporally aligned and fused to generate an initial spatiotemporally aligned correlation tensor;

[0020] Principal component analysis is performed on the initial correlation tensor to extract key coupling modes, and the key coupling modes are orthogonalized by singular value decomposition to obtain the full-chain spatiotemporal correlation matrix.

[0021] In a further implementation, the step of constructing a dynamic topology map based on the topological connections of the power system's generation, transmission, distribution, and consumption links includes:

[0022] Based on the topological connections of the power system's generation, transmission, distribution, and consumption links, a dynamic topology graph is constructed with generator sets, transmission nodes, distribution buses, and user loads as vertices and power transmission paths as edges.

[0023] Based on the full-chain association quantization representation, dynamic weight features are assigned to each vertex and edge in the dynamic topology graph.

[0024] In a further embodiment, the multi-layer graph convolutional network includes at least three graph convolutional networks and an output layer, wherein the three-layer graph convolutional network includes a first-layer graph convolutional network, a second-layer graph convolutional network, and a third-layer graph convolutional network;

[0025] The first layer of graph convolutional network is used to calculate the instantaneous power transfer association strength between vertices based on the features of adjacent vertices in the dynamic topology graph and the dynamic weight features of the edges, so as to obtain the vertex-level association strength matrix.

[0026] The second layer graph convolutional network is used to perform regional spatial clustering of the vertices of the power system generation and transmission links based on the vertex-level correlation strength matrix, so as to obtain the local correlation matrix of energy interaction in the power system generation and transmission links;

[0027] The third-layer graph convolutional network is used to perform power supply path traversal on the vertices of the power system distribution link based on the vertex-level correlation strength matrix, so as to obtain the global correlation matrix of the load response of the power system distribution link.

[0028] The output layer is used to expand and concatenate the local correlation matrix and the global correlation matrix to obtain a global-local interactive correlation matrix.

[0029] In a further implementation, the step of learning the spatiotemporal mapping relationship between the power grid operating state and carbon emissions using a multi-branch convolutional neural network based on the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system's generation, transmission, distribution, and consumption links, to obtain the spatiotemporal mapping prediction result of power grid carbon emissions includes:

[0030] The coupling strength distribution of the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system generation, transmission, distribution and consumption links are mapped in three dimensions to construct an electric carbon spatiotemporal feature cube;

[0031] The electric carbon spatiotemporal feature cube is input into a multi-branch convolutional neural network for feature learning to obtain an electric carbon spatiotemporal fusion feature tensor, and the power grid is divided into zones according to the generation, transmission, distribution and consumption links of the power system.

[0032] The spatial resolution of the electrocarbon spatiotemporal fusion feature tensor is restored to the power grid partition scale by deconvolution operation, resulting in the electrocarbon spatiotemporal feature restoration tensor.

[0033] Based on the reconstructed tensor of the electrical carbon spatiotemporal characteristics, carbon emission intensity regression calculation is performed on each power grid partition to obtain the spatiotemporal mapping prediction result of power grid carbon emissions for the target period.

[0034] In a further embodiment, the multi-branch convolutional neural network includes a spatial feature extraction branch network, a temporal feature extraction branch network, and a cross-scale feature fusion branch network.

[0035] In a further implementation, the step of dynamically optimizing the entire-chain spatiotemporal correlation matrix by introducing an attention mechanism based on the target region association data where the carbon emission deviation exceeds a preset deviation threshold in the spatiotemporal mapping prediction results of the power grid carbon emissions, to obtain the power-carbon synergistic coupling matrix, includes:

[0036] Based on the spatiotemporal mapping prediction results of power grid carbon emissions and the benchmark carbon emission intensity, the target area associated data whose carbon emission deviation exceeds a preset deviation threshold is identified.

[0037] Dynamic attention weights are calculated on the associated data of the target region to obtain attention weight coefficients, and the attention weight coefficients are multiplied element-wise with the spatiotemporal association matrix of the whole chain to obtain the initial electrocarbon association matrix;

[0038] By introducing an electrocarbon coupling factor to weight and fuse the initial electrocarbon correlation matrix, an electrocarbon cooperative coupling matrix is ​​generated.

[0039] In a further implementation, the step of using a generative adversarial network to quantify the spatial distribution of the low-carbon effect of power grid planning based on the electric-carbon synergistic coupling matrix to obtain the dynamic synergistic analysis results of electric-carbon includes:

[0040] Based on the aforementioned electric-carbon synergistic coupling matrix and the pre-acquired variables to be planned, a low-carbon planning feature vector is constructed.

[0041] Using the low-carbon planning feature vector as input, a spatial distribution map of low-carbon effects is generated through a generative adversarial network;

[0042] An anisotropic diffusion algorithm is used to smooth the spatial distribution map of the low-carbon effect to obtain a smoothed spatial distribution of the low-carbon effect.

[0043] The element-wise multiplication of the smooth low-carbon effect spatial distribution with the electric carbon synergistic coupling matrix yields the power grid zone electric carbon synergistic efficiency index, which is then used as the result of the electric carbon dynamic synergistic analysis.

[0044] Secondly, the present invention provides an electrocarbon dynamic synergistic analysis system, the system comprising:

[0045] The feature extraction module is used to extract features from real-time operational data of the power system's generation, transmission, distribution and consumption links to obtain a quantitative representation of the entire chain of correlations;

[0046] The data analysis module is used to perform multi-dimensional coupling relationship modeling on the nonlinear coupling relationship between the generation, transmission, distribution and consumption links of the power system based on the full-chain correlation quantification representation, and generate a full-chain spatiotemporal correlation matrix;

[0047] The model prediction module is used to learn the spatiotemporal mapping relationship between the power grid operating status and carbon emissions based on the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system's generation, transmission, distribution and consumption links, and to obtain the spatiotemporal mapping prediction result of power grid carbon emissions.

[0048] The collaborative coupling module is used to dynamically optimize the whole-chain spatiotemporal correlation matrix based on the target area association data where the carbon emission deviation exceeds a preset deviation threshold in the spatiotemporal mapping prediction results of the power grid carbon emissions, by introducing an attention mechanism, so as to obtain the power-carbon collaborative coupling matrix.

[0049] The collaborative analysis module is used to quantify the spatial distribution of the low-carbon effect of power grid planning based on the aforementioned electric carbon collaborative coupling matrix, and obtain the dynamic collaborative analysis results of electric carbon.

[0050] This invention provides a method and system for dynamic collaborative analysis of power generation and carbon emissions. The method extracts features from real-time operational data of the power system's generation, transmission, distribution, and consumption links to obtain a full-chain correlation quantification representation. Based on this full-chain correlation quantification representation, it models the nonlinear coupling relationships between the power system's generation, transmission, distribution, and consumption links in a multi-dimensional manner, generating a full-chain spatiotemporal correlation matrix. Based on the full-chain spatiotemporal correlation matrix and carbon emission parameters of the power system's generation, transmission, distribution, and consumption links, it uses a multi-branch convolutional neural network to learn the spatiotemporal mapping relationship between the power grid's operating state and carbon emissions, obtaining a spatiotemporal mapping prediction result for power grid carbon emissions. Based on the target area correlation data in the power grid carbon emission spatiotemporal mapping prediction result where the carbon emission deviation exceeds a preset deviation threshold, it introduces an attention mechanism to dynamically optimize the full-chain spatiotemporal correlation matrix, obtaining a power generation and carbon emissions collaborative coupling matrix. Based on this power generation and carbon emissions collaborative coupling matrix, it employs a generative adversarial network to quantify the spatial distribution of the low-carbon effects of power grid planning, obtaining the dynamic collaborative analysis result of power generation and carbon emissions. Compared with existing technologies, this method uses spatiotemporal correlation modeling and dynamic optimization mechanisms to accurately map the power grid operation status and carbon emissions, realizing dynamic collaborative analysis of the power grid system and carbon emissions, and providing a scientific basis for the optimization of power grid planning. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the electrocarbon dynamic synergistic analysis method provided in the embodiments of the present invention;

[0052] Figure 2 This is a block diagram of the electrocarbon dynamic collaborative analysis system provided in an embodiment of the present invention.

[0053] Figure labeling: 101, Feature extraction module; 102, Data analysis module; 103, Model prediction module; 104, Collaborative coupling module; 105, Collaborative analysis module. Detailed Implementation

[0054] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are given for illustrative purposes only and should not be construed as limiting the present invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0055] refer to Figure 1 This invention provides a method for dynamic synergistic analysis of electrocarbon, such as... Figure 1 As shown, the method includes the following steps:

[0056] S1. Extract features from real-time operational data of the power system's generation, transmission, distribution, and consumption links to obtain a quantitative representation of the entire chain of connections.

[0057] In some implementations, the step of extracting features from real-time operational data of the power system's generation, transmission, distribution, and consumption links to obtain a full-chain correlation quantification representation includes:

[0058] Collect real-time and historical operating data from the power system's generation, transmission, distribution, and consumption stages to form multi-source operating data encompassing these stages.

[0059] The multi-source operational data is standardized to obtain standardized multi-source operational data, and the standardized multi-source operational data is segmented using a sliding window to construct standardized time-series data slices.

[0060] Spatiotemporal feature extraction is performed on the standardized time-series data slices to obtain a hybrid feature vector;

[0061] Based on the hybrid feature vector, feature compression and association encoding are performed using a fully connected layer to obtain a full-chain association quantization representation.

[0062] Specifically, this embodiment utilizes intelligent sensing devices deployed in the power system's generation, transmission, distribution, and consumption links to collect real-time operational data. This real-time power grid operational data includes multi-source heterogeneous data from generation, transmission, distribution, and consumption links. For example, real-time operational data from these links may include multi-dimensional operational parameters such as generator output, transmission line load rate, distribution node voltage, and user power consumption. Simultaneously, this embodiment retrieves historical operational data from a historical operational database, forming multi-source operational data encompassing all aspects of the power system's generation, transmission, distribution, and consumption. This multi-source operational data is then standardized, specifically using a mean-variance normalization method to eliminate dimensional differences. For example, voltage parameters, after standardization, conform to a mean of zero and a standard deviation of zero. The data follows a normal distribution with a difference of 1. In this embodiment, the standardized multi-source operational data is segmented using a sliding window to construct standardized time-series data slices. For example, this embodiment uses a 15-minute time window and a 5-minute sliding step to process the standardized multi-source operational data into equal-length slices, generating standardized time-series data slices. These standardized time-series data slices are then input into a pre-constructed spatiotemporal feature encoder. The spatiotemporal feature encoder includes a spatial feature extraction module and a temporal feature extraction module. The spatial feature extraction module uses a three-layer convolutional network to extract the spatial correlation features between adjacent nodes in a hierarchical manner, extracting local spatial distribution patterns. For example, the three-layer convolutional network uses a 5×5 convolutional kernel group to extract the spatial distribution features of generator sets and transmission lines in a hierarchical manner, outputting a 128-dimensional spatial feature vector.

[0063] Meanwhile, in this embodiment, the standardized time-series data slices are input into the time-series feature extraction module. The time-series feature extraction module uses a dual-channel long short-term memory network to model the dynamic change law of power parameters. In this embodiment, 128 memory units can be set to capture minute-level time-series dependencies. The time-series evolution law of distribution node voltage and user load is modeled through 128 memory units, and a 128-dimensional time-series feature vector is output. Then, the spatial feature vector and the time-series feature vector are concatenated in dimensions to form a 256-dimensional hybrid feature vector. Feature compression and correlation encoding are performed through a fully connected layer. In this embodiment, two fully connected layers can be set. The first fully connected layer uses a linear rectification activation function to realize the nonlinear transformation of features. The second fully connected layer compresses the dimension to 256 dimensions through linear projection. Finally, a full-chain correlation quantization representation of the dynamic coupling relationship of the entire power system generation, transmission, distribution and consumption is output.

[0064] S2. Based on the full-chain correlation quantization representation, perform multi-dimensional coupling relationship modeling on the nonlinear coupling relationship between the power system generation, transmission, distribution and consumption links, and generate a full-chain spatiotemporal correlation matrix.

[0065] In some implementations, the step of performing multi-dimensional coupling relationship modeling on the nonlinear coupling relationship between the power system's generation, transmission, distribution, and consumption links based on the full-chain correlation quantization representation, and generating a full-chain spatiotemporal correlation matrix, includes:

[0066] A dynamic topology diagram is constructed based on the topological connections of the power system's generation, transmission, distribution, and consumption links;

[0067] The dynamic topology graph is input into a multi-layer graph convolutional network to extract cross-link coupling features layer by layer, generating a global-local interaction correlation matrix.

[0068] The global-local interaction correlation matrix is ​​spatiotemporally aligned and fused to generate an initial spatiotemporally aligned correlation tensor;

[0069] Principal component analysis is performed on the initial correlation tensor to extract key coupling modes, and the key coupling modes are orthogonalized by singular value decomposition to obtain the full-chain spatiotemporal correlation matrix.

[0070] Specifically, this embodiment constructs a dynamic topology graph based on the topological connections of the power system's generation, transmission, distribution, and consumption links. This graph uses generator sets, transmission nodes, distribution buses, and user loads as vertices and power transmission paths as edges. Specifically, the dynamic topology graph uses generator sets as source points, transmission nodes as relay points, distribution buses as distribution points, and user loads as endpoints. This embodiment assigns dynamic weight features to each vertex and edge in the dynamic topology graph based on the full-chain correlation quantization representation. For example, based on the power, voltage, and load rate parameters in the full-chain correlation quantization representation, dynamic weight features are assigned to each vertex and edge in the topology graph. The vertex weights are normalized values ​​of real-time operating parameters, and the edge weights are the line transmission capacity and load rate. The ratio is then used to input the dynamic topology graph into a multi-layer graph convolutional network. Cross-link coupling features are extracted layer by layer through a message passing mechanism. In this embodiment, the multi-layer graph convolutional network includes at least three graph convolutional networks and an output layer. The three-layer graph convolutional network includes a first-layer graph convolutional network, a second-layer graph convolutional network, and a third-layer graph convolutional network. Specifically, the first-layer graph convolutional network aggregates features from adjacent vertices and calculates the power transfer correlation strength between vertices; the second-layer graph convolutional network fuses regional topology features to generate a local correlation matrix representing the energy interaction between the transmission and distribution links; and the third-layer graph convolutional network integrates the topology features of the entire network to generate a global correlation matrix representing the load response of the distribution and application links.

[0071] The first layer of graph convolutional network is used to calculate the instantaneous power transfer association strength between vertices based on the features of adjacent vertices in the dynamic topology graph and the dynamic weight features of the edges, so as to obtain the vertex-level association strength matrix.

[0072] The second layer graph convolutional network is used to perform regional spatial clustering of the vertices of the power system's generation and transmission links based on the vertex-level correlation strength matrix, to obtain the local correlation matrix of energy interaction in the power system's generation and transmission links;

[0073] The third-layer graph convolutional network is used to perform power supply path traversal on the vertices of the power system distribution link based on the vertex-level correlation strength matrix, so as to obtain the global correlation matrix of the load response of the power system distribution link.

[0074] The output layer is used to expand and concatenate the local correlation matrix and the global correlation matrix to obtain a global-local interactive correlation matrix.

[0075] In a specific embodiment, in the first-layer graph convolutional network, this embodiment can perform weighted aggregation of neighborhood features based on the power and voltage characteristics of adjacent vertices in the dynamic topology graph, using the dynamic weight features of edges. Specifically, this can be achieved by multiplying the features of adjacent vertices by the dynamic weight features of edges, summing the results, and then dividing by the number of adjacent vertices to generate a vertex-level association strength matrix representing the instantaneous power transmission association strength between vertices. The element values ​​of the vertex-level association strength matrix are the standardized instantaneous power transmission association strength between vertices. In the second-layer graph convolutional network, this embodiment performs electrical distance clustering on the vertex group of the generation-transmission link based on the vertex-level association matrix. For example, using the transmission node as the cluster center, the mean association strength between the generator set and the transmission node is calculated to generate a local association matrix representing the energy interaction of the generation-transmission link. The dimensions of the local association matrix include the number of generator sets and the number of transmission nodes. In the third-layer graph convolutional network... This embodiment performs topological path traversal on the vertex group of the distribution-consumption link based on the vertex-level correlation matrix. For example, this embodiment accumulates the correlation strength values ​​of each vertex on the power supply path from the distribution bus to the user load to generate a global correlation matrix characterizing the load response of the distribution-consumption link. The dimensions of the global correlation matrix may include the number of distribution buses and the number of user load groups. In the output layer, this embodiment expands and concatenates the dimensions of the local correlation matrix and the global correlation matrix. The local correlation matrix is ​​expanded into a four-dimensional tensor of generation, transmission, distribution and user by adding a distribution-consumption link dimension and filling it with zero values. At the same time, the global correlation matrix is ​​expanded into a four-dimensional tensor of the same dimension by adding a generation-transmission link dimension and filling it with zero values. The two four-dimensional tensors are weighted and superimposed along the time dimension. The weight value can be set as the harmonic mean of the real-time power ratio of the generation-transmission link and the load ratio of the distribution-consumption link to generate a spatiotemporally fused global-local interactive correlation matrix.

[0076] Next, this embodiment aligns the local and global correlation matrices in terms of spatiotemporal dimensions. In the time dimension, a 10-minute time window is used, and the matrix elements are smoothed using a moving average method. In the spatial dimension, based on the power grid zoning structure, a weighted fusion operation is performed on the local and global correlation matrices. The weight values ​​can be the load proportion of each power grid zoning area. After fusion, an initial correlation tensor with spatiotemporal continuity is generated. This embodiment uses principal component analysis to analyze the initial correlation tensor, retaining principal components with a cumulative variance contribution rate exceeding 85%, extracting key coupling patterns that characterize the essential features of cross-link coupling, and orthogonalizing these key coupling patterns through singular value decomposition to eliminate idiosyncratic features. The singular value decomposition (SVD) results include a left singular vector, a singular value matrix, and a right singular vector. The left singular vector represents the spatial distribution pattern, i.e., the distribution characteristics at different spatial locations. The singular value matrix represents the importance of the pattern, and the right singular vector represents the temporal evolution pattern, i.e., the change characteristics at different times. In this embodiment, the number of principal components to be retained is determined based on the magnitude of the singular values. The number of principal components is determined by the singular value decay threshold. After retaining the first few principal components based on the determined number of principal components, the orthogonalized key coupling pattern matrix is ​​reconstructed. Specifically, matrix multiplication is performed on the retained left singular vector, singular value matrix, and right singular vector to obtain the orthogonalized coupling pattern matrix.

[0077] Finally, in this embodiment, the orthogonalized coupling mode matrix is ​​reorganized into a four-dimensional tensor according to the generation, transmission, distribution, and consumption links. The dimensions of the four-dimensional tensor are the number of generator sets, the number of transmission nodes, the number of distribution buses, and the number of user loads. For each quadruple (i.e., a combination of one generator set, one transmission node, one distribution bus, and one user load) in the generation, transmission, distribution, and consumption links, its coupling strength value is calculated as follows: the element values ​​at the corresponding positions in the orthogonalized coupling mode matrix are weighted and summed according to the variance contribution rate weight of each principal component, where the variance contribution rate is... The rate weight represents the contribution of each principal component to the overall coupling mode. The resulting four-dimensional tensor is expanded into a two-dimensional matrix according to the power grid partitions of generation, transmission, distribution and consumption links. The rows and columns of the two-dimensional matrix correspond to the power grid partitions of generation, transmission, distribution and consumption links, respectively. The element values ​​represent the coupling strength between different power grid partitions, such as the coupling strength between the generation partition and the consumption partition. Finally, the full-chain spatiotemporal correlation matrix is ​​output. The element values ​​of the full-chain spatiotemporal correlation matrix represent the quantified value of the coupling strength between each link. The rows and columns correspond to the topological partitions of generation, transmission, distribution and consumption links, respectively.

[0078] S3. Based on the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system's generation, transmission, distribution and consumption links, a multi-branch convolutional neural network is used to learn the spatiotemporal mapping relationship between the power grid's operating status and carbon emissions, and the spatiotemporal mapping prediction result of the power grid's carbon emissions is obtained.

[0079] In some implementations, the step of learning the spatiotemporal mapping relationship between the power grid operating state and carbon emissions using a multi-branch convolutional neural network based on the full-chain spatiotemporal correlation matrix and carbon emission parameters of the power system's generation, transmission, distribution, and consumption links, to obtain the spatiotemporal mapping prediction result of power grid carbon emissions includes:

[0080] The coupling strength distribution of the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system generation, transmission, distribution and consumption links are mapped in three dimensions to construct an electric carbon spatiotemporal feature cube;

[0081] The electric carbon spatiotemporal feature cube is input into a multi-branch convolutional neural network for feature learning to obtain an electric carbon spatiotemporal fusion feature tensor, and the power grid is divided into zones according to the generation, transmission, distribution and consumption links of the power system.

[0082] The spatial resolution of the electrocarbon spatiotemporal fusion feature tensor is restored to the power grid partition scale by deconvolution operation, resulting in the electrocarbon spatiotemporal feature restoration tensor.

[0083] Based on the reconstructed tensor of the electrical carbon spatiotemporal characteristics, carbon emission intensity regression calculation is performed on each power grid partition to obtain the spatiotemporal mapping prediction result of power grid carbon emissions for the target period.

[0084] Specifically, this embodiment constructs an electric carbon spatiotemporal feature cube by performing three-dimensional spatial mapping based on the coupling strength distribution of the full-chain spatiotemporal correlation matrix and carbon emission parameters such as the carbon emission intensity per unit power generation of generator sets, the carbon emission coefficient of transmission line losses, the carbon flow density of distribution areas, and the carbon footprint factor on the user side. The electric carbon spatiotemporal feature cube includes a spatial dimension, a temporal dimension, and feature channels. The spatial dimension corresponds to the power grid partition, which can be a physical region obtained by dividing the power grid topology and dispatch management boundaries. For example, the power grid partition can include a power generation cluster area, a transmission grid, a distribution grid area, and a load aggregation area. The temporal dimension is a continuous time series with a resolution of 1 hour. The feature channels are fused feature vectors containing coupling strength values, carbon emission parameters, and operating status parameters.

[0085] In this embodiment, the spatiotemporal feature cube of electricity carbon is input into a multi-branch convolutional neural network for feature learning. The multi-branch convolutional neural network includes a spatial feature extraction branch network, a temporal feature extraction branch network, and a cross-scale feature fusion branch network. The spatial feature extraction branch network uses a 3×3 convolutional kernel group to extract the spatial correlation features of the power generation-transmission link along the power grid partition dimension through three layers of convolutional operations, and outputs a spatial feature map. Each pixel of the spatial feature map represents the spatial coordination pattern of the coupling strength of the power generation-transmission link and carbon emissions within a specific partition. The temporal feature extraction branch network uses an dilated convolutional structure to capture the temporal evolution law of carbon emissions in the distribution-consumption link along the time slice dimension with an exponentially increasing expansion coefficient, and outputs a temporal feature map. Each pixel of the temporal feature map represents the dynamic correlation law between the feedback strength of the distribution-consumption link and the carbon emission factor within a specific time period. The cross-scale feature fusion branch network concatenates the spatial feature map and the temporal feature map in the channel dimension, and performs feature compression through a 1×1 convolutional kernel to generate an electricity carbon spatiotemporal fusion feature tensor with half the number of channels of the original features.

[0086] Next, this embodiment restores the spatial resolution of the spatiotemporal fusion feature tensor of electric carbon to the original power grid partition scale through deconvolution operation. The deconvolution stride can be set as the reciprocal of the physical diameter of the power grid partition to ensure that the output tensor corresponds to the power grid topology. Based on the spatiotemporal features of electric carbon, the tensor is restored, and a pixel-by-pixel fully connected layer is used to perform carbon emission intensity regression calculation for each power grid partition. Specifically, according to the coordinate position of the partition in the power grid topology, the feature vector corresponding to the position is extracted from the feature tensor. The feature vector is input into the fully connected layer, and the dot product of the feature vector and the weight matrix is ​​added with a bias term to output the predicted value of carbon emission intensity of the power grid partition in the target time period. All the predicted values ​​of the power grid partition are arranged according to the topological position to generate the spatiotemporal mapping matrix of power grid carbon emissions. The matrix element value represents the predicted value of carbon emission intensity of the corresponding partition in the specified time period.

[0087] S4. Based on the target area association data in the spatiotemporal mapping prediction results of the power grid carbon emissions, where the carbon emission deviation exceeds a preset deviation threshold, an attention mechanism is introduced to dynamically optimize the full-chain spatiotemporal association matrix to obtain the power-carbon synergistic coupling matrix.

[0088] In some implementations, the step of dynamically optimizing the entire-chain spatiotemporal correlation matrix by introducing an attention mechanism based on the target region association data where the carbon emission deviation exceeds a preset deviation threshold in the spatiotemporal mapping prediction results of the power grid carbon emissions, to obtain the power-carbon synergistic coupling matrix, includes:

[0089] Based on the spatiotemporal mapping prediction results of power grid carbon emissions and the benchmark carbon emission intensity, the target area associated data whose carbon emission deviation exceeds a preset deviation threshold is identified.

[0090] Dynamic attention weights are calculated on the associated data of the target region to obtain attention weight coefficients, and the attention weight coefficients are multiplied element-wise with the spatiotemporal association matrix of the whole chain to obtain the initial electrocarbon association matrix;

[0091] By introducing an electrocarbon coupling factor to weight and fuse the initial electrocarbon correlation matrix, an electrocarbon cooperative coupling matrix is ​​generated.

[0092] Specifically, this embodiment calculates the deviation between the predicted carbon emission intensity and the benchmark carbon emission intensity based on the spatiotemporal mapping prediction results of the power grid, compares them with historical benchmark carbon emission curves, identifies target areas where the carbon emission intensity deviation exceeds a preset deviation threshold, and extracts the associated data corresponding to the target areas. The associated data can include spatial associated data, temporal associated data, and operational associated data. Spatial associated data refers to the row and column coordinate range of the target area in the full-chain spatiotemporal associated matrix. For example, in a power grid containing multiple substations and transmission lines, the power supply range of each substation is a spatial area, and its coordinate range can be represented by the substation's geographical location and power supply radius. Temporal associated data is the rate of change of coupling strength in the generation-transmission-distribution-consumption links of the target area during the deviation period. Operational associated data is the real-time power transmission characteristics of the generation-transmission-distribution-consumption links of the target area. For example, real-time power transmission characteristics can be characterized as the ratio of real-time power transmission volume to rated capacity. The value reflects the actual state of the power grid during operation. Then, in this embodiment, the attention weight coefficient of each target area at different time points is calculated based on the target area correlation data. The specific calculation process is as follows: the spatial attention weight is obtained according to the ratio of the coupling strength of the target area to the average coupling strength of the entire network; the temporal attention weight is obtained according to the percentage of the deviation duration to the analysis period; and the operational attention weight is obtained according to the ratio of the real-time power transmission to the rated capacity. The three types of weights, namely spatial weight, temporal weight and operational weight, are geometrically averaged to generate the attention weight coefficient. In this embodiment, the attention weight coefficient is multiplied element-wise with the full-chain spatiotemporal correlation matrix. The element value of the full-chain spatiotemporal correlation matrix is ​​multiplied by the attention weight coefficient of the corresponding target area at the corresponding time point to obtain the optimized full-chain spatiotemporal correlation matrix. The optimized full-chain spatiotemporal correlation matrix is ​​then normalized to keep the row and column vector magnitudes in units, generating the initial electric carbon correlation matrix.

[0093] This embodiment can extract the identification information of key carbon reduction areas to obtain regional constraint factors. For example, the element values ​​of the initial electrocarbon correlation matrix corresponding to key carbon reduction areas can be set to 1.2, and non-key areas to 1, based on the regional constraint factors. Simultaneously, based on the distribution characteristics of high-carbon emission areas statistically analyzed from the historical carbon emission database, a carbon emission constraint factor matrix is ​​generated. For example, in the carbon emission constraint factor matrix, the element value corresponding to high-carbon emission areas is 0.7, and other areas are 1. The standard deviation multiple of the carbon emission intensity of the target area and the mean of the entire network is calculated to generate a sensitivity factor matrix distributed along the diagonal of the matrix. This embodiment implements a hierarchical weighted fusion of the initial electrocarbon correlation matrix and three types of electrocarbon coupling factors. The initial electrocarbon correlation matrix is ​​multiplied element-wise by the regional constraint factors, the first-level output matrix is ​​multiplied element-wise by the carbon emission constraint factor matrix, and the second-level output matrix is ​​added element-wise by the sensitivity factor matrix. The resulting matrix is ​​then processed. Normalization is performed to ensure the sum of elements in each row is 1. The initial weighted and fused electric-carbon correlation matrix is ​​then subjected to time-series smoothing, for example, by applying a moving average filter to the matrix elements in a 10-minute window to eliminate instantaneous fluctuation noise, resulting in an electric-carbon synergistic coupling matrix. In a specific embodiment, this embodiment can enhance the characteristics of the initial electric-carbon correlation matrix based on the constraints of low-carbon power grid planning. Specifically, the element values ​​of the matrix row vectors corresponding to key low-carbon planning areas are increased to 1.2 to 1.5 times their original values; the element values ​​of the matrix column vectors corresponding to historically high carbon emission load areas are reduced to 0.7 to 0.9 times their original values; and a carbon emission sensitivity factor is superimposed along the diagonal of the matrix. The carbon emission sensitivity factor is the standard deviation multiple of the carbon emission intensity of the target area and the average of the entire network, generating an electric-carbon synergistic coupling matrix with directional enhancement of electric-carbon coupling characteristics. The matrix element values ​​characterize the optimized electric-carbon dynamic coupling strength.

[0094] S5. Based on the aforementioned electric-carbon synergistic coupling matrix, a generative adversarial network is used to quantify the spatial distribution of the low-carbon effect of power grid planning, and the results of dynamic synergistic analysis of electric-carbon are obtained.

[0095] In some embodiments, the step of using a generative adversarial network to quantify the spatial distribution of low-carbon effects in power grid planning based on the electro-carbon synergistic coupling matrix to obtain the dynamic synergistic analysis results of electro-carbon includes:

[0096] Based on the aforementioned electric-carbon synergistic coupling matrix and the pre-acquired variables to be planned, a low-carbon planning feature vector is constructed.

[0097] Using the low-carbon planning feature vector as input, a spatial distribution map of low-carbon effects is generated through a generative adversarial network;

[0098] An anisotropic diffusion algorithm is used to smooth the spatial distribution map of the low-carbon effect to obtain a smoothed spatial distribution of the low-carbon effect.

[0099] The element-wise multiplication of the smooth low-carbon effect spatial distribution with the electric carbon synergistic coupling matrix yields the power grid zone electric carbon synergistic efficiency index, which is then used as the result of the electric carbon dynamic synergistic analysis.

[0100] Specifically, this embodiment integrates the row and column coupling strength distribution of the electric-carbon synergistic coupling matrix with the variables to be planned, and constructs a low-carbon planning feature tensor according to the number of power grid partitions and the dimensions of the planning variables. Specifically, for each power grid partition, this embodiment combines the electric-carbon synergistic coupling strength value corresponding to that partition with the planning variable value to form a feature vector for that partition. Then, the feature vectors of all partitions are combined into a feature tensor. Here, the power grid partition is the topological partition corresponding to the electric-carbon synergistic coupling matrix, and the variables to be planned can include planning variables such as energy penetration rate, energy storage configuration ratio, and demand response target parameters. Simultaneously, this embodiment constructs a generative adversarial network with power grid topology constraints. The generative adversarial network includes a generator network and a discriminator network. The generator network adopts a deconvolutional neural network structure. In this embodiment, the low-carbon planning feature tensor is input into the generator network, and the deconvolution operation is used to gradually restore the low-carbon planning feature tensor. The system generates a low-carbon effect spatial distribution map with dimensions consistent with the power grid partitions. During the generation process, each layer of the deconvolutional neural network performs nonlinear transformations and spatial expansions on the input features, ensuring that each pixel value in the output low-carbon effect spatial distribution map reflects the carbon emission reduction potential coefficient of the corresponding power grid partition. This low-carbon effect spatial distribution map intuitively displays the carbon emission reduction potential of each power grid partition under different planning variables. The discriminator network adopts a multi-scale convolutional neural network structure, which can capture feature information at different scales. By calculating the degree of matching between the generated low-carbon effect spatial distribution map and the power grid physical constraints and historical measured low-carbon effect spatial distribution maps, a authenticity probability value is output. This authenticity probability value represents the similarity between the generated low-carbon effect spatial distribution map and the real situation, and is used for subsequent training and optimization of the generator.

[0101] This embodiment trains the generator and discriminator networks using an alternating training method, defining loss functions for both networks during training. The generator's loss function includes adversarial loss and planning constraint loss. Adversarial loss minimizes the discriminator's error in judging the authenticity of the generated distribution map, making it as close to the true distribution as possible. Planning constraint loss ensures the generated results meet the constraints of power grid planning, such as a certain proportion of renewable energy generation. The discriminator's loss function is the classification loss between the true and generated distributions. Classification loss continuously adjusts the parameters of the generator and discriminator, making the generated low-carbon effect spatial distribution map increasingly closer to reality, while ensuring the discriminator accurately judges the authenticity of the distribution map. Feasibility constraints for planning variables can be introduced when calculating the loss, resulting in a stable generator capable of generating accurate low-carbon effect spatial distribution maps. It should be noted that during training, when the carbon reduction potential coefficient in the low-carbon effect spatial distribution map changes by less than 1% over 10 consecutive iterations, the learning rate is automatically reduced to prevent overfitting. This embodiment employs anisotropic diffusion based on the topological connections between power grid zones. The algorithm smooths the low-carbon effect spatial distribution map output by the trained and stable generator. The anisotropic diffusion algorithm, based on the spatial structure characteristics of power grid partitions, smoothly transitions the carbon reduction potential coefficients at partition boundaries, eliminating abrupt boundary changes and making the distribution map more consistent with the actual operation of the power grid. Specifically, the diffusion coefficient can be set as the reciprocal of the impedance values ​​of adjacent partitions, thereby generating a smooth low-carbon effect spatial distribution map, eliminating abrupt boundary changes, and then performing element-wise multiplication of the smooth low-carbon effect spatial distribution map with the power grid-carbon co-coupling matrix. Specifically, this involves multiplying the corresponding power grid partition carbon reduction potential coefficients... Multiplying by the power-carbon synergy coupling strength value generates a power grid zone power-carbon synergy efficiency index matrix. The element values ​​of the power grid zone power-carbon synergy efficiency index matrix reflect the comprehensive efficiency level of the power grid zone under the consideration of power-carbon synergy effect. This comprehensive efficiency level takes into account the carbon emission reduction potential coefficient and the power-carbon synergy coupling strength, and to a certain extent can indirectly reflect the relative advantage of the power grid zone in achieving the power-carbon synergy goal. In this embodiment, the power grid zone can be divided into priority implementation zone, optimization adjustment zone and basic maintenance zone according to the power grid zone power-carbon synergy efficiency index, and power grid zone dynamic synergy analysis results with hierarchical labels can be generated.

[0102] This invention provides a method for dynamic collaborative analysis of power generation and carbon emissions. The method extracts features from real-time operational data of the power system's generation, transmission, distribution, and consumption links to obtain a full-chain correlation quantification representation. Based on this full-chain correlation quantification representation, it models the nonlinear coupling relationships between the power system's generation, transmission, distribution, and consumption links in a multi-dimensional manner, generating a full-chain spatiotemporal correlation matrix. Using the full-chain spatiotemporal correlation matrix and carbon emission parameters of the power system's generation, transmission, distribution, and consumption links, it employs a multi-branch convolutional neural network to learn the spatiotemporal mapping relationship between the power grid's operating state and carbon emissions, obtaining a spatiotemporal mapping prediction result for power grid carbon emissions. Based on the target area correlation data in the power grid carbon emission spatiotemporal mapping prediction result where carbon emission deviation exceeds a preset deviation threshold, it introduces an attention mechanism to dynamically optimize the full-chain spatiotemporal correlation matrix, obtaining a power generation and carbon emissions collaborative coupling matrix. Based on this power generation and carbon emissions collaborative coupling matrix, it uses a generative adversarial network to quantify the spatial distribution of the low-carbon effects of power grid planning, obtaining the dynamic collaborative analysis result of power generation and carbon emissions. Compared with existing technologies, this method uses spatiotemporal correlation modeling and dynamic optimization mechanisms to accurately map the power grid operation status and carbon emissions, realizing dynamic collaborative analysis of the power grid system and carbon emissions. This provides a scientific basis for the optimization of power grid planning and strong support for the low-carbon development of the power system.

[0103] It should be noted that the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0104] In one embodiment, such as Figure 2 As shown, this embodiment of the invention provides an electrocarbon dynamic synergistic analysis system, the system comprising:

[0105] The feature extraction module 101 is used to extract features from the real-time operation data of the power system's generation, transmission, distribution and consumption links to obtain a full-chain correlation quantitative representation;

[0106] Data analysis module 102 is used to perform multi-dimensional coupling relationship modeling on the nonlinear coupling relationship between the generation, transmission, distribution and consumption links of the power system based on the full-chain correlation quantification representation, and generate a full-chain spatiotemporal correlation matrix;

[0107] The model prediction module 103 is used to learn the spatiotemporal mapping relationship between the power grid operating status and carbon emissions based on the full-chain spatiotemporal correlation matrix and the carbon emission parameters of the power system generation, transmission, distribution and consumption links, and to obtain the spatiotemporal mapping prediction result of power grid carbon emissions by using a multi-branch convolutional neural network.

[0108] The collaborative coupling module 104 is used to dynamically optimize the whole-chain spatiotemporal correlation matrix by introducing an attention mechanism based on the target area association data in the spatiotemporal mapping prediction results of the power grid carbon emissions, where the carbon emission deviation exceeds a preset deviation threshold, to obtain the power-carbon collaborative coupling matrix.

[0109] The collaborative analysis module 105 is used to quantify the spatial distribution of the low-carbon effect of power grid planning based on the electric carbon collaborative coupling matrix and to obtain the dynamic collaborative analysis results of electric carbon.

[0110] For specific limitations regarding the electrocarbon dynamic synergistic analysis system, please refer to the above-described limitations regarding the electrocarbon dynamic synergistic analysis method, which will not be repeated here. Those skilled in the art will recognize that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0111] This invention provides a dynamic collaborative analysis system for power grid carbon emissions. The system uses a feature extraction module to extract features from real-time operational data of the power system's generation, transmission, distribution, and consumption links, obtaining a full-chain correlation quantification representation. A data analysis module models the nonlinear coupling relationships between these links using a multi-dimensional model, generating a full-chain spatiotemporal correlation matrix. A model prediction module uses a multi-branch convolutional neural network to learn the spatiotemporal mapping relationship between the power grid's operating state and carbon emissions, based on the full-chain spatiotemporal correlation matrix and carbon emission parameters from the power system's generation, transmission, distribution, and consumption links, to obtain a spatiotemporal mapping prediction result for power grid carbon emissions. A collaborative coupling module uses an attention mechanism to dynamically optimize the full-chain spatiotemporal correlation matrix based on the target area correlation data where carbon emission deviations exceed a preset deviation threshold in the spatiotemporal mapping prediction result, obtaining a power grid carbon collaborative coupling matrix. Finally, a collaborative analysis module uses a generative adversarial network to quantify the spatial distribution of low-carbon effects in power grid planning based on the power grid carbon collaborative coupling matrix, obtaining dynamic collaborative analysis results for power grid carbon emissions. Compared with existing technologies, this system uses spatiotemporal correlation modeling and dynamic optimization mechanisms to accurately map the power grid operation status and carbon emissions, realizing dynamic collaborative analysis of the power grid system and carbon emissions. This provides a scientific basis for the optimization of power grid planning and strong support for the low-carbon development of the power system.

[0112] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this invention, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. An electrocarbon dynamic co-analysis method, characterized in that, The method comprises the following steps: characteristic extraction is performed on real-time operation data of power system generation, transmission, distribution and utilization links to obtain a full-chain correlation quantitative representation; a multi-dimensional coupling relationship model is established for a nonlinear coupling relationship between the power system generation, transmission, distribution and utilization links according to the full-chain correlation quantitative representation, and a full-chain space-time correlation matrix is generated; specifically, a dynamic topology graph is constructed according to a topological connection relationship of the power system generation, transmission, distribution and utilization links; cross-link coupling features are extracted layer by layer in a multi-layer graph convolution network by inputting the dynamic topology graph, and a global-local interaction correlation matrix is generated; the global-local interaction correlation matrix is subjected to space-time dimension alignment fusion to generate an initial correlation tensor subjected to space-time alignment; principal component analysis is performed on the initial correlation tensor to extract key coupling modes, and the key coupling modes are subjected to orthogonalization processing through singular value decomposition to obtain the full-chain space-time correlation matrix; a multi-branch convolutional neural network is used to learn a space-time mapping relationship between a power grid operation state and carbon emissions according to the full-chain space-time correlation matrix and carbon emission parameters of the power system generation, transmission, distribution and utilization links, and a power grid carbon emission space-time mapping prediction result is obtained; a dynamic optimization is performed on the full-chain space-time correlation matrix by introducing an attention mechanism according to target region correlation data in which a carbon emission amount deviation exceeds a preset deviation threshold in the power grid carbon emission space-time mapping prediction result, and an electricity-carbon collaborative coupling matrix is obtained. Based on the electricity-carbon collaborative coupling matrix, a generative adversarial network is used to quantize a low-carbon effect spatial distribution of a power grid plan, and an electricity-carbon dynamic collaborative analysis result is obtained.

2. The electrocarbon dynamic co-analysis method of claim 1, wherein, The step of performing characteristic extraction on real-time operation data of power system generation, transmission, distribution and utilization links to obtain a full-chain correlation quantitative representation comprises: collecting real-time operation data and historical operation data of power system generation, transmission, distribution and utilization links to form multi-source operation data containing the power system generation, transmission, distribution and utilization links; standardizing the multi-source operation data to obtain standardized multi-source operation data, and segmenting the standardized multi-source operation data by using a sliding window to construct standardized time series data slices; performing space-time feature extraction on the standardized time series data slices to obtain a mixed feature vector; performing feature compression and correlation coding by using a full connection layer according to the mixed feature vector to obtain a full-chain correlation quantitative representation.

3. The electrocarbon dynamic co-analysis method of claim 1, wherein, The step of constructing a dynamic topology graph according to a topological connection relationship of power system generation, transmission, distribution and utilization links comprises: constructing a dynamic topology graph with generators, transmission nodes, distribution busbars and user loads as vertices and power transmission paths as edges according to a topological connection relationship of power system generation, transmission, distribution and utilization links; allocating dynamic weight features to each vertex and edge in the dynamic topology graph based on the full-chain correlation quantitative representation.

4. The electrocarbon dynamic co-analysis method of claim 3, wherein: The multi-layer graph convolution network at least comprises a three-layer graph convolution network and an output layer, and the three-layer graph convolution network comprises a first layer graph convolution network, a second layer graph convolution network and a third layer graph convolution network; the first layer graph convolution network is configured to calculate vertex-level correlation strength by calculating an instantaneous power transmission correlation strength between adjacent vertices in the dynamic topology graph through a dynamic weight feature of an edge to obtain a vertex-level correlation strength matrix; The second layer graph convolutional network is configured to perform regional level spatial clustering on the vertices of the power system generation and transmission link based on the vertex level correlation strength matrix, so as to obtain a local correlation matrix of energy interaction of the power system generation and transmission link. The third layer graph convolutional network is configured to perform power supply path traversal on the vertices of the power system distribution and utilization link based on the vertex level correlation strength matrix, so as to obtain a global correlation matrix of load response of the power system distribution and utilization link. The output layer is configured to perform dimension expansion and splicing on the local correlation matrix and the global correlation matrix, so as to obtain a global-local interaction correlation matrix.

5. The electrocarbon dynamic co-analysis method of claim 1, wherein, The step of learning the spatio-temporal mapping relationship between the power grid operation state and the carbon emission amount by using a multi-branch convolutional neural network according to the full-chain spatio-temporal correlation matrix and the carbon emission parameters of the power system generation, transmission, distribution and utilization links to obtain the power grid carbon emission spatio-temporal mapping prediction result comprises: performing three-dimensional spatial mapping on the coupling strength distribution of the full-chain spatio-temporal correlation matrix and the carbon emission parameters of the power system generation, transmission, distribution and utilization links to construct an electric-carbon spatio-temporal feature cube; inputting the electric-carbon spatio-temporal feature cube into a multi-branch convolutional neural network for feature learning to obtain an electric-carbon spatio-temporal fusion feature tensor, and dividing the power grid into partitions according to the power system generation, transmission, distribution and utilization links; restoring the spatial resolution of the electric-carbon spatio-temporal fusion feature tensor to the power grid partition scale by deconvolution operation to obtain an electric-carbon spatio-temporal feature restoration tensor; performing carbon emission intensity regression calculation on each power grid partition based on the electric-carbon spatio-temporal feature restoration tensor to obtain the power grid carbon emission spatio-temporal mapping prediction result of the power grid partition in a target period.

6. A method of electrocarbon dynamic co-analysis as claimed in claim 5, characterized by: The multi-branch convolutional neural network comprises a spatial feature extraction branch network, a time sequence feature extraction branch network and a cross-scale feature fusion branch network.

7. The electrocarbon dynamic co-analysis method of claim 1, wherein, The step of introducing an attention mechanism to dynamically optimize the full-chain spatio-temporal correlation matrix according to the target region associated data in the power grid carbon emission spatio-temporal mapping prediction result whose carbon emission amount deviation exceeds a preset deviation threshold to obtain an electric-carbon collaborative coupling matrix comprises: identifying the target region associated data whose carbon emission amount deviation exceeds the preset deviation threshold according to the power grid carbon emission spatio-temporal mapping prediction result and the benchmark carbon emission intensity; performing dynamic attention weight calculation on the target region associated data to obtain an attention weight coefficient, and multiplying the attention weight coefficient with the full-chain spatio-temporal correlation matrix element by element to obtain an initial electric-carbon correlation matrix; generating an electric-carbon collaborative coupling matrix by introducing an electric-carbon coupling factor to weight and fuse the initial electric-carbon correlation matrix.

8. The electrocarbon dynamic co-analysis method of claim 5, wherein, The step of quantifying the spatial distribution of the low-carbon effect of the power grid planning based on the electric-carbon collaborative coupling matrix by using a generative adversarial network to obtain an electric-carbon dynamic collaborative analysis result comprises: constructing a low-carbon planning feature vector based on the electric-carbon collaborative coupling matrix and the pre-acquired to-be-planned variables; generating a low-carbon effect spatial distribution map by the generative adversarial network with the low-carbon planning feature vector as the input; performing smoothing processing on the low-carbon effect spatial distribution map by using an anisotropic diffusion algorithm to obtain a smoothed low-carbon effect spatial distribution; Element-level multiplication of the smooth low-carbon effect spatial distribution and the electric-carbon synergistic coupling matrix obtains an electric grid partition electric-carbon synergistic efficiency index, and the electric grid partition electric-carbon synergistic efficiency index is taken as an electric-carbon dynamic synergistic analysis result.

9. An electrocarbon dynamic co-analysis system, characterized in that, The system comprises: a feature extraction module configured to perform feature extraction on real-time operation data of power system generation, transmission, distribution and utilization links to obtain a full-chain correlation quantitative representation; a data analysis module configured to perform multi-dimensional coupling relationship modeling on a nonlinear coupling relationship between the power system generation, transmission, distribution and utilization links according to the full-chain correlation quantitative representation to generate a full-chain space-time correlation matrix; specifically, a dynamic topology graph is constructed according to a topological connection relationship of the power system generation, transmission, distribution and utilization links; cross-link coupling features are extracted layer by layer in a multi-layer graph convolution network by inputting the dynamic topology graph into the multi-layer graph convolution network to generate a global-local interactive correlation matrix; the global-local interactive correlation matrix is subjected to space-time dimension alignment fusion to generate an initial correlation tensor subjected to space-time alignment; principal component analysis is performed on the initial correlation tensor to extract key coupling modes, and the key coupling modes are subjected to orthogonalization processing through singular value decomposition to obtain the full-chain space-time correlation matrix; a model prediction module configured to learn a space-time mapping relationship between a power grid operation state and carbon emissions by using a multi-branch convolutional neural network according to the full-chain space-time correlation matrix and carbon emission parameters of the power system generation, transmission, distribution and utilization links to obtain a power grid carbon emission space-time mapping prediction result; a synergistic coupling module configured to introduce an attention mechanism to dynamically optimize the full-chain space-time correlation matrix according to target region correlation data in which a carbon emission amount deviation exceeds a preset deviation threshold in the power grid carbon emission space-time mapping prediction result to obtain an electric-carbon synergistic coupling matrix; a synergistic analysis module configured to quantize a power grid planning low-carbon effect spatial distribution by using a generative adversarial network based on the electric-carbon synergistic coupling matrix to obtain an electric-carbon dynamic synergistic analysis result.

Images