A carbon metering terminal based on edge intelligence collaboration

Abstract

The present invention relates to the field of carbon management technology, and specifically to a carbon metering terminal based on edge intelligent collaboration, the terminal comprising: a model partition update module, based on real-time collected carbon emission data, constructs a carbon metering model according to the carbon emission data, and generates partition-optimized carbon metering information by adjusting and dividing the carbon metering model. In the present invention, the accuracy and efficiency of carbon emission prediction are significantly enhanced through the real-time collection and partition optimization of carbon emission data, combined with the extraction of time series features and statistical features. In the data processing process, by establishing a filtering threshold range, abnormal data and repetitive data are effectively eliminated, the data quality is improved, and inefficient data is avoided from interfering with subsequent analysis. In the collaborative optimization of edge nodes, the communication delay and resource waste between nodes are reduced by means of dynamic weight adjustment and path optimization, and the transmission efficiency and collaboration effect are improved.

Description

Technical Field

[0001] The present invention relates to the field of carbon management technology, and in particular to a carbon metering terminal based on edge intelligent collaboration. Background Art

[0002] The field of carbon management technology includes technical methods and equipment related to carbon emissions, carbon sink quantification, carbon asset management and carbon trading. In the field of carbon management technology, its core content is to measure, analyze and manage carbon emissions and carbon absorption in a scientific and systematic way, so as to provide accurate carbon emission data support for enterprises, institutions and individuals.

[0003] Among them, the carbon metering terminal refers to a terminal system used to collect, process and transmit carbon emission and carbon absorption data. The terminal mainly collects data accurately for carbon emission sources and carbon sink areas, and completes the acquisition of parameters including carbon dioxide concentration and emission flow.

[0004] It is difficult for existing technologies to perform partition optimization and feature extraction for carbon emission data collected in real time. Fixed models are usually used, which cannot cope with the insufficient processing accuracy caused by data changes. For example, when there are time series fluctuations in carbon emission data, existing technologies cannot accurately predict future trends, which can easily lead to data analysis distortion. For data quality control, it is difficult for existing technologies to eliminate abnormal or duplicate data through precise filtering, resulting in high data noise, which may interfere with subsequent model calculations and analysis results. In terms of edge node collaboration, existing technologies usually lack dynamic analysis of task collaboration requirements and data interaction frequency, and cannot flexibly adjust the transmission path or task allocation rules between nodes. For example, in scenarios where the communication volume or computing load between nodes increases suddenly, existing technologies often cause data transmission delays due to insufficient path optimization, thereby affecting the overall task execution efficiency. In addition, existing technologies cannot dynamically adjust task allocation according to the computing performance of real-time monitoring, resulting in uneven allocation of task resources, which restricts the overall performance of the system. Summary of the invention

[0005] The present invention provides a carbon metering terminal based on edge intelligent collaboration.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme:

[0007] A carbon metering terminal based on edge intelligence collaboration includes:

[0008] The model partition update module builds a carbon measurement model based on the real-time collected carbon emission data, and generates carbon measurement information after partition optimization by adjusting and dividing the carbon measurement model;

[0009] The carbon emission data filtering module extracts the processed carbon emission data from the carbon measurement information after the partition optimization, establishes a data filtering threshold range, compares the data filtering threshold range with all carbon emission data, filters the corresponding carbon emission data according to the comparison result, and generates a filtered carbon emission data set;

[0010] The collaborative path optimization module identifies the edge nodes of the carbon emission data in the filtered carbon emission data set, constructs a collaborative dependency graph between the edge nodes according to the identification results, analyzes the task collaboration requirements and data interaction frequency between the edge nodes in the collaborative dependency graph to determine the weights between the edge nodes, optimizes the transmission paths between the nodes according to the weights between the edge nodes, and generates an optimized collaborative path;

[0011] The task dynamic adjustment module monitors the computing load, task execution frequency and data transmission delay of the optimized collaborative path in real time, dynamically adjusts the weights between edge nodes according to the monitoring results, updates the collaborative task grouping between the adjusted edge nodes, and obtains the collaborative task grouping record.

[0012] As a further solution of the present invention, the steps for obtaining the carbon measurement information after the partition optimization are specifically as follows:

[0013] Based on the real-time collected carbon emission data, a carbon measurement model is constructed according to the carbon emission data, and the input layer partition of the carbon measurement model is generated by preprocessing the carbon emission data;

[0014] Extracting processed carbon emission data from the input layer partition of the carbon accounting model and converting it into a time series, extracting time series and statistical features for each data point in the data set, and generating a feature extraction partition of the carbon accounting model;

[0015] Based on the feature extraction of the carbon measurement model, the time series characteristics and statistical characteristics in the partition are adopted:

[0016] ;

[0017] Calculate the predicted carbon emissions at future times , get the carbon measurement information after partition optimization;

[0018] in, Represents the change in carbon emissions between two consecutive time points. represents the average carbon emissions within the selected time window, is the standard deviation of carbon emission data, express The impact on future carbon emissions forecasts, express The impact on future carbon emissions forecasts, express The impact on future carbon emissions forecasts, Represents the bias term.

[0019] As a further solution of the present invention, the step of establishing a data filtering threshold range is specifically:

[0020] Based on the processed carbon emission data in the carbon measurement information after the partition optimization, the formula is adopted:

[0021] ;

[0022] Calculating quantiles of carbon emissions data , according to the quantile Divide the upper and lower bounds of the distribution value;

[0023] in, is the percentile, is the total number of data points, Represents the total number of data points plus , It is the carbon emission data after sorting. The value at the index position, It is the carbon emission data after sorting. The value at the index position, is the target position of the quantile, Represents the distance between the quantile target position and the lower bound index, Represents the difference between the upper and lower bound indexes of the sorted data;

[0024] Based on the quantiles and the upper and lower bound distribution values, the selected quantiles are used as initial boundaries, the upper and lower limits are expanded, and a data filtering threshold range is established.

[0025] As a further solution of the present invention, the steps of obtaining the filtered carbon emission data set are specifically:

[0026] Based on the data filtering threshold range, the carbon emission data points are compared with the upper and lower limit values ​​one by one, abnormal data points, repetitive data points and data points with a change amplitude lower than a preset amplitude threshold are screened and marked, and the screened data point classification information is generated;

[0027] Based on the filtered data point classification information, the marked data points are removed, and the remaining data points are reordered and sorted to generate a filtered carbon emission data set.

[0028] As a further solution of the present invention, the step of constructing a collaborative dependency graph between edge nodes is specifically as follows:

[0029] Based on the filtered carbon emission data, extract the source device information of the carbon emission data, analyze the device's collection frequency, data transmission performance, and computing load, identify the edge nodes in the area according to the device location and function, mark them, and generate an edge node list;

[0030] Based on the edge node list, data interaction records and task collaboration requirements between edge nodes are extracted, the collaboration strength between nodes is analyzed, and a collaborative dependency graph is constructed through node and collaboration relationships.

[0031] As a further solution of the present invention, the step of obtaining the optimized collaborative path is specifically:

[0032] Based on the task collaboration requirements and data interaction frequency between edge nodes in the collaborative dependency graph, the formula is adopted:

[0033] ;

[0034] Computing edge nodes and The weight between ;

[0035] in, is the first adjustment coefficient, is the second adjustment factor, Is a node and The frequency of data interaction between Is a node and The demand for task collaboration between

[0036] Based on the weight values ​​between edge nodes and , all edge nodes are weighted and path optimization is performed, transmission paths are preferentially selected according to the weights, path information is recorded, and an optimized collaborative path is generated.

[0037] As a further solution of the present invention, the step of dynamically adjusting the weights between edge nodes is specifically:

[0038] Based on the optimized collaborative path, the computing load, task execution frequency and data transmission delay in the path are monitored in real time, the performance monitoring data of each node in each path is obtained, and a path performance monitoring data set is generated;

[0039] Based on the path performance monitoring data set, the weights between edge nodes in the path are dynamically adjusted, the path performance priority is evaluated and the weight value is adjusted according to the performance data normalization result, and a dynamically adjusted path weight record is generated.

[0040] As a further solution of the present invention, the steps of obtaining the collaborative task group record are specifically as follows:

[0041] Based on the dynamically adjusted path weight records, the computing load utilization, task execution frequency and data transmission delay value of the edge node are collected, the node performance data is analyzed and the task allocation priority is adjusted, the task allocation rules are reconfigured, and the adjusted task collaboration grouping is generated;

[0042] Based on the adjusted task collaboration grouping, the task dependency and data interaction frequency within the collaboration group are optimized according to the node weight allocation result, the task allocation rule is updated, the node and task information of each collaboration group is recorded, and the collaborative task grouping record is generated.

[0043] Compared with the prior art, the advantages and positive effects of the present invention are:

[0044] In the present invention, the accuracy and efficiency of carbon emission prediction are significantly enhanced through real-time collection and partition optimization of carbon emission data, combined with the extraction method of time series features and statistical features. In the data processing process, by establishing a filtering threshold range, abnormal data and repetitive data are effectively eliminated, the data quality is improved, and inefficient data is avoided from interfering with subsequent analysis. In the collaborative optimization of edge nodes, the communication delay and resource waste between nodes are reduced by means of dynamic weight adjustment and path optimization, and the transmission efficiency and collaboration effect are improved. The mechanism of dynamically adjusting task grouping rules ensures that edge nodes can achieve efficient division of labor and collaboration in complex task scenarios, and ensures the stability and efficiency of the overall operation of the system. By constructing an edge node collaborative dependency graph and a task collaboration optimization path, the bottleneck problem in the data processing and task allocation process is solved, providing reliable support for the accurate monitoring and analysis of carbon emission data. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. The drawings are only used to illustrate the implementation methods and are not to be considered as limitations of the present invention.

[0046] Figure 1 It is a terminal flow chart of the present invention;

[0047] Figure 2 A flow chart for obtaining carbon measurement information after partition optimization for the present invention;

[0048] Figure 3 A flow chart for establishing data filtering threshold ranges for the present invention;

[0049] Figure 4 A flow chart for obtaining a filtered carbon emission data set for the present invention;

[0050] Figure 5 A flowchart for constructing a collaborative dependency graph between edge nodes for the present invention;

[0051] Figure 6 A flow chart for obtaining an optimized collaborative path for the present invention;

[0052] Figure 7 A flow chart of dynamically adjusting weights between edge nodes according to the present invention;

[0053] Figure 8 A flowchart of obtaining collaborative task grouping records according to the present invention. DETAILED DESCRIPTION

[0054] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as commonly understood by technicians in the field of the present invention; the terms used in the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings, and are intended to cover non-exclusive inclusions.

[0056] In the description of the embodiments of the present invention, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present invention, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0057] In the description of the embodiments of the present invention, the term "and / or" is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0058] In the description of the embodiments of the present invention, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0059] In the description of the embodiments of the present invention, the technical terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the embodiments of the present invention.

[0060] In the description of the embodiments of the present invention, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the embodiments of the present invention can be understood according to the specific circumstances.

[0061] Example

[0062] The embodiment of the present invention provides a carbon metering terminal based on edge intelligence collaboration, such as Figure 1 As shown, including:

[0063] The model partition update module builds a carbon measurement model based on the real-time collected carbon emission data, and generates carbon measurement information after partition optimization by adjusting and dividing the carbon measurement model;

[0064] The carbon emission data filtering module extracts the processed carbon emission data from the carbon measurement information after partition optimization, establishes a data filtering threshold range, compares the data filtering threshold range with all carbon emission data, filters the corresponding carbon emission data according to the comparison result, and generates a filtered carbon emission data set;

[0065] The collaborative path optimization module identifies the edge nodes of the carbon emission data in the filtered carbon emission data set, builds a collaborative dependency graph between the edge nodes based on the identification results, analyzes the task collaboration requirements and data interaction frequency between the edge nodes in the collaborative dependency graph to determine the weights between the edge nodes, optimizes the transmission paths between the nodes based on the weights between the edge nodes, and generates an optimized collaborative path;

[0066] The task dynamic adjustment module monitors the computing load, task execution frequency and data transmission delay of the optimized collaborative path in real time, dynamically adjusts the weights between edge nodes according to the monitoring results, updates the collaborative task grouping between the adjusted edge nodes, and obtains the collaborative task grouping record;

[0067] The carbon metering information after partition optimization includes input layer partitions, feature extraction partitions, and parameter output partitions. The filtered carbon emission data set includes processed carbon emission data, repetitive data markers, and data with a change range lower than the filtering threshold range. The optimized collaborative path specifically includes the transmission path of the edge nodes, the weight distribution between nodes, and path sharing information. The collaborative task grouping records include adjusted task grouping information, dynamic collaborative relationships between edge nodes, and task execution frequency records.

[0068] See also Figure 2 , the specific steps for obtaining carbon measurement information after partition optimization are:

[0069] Based on the real-time collected carbon emission data, a carbon measurement model is constructed according to the carbon emission data, and the input layer partition of the carbon measurement model is generated by preprocessing the carbon emission data;

[0070] The collected data include the original emission values ​​recorded by multiple carbon emission sensors, data collection timestamps and equipment numbers. First, the collected data are formatted to unify all data into a standard structured table format. Abnormal data and missing values ​​are removed through data cleaning methods. Abnormal data can be eliminated by setting the upper and lower threshold ranges of carbon emissions (such as a minimum value of 5ppm and a maximum value of 500ppm). The cleaned data is used as the processed data for the input layer partition to generate a processed carbon emission data set.

[0071] Extract the processed carbon emission data in the input layer partition of the carbon accounting model and convert it into a time series, extract the time series and statistical features of each data point in the data set, and generate the feature extraction partition of the carbon accounting model;

[0072] The feature extraction partition calls the carbon emission data set after input layer partition processing, and extracts time series and statistical features for each data point in the data set. The time series features are processed by the sliding window method, and the width of the sliding window is set to 10 time units. The mean, maximum and minimum values ​​of the carbon emission data in each window are calculated as time series features. The extraction of statistical features is based on data analysis software such as Python's Pandas library or MATLAB software. The mean, standard deviation and skewness of the data are quickly calculated through built-in functions. After the feature extraction is completed, all time series features and statistical features are organized into a feature matrix. Each row in the feature matrix represents all the features of a data point, and a feature set containing time series features and statistical features is generated for subsequent carbon emission trend prediction.

[0073] The time series characteristics and statistical characteristics in the feature extraction partition based on the carbon measurement model are adopted by the formula:

[0074] ;

[0075] Calculate the predicted carbon emissions at future times , get the carbon measurement information after partition optimization;

[0076] in, is the time series change rate, which indicates the change in carbon emissions between two consecutive time points. The calculation formula is: , Is the current time point Carbon emissions are obtained through real-time data recorded by sensors. It's the previous time point The carbon emissions are also obtained through the data recorded by sensors. represents the average carbon emissions within the selected time window, is the standard deviation of carbon emission data, indicating the degree of fluctuation of carbon emissions. represents the time series change rate ( ) on the future carbon emissions forecast results. The higher the value, the greater the weight of the change rate on the forecast. It is suitable for scenarios with drastic changes in short-term carbon emissions. Through experiments, historical carbon emission data are collected, and the time series change rate and the corresponding future carbon emission actual value are extracted. The regression analysis method is used to calculate the correlation between the time series change rate and the future carbon emission value. For example, the Pearson correlation coefficient is used to judge the correlation strength. A sample data is selected, such as the carbon emission data of the past month, and the least squares method is used to fit the formula The coefficients of , represents the moving average ( ) on the future carbon emissions forecast results. The higher the value, the greater the weight of the moving average on the forecast results. It is suitable for scenarios where the long-term carbon emissions trend is relatively stable. Through experiments, historical carbon emissions data are collected, and the moving average and the corresponding future carbon emissions actual value are extracted. The sliding window technology is used to calculate the moving average within a time window. For example, the carbon emissions data of the past 10 time units are selected, and the regression analysis method is used to calculate the correlation between the moving average and the future carbon emissions value. The least squares fitting formula is used The coefficients of , represents the standard deviation of the data ( ) on the future carbon emissions forecast results. The higher the value, the greater the weight of the fluctuation range on the forecast results. It is suitable for scenarios where carbon emissions data fluctuate violently. Through experiments, historical carbon emissions data are collected, standard deviations and corresponding future carbon emissions actual values ​​are extracted, and the standard deviation of carbon emissions data within a time window is calculated. Regression analysis is used to analyze the correlation between standard deviations and future carbon emissions values. The least squares method is used to fit the formula The coefficients are determined based on the fitting results. , Represents the bias term, which indicates the basic prediction value of the model for carbon emissions when there is no feature data involved in the calculation. It is usually used to adjust the model output to a reasonable range, collect historical carbon emission data, and extract features. , , The actual carbon emission value when is zero, the minimum or median carbon emission in the statistical sample is used as the initial value of the bias term, and the bias term is adjusted through experiments Make the prediction closer to the actual value by increasing or decreasing The value of minimizes the error between the prediction and the actual value. The final value of .

[0077] Set parameters, time series change rate: ppm (50ppm at the current time point, 45ppm at the previous time point); moving average: ppm (average of the first 10 time points); standard deviation: ppm (based on the fluctuation calculated at the first 10 time points). The model parameters obtained by fitting are , , , .

[0078] Calculation process:

[0079] ;

[0080] ;

[0081] The result shows that the predicted value of carbon emissions at a future moment is 47ppm, and the predicted value of 47ppm is compared with the average value and change trend of carbon emissions in the historical time window. For example, if the average value of historical data in the past 10 time units is 45ppm, and the rate of change in the past period of time is stable or slowly increasing, the result of 47ppm indicates that future carbon emissions may show a slight upward trend. If the predicted value is significantly higher than the mean of historical data (such as more than 5% higher), it can be judged that future carbon emissions may enter a rapid growth stage; if the predicted value is close to the historical mean (such as the difference is within 5%), it can be judged that future carbon emissions will maintain a relatively stable trend; if the predicted value is lower than the historical mean, it can be judged that future carbon emissions may decrease. Based on the collaborative calculation and optimization of input layer partitioning, feature extraction partitioning and parameter output partitioning, the partition-optimized carbon measurement information is generated.

[0082] See also Figure 3 , the steps to establish the data filtering threshold range are as follows:

[0083] Based on the processed carbon emission data in the carbon measurement information after partition optimization, the formula is used:

[0084] ;

[0085] Calculating quantiles of carbon emissions data , according to the quantile Divide the upper and lower bounds of the distribution value;

[0086] in, It represents the value of the percentage position in the data, such as the 25th percentile (lower quartile), the 50th percentile (median), and the 75th percentile (upper quartile). It is calculated by arranging the standardized carbon emission data in ascending order and combining the percentage position to describe the distribution characteristics of the data. is the percentile, ranging from , represents the location ratio of the quantile, for example, Corresponding to the 25th quantile, Corresponding to the 75th percentile, this value is a preset parameter used to specify the target position for calculating the quantile. is the total number of data points, indicating the total amount of carbon emission data after standardization. It is directly obtained after the data collected by the carbon emission sensor is processed. For example, if the collected data points are 100, then , Represents the total number of data points plus , It is the carbon emission data after sorting. The value at the index position represents the data point at the index position of the lower bound of the quantile. The data is extracted after being sorted in ascending order. For example, the first data point is recorded as , the second data point is recorded as , It is the carbon emission data after sorting. The value at the index position represents the data point at the upper bound index position of the quantile. is the actual target position of the quantile, the value when it is not rounded, used to locate the upper and lower bound indexes of the sorted data. Represents the distance between the quantile target position and the lower bound index, which is used to determine the linear weight ratio of the quantile between the upper and lower bound values. Represents the difference between the upper and lower bounds of the sorted data, which is used to calculate the linear interpolation of the quantile between the upper and lower bounds.

[0087] Suppose the data set is [12, 14, 15, 17, 19, 21, 24, 25, 30], and the total number of data is , in ascending order;

[0088] Calculate the 25th quantile ( ):

[0089] ;

[0090] ;

[0091] ;

[0092] Calculate the 75th quantile ( ):

[0093] ;

[0094] ;

[0095] ;

[0096] The results show that the calculated 25th quantile is 14.5 and the 75th quantile is 27.5. These two sets of values ​​define the upper and lower bounds of the data, indicating the boundary points of the data distribution within this range, which can be further used to establish the threshold range for data filtering.

[0097] Based on the quantile and upper and lower bound distribution values, the selected quantile is used as the initial boundary, the upper and lower limits are expanded, and the data filtering threshold range is established;

[0098] The data filtering threshold range is established based on the quantile and upper and lower bound distribution values. First, the 25th quantile (14.5) is used as the initial lower limit value and the 75th quantile (27.5) is used as the initial upper limit value. The quantile interval (i.e., the 75th quantile minus the 25th quantile) is calculated. The quantile interval is multiplied by the adjustment coefficient (such as 0.1) to expand the upper and lower bounds. The expanded lower limit is calculated as , the expansion upper limit is calculated as , through this operation, the expanded data filtering threshold range is obtained.

[0099] See also Figure 4 , the steps to obtain the filtered carbon emission dataset are as follows:

[0100] Based on the data filtering threshold range, the carbon emission data points are compared with the upper and lower limit values ​​one by one, abnormal data points, repetitive data points and data points with a change range lower than the preset range threshold are screened and marked, and the filtered data point classification information is generated;

[0101] First, all carbon emission data points are extracted based on the filter threshold range (boundary values ​​after the upper and lower bounds are expanded) obtained by the previous calculation, and each data point is screened and analyzed one by one. The judgment process includes the following steps: for each carbon emission data point, its value is extracted and compared with the upper and lower bounds. If the data point is less than the lower limit or greater than the upper limit, it is directly marked as an abnormal data point; if the data point is within the upper and lower bounds, its continuity and variation range are further analyzed. The continuity analysis compares whether the value difference of adjacent data points is less than the set repeatability threshold. For example, the value difference of 5 consecutive data points that are all less than a specific threshold (such as 1ppm) is defined as a repeatable data point; the variation range analysis extracts the change value between the data point and the adjacent data points before and after. If the change value is less than the set amplitude threshold (such as 0.5ppm), it is marked as a low variation range data point, and the screened abnormal data points, repeatable data points and low variation range data points are recorded separately.

[0102] Based on the classification information of the filtered data points, the marked data points are removed, and the remaining data points are reordered and sorted to generate a filtered carbon emission data set;

[0103] After removing the abnormal points, the data points are re-sorted in chronological order to ensure that the timing characteristics of the data points are retained. At the same time, the corresponding timestamp, equipment number and other meta-information of each data point are retained to facilitate the subsequent tracing of the data source. After the sorting is completed, a new data table or matrix structure is generated, in which each row corresponds to a data point, and the columns include timestamps, carbon emission values ​​and related meta-information. The integrated data set is stored as a filtered carbon emission data set, and the filtered data set is used as input data for subsequent carbon emission analysis and prediction.

[0104] See also Figure 5 ,The specific steps of constructing the collaborative dependency graph between edge nodes are as follows:

[0105] Based on the filtered carbon emission data, extract the source equipment information of carbon emission data, analyze the equipment's collection frequency, data transmission performance and computing load, identify the edge nodes in the area according to the equipment location and function, mark them, and generate an edge node list;

[0106] First, information related to the device number and data collection timestamp is extracted from the filtered carbon emission data set, and the data collection source device is confirmed in combination with the device deployment list in the region. These devices are classified as edge nodes according to their physical location and task function. Then, the collection frequency, data transmission performance, and computing load of each device are analyzed, and devices with higher collection frequency, shorter data transmission path, and computing capability are marked as candidate edge nodes. Finally, according to the edge node deployment rules in the region, eligible devices are included in the edge node list, and their device number, location identification, and functional information are recorded.

[0107] Based on the edge node list, extract the data interaction records and task collaboration requirements between edge nodes, analyze the collaboration strength between nodes, and build a collaborative dependency graph through the node and collaboration relationship;

[0108] Take the task collaboration requirements and data interaction records of each edge node. The task collaboration requirements are obtained by analyzing the functional type of the node (such as data collection, preprocessing, calculation, etc.) and the task execution frequency. The data interaction records are obtained by extracting the amount of data transmitted between nodes, transmission delay and interaction frequency. Then, according to the extracted collaboration requirements and data interaction records, determine the collaboration intensity between each pair of edge nodes. The collaboration intensity is calculated by standardizing the data interaction volume and task collaboration frequency. Then, use the nodes as points in the graph and the collaboration intensity as the weight of the edge to construct a collaborative dependency graph. Each edge node in the graph is identified by its unique device number, and the weight of each edge represents the collaboration intensity between the corresponding nodes.

[0109] See also Figure 6 , the steps for obtaining the optimized collaborative path are as follows:

[0110] Based on the task collaboration requirements and data interaction frequency between edge nodes in the collaborative dependency graph, the formula is adopted:

[0111] ;

[0112] Computing edge nodes and The weight between ;

[0113] in, is the first adjustment coefficient, which is used to balance the impact of the data interaction frequency between nodes on the weight. It is determined based on the importance of the data interaction frequency in historical data analysis. For example, if data interaction has a greater impact on collaboration efficiency, then Higher value, is the second adjustment coefficient, which is used to balance the impact of task collaboration requirements between nodes on the weight. Similarly, the impact of task collaboration requirements on the overall system performance is determined and meets , Is a node and The frequency of data interaction between nodes, in units of transmission times / minute, is recorded by recording the data transmission records between nodes (such as the number of packets sent and received), and extracted using network monitoring tools (such as Wireshark) or data acquisition modules. Is a node and The demand for task collaboration between nodes is expressed in the number of tasks per hour. By analyzing the task allocation and execution records of edge nodes, the number of tasks that need to be completed by collaboration between each pair of nodes is extracted. It is the normalized denominator, which is used to limit the weight values ​​between nodes to the range of 0-1. It is the maximum value of the unnormalized weight values ​​among all node pairs.

[0114] If the node and If data is sent 30 times and received 20 times in one minute, then the data interaction frequency is times / minute. and If 15 tasks are completed collaboratively within one hour, tasks / hour. The analysis found that data interaction contributes 60% to system efficiency, and task collaboration contributes 40%. , If the node - and - The unnormalized weights of are 51 and 34 respectively, so the normalized denominator is .

[0115] Data interaction frequency: times / minute; task collaboration demand: Tasks / hour; adjustment factor: , . Unnormalized weight value calculation:

[0116] ;

[0117] Normalized denominator calculation:

[0118] If another node The unnormalized weight value of is 40, so the normalized denominator is . Normalized weight calculation:

[0119] ;

[0120] ;

[0121] This result shows that the node - The weight value of is 1, indicating the most closely cooperating and interacting node pair in the network, while the node - The weight value is 0.784, indicating that its collaboration is slightly lower than - .

[0122] Based on the weight values ​​between edge nodes and paths, all edge nodes are weighted and path optimization is performed. The transmission path is selected first according to the weight, the path information is recorded, and the optimized collaborative path is generated.

[0123] According to the calculated node weight value , the minimum path search algorithm is used to optimize the transmission path between edge nodes. The specific process is as follows: First, the weight values ​​between edge nodes are used as input data to construct a weighted graph. The nodes of the graph represent edge nodes, and the edge weight values ​​are the weights between each node. ; Then, the weight-based Dijkstra algorithm is used to search for the minimum path for each node as the starting point. Starting from the starting node, the path cost to all other nodes is calculated. The path cost is the cumulative value of the weights of each edge. In each iteration, the unvisited node with the smallest current path cost is selected as the next expansion node, and its path information is updated. The above operation is repeated until all nodes are visited. After the path search is completed, the optimal path set from the starting node to the end node is extracted, and all the intermediate nodes passed in the path are recorded. For example, if the node and nodes Weight ,node and nodes Weight , according to the search algorithm, the path with higher weight is selected first, and the path As the main transmission path, as an alternative path; the final output collaborative path set contains the optimal transmission path between all edge nodes, providing a basis for subsequent task allocation and data transmission, and generating optimized collaborative paths.

[0124] See also Figure 7 ,The steps for dynamically adjusting the weights between edge nodes are as follows:

[0125] Based on the optimized collaborative path, the computing load, task execution frequency and data transmission delay in the path are monitored in real time, the performance monitoring data of each node in each path is obtained, and the path performance monitoring data set is generated;

[0126] By obtaining the CPU occupancy rate and memory utilization rate of each edge node in the path, we can obtain the information of the path calculation load. Through the task management log, we analyze the number of tasks completed and the processing interval of each edge node in unit time, extract the task execution frequency data, and use the network traffic monitoring software to record the data packet transmission time between each node in the path to monitor the data transmission delay. For example, in the monitoring process of path ABC, the CPU occupancy rate of node A is 75%, the memory utilization rate is 60%, the task execution frequency is 15 times / minute, and the data transmission delay is recorded as 12 milliseconds. The CPU occupancy rate of node B is 65%, the memory utilization rate is 50%, the task execution frequency is 20 times / minute, and the data transmission delay is 10 milliseconds, forming a performance monitoring data set for each path.

[0127] Based on the path performance monitoring data set, the weights between edge nodes in the path are dynamically adjusted, the path performance priority is evaluated and the weight value is adjusted according to the performance data standardization results, and the dynamically adjusted path weight record is generated;

[0128] First, analyze the priority of each path and the node load distribution in the path weight record. For example, the weight priority of path AB is 0.85, and the task execution frequency is 25 times / minute. The weight priority of path AC is 0.65, and the task execution frequency is 18 times / minute. By comparison, it is found that path AB has a higher priority and better task processing capability. The task grouping is updated through scheduling tools such as Kubernetes, and high-priority tasks are preferentially assigned to path AB, and the task load of path AC is reduced. Further combined with node status data, for example, the CPU occupancy rate of node B is 60%, and the memory utilization rate is 75%, and the CPU occupancy rate of node C is 80%, and the memory utilization rate is 85%. The load balancing algorithm is used to distribute and adjust the number of task executions to reduce the task pressure of node C. Finally, the update of the collaborative task grouping is completed to form a new task allocation table.

[0129] See also Figure 8 , the specific steps for obtaining collaborative task group records are:

[0130] Based on the dynamically adjusted path weight records, the edge node computing load utilization, task execution frequency, and data transmission delay value are collected, the node performance data is analyzed and the task allocation priority is adjusted, the task allocation rules are reconfigured, and the adjusted task collaboration grouping is generated;

[0131] The computing load utilization data is extracted through node performance log records. For example, the current load of the node is calculated based on the ratio of the number of tasks processed per second to the maximum processing capacity of the node. The task execution frequency is extracted by generating a frequency distribution diagram by counting the number of tasks completed per minute. The transmission delay is generated by measuring the sending and returning time of data packets between nodes to generate a network delay data table. These data are compared and analyzed with the current task collaboration weight configuration. When the computing load utilization of a node exceeds 80% or the data transmission delay exceeds 50 milliseconds, its task allocation priority is adjusted, and the allocation weight of low-frequency tasks is increased. For example, node A with lower delay is assigned to the high-frequency task path, while reducing the task allocation priority of node B with higher delay. Finally, the task collaboration grouping between nodes is readjusted according to the analysis results.

[0132] Based on the adjusted task collaboration grouping, optimize the task dependency and data interaction frequency within the collaboration group according to the node weight allocation result, update the task allocation rules and record the node and task information of each collaboration group, and generate collaborative task grouping records;

[0133] The nodes with updated weights are preferentially assigned to the corresponding task groups. For example, when the weight allocation of node A is higher than that of node B, A is preferentially assigned to the high-frequency task collaboration group, and B is assigned to the collaboration group with low computing priority. By parsing the task dependencies in the weight adjustment file, the data interaction frequency and close task association of the nodes in the same collaboration group are ensured. A data record file containing each collaboration group node member, task dependency information and collaboration priority is dynamically generated, and finally a complete collaborative task grouping record is formed for task allocation and execution monitoring of subsequent collaborative paths.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, not to limit them; although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention. In particular, as long as there is no structural conflict, the various technical features mentioned in each embodiment can be combined in any way. The present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.

Claims

1. A carbon metering terminal based on edge intelligence collaboration, characterized in that: The terminal comprises: The model partition update module builds a carbon measurement model based on the real-time collected carbon emission data, and generates carbon measurement information after partition optimization by adjusting and dividing the carbon measurement model; The carbon emission data filtering module extracts the processed carbon emission data from the carbon measurement information after the partition optimization, establishes a data filtering threshold range, compares the data filtering threshold range with all carbon emission data, filters the corresponding carbon emission data according to the comparison result, and generates a filtered carbon emission data set; The collaborative path optimization module identifies the edge nodes of the carbon emission data in the filtered carbon emission data set, constructs a collaborative dependency graph between the edge nodes according to the identification results, analyzes the task collaboration requirements and data interaction frequency between the edge nodes in the collaborative dependency graph to determine the weights between the edge nodes, optimizes the transmission paths between the nodes according to the weights between the edge nodes, and generates an optimized collaborative path; The task dynamic adjustment module monitors the computing load, task execution frequency and data transmission delay of the optimized collaborative path in real time, dynamically adjusts the weights between edge nodes according to the monitoring results, updates the collaborative task grouping between the adjusted edge nodes, and obtains the collaborative task grouping record.

2. The carbon metering terminal based on edge intelligence collaboration according to claim 1 is characterized in that: The steps for obtaining the carbon measurement information after the partition optimization are specifically as follows: Based on the real-time collected carbon emission data, a carbon measurement model is constructed according to the carbon emission data, and the input layer partition of the carbon measurement model is generated by preprocessing the carbon emission data; Extracting processed carbon emission data from the input layer partition of the carbon accounting model and converting it into a time series, extracting time series and statistical features for each data point in the data set, and generating a feature extraction partition of the carbon accounting model; Based on the feature extraction of the carbon measurement model, the time series characteristics and statistical characteristics in the partition are adopted: ； Calculate the predicted carbon emissions at future times , get the carbon measurement information after partition optimization; in, Represents the change in carbon emissions between two consecutive time points. represents the average carbon emissions within the selected time window, is the standard deviation of carbon emission data, express The impact on future carbon emissions forecasts, express The impact on future carbon emissions forecasts, express The impact on future carbon emissions forecasts, Represents the bias term.

3. The carbon metering terminal based on edge intelligence collaboration according to claim 2 is characterized in that: The steps of establishing the data filtering threshold range are specifically as follows: Based on the processed carbon emission data in the carbon measurement information after the partition optimization, the formula is adopted: ； Calculating quantiles of carbon emissions data , according to the quantile Divide the upper and lower bounds of the distribution value; in, is the percentile, is the total number of data points, Represents the total number of data points plus , It is the carbon emission data after sorting. The value at the index position, It is the carbon emission data after sorting. The value at the index position, is the target position of the quantile, Represents the distance between the quantile target position and the lower bound index, Represents the difference between the upper and lower bound indexes of the sorted data; Based on the quantiles and the upper and lower bound distribution values, the selected quantiles are used as initial boundaries, the upper and lower limits are expanded, and a data filtering threshold range is established.

4. The carbon metering terminal based on edge intelligence collaboration according to claim 3 is characterized in that: The steps for obtaining the filtered carbon emission data set are specifically as follows: Based on the data filtering threshold range, the carbon emission data points are compared with the upper and lower limit values ​​one by one, abnormal data points, repetitive data points and data points with a change amplitude lower than a preset amplitude threshold are screened and marked, and the filtered data point classification information is generated; Based on the filtered data point classification information, the marked data points are removed, and the remaining data points are reordered and sorted to generate a filtered carbon emission data set.

5. The carbon metering terminal based on edge intelligence collaboration according to claim 4 is characterized in that: The steps of constructing the collaborative dependency graph between edge nodes are specifically as follows: Based on the filtered carbon emission data, extract the source device information of the carbon emission data, analyze the device's collection frequency, data transmission performance, and computing load, identify the edge nodes in the area according to the device location and function, mark them, and generate an edge node list; Based on the edge node list, data interaction records and task collaboration requirements between edge nodes are extracted, the collaboration strength between nodes is analyzed, and a collaborative dependency graph is constructed through node and collaboration relationships.

6. The carbon metering terminal based on edge intelligence collaboration according to claim 5 is characterized in that: The steps for obtaining the optimized collaborative path are specifically as follows: Based on the task collaboration requirements and data interaction frequency between edge nodes in the collaborative dependency graph, the formula is adopted: ； Computing edge nodes and The weight between ; in, is the first adjustment coefficient, is the second adjustment factor, Is a node and The frequency of data interaction between Is a node and The demand for task collaboration between Based on the weight values ​​between edge nodes and , all edge nodes are weighted and path optimization is performed, transmission paths are preferentially selected according to the weights, path information is recorded, and an optimized collaborative path is generated.

7. The carbon metering terminal based on edge intelligence collaboration according to claim 6 is characterized in that: The step of dynamically adjusting the weights between edge nodes is specifically as follows: Based on the optimized collaborative path, the computing load, task execution frequency and data transmission delay in the path are monitored in real time, the performance monitoring data of each node in each path is obtained, and a path performance monitoring data set is generated; Based on the path performance monitoring data set, the weights between edge nodes in the path are dynamically adjusted, the path performance priority is evaluated and the weight value is adjusted according to the performance data normalization result, and a dynamically adjusted path weight record is generated.

8. The carbon metering terminal based on edge intelligence collaboration according to claim 7 is characterized in that: The steps for obtaining the collaborative task group record are specifically as follows: Based on the dynamically adjusted path weight records, the computing load utilization, task execution frequency and data transmission delay value of the edge node are collected, the node performance data is analyzed and the task allocation priority is adjusted, the task allocation rules are reconfigured, and the adjusted task collaboration grouping is generated; Based on the adjusted task collaboration grouping, the task dependency and data interaction frequency within the collaboration group are optimized according to the node weight allocation result, the task allocation rule is updated, the node and task information of each collaboration group is recorded, and the collaborative task grouping record is generated.

Images