Carbon energy flow tracing optimization method and system

Abstract

The invention relates to the technical field of carbon emission management, in particular to a carbon energy flow tracing optimization method and system, and the method comprises the steps: constructing a carbon emission network topological structure according to an original carbon emission data set, and generating a carbon emission tracing path through a Markov chain model; a kernel density estimation method based on the Bayesian theory is used to obtain a carbon emission confidence interval of the traceability path node; according to the carbon emission confidence interval, performing different space-time granularity division on the original carbon emission data set through a quadtree structure to obtain carbon emission space-time granularity candidate sets of different space-time granularities; carrying out optimization analysis on the carbon emission space-time granularity candidate set by utilizing a genetic algorithm, and determining optimal fine-granularity carbon emission data; and according to the optimal fine-grained carbon emission data, a carbon emission traceability model is utilized to obtain a carbon energy flow traceability tracking result. According to the invention, accurate carbon energy flow traceability tracking is realized through the Markov chain model, the genetic algorithm and other technologies, the carbon emission traceability precision and traceability efficiency are improved, and the reliability of the traceability result is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission management, and in particular to a carbon energy flow tracing optimization method and system. Background Art

[0002] As the global climate change problem becomes increasingly serious, the monitoring and management of carbon emissions has become an important topic in scientific research and engineering practice. As a key tool for achieving refined management of carbon emissions, the core task of the carbon emission traceability system is to track the source and destination of carbon emissions by constructing a carbon emission network topology, thereby providing a scientific basis for the formulation of emission reduction strategies. However, the current carbon energy flow traceability system faces many technical challenges in practical applications, especially when processing large-scale carbon emission data with different spatiotemporal granularities. These data come from emission sources of various energy types and industries, and contain key information such as geographical location, energy consumption, and emission factors. With the continuous growth of data scale, the complexity of the traceability algorithm has also increased sharply. Traditional data processing methods perform poorly when processing such a large data set. There is a complex trade-off between data granularity and traceability accuracy, which leads to low traceability efficiency and difficulty in accurately tracking the flow of carbon. Therefore, how to implement an adaptive traceability strategy and ensure the reliability of traceability results under the premise of ensuring traceability accuracy is a key issue that needs to be urgently solved in the current research on carbon emission traceability systems. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a carbon energy flow tracing optimization method and system.

[0004] In a first aspect, the present invention provides a method for optimizing carbon energy flow tracing, the method comprising the following steps: Collecting an original carbon emission data set of a target energy type, and constructing a carbon emission network topology structure based on the original carbon emission data set; Based on the carbon emission network topology, the energy transmission relationship between different emission source nodes is modeled using the Markov chain model to generate a carbon emission traceability path; According to the energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path, the kernel density estimation method based on Bayesian theory is used to estimate the confidence interval of carbon emissions at the traceability path nodes. According to the carbon emission confidence interval, the original carbon emission data set is divided into different spatiotemporal granularities through a quadtree structure to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities; Using a genetic algorithm to optimize and analyze the carbon emission spatiotemporal granularity candidate set, find the optimal balance between data granularity and traceability accuracy, and determine the optimal fine-grained carbon emission data; Based on the optimal fine-grained carbon emission data, the carbon emission traceability model is used to obtain the carbon energy flow traceability results.

[0005] In a further embodiment, the step of constructing a carbon emission network topology structure according to the carbon emission data set comprises: Obtaining energy flow information of each emission source from the original carbon emission data set, and determining the energy transmission relationship between different emission sources based on the energy flow information of each emission source; the energy transmission relationship includes the energy flow direction and energy flow intensity between different emission sources; The emission sources in the original carbon emission data set are defined as emission source nodes, the energy transmission relationship is defined as edges, and the carbon emission network topology structure is constructed using a directed graph.

[0006] In a further embodiment, the step of modeling the energy transmission relationship between different emission source nodes based on the carbon emission network topology using a Markov chain model to generate a carbon emission traceability path includes: All emission source nodes in the carbon emission network topology structure are traversed using a graph traversal method, and each traversed emission source node is defined as a single state in a Markov chain model to construct a state space; Based on the energy transmission data between each emission source node, the transfer probability from each emission source node to other emission source nodes is calculated to form a transfer probability matrix; Determine the initial state probability of each emission source node based on energy consumption data and obtain the initial state probability distribution; According to the initial state probability distribution and the transition probability matrix, the Markov chain process is simulated by a random walk algorithm to perform multiple state transitions to obtain the potential traceability path of each emission source node; The cumulative probability of each potential tracing path is calculated according to the transfer probability matrix, and all potential tracing paths are sorted according to the cumulative probability, and the potential tracing path with the highest cumulative probability is taken as the carbon emission tracing path.

[0007] In a further embodiment, the sum of the initial state probabilities of all emission source nodes is one.

[0008] In a further implementation scheme, the step of estimating the confidence interval of carbon emissions of the nodes on the carbon emission tracing path using the kernel density estimation method based on Bayesian theory according to the energy consumption data and carbon emission factors of each emission source node on the carbon emission tracing path includes: Obtaining energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path; the energy consumption data includes energy consumption type, energy consumption amount and energy consumption time; Calculate the carbon emissions of each emission source node on the carbon emission traceability path based on the energy consumption data and the carbon emission factor; According to the carbon emissions of each emission source node on the carbon emission traceability path, the probability density function of the carbon emissions of each emission source node is established using the kernel density estimation method; According to the prior distribution and probability density function of carbon emissions, the posterior distribution of carbon emissions is calculated by Bayesian theorem; The Markov Monte Carlo method is used to sample the posterior distribution to obtain the carbon emission posterior samples, and the confidence interval of carbon emissions for each traceability path node is calculated based on the carbon emission posterior samples.

[0009] In a further embodiment, the step of dividing the original carbon emission data set into different spatiotemporal granularities by a quadtree structure according to the carbon emission confidence interval to obtain a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities includes: The original carbon emission data set is decomposed into multiple scales using discrete wavelet transform to generate carbon emission temporal characteristics at different time scales. According to the longitude and latitude coordinates of the boundaries of the area to be traced, the area to be traced is spatially recursively divided using a quadtree structure to obtain a number of spatial sub-areas with different spatial granularities; After all spatial sub-regions are divided, the temporal characteristics of carbon emissions in each spatial sub-region are extracted to obtain the spatiotemporal data series of carbon emissions at different spatiotemporal granularities; Inputting the carbon emission confidence interval into a convolutional neural network, and using the convolutional neural network to perform traceability calculation on the carbon emission spatiotemporal data sequence to obtain traceability confidence; According to the traceability confidence, a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities is screened out from the carbon emission spatiotemporal data sequence.

[0010] In a further embodiment, the step of recursively dividing the area to be traced by using a quadtree structure to obtain a plurality of spatial sub-areas of different spatial granularities further includes: Traversing all the spatial sub-regions, calculating the weighted average of the carbon emission confidence intervals of all traceability path nodes in the current spatial sub-region, and obtaining the spatial distribution characteristics of the confidence intervals; If the spatial distribution characteristics of the confidence interval are greater than the preset confidence threshold, the quadtree structure is continued to be used to recursively divide the current spatial sub-region into spaces until the weighted average values ​​of the carbon emission confidence intervals of all spatial sub-regions are lower than the preset confidence threshold or the preset maximum division depth is reached.

[0011] In a further embodiment, the step of optimizing and analyzing the carbon emission spatiotemporal granularity candidate set using a genetic algorithm to find the optimal balance between data granularity and traceability accuracy and determining the optimal fine-grained carbon emission data includes: Simulate the candidate sets of carbon emission spatiotemporal granularity at different spatiotemporal granularity to evaluate the corresponding traceability accuracy and computational efficiency, where the computational efficiency includes the time and resource consumption required for the simulation; From the plurality of carbon emission spatiotemporal granularity candidate sets, a carbon emission spatiotemporal granularity candidate set having a computational efficiency within a target computational efficiency range is selected as a primary spatiotemporal granularity; Using an information entropy algorithm to measure the spatiotemporal fine granularity of the preliminary spatiotemporal granularity to obtain data fine granularity; Calculate the weighted average of the data granularity and the traceability accuracy to obtain a balance index of the preliminary spatiotemporal granularity; Encoding each group of preliminary spatiotemporal granularity as an individual in a genetic algorithm, and iteratively optimizing the individuals using the genetic algorithm; In each iterative optimization, the fitness of the individual is calculated according to the balance index, and the population is continuously iterated and optimized through multiple generations of genetic operations until the optimal balance point between data granularity and traceability accuracy is found; The individual with the highest fitness is selected as the optimal solution, and the optimal fine-grained carbon emission data is obtained based on the optimal solution.

[0012] In a further embodiment, the process of constructing the carbon emission tracing model includes: Extract the basic features of carbon emissions from the optimal fine-grained carbon emission data, and use the maximum information coefficient to calculate the feature correlation strength value between the basic features of carbon emissions and the traceability accuracy; According to the feature association strength value, a recursive feature elimination method is used to perform feature selection to screen out the retained basic features of carbon emissions; Based on the retained basic characteristics of carbon emissions, a carbon emission tracing model is established using support vector machines.

[0013] In a second aspect, the present invention provides a carbon energy flow tracing optimization system, the system comprising: A data collection module, used to collect the original carbon emission data set of the target energy type and construct a carbon emission network topology structure based on the original carbon emission data set; A path analysis module is used to model the energy transmission relationship between different emission source nodes based on the carbon emission network topology structure using a Markov chain model to generate a carbon emission tracing path; The confidence analysis module is used to estimate the confidence interval of carbon emissions at the nodes on the carbon emission traceability path based on the energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path using the kernel density estimation method based on Bayesian theory; A granularity division module is used to divide the original carbon emission data set into different spatiotemporal granularities through a quadtree structure according to the carbon emission confidence interval, so as to obtain a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities; An iterative optimization module is used to optimize and analyze the carbon emission spatiotemporal granularity candidate set using a genetic algorithm, find the optimal balance between data granularity and traceability accuracy, and determine the optimal fine-grained carbon emission data; The traceability module is used to obtain the carbon energy flow traceability results based on the optimal fine-grained carbon emission data and the carbon emission traceability model.

[0014] The present invention provides a carbon energy flow tracing optimization method and system, the method collects the original carbon emission data set of the target energy type, and constructs a carbon emission network topology structure according to the original carbon emission data set; based on the carbon emission network topology structure, the energy transmission relationship between different emission source nodes is modeled by using a Markov chain model to generate a carbon emission tracing path; according to the energy consumption data and carbon emission factors of each emission source node on the carbon emission tracing path, a kernel density estimation method based on Bayesian theory is used to estimate the carbon emission confidence interval of the tracing path node; according to the carbon emission confidence interval, the original carbon emission data set is divided into different spatiotemporal granularities through a quadtree structure to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities; the carbon emission spatiotemporal granularity candidate sets are optimized and analyzed by using a genetic algorithm to find the optimal balance between data granularity and tracing accuracy, and determine the optimal fine-grained carbon emission data; according to the optimal fine-grained carbon emission data, a carbon emission tracing model is used to obtain a carbon energy flow tracing result. Compared with existing technologies, this method combines Markov chain models and genetic algorithms to not only achieve accurate carbon energy flow tracing, but also realizes a trade-off between data granularity and traceability accuracy. It can effectively cope with the growth of data scale, significantly improve the carbon emission traceability accuracy and efficiency, ensure the reliability of traceability results, and provide a scientific basis for emission reduction strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the process flow of the carbon energy flow tracing optimization method provided by an embodiment of the present invention; Figure 2 It is a block diagram of a carbon energy flow tracing optimization system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The following specifically illustrates the implementation mode of the present invention in conjunction with the accompanying drawings. The embodiments are provided for illustrative purposes only and cannot be understood as limiting the present invention. The accompanying drawings are provided 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.

[0017] refer to Figure 1 , the embodiment of the present invention provides a carbon energy flow tracing optimization method, such as Figure 1 As shown, the method comprises the following steps: S1. Collecting an original carbon emission data set of a target energy type, and constructing a carbon emission network topology structure based on the original carbon emission data set.

[0018] In this embodiment, the step of constructing a carbon emission network topology structure according to the carbon emission data set includes: Obtaining energy flow information of each emission source from the original carbon emission data set, and determining the energy transmission relationship between different emission sources based on the energy flow information of each emission source; the energy transmission relationship includes the energy flow direction and energy flow intensity between different emission sources; The emission sources in the original carbon emission data set are defined as emission source nodes, the energy transmission relationship is defined as edges, and the carbon emission network topology structure is constructed using a directed graph.

[0019] Specifically, this embodiment can obtain the original carbon emission data set of the target energy type from the carbon emission database. The original carbon emission data set may include information such as carbon emissions, energy types, and energy consumption at different time points. The original carbon emission data set is cleaned to remove outliers and fill in missing data to ensure the integrity and accuracy of the data. The energy flow information of each emission source is extracted from the cleaned original carbon emission data set. The energy flow information includes information such as energy type, consumption, source and destination. The energy flow direction between different emission sources is determined based on the extracted energy flow information. For example, if a power plant supplies electricity to a factory, there is an energy flow direction from the power plant to the factory. Based on data such as the energy transmission amount in the energy flow information, the energy flow intensity between different emission sources is calculated, and the energy flow direction and energy flow intensity between different emission sources are integrated into an energy transmission relationship.

[0020] This embodiment defines each emission source in the original carbon emission data set as a node, namely, an emission source node. These nodes represent entities in the network, and define energy transmission relationships as edges, namely, paths connecting different emission source nodes. A weight is assigned to each edge according to the energy flow intensity to represent the amount of energy transmission on the path. The emission source nodes and edges are combined using the directed graph structure in graph theory to construct a carbon emission network topology. In this directed graph, each node represents an emission source, each directed edge represents an energy transmission relationship, and the weight of the edge represents the energy flow intensity.

[0021] S2. Based on the carbon emission network topology structure, the Markov chain model is used to model the energy transmission relationship between different emission source nodes to generate a carbon emission tracing path.

[0022] In this embodiment, based on the carbon emission network topology structure, the energy transmission relationship between different emission source nodes is modeled using a Markov chain model to generate a carbon emission tracing path, including: All emission source nodes in the carbon emission network topology structure are traversed using a graph traversal method, and each traversed emission source node is defined as a single state in a Markov chain model to construct a state space; Based on the energy transmission data between each emission source node, the transfer probability from each emission source node to other emission source nodes is calculated to form a transfer probability matrix; Determine the initial state probability of each emission source node based on energy consumption data and obtain the initial state probability distribution; According to the initial state probability distribution and the transition probability matrix, the Markov chain process is simulated by a random walk algorithm to perform multiple state transitions to obtain the potential traceability path of each emission source node; The cumulative probability of each potential tracing path is calculated according to the transfer probability matrix, and all potential tracing paths are sorted according to the cumulative probability, and the potential tracing path with the highest cumulative probability is taken as the carbon emission tracing path.

[0023] Specifically, this embodiment can use a graph traversal algorithm such as a depth-first search algorithm (DFS) or a breadth-first search algorithm (BFS) to traverse each node in the carbon emission network topology structure. The depth-first search algorithm (DFS) or the breadth-first search algorithm (BFS) can ensure that each node is visited once and only once. Then, this embodiment defines each emission source node traversed as a state in the Markov chain model. For example, if there are N emission source nodes in the network, the Markov chain model will have N states, each state corresponds to an emission source node, and a state space is constructed. The state space is a collection of all emission source nodes, and energy transmission data between emission source nodes is collected. The energy transmission data includes information such as energy transmission amount and transmission frequency. For each emission source node (state i), the probability of its transfer to other emission source nodes (state j) is calculated. This probability can be estimated by analyzing the proportion of energy transmission from state i to state j in the total transmission amount within a preset time window to form a transfer probability matrix P, where, represents the probability of transitioning from state i to state j. It should be noted that for any state i, the sum of all possible transition probabilities should be 1.

[0024] Next, this embodiment collects energy consumption data of each emission source node. The energy consumption data includes information such as the energy consumption amount and energy consumption type of each emission source node. Based on the energy consumption data, an initial state probability is assigned to each emission source node (state). This probability can reflect the degree of activity or importance of each node at the initial moment. For example, proportional distribution is performed according to the size of energy consumption to ensure that the sum of the initial state probabilities of all emission source nodes is one. This embodiment initializes the current state distribution to the initial state probability distribution. According to the current state distribution and the transition probability matrix P, a state is randomly selected as the next state. This process is repeated multiple times to simulate multiple state transitions of the Markov chain and record the path of each state transition. These paths are potential traceability paths. For each potential traceability path In this embodiment, the cumulative probability is calculated through the transfer probability matrix P. Specifically, the transfer probability of each step on the potential traceability path is multiplied to calculate the cumulative probability, so as to sort all potential traceability paths according to the cumulative probability. Among them, the potential traceability path with higher cumulative probability indicates that the probability of its energy transmission is greater, and the probability that it belongs to the real traceability path of carbon emissions is greater. Therefore, in this embodiment, the potential traceability path with the highest cumulative probability is used as the final carbon emission traceability path. The final carbon emission traceability path represents the path that is most likely to reach the current carbon emission state after a series of state transfers starting from the initial state. In this embodiment, by constructing a carbon emission network topology structure and using the Markov chain model to model the energy transmission relationship, it can efficiently track the source and destination of carbon emissions and significantly improve the traceability accuracy.

[0025] S3. According to the energy consumption data and carbon emission factors of each emission source node on the carbon emission tracing path, the confidence interval of carbon emissions at the tracing path nodes is estimated using the kernel density estimation method based on Bayesian theory.

[0026] In this embodiment, the step of estimating the confidence interval of carbon emissions of the nodes on the carbon emission tracing path using the kernel density estimation method based on Bayesian theory according to the energy consumption data and carbon emission factors of each emission source node on the carbon emission tracing path includes: Obtaining energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path; the energy consumption data includes energy consumption type, energy consumption amount and energy consumption time; Calculate the carbon emissions of each emission source node on the carbon emission traceability path based on the energy consumption data and the carbon emission factor; According to the carbon emissions of each emission source node on the carbon emission traceability path, the probability density function of the carbon emissions of each emission source node is established using the kernel density estimation method; According to the prior distribution and probability density function of carbon emissions, the posterior distribution of carbon emissions is calculated by Bayesian theorem; The Markov Monte Carlo method is used to sample the posterior distribution to obtain the carbon emission posterior samples, and the confidence interval of carbon emissions for each traceability path node is calculated based on the carbon emission posterior samples.

[0027] Specifically, this embodiment collects the energy consumption type, energy consumption and energy consumption time of each emission source node to form energy consumption data. For each emission source node, this embodiment searches for the corresponding carbon emission factor according to its energy consumption type. The carbon emission factor is the carbon emissions generated by unit energy consumption. For each emission source node, this embodiment multiplies the energy consumption by the corresponding carbon emission factor to obtain the carbon emissions of the emission source node. The above steps are repeated to calculate the carbon emissions of all emission source nodes. According to the calculated carbon emission data, the probability density of the carbon emissions of each emission source node is estimated by the kernel density estimation method (KDE) to obtain its probability density function. KDE is a non-parametric estimation method used to estimate the probability density function of a random variable. The kernel density estimation method estimates the distribution of the overall data by placing a kernel function around each data point and superimposing these kernel functions. This embodiment selects a Gaussian kernel function as the kernel function and estimates the density function by placing the Gaussian kernel function on each data point.

[0028] This embodiment assumes a prior distribution of carbon emissions (such as a normal distribution) based on historical data, and calculates the posterior distribution of carbon emissions based on the Bayesian theorem, combining the prior distribution and the probability density function (likelihood function). Then, the posterior distribution is sampled using the Markov Chain Monte Carlo (MCMC) method (such as the Metropolis-Hastings algorithm or Gibbs sampling) to obtain a large number of carbon emission posterior samples. The confidence interval (such as a 95% confidence interval) of the carbon emissions of each traceability path node is estimated based on statistics such as the central tendency (such as the mean) and variability (such as the standard deviation) of the posterior samples. For example, the posterior distribution of node A is sampled 10,000 times using the MCMC method to obtain 10,000 carbon emission samples. The 95% confidence interval of the carbon emissions of node A can be obtained by the sample mean ±1.96 times the standard deviation. To approximate, this embodiment is based on Bayesian theory and kernel density estimation method, combining prior information and observation data to estimate the confidence interval of carbon emissions of each emission source node on the carbon emission tracing path, which can more comprehensively reflect the uncertainty of carbon emissions. In summary, with the continuous growth of data scale, the complexity of traditional tracing algorithms has also increased sharply, resulting in low tracing efficiency. This embodiment reduces the complexity of the tracing algorithm and improves the tracing efficiency by adopting Markov chain model to model the energy transmission relationship and using the kernel density estimation method of Bayesian theory to estimate the confidence interval of carbon emissions.

[0029] S4. According to the carbon emission confidence interval, the original carbon emission data set is divided into different spatiotemporal granularities through a quadtree structure to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities.

[0030] In this embodiment, the step of dividing the original carbon emission data set into different spatiotemporal granularities by a quadtree structure according to the carbon emission confidence interval to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities includes: The original carbon emission data set is decomposed into multiple scales using discrete wavelet transform to generate carbon emission temporal characteristics at different time scales. According to the longitude and latitude coordinates of the boundaries of the area to be traced, the area to be traced is spatially recursively divided using a quadtree structure to obtain a number of spatial sub-areas with different spatial granularities; After all spatial sub-regions are divided, the temporal characteristics of carbon emissions in each spatial sub-region are extracted to obtain the spatiotemporal data series of carbon emissions at different spatiotemporal granularities; Inputting the carbon emission confidence interval into a convolutional neural network, and using the convolutional neural network to perform traceability calculation on the carbon emission spatiotemporal data sequence to obtain traceability confidence; According to the traceability confidence, a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities is screened out from the carbon emission spatiotemporal data sequence.

[0031] Specifically, this embodiment cleans and preprocesses the original carbon emission data set, including removing outliers, filling missing values, etc., to obtain a preprocessed carbon emission data set to ensure the quality of the data. According to the characteristics of the preprocessed carbon emission data set, a wavelet base such as Daubechies wavelet or Haar wavelet is selected, and then a discrete wavelet transform is performed on the preprocessed carbon emission data set according to the wavelet base to decompose it into low-frequency components and high-frequency components. The low-frequency components represent the long-term trend of carbon emissions, and the high-frequency components represent the short-term fluctuations of carbon emissions. The low-frequency and high-frequency components of the time scale are reconstructed into carbon emission time characteristics of different time scales. For example, this embodiment can reconstruct the characteristics of different time scales such as day, week, month, season, year, etc. At the same time, this embodiment extracts the characteristics of different time scales from geographic information. The boundary longitude and latitude coordinates of the area to be traced are obtained from the geographic information system (GIS). According to the specific boundary longitude and latitude coordinates of the area to be traced, the coverage range of the root node of the quadtree is determined. The area to be traced is taken as the root node of the quadtree. According to the set spatial granularity threshold (such as area size, grid size, etc.), the root node is recursively divided into four equal parts using the quadtree division rule. Each division divides the current node into four child nodes, and each child node represents a smaller spatial sub-region. When the division reaches the set minimum spatial granularity or meets other termination conditions (such as the number of sub-regions, division depth, etc.), the recursive division is stopped. Several spatial sub-regions with different spatial granularities are generated through recursive division, and each spatial sub-region corresponds to a leaf node in the quadtree.

[0032] After all the spatial sub-regions are divided, the corresponding carbon emission time features are extracted from the carbon emission time features at different time scales according to the position of each spatial sub-region in the quadtree, and the time features and spatial position information of each spatial sub-region are combined into a carbon emission spatiotemporal data sequence. Each carbon emission spatiotemporal data sequence contains the carbon emission data of a specific spatial sub-region at different time scales. The carbon emission spatiotemporal data sequence is standardized to improve the training efficiency and performance of the convolutional neural network. The standardized spatiotemporal data sequence and carbon emission confidence interval are converted into an input matrix form acceptable to the convolutional neural network. The convolutional neural network structure includes the following steps: The connected input layer, multiple convolutional layers, pooling layers and fully connected layers, this embodiment uses a training data set of known carbon emission sources and corresponding confidence levels to train the convolutional neural network to ensure that the network can learn the relationship between the confidence interval and the carbon emissions. During the training process, the network is trained by optimization methods such as back propagation algorithm and gradient descent. At the same time, the influence of the confidence interval on the weight update is considered, and the network parameters are adjusted to minimize the prediction error, so that the network can accurately predict the traceability confidence of the carbon emission source. The spatiotemporal data sequence of carbon emissions to be traced and the confidence interval of carbon emissions are input into the trained convolutional neural network, and the network outputs the traceability confidence of each candidate carbon emission source.

[0033] It should be noted that the carbon emission confidence interval provides additional information about the uncertainty of carbon emission data. In this embodiment, the carbon emissions and their confidence intervals are used as inputs to the network, and the confidence intervals are integrated into the network as additional input features. In the weight update process of the convolution layer, the carbon emission confidence interval is used as a weight adjustment factor of the convolution layer to affect the weight update of the convolution kernel. If a carbon emission confidence interval is wide, it indicates that the uncertainty of the data point is high. The convolutional neural network will consider the influence of the confidence interval in the loss function and reduce the weight of the data point in the traceability calculation to reduce the impact of its uncertainty on the final result. For example, this embodiment adopts a weighted loss function, in which the weight is inversely proportional to the width of the confidence interval. In this way, data points with narrower confidence intervals can be given higher attention during the training process. The weighted loss function can be expressed as: In the formula, is the weighted loss function; is the true value; is the predicted value; is the width of the confidence interval for carbon emissions at the mth data point.

[0034] This embodiment introduces a carbon emission confidence interval so that the convolutional neural network can simultaneously consider carbon emissions and their uncertainty when extracting spatiotemporal features, so that the convolutional neural network can use this information to more accurately evaluate the possibility of different carbon emission sources. This method not only captures the spatiotemporal characteristics of carbon emission data, but also fully considers the uncertainty of the data, thereby improving the accuracy and reliability of carbon emission traceability calculations. Finally, a traceability confidence threshold is set according to actual needs, and the threshold is used to screen carbon emission spatiotemporal granularity candidate sets with high confidence. Specifically, the traceability confidence of each carbon emission spatiotemporal data sequence is compared with the set traceability confidence threshold, and the carbon emission spatiotemporal data sequence with a traceability confidence higher than the traceability confidence threshold is selected as the carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities. This embodiment divides the original carbon emission data set into different spatiotemporal granularities, and screens out a carbon emission spatiotemporal granularity candidate set with high confidence according to the carbon emission confidence interval, thereby realizing accurate traceability and spatiotemporal granularity division of carbon emission data, and providing strong support for carbon emission management.

[0035] In this embodiment, after the step of recursively dividing the area to be traced by using a quadtree structure to obtain a plurality of spatial sub-areas of different spatial granularities, the following steps are further included: Traversing all the spatial sub-regions, calculating the weighted average of the carbon emission confidence intervals of all traceability path nodes in the current spatial sub-region, and obtaining the spatial distribution characteristics of the confidence intervals; If the spatial distribution characteristics of the confidence interval are greater than the preset confidence threshold, the quadtree structure is continued to be used to recursively divide the current spatial sub-region into spaces until the weighted average values ​​of the carbon emission confidence intervals of all spatial sub-regions are lower than the preset confidence threshold or the preset maximum division depth is reached.

[0036] Specifically, this embodiment traverses all spatial sub-regions obtained by quadtree division. During the traversal process, each spatial sub-region (i.e., the node of the quadtree) is visited. For the currently visited spatial sub-region, the carbon emission data and confidence intervals of all traceability path nodes therein are collected. According to the importance of the traceability path nodes (such as assigning weights according to factors such as emission size and location), the weighted average of the carbon emission confidence intervals of all traceability path nodes in the current spatial sub-region is calculated. The calculation formula of the weighted average can be obtained by calculating the weighted average of the upper and lower limits of the confidence interval. The calculated weighted average of the confidence interval is compared with a preset confidence threshold. If the weighted average is greater than the preset confidence threshold, it means that the uncertainty of carbon emissions in the current spatial sub-region is high, and the spatial division needs to be further refined to reduce the uncertainty. If the weighted average is greater than the preset confidence threshold and does not reach the preset maximum division depth, the quadtree structure is continued to be used for spatial recursive division of the current spatial sub-region. If the weighted average is less than or equal to the preset If the confidence threshold is reached, or the preset maximum division depth is reached, further division of the current spatial sub-region is stopped. For example, when the entire area is traversed for the first time, the weighted average of its carbon emission confidence interval is calculated. Assuming that the weighted average is 0.2, which is greater than the preset confidence threshold 0.1, and the maximum division depth of 5 layers is not reached, the first recursive division is performed, and the 100km×100km area to be traced is divided into four 25km×25km sub-regions, and the weighted average of their carbon emission confidence intervals is calculated respectively. The above traversal, calculation and division process is repeated for each sub-region until the weighted average of all sub-regions is less than or equal to 0.1, or the maximum division depth of 5 layers is reached. By using the quadtree structure to recursively divide the carbon emission data in space, and evaluating and adjusting the division strategy according to the weighted average of the carbon emission confidence interval, a more spatially refined and more controllable traceability area division scheme in terms of carbon emission uncertainty can be obtained, thereby realizing refined management and traceability analysis of carbon emission sources.

[0037] S5. Use a genetic algorithm to optimize and analyze the carbon emission spatiotemporal granularity candidate set, find the optimal balance between data granularity and traceability accuracy, and determine the optimal fine-grained carbon emission data.

[0038] In this embodiment, the step of optimizing and analyzing the carbon emission spatiotemporal granularity candidate set by using a genetic algorithm to find the optimal balance between data granularity and traceability accuracy and determining the optimal fine-grained carbon emission data includes: Simulate the candidate sets of carbon emission spatiotemporal granularity at different spatiotemporal granularity to evaluate the corresponding traceability accuracy and computational efficiency, where the computational efficiency includes the time and resource consumption required for the simulation; From the plurality of carbon emission spatiotemporal granularity candidate sets, a carbon emission spatiotemporal granularity candidate set having a computational efficiency within a target computational efficiency range is selected as a primary spatiotemporal granularity; Using an information entropy algorithm to measure the spatiotemporal fine granularity of the preliminary spatiotemporal granularity to obtain data fine granularity; Calculate the weighted average of the data granularity and the traceability accuracy to obtain a balance index of the preliminary spatiotemporal granularity; Encoding each group of preliminary spatiotemporal granularity as an individual in a genetic algorithm, and iteratively optimizing the individuals using the genetic algorithm; In each iterative optimization, the fitness of the individual is calculated according to the balance index, and the population is continuously iterated and optimized through multiple generations of genetic operations until the optimal balance point between data granularity and traceability accuracy is found; The individual with the highest fitness is selected as the optimal solution, and the optimal fine-grained carbon emission data is obtained based on the optimal solution.

[0039] Specifically, this embodiment uses a system dynamics (SD) model to input carbon emission data into a simulation model at different spatiotemporal granularities, simulates the carbon emission data at each spatiotemporal granularity, and evaluates the traceability accuracy of the simulation results of each candidate set, that is, determines the degree of match between the simulation results and the actual carbon emission observation data. At the same time, the time and resource consumption required for the simulation are recorded as indicators of computing efficiency, and a target computing efficiency range is set according to actual needs. The target computing efficiency range includes a time upper limit and a resource consumption upper limit. The computing efficiency obtained by the previous evaluation is compared with the target computing efficiency range, and the spatiotemporal granularity candidate set that meets the conditions is screened out as the primary spatiotemporal granularity.

[0040] Information entropy is an indicator for measuring data uncertainty and can be used to evaluate the granularity of spatiotemporal data. For each preliminary spatiotemporal granularity, the information entropy of its data is calculated as a measure of data granularity. Then, corresponding weights are set for data granularity and traceability accuracy according to actual needs. The data granularity and traceability accuracy of each preliminary spatiotemporal granularity are weighted averaged according to the set weights to obtain the balance index. The balance index will be used to evaluate the comprehensive performance of different preliminary spatiotemporal granularities. Each group of preliminary spatiotemporal granularities is encoded as an individual. Each individual contains the encoded information of the spatiotemporal granularity. A group of individuals is randomly generated as the initial population. The population is iteratively optimized through genetic operations such as selection, crossover, and mutation. In each iteration, the fitness of each individual is calculated according to the balance index, and excellent individuals are selected as parents according to the fitness to generate the next generation. , perform a crossover operation on the selected parent generation to generate new individuals, perform a mutation operation on the newly generated individuals to increase the diversity of the population, and the newly generated individuals replace some or all of the old individuals to form a new population. After the iterative optimization is completed, the individual with the highest fitness is selected as the optimal solution, and the encoded information of the optimal solution is decoded into the optimal fine-grained carbon emission data. When processing large-scale data, traditional carbon emission traceability systems often find it difficult to balance the relationship between data granularity and traceability accuracy, resulting in low traceability efficiency. This embodiment introduces a genetic algorithm and a quadtree structure to find the optimal balance between data granularity and traceability accuracy, thereby achieving optimal selection of data granularity, which can effectively solve the problem of spatiotemporal granularity optimization of carbon emission data, and effectively respond to the growth of data scale, thereby improving traceability efficiency while ensuring traceability accuracy.

[0041] S6. Based on the optimal fine-grained carbon emission data, the carbon emission traceability model is used to obtain the carbon energy flow traceability results.

[0042] In this embodiment, the process of constructing the carbon emission tracing model includes: Extract the basic features of carbon emissions from the optimal fine-grained carbon emission data, and use the maximum information coefficient to calculate the feature correlation strength value between the basic features of carbon emissions and the traceability accuracy; According to the feature association strength value, a recursive feature elimination method is used to perform feature selection to screen out the retained basic features of carbon emissions; Based on the retained basic characteristics of carbon emissions, a carbon emission tracing model is established using support vector machines.

[0043] Specifically, this embodiment preprocesses the collected optimal fine-grained carbon emission data. The preprocessing includes data formatting, missing value filling, etc., and based on the preprocessed optimal fine-grained carbon emission data, the principal component analysis method is used to extract the basic characteristics of carbon emissions. These basic characteristics of carbon emissions include carbon emissions, carbon emission intensity, energy consumption data, etc. Then, this embodiment uses the maximum information coefficient (MIC) method to quantify the correlation strength value between each basic carbon emission characteristic and the traceability accuracy. The maximum information coefficient is a statistic that measures the mutual information between two variables. It can evaluate the linear and nonlinear relationship between the basic carbon emission characteristic and the target variable (i.e., traceability accuracy). For each basic carbon emission characteristic, this embodiment calculates the mutual information between it and the traceability accuracy, and normalizes the mutual information to obtain the correlation strength value between the basic carbon emission characteristic and the traceability accuracy. The higher the MIC value, the stronger the correlation between the characteristic and the traceability accuracy. Then, the recursive feature elimination method (RFE) is used to quantify the correlation strength value between the basic carbon emission characteristic and the traceability accuracy. The basic features are screened. The recursive feature elimination method is an iterative method, which repeatedly trains the model and eliminates the least important features until the specified number of features is reached or a certain stopping criterion is met. In each iteration of the recursive feature elimination method, this embodiment can use a machine learning model as a base model to evaluate the importance score of each basic feature of carbon emissions. In each iteration, the feature with the lowest importance score is eliminated from the feature set, and the model is retrained. The above process is repeated until a predetermined number of features is reached or the importance scores of all features exceed a preset importance threshold. Finally, this embodiment uses the basic features of carbon emissions retained by the recursive feature elimination method (RFE) as input variables, and the carbon emission source as the output variable to train a support vector machine (SVM) model. In carbon emission tracing, the trained SVM model is used to perform traceability analysis on new carbon emission data, and outputs the traceability tracking results of the carbon energy flow. The traceability tracking results include information such as the source, path and intensity of carbon emissions.

[0044] An embodiment of the present invention provides a carbon energy flow tracing optimization method, which collects the original carbon emission data set of the target energy type and constructs a carbon emission network topology structure based on the original carbon emission data set; based on the carbon emission network topology structure, the energy transmission relationship between different emission source nodes is modeled using a Markov chain model to generate a carbon emission tracing path; according to the energy consumption data and carbon emission factors of each emission source node on the carbon emission tracing path, a kernel density estimation method based on Bayesian theory is used to estimate the carbon emission confidence interval of the tracing path node; according to the carbon emission confidence interval, the original carbon emission data set is divided into different spatiotemporal granularities through a quadtree structure to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities; the carbon emission spatiotemporal granularity candidate sets are optimized and analyzed using a genetic algorithm to find the optimal balance between data granularity and tracing accuracy, and determine the optimal fine-grained carbon emission data; based on the optimal fine-grained carbon emission data, a carbon emission tracing model is used to obtain a carbon energy flow tracing result. Compared with existing technologies, this method combines Markov chain models and genetic algorithms to not only achieve accurate carbon energy flow tracing, but also realizes a trade-off between data granularity and traceability accuracy. It can effectively cope with the growth of data scale, significantly improve the carbon emission traceability accuracy and efficiency, ensure the reliability of traceability results, and provide a scientific basis for emission reduction strategies.

[0045] It should be noted that the size of the serial numbers of the above-mentioned processes does not mean 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 the present application.

[0046] In one embodiment, Figure 2 As shown, an embodiment of the present invention provides a carbon energy flow tracing optimization system, the system comprising: The data collection module 101 is used to collect the original carbon emission data set of the target energy type and construct a carbon emission network topology structure according to the original carbon emission data set; A path analysis module 102 is used to model the energy transmission relationship between different emission source nodes based on the carbon emission network topology structure using a Markov chain model to generate a carbon emission tracing path; The confidence analysis module 103 is used to estimate the carbon emission confidence interval of the traceability path node using the kernel density estimation method based on Bayesian theory according to the energy consumption data and carbon emission factor of each emission source node on the carbon emission traceability path; The granularity division module 104 is used to divide the original carbon emission data set into different spatiotemporal granularities through a quadtree structure according to the carbon emission confidence interval, so as to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities; Iterative optimization module 105, used to optimize and analyze the carbon emission spatiotemporal granularity candidate set using a genetic algorithm, find the optimal balance between data granularity and traceability accuracy, and determine the optimal fine-grained carbon emission data; The traceability module 106 is used to obtain the carbon energy flow traceability results based on the optimal fine-grained carbon emission data and using the carbon emission traceability model.

[0047] For the specific definition of a carbon energy flow tracing optimization system, please refer to the above-mentioned definition of a carbon energy flow tracing optimization method, which will not be repeated here. A person of ordinary skill in the art can appreciate 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 performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0048] An embodiment of the present invention provides a carbon energy flow tracing optimization system, which collects the original carbon emission data set of the target energy type through a data acquisition module, and constructs a carbon emission network topology structure based on the original carbon emission data set; the path analysis module uses a Markov chain model to model the energy transmission relationship between different emission source nodes based on the carbon emission network topology structure, and generates a carbon emission tracing path; the confidence analysis module estimates the carbon emission confidence interval of the traceability path node based on the energy consumption data and carbon emission factors of each emission source node on the carbon emission tracing path using a kernel density estimation method based on Bayesian theory; the granularity division module divides the original carbon emission data set into different spatiotemporal granularities through a quadtree structure according to the carbon emission confidence interval, and obtains a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities; the iterative optimization module uses a genetic algorithm to optimize and analyze the carbon emission spatiotemporal granularity candidate set, finds the optimal balance between data granularity and traceability accuracy, and determines the optimal fine-grained carbon emission data; the traceability tracing module obtains the carbon energy flow traceability tracing result based on the optimal fine-grained carbon emission data using a carbon emission traceability model. Compared with existing technologies, this system combines Markov chain models, genetic algorithms and other technologies to not only achieve accurate carbon energy flow tracing, but also realizes a trade-off between data granularity and traceability accuracy. It can effectively cope with the growth of data scale, significantly improve the carbon emission traceability accuracy and efficiency, ensure the reliability of traceability results, and provide a scientific basis for emission reduction strategies.

[0049] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A carbon energy flow tracing optimization method, characterized in that: The following steps are involved: Collecting an original carbon emission data set of a target energy type, and constructing a carbon emission network topology structure based on the original carbon emission data set; Based on the carbon emission network topology, the energy transmission relationship between different emission source nodes is modeled using the Markov chain model to generate a carbon emission traceability path; According to the energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path, the kernel density estimation method based on Bayesian theory is used to estimate the confidence interval of carbon emissions at the traceability path nodes. According to the carbon emission confidence interval, the original carbon emission data set is divided into different spatiotemporal granularities through a quadtree structure to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities; Using a genetic algorithm to optimize and analyze the carbon emission spatiotemporal granularity candidate set, find the optimal balance between data granularity and traceability accuracy, and determine the optimal fine-grained carbon emission data; Based on the optimal fine-grained carbon emission data, the carbon emission traceability model is used to obtain the carbon energy flow traceability results.

2. A carbon energy flow tracing optimization method as claimed in claim 1, characterized in that: The step of constructing a carbon emission network topology structure according to the carbon emission data set comprises: Obtaining energy flow information of each emission source from the original carbon emission data set, and determining the energy transmission relationship between different emission sources based on the energy flow information of each emission source; the energy transmission relationship includes the energy flow direction and energy flow intensity between different emission sources; The emission sources in the original carbon emission data set are defined as emission source nodes, the energy transmission relationship is defined as edges, and the carbon emission network topology structure is constructed using a directed graph.

3. A carbon energy flow tracing optimization method as claimed in claim 1, characterized in that: The step of modeling the energy transmission relationship between different emission source nodes based on the carbon emission network topology using a Markov chain model to generate a carbon emission tracing path includes: Using a graph traversal method to traverse all emission source nodes in the carbon emission network topology structure, and defining each traversed emission source node as a single state in a Markov chain model to construct a state space; Based on the energy transmission data between each emission source node, the transfer probability from each emission source node to other emission source nodes is calculated to form a transfer probability matrix; Determine the initial state probability of each emission source node based on energy consumption data and obtain the initial state probability distribution; According to the initial state probability distribution and the transition probability matrix, the Markov chain process is simulated by a random walk algorithm to perform multiple state transitions to obtain the potential traceability path of each emission source node; The cumulative probability of each potential tracing path is calculated according to the transfer probability matrix, and all potential tracing paths are sorted according to the cumulative probability, and the potential tracing path with the highest cumulative probability is taken as the carbon emission tracing path.

4. A carbon energy flow tracing optimization method as claimed in claim 3, characterized in that: The sum of the initial state probabilities of all emission source nodes is one.

5. A carbon energy flow tracing optimization method as claimed in claim 1, characterized in that: The step of estimating the carbon emission confidence interval of the traceability path node using the Bayesian kernel density estimation method based on the energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path includes: Obtaining energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path; the energy consumption data includes energy consumption type, energy consumption amount and energy consumption time; Calculate the carbon emissions of each emission source node on the carbon emission traceability path based on the energy consumption data and the carbon emission factor; According to the carbon emissions of each emission source node on the carbon emission traceability path, the probability density function of the carbon emissions of each emission source node is established using the kernel density estimation method; According to the prior distribution and probability density function of carbon emissions, the posterior distribution of carbon emissions is calculated by Bayesian theorem; The Markov Monte Carlo method is used to sample the posterior distribution to obtain the carbon emission posterior samples, and the confidence interval of carbon emissions for each traceability path node is calculated based on the carbon emission posterior samples.

6. A carbon energy flow tracing optimization method as claimed in claim 1, characterized in that: The step of dividing the original carbon emission data set into different spatiotemporal granularities by a quadtree structure according to the carbon emission confidence interval to obtain carbon emission spatiotemporal granularity candidate sets of different spatiotemporal granularities includes: The original carbon emission data set is decomposed into multiple scales using discrete wavelet transform to generate carbon emission temporal characteristics at different time scales. According to the longitude and latitude coordinates of the boundaries of the area to be traced, the area to be traced is spatially recursively divided using a quadtree structure to obtain a number of spatial sub-areas with different spatial granularities; After all spatial sub-regions are divided, the temporal characteristics of carbon emissions in each spatial sub-region are extracted to obtain the spatiotemporal data series of carbon emissions at different spatiotemporal granularities; Inputting the carbon emission confidence interval into a convolutional neural network, and using the convolutional neural network to perform traceability calculation on the carbon emission spatiotemporal data sequence to obtain traceability confidence; According to the traceability confidence, a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities is screened out from the carbon emission spatiotemporal data sequence.

7. A carbon energy flow tracing optimization method as claimed in claim 6, characterized in that: After the step of recursively dividing the area to be traced by using a quadtree structure to obtain a plurality of spatial sub-areas of different spatial granularities, the following step is further included: Traversing all the spatial sub-regions, calculating the weighted average of the carbon emission confidence intervals of all traceability path nodes in the current spatial sub-region, and obtaining the spatial distribution characteristics of the confidence intervals; If the spatial distribution characteristics of the confidence interval are greater than the preset confidence threshold, the quadtree structure is continued to be used to recursively divide the current spatial sub-region into spaces until the weighted average values ​​of the carbon emission confidence intervals of all spatial sub-regions are lower than the preset confidence threshold or the preset maximum division depth is reached.

8. A carbon energy flow tracing optimization method as claimed in claim 1, characterized in that: The step of optimizing and analyzing the carbon emission spatiotemporal granularity candidate set by using a genetic algorithm to find the optimal balance between data granularity and traceability accuracy and determining the optimal fine-grained carbon emission data includes: Simulate the candidate sets of carbon emission spatiotemporal granularity at different spatiotemporal granularity to evaluate the corresponding traceability accuracy and computational efficiency, where the computational efficiency includes the time and resource consumption required for the simulation; From the plurality of carbon emission spatiotemporal granularity candidate sets, a carbon emission spatiotemporal granularity candidate set having a computational efficiency within a target computational efficiency range is selected as a primary spatiotemporal granularity; Using an information entropy algorithm to measure the spatiotemporal fine granularity of the preliminary spatiotemporal granularity to obtain data fine granularity; Calculate the weighted average of the data granularity and the traceability accuracy to obtain a balance index of the preliminary spatiotemporal granularity; Encoding each group of preliminary spatiotemporal granularity as an individual in a genetic algorithm, and iteratively optimizing the individuals using the genetic algorithm; In each iterative optimization, the fitness of the individual is calculated according to the balance index, and the population is continuously iterated and optimized through multiple generations of genetic operations until the optimal balance point between data granularity and traceability accuracy is found; The individual with the highest fitness is selected as the optimal solution, and the optimal fine-grained carbon emission data is obtained based on the optimal solution.

9. A carbon energy flow tracing optimization method as claimed in claim 1, characterized in that: The construction process of the carbon emission tracing model includes: Extract the basic features of carbon emissions from the optimal fine-grained carbon emission data, and use the maximum information coefficient to calculate the feature correlation strength value between the basic features of carbon emissions and the traceability accuracy; According to the feature association strength value, a recursive feature elimination method is used to perform feature selection to screen out the retained basic features of carbon emissions; Based on the retained basic characteristics of carbon emissions, a carbon emission tracing model is established using support vector machines.

10. A carbon energy flow tracing optimization system, characterized in that: The system comprises: A data collection module, used to collect the original carbon emission data set of the target energy type and construct a carbon emission network topology structure based on the original carbon emission data set; A path analysis module is used to model the energy transmission relationship between different emission source nodes based on the carbon emission network topology structure using a Markov chain model to generate a carbon emission tracing path; The confidence analysis module is used to estimate the confidence interval of carbon emissions at the nodes on the carbon emission traceability path based on the energy consumption data and carbon emission factors of each emission source node on the carbon emission traceability path using the kernel density estimation method based on Bayesian theory; A granularity division module is used to divide the original carbon emission data set into different spatiotemporal granularities through a quadtree structure according to the carbon emission confidence interval, so as to obtain a carbon emission spatiotemporal granularity candidate set of different spatiotemporal granularities; An iterative optimization module is used to optimize and analyze the carbon emission spatiotemporal granularity candidate set using a genetic algorithm, find the optimal balance between data granularity and traceability accuracy, and determine the optimal fine-grained carbon emission data; The traceability module is used to obtain the carbon energy flow traceability results based on the optimal fine-grained carbon emission data and the carbon emission traceability model.

Images