A power system carbon emission metering method based on dynamic electric carbon emission factor

Abstract

The application discloses a kind of power system carbon emission metering methods based on dynamic electric carbon emission factor, it is related to environmental cross technical field, including, according to carbon potential value, hierarchical aggregation node carbon potential weight is divided according to voltage grade, time series carbon potential curve is generated, and carbon emission total amount is obtained by dynamic carbon flow density accumulation method, while based on grid topological relation determination carbon flow propagation direction and obtain tracking total amount;The relative deviation of tracking total amount and carbon emission total amount is calculated, relative deviation is compared and analyzed with historical carbon emission error threshold value, and carbon emission metering report is generated;By inputting three-dimensional real-time operation data cube into neural network model, dynamic carbon emission factor is predicted through space-time feature double channel, the spatial dependence and time dynamics in the grid multi-source heterogeneous data are collaboratively modeled, the instantaneous change characteristics of carbon emission under complex operating conditions are effectively captured, and the prediction accuracy and timeliness are improved.

Description

Technical Field

[0001] This invention relates to the field of environmental cross-technology, and in particular to a method for measuring carbon emissions from power systems based on dynamic electrical carbon emission factors. Background Technology

[0002] In power systems, carbon emission measurement is a crucial step in assessing energy consumption and environmental impact. Traditional carbon emission measurement methods primarily rely on static carbon emission factors, combined with grid load data for estimation. These methods typically involve collecting historical data from both the generation and load sides, and then estimating carbon emissions using simple linear models or statistical methods. For example, based on historical average generation structure and load curves, a fixed carbon emission factor is used to calculate the total carbon emissions for each time period. Especially with the increasing proportion of renewable energy and frequent load fluctuations, static methods struggle to accurately reflect real-time carbon emissions.

[0003] However, the aforementioned conventional methods exhibit certain limitations when faced with complex and ever-changing power grid operating conditions. First, in processing high-frequency data updates, traditional methods often fail to respond promptly to rapid changes in power grid conditions, leading to lags in carbon emission estimation. Second, due to the lack of effective capture of spatiotemporal characteristics, conventional methods struggle to accurately describe the distribution of carbon emissions across different regions and time periods. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a power system carbon emission metering method based on dynamic electrical carbon emission factors, which solves the problem of minute-level accurate tracking of grid carbon emissions and rapid response to grid status.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for measuring carbon emissions from power systems based on dynamic electrical carbon emission factors, comprising:

[0008] Collect real-time operating data of the power system and preprocess it to generate a three-dimensional real-time operating data cube;

[0009] The three-dimensional real-time running data cube is input into the neural network model to predict dynamic carbon emission factors through a dual-pathway of spatiotemporal features.

[0010] Based on historical power grid equipment status data, the power grid topology is obtained, and a dynamic carbon emission graph structure is constructed through a dynamic graph convolutional network.

[0011] The dynamic carbon emission factor is defined as a node in the dynamic carbon emission graph structure, and aggregated calculations are performed using historical line resistance parameters to obtain the carbon potential value of the node.

[0012] Based on the carbon potential value, the node carbon potential weights are aggregated in layers according to voltage level to generate a time series carbon potential curve. The total carbon emissions are obtained through the dynamic carbon flow density accumulation method. At the same time, the carbon flow propagation direction is determined based on the power grid topology and the total tracking amount is obtained.

[0013] The relative deviation between the total tracked amount and the total carbon emissions is calculated, and the relative deviation is compared and analyzed with historical carbon emission error thresholds to generate a carbon emission measurement report.

[0014] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electric carbon emission factor described in this invention, the real-time operation data of the power system includes dynamic data of the generation side, spatiotemporal data of the load side, and environmental coupling data.

[0015] The preprocessing includes data cleaning, time alignment, normalization, feature extraction, and quality verification.

[0016] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electric carbon emission factor described in this invention, the generation of the three-dimensional real-time operation data cube refers to the fusion and organization of the preprocessed real-time operation data of the power system according to spatial dimension, time dimension and parameter dimension to generate a three-dimensional real-time operation data cube.

[0017] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factors described in this invention, the specific steps of inputting a three-dimensional real-time operating data cube into a neural network model and predicting the dynamic carbon emission factor through a dual-pathway spatiotemporal feature analysis are as follows:

[0018] The three-dimensional real-time running data cube is time-sliding segmented according to a fixed time window to generate spatiotemporal data of power grid nodes;

[0019] Spatiotemporal data are input into a neural network model to capture the spatial characteristics of the interaction between power grid nodes through a spatial feature path, while the temporal feature path is used to obtain the temporal data sequence of each node as time changes.

[0020] Spatial features are fused with temporal data sequences to generate spatial-temporal fused features, which are then encoded and mapped to generate spatial-temporal fused feature vectors.

[0021] Based on the spatial-temporal fusion feature vector, the evolution law of carbon potential is captured by the spatiotemporal feature correlation analysis method, and dynamic analysis is performed to predict dynamic carbon emission factors.

[0022] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factor described in this invention, the step of obtaining the power grid topology relationship based on historical power grid equipment status data refers to extracting the physical connection relationship of the power grid through time series analysis based on historical power grid equipment status data, and fusing it with the time under the reference clock to obtain the power grid topology relationship.

[0023] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factors described in this invention, the specific steps for constructing a dynamic carbon emission graph structure using a dynamic graph convolutional network are as follows:

[0024] Based on the power grid topology, a mapping operation is performed on the connection relationships between nodes to generate the association strength between nodes;

[0025] Based on the strength of the association between nodes, a dynamic carbon emission graph structure is constructed using a dynamic graph convolutional network.

[0026] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factors described in this invention, the specific steps of defining the dynamic carbon emission factors as nodes in a dynamic carbon emission graph structure and performing aggregation calculations based on historical line resistance parameters to obtain the carbon potential value of the nodes are as follows.

[0027] Based on the dynamic carbon emission factor, calculate the carbon emission data at each time point and at each power grid node and define them as nodes in the dynamic carbon emission graph structure;

[0028] The historical line resistance parameters are defined as edges of the carbon emission graph structure, and a traversal operation is performed to generate a carbon emission adjacency matrix.

[0029] The carbon emission adjacency matrix is ​​aggregated to obtain the losses and impacts generated when adjacent edges interact with other nodes, and the carbon potential of the nodes is obtained through power flow analysis.

[0030] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factors described in this invention, the steps of aggregating node carbon potential weights hierarchically according to voltage levels based on carbon potential values, generating time-series carbon potential curves, and obtaining the total carbon emissions through the dynamic carbon flux density accumulation method are as follows.

[0031] The voltage level is mapped to the carbon potential value of the node, and the carbon potential weight of the node is aggregated hierarchically according to the voltage level to generate the overall carbon potential of the voltage level.

[0032] The overall carbon potential of the voltage level is fused with the time of a fixed interval to generate carbon potential data of the voltage level nodes as a function of time, and a time series carbon potential curve is generated by the moving average method.

[0033] Based on the time-series carbon potential curve, a carbon flux density index is defined, and the cumulative carbon flux density value at each load point is calculated step by step using the dynamic carbon flux density accumulation method to generate the total carbon emissions.

[0034] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factor described in this invention, the step of determining the carbon flow propagation direction and obtaining the total tracking amount based on the power grid topology relationship refers to tracking the propagation direction of carbon flow from the power source to each consumption point based on the power grid topology relationship, calculating the carbon emission amount corresponding to each propagation direction, and simultaneously summarizing and analyzing the data to obtain the total tracking amount.

[0035] As a preferred embodiment of the power system carbon emission measurement method based on dynamic electrical carbon emission factors described in this invention, the steps of calculating the relative deviation between the total tracking amount and the total carbon emissions, comparing and analyzing the relative deviation with historical carbon emission error thresholds, and generating a carbon emission measurement report are as follows.

[0036] The total amount tracked is compared with the total amount of carbon emissions, the relative deviation between the two is calculated, and the relative deviation is compared and analyzed with the historical carbon emission error threshold to verify the rationality of carbon emissions.

[0037] Based on the rationality results of carbon emissions, identify abnormal areas and generate a carbon emission metering report for the power system.

[0038] The beneficial effects of this invention are as follows: by inputting a three-dimensional real-time running data cube into a neural network model, dynamic carbon emission factors are predicted through a dual-pathway spatiotemporal feature model, thereby achieving collaborative modeling of spatial dependence and temporal dynamics in multi-source heterogeneous data of the power grid. This effectively captures the instantaneous change characteristics of carbon emissions under complex operating conditions and improves prediction accuracy and timeliness. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a power system carbon emission measurement method based on dynamic electrical carbon emission factors.

[0041] Figure 2 A flowchart for generating total carbon emissions.

[0042] Figure 3 This is a flowchart for predicting dynamic carbon emission factors.

[0043] Figure 4This is a flowchart for calculating and verifying carbon potential. Detailed Implementation

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0045] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0047] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for measuring carbon emissions from a power system based on a dynamic electrical carbon emission factor, comprising the following steps:

[0048] S1. Collect real-time operating data of the power system and preprocess it to generate a three-dimensional real-time operating data cube;

[0049] Real-time operational data of a power system includes dynamic data from the generation side, spatiotemporal data from the load side, and environmental coupling data.

[0050] It should be noted that the dynamic data on the power generation side is collected through terminals (such as PLC controllers and condition monitoring devices), and the unit output, fuel carbon emission coefficient and regulation response dynamics are captured in real time at a frequency of 1-5 seconds according to the IEC 61850 protocol; the spatiotemporal data on the load side is uploaded to the user's electricity consumption in 15-minute granularity, and is collected by the photovoltaic inverter in real time through the distributed power monitoring system; the environmental coupling data is obtained through the meteorological monitoring grid (1km resolution) to obtain wind speed / sunlight / temperature parameters.

[0051] Preprocessing includes data cleaning, time alignment, normalization, feature extraction, and quality verification;

[0052] It should be noted that data cleaning involves removing outliers exceeding the standard deviation range. Missing data is filled using a combination of linear interpolation in the time dimension and the KNN algorithm in the spatial dimension, for example, by using the mean of five adjacent similar nodes. Time alignment follows the PTP precision clock protocol to unify timestamps, setting the timestamp alignment precision to one millisecond. A sliding window alignment method is used to process time series data, for example, with a window size of twelve sampling points. Normalization distinguishes between electrical quantity data and environmental data, applying different operations. Electrical quantity data uses Min-Max normalization to map to the zero-to-one interval, while environmental data uses Z-s... Core standardization processing; feature extraction is performed through sliding segmentation, including time-domain features, frequency-domain features, and topological features. Time-domain features include the calculation of RMS voltage, RMS current, and power factor; frequency-domain features are obtained by performing FFT transformation to obtain harmonic distortion rate parameters; and topological features are calculated by calculating node degree and betweenness centrality, for example, using a sliding window to extract 24-hour trend features. Quality verification performs four verification operations: integrity check, accuracy verification, timeliness detection, and consistency comparison. Integrity check counts the proportion of missing data, and accuracy verification uses Kirchhoff's laws to perform energy conservation verification.

[0053] It should be noted that the one-millisecond alignment accuracy setting is based on the power system synchronization phasor measurement device and industry standards.

[0054] The preprocessed real-time operation data of the power system is fused and organized according to spatial, temporal, and parameter dimensions to generate a three-dimensional real-time operation data cube.

[0055] Furthermore, based on task requirements, a three-dimensional matrix is ​​defined. Then, spatial dimension organization operations establish a spatial grid using substations, lines, and nodes as coordinate indices, for example, dividing a provincial power grid into 5,000 spatial units. Time dimension organization operations establish time axis indices at fixed time intervals, for example, 388 time point indices corresponding to a five-minute interval. Parameter dimension organization operations classify electrical quantities, environmental quantities, and other parameters to establish feature axis indices, such as 120 parameter types including voltage, current, and power factor. Subsequently, the preprocessed real-time operation data records of the power system are filled into the corresponding positions in the three-dimensional matrix according to the spatial dimension index, time dimension index, and parameter dimension index, for example, the voltage value of a substation node at a specific time point is filled into the corresponding matrix coordinates. The three-dimensional real-time operation data cube verification operation performs an integrity check to confirm that each dimension index position is filled with data, for example, the spatial dimension index coverage must reach 100%. Finally, a three-dimensional real-time operation data cube containing spatial, time, and parameter dimensions is formed. The three-dimensional real-time operation data cube supports slicing queries by any dimension, for example, extracting the voltage parameter time series of a line node for the entire day.

[0056] S2. Input the three-dimensional real-time running data cube into the neural network model, and predict the dynamic carbon emission factor through the spatiotemporal feature dual pathway;

[0057] The three-dimensional real-time running data cube is time-sliding segmented according to a specific time to generate spatiotemporal data of power grid nodes;

[0058] Furthermore, a time sliding window parameter operation is defined, and a sliding window extraction operation is performed on the time axis to move the window to cover the time index range of the three-dimensional real-time running data cube according to a specific time and a fixed step size. For each window position, a data slicing operation is performed to extract the spatial dimension data and parameter dimension data of the three-dimensional real-time running data cube within the corresponding time window. Then, based on the spatial dimension data and parameter dimension data, a multi-scale scanning operation is performed through the dynamic feature tensor reconstruction method to generate the spatiotemporal data of the power grid nodes.

[0059] It should be noted that a fixed time window refers to a dynamic time period benchmark defined by a time sliding window.

[0060] It should be noted that the time sliding window parameters, namely the window length, sliding step size, and topology change triggering mechanism, are defined according to the functional attributes of the power system (such as the power system fluctuation cycle).

[0061] Spatiotemporal data are input into a neural network model to capture the spatial characteristics of the interaction between power grid nodes through a spatial feature path, while the temporal feature path is used to obtain the temporal data sequence of each node as time changes.

[0062] Furthermore, spatiotemporal data are input into the neural network model. In the spatial feature path, an adjacency matrix is ​​constructed using the power grid topology. Based on the adjacency matrix, multi-layer aggregation of parameters such as the dynamic data of the generation side of each power grid node is performed through graph convolution operations to capture the spatial features of the mutual influence of power flow between adjacent nodes. For example, a 5-layer graph convolutional network is used to extract the correlation features between a node and its neighbors within two hops. In the temporal feature path, the sequence of parameters such as the spatiotemporal data of the load side of each power grid node within a continuous time step is input into the Long Short-Term Memory network. The hidden state is updated time-by-time through a gating mechanism to obtain the temporal data sequence of each node changing over time.

[0063] The training process of the Long Short-Term Memory (LSTM) network is as follows: The spatiotemporal data of power grid nodes, generated by time-sliding partitioning of historical 3D real-time running data cubes, is used as input samples. The actual dynamic carbon emission factors of each node at the corresponding time point are used as the target output, and the data is divided into training, validation, and test sets in a 7:2:1 ratio. Next, the network parameters are initialized, and the initial values ​​of the weight matrices and bias vectors of the forget gate, input gate, and output gate are set using the Xavier method. Then, forward propagation is performed, inputting the input sequence data into the LTM network units in time-step order, calculating the forget gate activation value, input gate activation value, cell state update value, and output gate activation value sequentially, calculating the loss function value, and using the mean squared error function to quantify the deviation between the predicted dynamic carbon emission factors and the actual values. Then, backpropagation is performed, calculating the gradient of the loss function with respect to all weight parameters using the time backpropagation algorithm, and using gradient pruning techniques to limit the gradient norm to no more than 5.0 to prevent gradient explosion. Finally, the network parameters are updated, completing the training of the LTM network.

[0064] It should be noted that the training process of the neural network model is as follows: Historical three-dimensional real-time running data cubes are used as input samples, generated by time-sliding segmentation. The actual dynamic carbon emission factors of each power grid node at the corresponding time are used as the target output. The spatiotemporal dataset is divided into training, validation, and test sets. First, the weights and bias parameters of the neural network model are initialized. The gradient of the loss function relative to the neural network model parameters is calculated using the backpropagation algorithm. The loss function is the mean squared error between the predicted dynamic carbon emission factor and the actual dynamic carbon emission factor. For example, the parameters are updated using the Adam optimizer with an initial learning rate of 0.001. During training, iterations are performed with a fixed batch size. The loss change is monitored through the validation set to prevent overfitting. Training stops when the loss no longer decreases after several consecutive training cycles, ultimately obtaining a neural network model that can accurately predict dynamic carbon emission factors.

[0065] The spatial features and temporal data sequences are fused using a multi-source data fusion method to generate spatial-temporal fused features, which are then encoded and mapped to generate spatial-temporal fused feature vectors.

[0066] Furthermore, the spatial features of the power grid nodes output from the spatial feature path and the temporal data sequences of each power grid node output from the temporal feature path are concatenated along the feature dimension to generate high-dimensional features. A multilayer perceptron is then used as a multi-source data fusion method to perform nonlinear transformation on the concatenated high-dimensional features. Spatial-temporal fusion features are extracted layer by layer through fully connected layers and activation functions. Subsequently, the spatial-temporal fusion features are encoded and mapped using fully connected layers to output a fusion feature vector. The fusion feature vector is then compressed to a fixed dimension, for example, mapped to a compact 64-dimensional representation, finally generating a spatial-temporal fusion feature vector.

[0067] The training process of the multilayer perceptron is as follows: The spatial-temporal fusion feature vector is used as the input sample, and the measured value of the dynamic carbon emission factor at the corresponding time step is used as the target output. These are divided into training, validation, and test sets in an 8:1:1 ratio. Next, the network parameters are initialized using the He initialization method to set the initial values ​​of the weight matrices and bias vectors of the input, hidden, and output layers. Then, forward propagation is performed, where the input layer receives a 256-dimensional feature vector, which undergoes linear transformation and activation function processing in a fully connected layer. The output values ​​of the hidden layers are calculated layer by layer to obtain the predicted dynamic carbon emission factor. Then, the loss function value is calculated, using the Huber loss function to quantify the deviation between the predicted and measured values. For example, a threshold δ=1.0 is set to balance the mean square error and absolute error characteristics. Next, the backpropagation process is performed, calculating the partial derivatives of the loss function with respect to the weights and biases layer by layer from the output layer to the input layer. Gradient clipping is used to limit the gradient norm to no more than 10.0. Finally, the network parameters are updated, completing the training process.

[0068] Based on the spatial-temporal fusion feature vector, the evolution law of carbon potential is captured by the spatiotemporal feature correlation analysis method, and then dynamic analysis is carried out through cross-modal attention mechanism to predict dynamic carbon emission factors.

[0069] Furthermore, based on the spatial-temporal fusion feature vector, the Pearson correlation coefficient or dynamic time warping distance between the spatial-temporal fusion feature vectors of different grid nodes within a continuous time step is first calculated using the spatiotemporal feature correlation analysis method. Then, a clustering analysis method is used to perform pattern segmentation to identify highly correlated spatiotemporal evolution patterns to capture the dynamic laws of carbon potential propagation in the grid. Subsequently, the spatial-temporal fusion feature vector and the dynamic laws of carbon potential propagation in the grid are used as inputs to a neural network model. Under the constraints of the spatial-temporal fusion feature vector and the dynamic laws of carbon potential propagation in the grid, a cross-modal attention mechanism is used to weighted aggregate the feature representations of different nodes and different times to predict the dynamic carbon emission factor of each grid node at the next time step.

[0070] It should be noted that the carbon potential evolution law refers to the spatiotemporal evolution pattern of the dynamic propagation, transfer and accumulation of carbon potential values ​​at each node in the power grid as they change over time and space.

[0071] S3. Based on historical power grid equipment status data, obtain the power grid topology and construct a dynamic carbon emission graph structure through a dynamic graph convolutional network;

[0072] Based on historical power grid equipment status data, the physical connection relationship of the power grid is extracted through time series analysis, and then fused with the time under the reference clock to obtain the power grid topology.

[0073] Furthermore, based on historical power grid equipment status data, the operating status sequences of equipment such as circuit breakers, disconnectors, transformers, and transmission lines are continuously monitored using time-series analysis methods. The on / off status changes of each device at different time steps are analyzed, and the physical connection topology is inferred using time-series correlation topology inference. At the same time, each set of equipment status observation records is aligned with the precise timestamp under the reference clock, such as using second-level time stamps under the UTC time standard, to ensure that the status information of all devices is synchronized under the same time base. Through time series matching and event sequence sorting, the physical connection relationship between devices is fused with the time information at the corresponding moment, and finally, the power grid topology relationship containing time-varying characteristics is obtained.

[0074] Based on the power grid topology, a precise mapping operation of the connection relationship between nodes is performed using the topology data parsing method to generate the correlation strength between nodes;

[0075] Furthermore, based on the power grid topology, the connection structure of each node in the power grid is traversed layer by layer and path analysis is performed using the topology data parsing method. The direct connection relationship and hierarchical transmission path between adjacent nodes are identified, and corresponding weight values ​​are assigned by the electrical parameter weighting method. For example, the reciprocal of the line resistance is used as the initial measure of connection strength. The weights are mapped to the interval between 0 and 1 through normalization. Then, the association strength calculation is performed on each pair of connected nodes to generate the association strength between nodes.

[0076] Based on the strength of the association between nodes, a dynamic carbon emission graph structure is constructed using a dynamic graph convolutional network.

[0077] Furthermore, based on the correlation strength between nodes, the electrical attributes of each node in the power grid are used as node features of the dynamic carbon emission graph structure. The correlation strength between nodes is used as the weight of the edges to construct an initial adjacency matrix. Subsequently, multi-layer graph convolution operations are set in the dynamic graph convolutional network. Each layer is updated by aggregating the feature information of the central node and its neighboring nodes and combining it with the edge weights to generate fused feature information. At the same time, the update frequency of the graph structure changes over time under the reference clock. Then, based on the fused feature information and under the constraint of the update frequency, the nonlinear transformation of the dynamic graph convolutional network is performed to output a node embedding representation containing spatiotemporal evolution characteristics. The dynamic carbon emission graph structure is then constructed using the graph structure reconstruction method.

[0078] The training process of the dynamic graph convolutional network should be explained as follows: Historical power grid equipment status data and corresponding dynamic carbon emission factors are used as training samples. The dynamic carbon emission graph structure, composed of node electrical attributes and inter-node correlation strength, is input into the graph convolutional network. The feature sequences of each node at multiple consecutive time steps are used as input, and the dynamic carbon emission factors for the next three time steps are used as prediction targets. The weight parameters of the graph convolutional network are initialized, and the mean squared error is used as the loss function to calculate the deviation between the predicted and true values. The network parameters are updated using the gradient descent optimizer through the backpropagation algorithm. During training, the model convergence is monitored by the loss value on the validation set. Training is terminated early when the loss does not decrease within 10 consecutive training cycles. Finally, a dynamic graph convolutional network that can capture the spatial dependence and temporal evolution of power grid nodes is obtained.

[0079] S4. Define the dynamic carbon emission factor as a node in the dynamic carbon emission graph structure, and perform aggregation calculations based on historical line resistance parameters to obtain the carbon potential value of the node.

[0080] Based on the dynamic carbon emission factor, calculate the carbon emission data at each time point and at each power grid node and define them as nodes in the dynamic carbon emission graph structure;

[0081] Furthermore, based on the dynamic carbon emission factor, the dynamic carbon emission factor value of the corresponding power grid node at each time point is obtained in a time series. Then, the power grid node at each time point is bound to the corresponding dynamic carbon emission factor value as the attribute feature of the power grid node in the dynamic carbon emission graph structure. Subsequently, in the dynamic carbon emission graph structure, the power grid node is defined as an entity and the dynamic carbon emission factor is defined as the representation of the node attribute. Finally, the carbon emission data at each time point and on the power grid node are defined as nodes in the dynamic carbon emission graph structure.

[0082] It should be noted that carbon emission data refers to the collection of carbon emission information across all aspects of the power system obtained through precise quantification technology.

[0083] The historical line resistance parameters are defined as edges of the carbon emission graph structure, and a carbon emission adjacency matrix is ​​generated by performing a depth-first traversal algorithm.

[0084] Furthermore, historical line resistance parameters are defined as edges in the carbon emission graph structure. The historical average resistance value corresponding to each transmission line in the power grid is used as the initial attribute of the edge. Then, the connection relationship between all nodes is determined based on the power grid topology. An edge entity is established for each pair of physically connected nodes in the carbon emission graph structure. Starting from any source node, the depth-first traversal algorithm recursively visits adjacent nodes along unvisited edges. For example, a stack structure is used to record the access path until all connected nodes are traversed. Each traversed edge is weighted according to its line resistance parameter. Finally, a carbon emission adjacency matrix describing the adjacency relationship of nodes in the carbon emission graph structure is generated based on the connection status between all nodes and the edge weights.

[0085] The carbon emission adjacency matrix is ​​aggregated to obtain the losses and impacts generated when adjacent edges interact with other nodes, and the carbon potential of the nodes is obtained through power flow analysis.

[0086] The specific expression for the carbon potential value of the generated node is:

[0087] ;

[0088] in, Represents a node The carbon potential value; Represents a node The set of adjacent nodes; Represents a node and nodes The edge weights (typically ranging from 0 to 1); Represents a node Dynamic carbon emission factors; Represents a node and nodes The reciprocal of the resistance parameter in the circuit; Represents a node and nodes The power flow in the circuit; This represents the carbon potential diffusion coefficient (typically ranging from 0.4 to 0.6). This represents the carbon potential Laplace operator.

[0089] It should be noted that the power flow is obtained from the parameter dimensions of the three-dimensional data cube; the set of adjacent nodes is obtained from the carbon emission adjacency matrix.

[0090] Furthermore, the carbon emission adjacency matrix is ​​aggregated. First, the weights of the direct adjacent edges of each node are calculated to represent the losses and impacts generated when adjacent edges interact with other nodes. Next, the impact of each node on carbon emissions when it acts as a power source or load point is evaluated using power flow analysis, and the carbon potential of the node is represented by a quantitative value.

[0091] S5. Based on the carbon potential value, the node carbon potential weights are aggregated in layers according to voltage level to generate a time series carbon potential curve. The total carbon emissions are obtained through the dynamic carbon flow density accumulation method. At the same time, the carbon flow propagation direction is determined based on the power grid topology and the total tracking amount is obtained.

[0092] The voltage level is mapped to the carbon potential value of the node through a topology analysis algorithm, and the carbon potential weight of the node is aggregated hierarchically according to the voltage level to generate the overall carbon potential of the voltage level.

[0093] Furthermore, the voltage level of each node in the power grid is identified through topology analysis algorithms. For example, the nodes are divided into different levels such as 500kV, 220kV, 110kV and 35kV, and the voltage level is used as a label to map to the carbon potential value of the corresponding node. Then, all nodes are grouped according to voltage level, the carbon potential value of the nodes in each group is assigned a weight, and the carbon potential value of all nodes in the same voltage level is weighted and summed to finally generate the overall carbon potential of the voltage level.

[0094] It should be noted that voltage level refers to the nominal voltage value used in a power system to distinguish different transmission and distribution levels, such as 500 kV, 220 kV, 110 kV, etc., which is used to characterize the voltage level and transmission capacity of each node or line in the power grid.

[0095] The overall carbon potential of the voltage level is fused with the time of a fixed interval to generate carbon potential data of the voltage level nodes as a function of time, and a time series carbon potential curve is generated by the moving average method.

[0096] Furthermore, the overall carbon potential of each voltage level is correlated with its corresponding timestamp. For example, with each 15-minute time interval, the overall carbon potential values ​​of each voltage level, such as 500kV and 220kV, are recorded at times such as 00:00, 00:15, and 00:30, forming a data sequence of carbon potential values ​​of voltage level nodes arranged in chronological order over time. Subsequently, the moving average method is applied to the carbon potential value data sequence of each voltage level, calculating the moving average value using data from multiple consecutive time points. For example, an 8-time window is used to calculate the average of every 8 consecutive carbon potential values, and the calculation is performed sequentially to eliminate the influence of short-term fluctuations. Finally, a smooth time series carbon potential curve of the voltage level is generated through curve fitting.

[0097] It should be noted that the fixed interval refers to the time unit divided for analyzing the dynamic changes of carbon potential at voltage levels, and is set according to the characteristics of power grid operation.

[0098] Based on the smoothed voltage level time series carbon potential curve, a carbon flux density index is defined. The cumulative carbon flux density value at each load point is calculated step by step using the dynamic carbon flux density accumulation method to generate the total carbon emissions.

[0099] Furthermore, based on the time-series carbon potential curve and the grid topology, starting from the power source point, the process proceeds radially along the transmission line to each load point. The carbon flux density value on each edge is integrated or accumulated over time to generate a carbon flux density value. The carbon flux density value at the current moment is then accumulated with the time interval to generate a cumulative carbon flux density value. Based on the cumulative carbon flux density value, a weighting method is used to assign weights, and the cumulative carbon flux density values ​​of all load points are weighted and summed according to their weights or electricity consumption ratios to ultimately generate the total carbon emissions.

[0100] Based on the power grid topology, the propagation direction of carbon flow from power generation to various consumption points is tracked using the power flow tracing method. The carbon emissions corresponding to each propagation direction are calculated, and the data are aggregated and analyzed simultaneously to obtain the total amount of carbon emissions tracked.

[0101] Furthermore, based on the power grid topology, the power flow tracing method is used to trace the power flow path downstream from each power generation node according to the actual power distribution ratio in the power network. This tracing method clarifies the direction of carbon flow propagation from the power generation node to each consumption point. For example, based on the distribution relationship between node injected power and branch power flow, it is determined that 70% of the power at a certain load point comes from power plant A and 30% from power plant B. Based on the amount of power transmitted in each propagation direction and its corresponding dynamic carbon emission factor, the carbon emission amount carried by each propagation path is calculated using the carbon flow allocation method. For example, if power plant A transmits 500 MWh of electricity to the load point, with a corresponding carbon emission factor of 0.8 kgCO2 / kWh, then the carbon emission amount of the propagation path is 400 tons. After calculating the carbon emission amounts from all power generation nodes to all consumption points according to the propagation path, the results are simultaneously summed and analyzed to obtain the total traced amount.

[0102] It should be noted that the actual power distribution ratio in a power network refers to the actual distribution ratio of electrical energy provided by each power source along different transmission paths according to the power flow distribution law during the operation of the power grid, reflecting the composition of the power source at each load point or branch.

[0103] S6. Calculate the relative deviation between the total tracking amount and the total carbon emissions, compare and analyze the relative deviation with the historical carbon emission error threshold, and generate a carbon emission measurement report.

[0104] The total amount tracked is compared with the total amount of carbon emissions, the relative deviation between the two is calculated, and the relative deviation is compared and analyzed with the historical carbon emission error threshold to verify the rationality of carbon emissions.

[0105] Furthermore, the total tracked amount is numerically compared with the total carbon emissions generated by the dynamic carbon flow density accumulation method. The ratio of the difference between the two to the total carbon emissions is calculated to obtain the relative deviation. The historical carbon emission error threshold (usually no more than ±3%) is compared with the calculated relative deviation. If the relative deviation is within the historical carbon emission error threshold, the carbon emission calculation result is deemed reasonable. If it exceeds the historical carbon emission error threshold, it is marked as abnormal. Finally, the rationality of carbon emissions is verified.

[0106] Based on the rationality results of carbon emissions, identify abnormal areas and generate a carbon emission metering report for the power system.

[0107] Furthermore, based on the reasonableness results of carbon emissions, if the relative deviation exceeds the historical carbon emission error threshold, an anomaly detection mechanism is activated. Combining the power grid topology and time-series carbon potential curves, the source region of the deviation is located. For example, if a 220kV power supply area is identified as having carbon potential values ​​that significantly deviate from the historical range at multiple consecutive time points during peak electricity consumption, and the local deviation between its tracking total and total carbon emissions exceeds 5%, the power generation structure, load changes, and line transmission conditions in the deviation source region are further analyzed to confirm the cause of the anomaly. For example, a sudden drop in renewable energy output may lead to an increase in thermal power supplementation. Based on the total carbon emissions, tracking total, relative deviation, carbon emission error threshold, location of the anomaly area, time of occurrence, and possible causes, a power system carbon emission metering report is generated.

[0108] This embodiment also provides a computer device applicable to the power system carbon emission measurement method based on dynamic electrical carbon emission factors, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the power system carbon emission measurement method based on dynamic electrical carbon emission factors as proposed in the above embodiment.

[0109] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0110] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the power system carbon emission measurement method based on dynamic electrical carbon emission factors as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0111] In summary, this invention achieves collaborative modeling of spatial dependence and temporal dynamics in multi-source heterogeneous data of the power grid by inputting a three-dimensional real-time operating data cube into a neural network model and predicting dynamic carbon emission factors through a dual-pathway spatiotemporal feature model. This effectively captures the instantaneous change characteristics of carbon emissions under complex operating conditions and improves prediction accuracy and timeliness.

[0112] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for measuring carbon emissions from a power system based on a dynamic electrical carbon emission factor, characterized in that: include, Collect real-time operating data of the power system and preprocess it to generate a three-dimensional real-time operating data cube; The three-dimensional real-time running data cube is input into the neural network model to predict dynamic carbon emission factors through a dual-pathway of spatiotemporal features. Based on historical power grid equipment status data, the power grid topology is obtained, and a dynamic carbon emission graph structure is constructed through a dynamic graph convolutional network. The dynamic carbon emission factor is defined as a node in the dynamic carbon emission graph structure, and aggregated calculations are performed using historical line resistance parameters to obtain the carbon potential value of the node. Based on the carbon potential value, the node carbon potential weights are aggregated in layers according to voltage level to generate a time series carbon potential curve. The total carbon emissions are obtained through the dynamic carbon flow density accumulation method. At the same time, the carbon flow propagation direction is determined based on the power grid topology and the total tracking amount is obtained. The relative deviation between the total tracked amount and the total carbon emissions is calculated, and the relative deviation is compared and analyzed with historical carbon emission error thresholds to generate a carbon emission measurement report.

2. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 1, characterized in that: The real-time operation data of the power system includes dynamic data on the generation side, spatiotemporal data on the load side, and environmental coupling data. The preprocessing includes data cleaning, time alignment, normalization, feature extraction, and quality verification.

3. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 2, characterized in that: The generation of the three-dimensional real-time operation data cube refers to the process of integrating and organizing the pre-processed real-time operation data of the power system according to spatial, temporal, and parameter dimensions to generate a three-dimensional real-time operation data cube.

4. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 3, characterized in that: The specific steps for inputting a three-dimensional real-time running data cube into a neural network model and predicting dynamic carbon emission factors through a dual-pathway spatiotemporal feature analysis are as follows. The three-dimensional real-time running data cube is time-sliding segmented according to a fixed time window to generate spatiotemporal data of power grid nodes; Spatiotemporal data are input into a neural network model to capture the spatial characteristics of the interaction between power grid nodes through a spatial feature path, while the temporal feature path is used to obtain the temporal data sequence of each node as time changes. Spatial features are fused with temporal data sequences to generate spatial-temporal fused features, which are then encoded and mapped to generate spatial-temporal fused feature vectors. Based on the spatial-temporal fusion feature vector, the evolution law of carbon potential is captured by the spatiotemporal feature correlation analysis method, and dynamic analysis is performed to predict dynamic carbon emission factors.

5. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 4, characterized in that: The process of obtaining the power grid topology relationship based on historical power grid equipment status data refers to extracting the physical connection relationship of the power grid through time series analysis based on historical power grid equipment status data, and then fusing it with the time under a reference clock to obtain the power grid topology relationship.

6. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 5, characterized in that: The specific steps for constructing a dynamic carbon emission graph structure using a dynamic graph convolutional network are as follows. Based on the power grid topology, a mapping operation is performed on the connection relationships between nodes to generate the association strength between nodes; Based on the strength of the association between nodes, a dynamic carbon emission graph structure is constructed using a dynamic graph convolutional network.

7. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 6, characterized in that: The process of defining dynamic carbon emission factors as nodes in a dynamic carbon emission graph structure and performing aggregation calculations based on historical line resistance parameters to obtain the carbon potential value of each node is as follows. Based on the dynamic carbon emission factor, calculate the carbon emission data of each power grid node at each time point and define it as a node in the dynamic carbon emission graph structure. The historical line resistance parameters are defined as edges of the carbon emission graph structure, and a traversal operation is performed to generate a carbon emission adjacency matrix. The carbon emission adjacency matrix is ​​aggregated to obtain the losses and impacts generated when adjacent edges interact with other nodes, and the carbon potential of the nodes is obtained through power flow analysis.

8. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 7, characterized in that: The process involves aggregating node carbon potential weights according to voltage levels based on carbon potential values, generating time-series carbon potential curves, and obtaining total carbon emissions using a dynamic carbon flux density accumulation method. The specific steps are as follows: The voltage level is mapped to the carbon potential value of the node, and the carbon potential weight of the node is aggregated hierarchically according to the voltage level to generate the overall carbon potential of the voltage level. The overall carbon potential of the voltage level is fused with the time of a fixed interval to generate carbon potential data of the voltage level nodes as a function of time, and a time series carbon potential curve is generated by the moving average method. Based on the time-series carbon potential curve, a carbon flux density index is defined, and the cumulative carbon flux density value at each load point is calculated step by step using the dynamic carbon flux density accumulation method to generate the total carbon emissions.

9. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 8, characterized in that: The process of determining the carbon flow propagation direction and obtaining the total tracking amount based on the power grid topology refers to tracing the propagation direction of carbon flow from power generation sources to various consumption points using the power flow tracing method, calculating the carbon emissions corresponding to each propagation direction, and simultaneously summarizing and analyzing the data to obtain the total tracking amount.

10. The power system carbon emission measurement method based on dynamic electrical carbon emission factor as described in claim 9, characterized in that: The calculation of the relative deviation between the total tracking amount and the total carbon emissions, the comparison and analysis of the relative deviation with historical carbon emission error thresholds, and the generation of a carbon emission measurement report are as follows. The total amount tracked is compared with the total amount of carbon emissions, the relative deviation between the two is calculated, and the relative deviation is compared and analyzed with the historical carbon emission error threshold to verify the rationality of carbon emissions. Based on the rationality results of carbon emissions, identify abnormal areas and generate a carbon emission metering report for the power system.

Images