Carbon emission data quality fluctuation identification method based on time sequence correlation analysis

Abstract

The invention discloses a carbon emission data quality fluctuation identification method based on time sequence correlation analysis, and relates to the technical field of data quality monitoring, and the method comprises the steps: interacting a plurality of carbon emission data sources, and obtaining a carbon emission long time sequence data set; performing on-time segmentation on the carbon emission long time sequence data set to obtain an on-time segmentation result; performing time sequence correlation analysis of the comparison time sequence data, determining a plurality of time sequence correlation characteristics of the plurality of carbon emission data sources, and performing longitudinal quality fluctuation evaluation of the plurality of carbon emission data sources; acquiring a plurality of synchronous correlation characteristics of the plurality of carbon emission data sources, performing transverse quality fluctuation evaluation, and outputting a transverse quality fluctuation coefficient; and carrying out data quality fluctuation identification on the plurality of carbon emission data sources. According to the method and the device, the technical problem of insufficient data consistency and accuracy caused by difficulty in effectively identifying quality fluctuation of the multi-source carbon emission data in time and space dimensions in the prior art is solved, and the technical effect of improving the reliability and consistency of the carbon emission data is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data quality monitoring, and in particular to a method for identifying carbon emission data quality fluctuations based on time series correlation analysis. Background Art

[0002] With the development of the carbon trading market, carbon emission data from different sources have become an important basis for carbon management and policy decisions. However, due to the large number of data sources and the different collection frequencies and data accuracy, data quality fluctuations are prone to occur in time and space dimensions, affecting the accuracy and consistency of monitoring results. Traditional quality assessment methods are difficult to capture the dynamic fluctuation characteristics of data and lack effective identification of multi-source data consistency and time series fluctuations. Summary of the invention

[0003] The present application provides a method for identifying quality fluctuations in carbon emission data based on time series correlation analysis, which is used to solve the technical problem that the existing technology is difficult to effectively identify quality fluctuations in multi-source carbon emission data in time and space dimensions, resulting in insufficient data consistency and accuracy.

[0004] In view of the above problems, the present application provides a method for identifying carbon emission data quality fluctuations based on time series correlation analysis.

[0005] The present application provides a method for identifying carbon emission data quality fluctuations based on time series correlation analysis, the method comprising:

[0006] Multiple carbon emission data sources in the interactive carbon emission trading market are monitored for long-term data series to obtain a long-term carbon emission data set, wherein the long-term carbon emission data set includes multiple groups of time series data corresponding to multiple carbon emission data sources; based on the data quality fluctuation identification requirements of the target scenario, a data identification time threshold is defined; the carbon emission long-term data set is segmented at the near-expiry date using the data identification time threshold to obtain a near-expiry segmentation result, wherein the near-expiry segmentation result includes multiple groups of comparison time series data and reference time series data; using the reference time series data as the benchmark data, a time series correlation analysis of the comparison time series data is performed in combination with a sliding window method to determine multiple time series correlation features of multiple carbon emission data sources, and based on the multiple time series correlation features, a longitudinal quality fluctuation evaluation of multiple carbon emission data sources is performed, and a longitudinal quality fluctuation coefficient is output; the multiple time series correlation features are aligned to perform fluctuation consistency analysis, multiple concurrent correlation features of multiple carbon emission data sources are obtained, and a transverse quality fluctuation evaluation of multiple carbon emission data sources is performed, and a transverse quality fluctuation coefficient is output; the longitudinal quality fluctuation coefficient and the transverse quality fluctuation coefficient are combined to identify data quality fluctuations of multiple carbon emission data sources.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The present application interacts with multiple carbon emission data sources in the carbon emission trading market, conducts long-term data monitoring, and obtains a long-term carbon emission data set, wherein the long-term carbon emission data set includes multiple groups of time series data corresponding to multiple carbon emission data sources; defines a data identification time threshold based on the data quality fluctuation identification requirements of the target scenario; performs near-expiry segmentation on the long-term carbon emission data set with the data identification time threshold to obtain near-expiry segmentation results, wherein the near-expiry segmentation results include multiple groups of comparative time series data and reference time series data; uses the reference time series data as the benchmark data, combines the sliding window method to perform time series correlation analysis on the comparative time series data, determines multiple time series correlation features of multiple carbon emission data sources, and performs longitudinal quality fluctuation evaluation on multiple carbon emission data sources based on multiple time series correlation features, and outputs a longitudinal quality fluctuation coefficient; aligns multiple time series correlation features to perform fluctuation consistency analysis, obtains multiple concurrent correlation features of multiple carbon emission data sources, and performs transverse quality fluctuation evaluation on multiple carbon emission data sources, and outputs a transverse quality fluctuation coefficient; combines the longitudinal quality fluctuation coefficient with the transverse quality fluctuation coefficient to identify data quality fluctuations of multiple carbon emission data sources. The present invention solves the technical problem that the prior art is difficult to effectively identify the quality fluctuations of multi-source carbon emission data in time and space dimensions, resulting in insufficient data consistency and accuracy. By monitoring the long time series of carbon emission data, the data identification time threshold and sliding window method are used to perform time series correlation analysis, and the longitudinal and lateral quality fluctuation coefficients are output respectively to comprehensively evaluate the fluctuation consistency of the data, ultimately achieving accurate identification of quality fluctuations of multi-source carbon emission data, and achieving the technical effect of improving the reliability and consistency of carbon emission data. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic flow chart of a method for identifying carbon emission data quality fluctuations based on time series correlation analysis provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the process of obtaining near-expiry segmentation results in the carbon emission data quality fluctuation identification method based on time series correlation analysis provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application provides a method for identifying quality fluctuations in carbon emission data based on time series correlation analysis, which is used to solve the technical problem that the existing technology is difficult to effectively identify the quality fluctuations of multi-source carbon emission data in time and space dimensions, resulting in insufficient data consistency and accuracy. By monitoring the long time series of carbon emission data, the data identification time threshold and sliding window method are used to perform time series correlation analysis, and the longitudinal and lateral quality fluctuation coefficients are output respectively to comprehensively evaluate the fluctuation consistency of the data, ultimately achieving accurate identification of quality fluctuations in multi-source carbon emission data, and achieving the technical effect of improving the reliability and consistency of carbon emission data.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0015] like Figure 1 As shown, the present application provides a method for identifying carbon emission data quality fluctuations based on time series correlation analysis, the method comprising:

[0016] Step S100: Interact with multiple carbon emission data sources in the carbon emission trading market, perform long-term data monitoring, and obtain a carbon emission long-term data set, wherein the carbon emission long-term data set includes multiple groups of time series data corresponding to multiple carbon emission data sources.

[0017] In the embodiment of the present application, in the interactive carbon emission trading market, carbon emission data is first obtained from multiple carbon emission data sources, which include different carbon emission monitoring points or reporting systems such as enterprises, industries, and regions. For these carbon emission data sources, long-term time series data is continuously monitored, that is, carbon emission data at each moment is continuously collected over a long time span, thereby forming a time change trend of carbon emissions.

[0018] During the data collection process, data from different sources are collected in chronological order into a long-time series data set for carbon emissions. This data set includes multiple sets of time series data from multiple carbon emission data sources, and each set of time series data records the carbon emissions of a data source at each time point. The long-time series data set consists of time series of multi-source data, providing time series records of different carbon emission sources.

[0019] Step S200: Based on the data quality fluctuation identification requirements of the target scenario, define a data identification time threshold.

[0020] In an embodiment of the present application, the target scenario of carbon emission monitoring and its related data quality fluctuation identification requirements are first extracted from the business demand database. These requirements are then parsed, including the required monitoring frequency, response time, accuracy requirements, and time range of interest, such as day, week, month, or quarter. After obtaining the requirements, the minimum feasible time interval is automatically calculated based on the update frequency of the data source. For example, if the data source update frequency is one day, the data identification time threshold is set to not less than one day to ensure that each time segment contains enough valid data points to represent the carbon emission fluctuations during that period. Next, the time threshold is automatically adjusted according to the monitoring frequency and accuracy requirements in the scenario requirements. For example, if the demand specifies monthly trend monitoring, the time threshold is set to one month so that fluctuations can be analyzed within the data segment of each month; if the demand is more refined, such as daily or weekly fluctuations, the time threshold is set to one day or one week accordingly.

[0021] Furthermore, the method provided in the application embodiment also includes:

[0022] Based on the data quality fluctuation identification requirements of the target scenario, a data identification time threshold is defined, wherein the data identification time threshold includes multiple time thresholds corresponding to the identification requirements of multiple carbon emission data sources.

[0023] In the embodiment of the present application, the data quality fluctuation identification requirements of the target scenario are first extracted from the business demand database, and the monitoring frequency, response speed and accuracy requirements of the different carbon emission data sources contained therein are analyzed. The data quality fluctuation identification requirements refer to the accuracy and response characteristics that the system needs to meet in identifying carbon emission data fluctuations in a specific scenario, such as requiring the identification of carbon emission fluctuations within a day, week, month or longer time range, or fine monitoring of fluctuation changes in different data sources.

[0024] After analyzing the demand, the specific identification requirements of different carbon emission data sources are identified. For example, some data sources may require daily monitoring to identify short-term changes, while other data sources may focus on monthly or quarterly trends. Therefore, according to the update frequency of different data sources and the identification accuracy required by the demand side, a separate time period monitoring threshold is set for each data source. This threshold defines the time interval based on which each data source should be segmented and monitored in the fluctuation analysis.

[0025] To this end, we first set the lower limit of the data identification time threshold according to the update frequency of each data source. For example, if the update frequency of a data source is every day, ensure that the time threshold of the data source is not less than one day to ensure that each segment contains enough data points to reflect the actual fluctuations of the data. Next, based on the specific identification requirements of each data source, multiple data identification time threshold values ​​are automatically generated, and these values ​​correspond to different data sources. For example, for data sources that require a higher monitoring frequency, the time threshold is set to a smaller time interval, such as one day or one week; and for data sources that focus on long-term trends, a larger time threshold is set, such as one month or one quarter.

[0026] After the above steps, multiple time thresholds are generated, corresponding to the identification requirements of different carbon emission data sources. These data identification time thresholds are called in the subsequent data segmentation and fluctuation identification process to ensure that each data source is segmented and analyzed within an appropriate time interval to meet the monitoring accuracy requirements in different scenarios.

[0027] Step S300: performing near-term segmentation on the carbon emission long time series data set based on the data identification time threshold to obtain near-term segmentation results, wherein the near-term segmentation results include multiple groups of comparison time series data and reference time series data.

[0028] In the embodiment of the present application, the carbon emission long time series data set is first cleaned to remove outliers and missing data to ensure the accuracy and integrity of the data. After the data cleaning is completed, the latest data node in the cleaned data set is selected as the segmentation starting point, that is, the starting position of the data segmentation.

[0029] Then, based on the segmentation starting point, the data is segmented according to the preset data identification time threshold. The data identification time threshold is used as the backtracking step to determine the time range of each data segment. Specifically, multiple data nodes within the step are extracted forward to generate the data for this period. This part of the data is the comparative time series data, which represents the carbon emissions that need to be identified for data quality fluctuations during this time period.

[0030] At the same time, the previous data adjacent to the comparison time series data outside the backtracking step is extracted as the reference time series data. The reference time series data provides a reference basis for fluctuation identification and is used for comparative analysis. It comes from a relatively stable historical data segment and provides a basis for identifying fluctuations in carbon emissions.

[0031] Finally, the entire cleaned carbon emission long time series data set is traversed, and the segmentation is repeated according to the above method to generate multiple sets of comparison time series data and corresponding reference time series data. The collection of these data segments is the result of the near-term segmentation, which contains carbon emission data in different time periods.

[0032] Further, such as Figure 2 As shown, in the method provided in the embodiment of the application, the carbon emission long time series data set is segmented according to the data identification time threshold to obtain the segmentation result, and further includes:

[0033] The carbon emission long time series data set is cleaned, and the latest data node of the carbon emission long time series data set after data cleaning is selected as the segmentation starting point; based on the segmentation starting point, the data identification time threshold is used as the backtracking step, and multiple data nodes within the backtracking step are extracted as the comparison time series data, and multiple data nodes outside the backtracking step are output as the reference time series data; the carbon emission long time series data set after data cleaning is traversed to obtain multiple groups of the comparison time series data and the reference time series data.

[0034] In an embodiment of the present application, the carbon emission long time series data set is first cleaned, and anomaly detection and interpolation techniques are used. For example, the Z-score method is used to identify outliers, and data beyond the standard range is regarded as abnormal; for missing values, linear interpolation or moving average method is applied to fill in. In addition, a time series smoothing algorithm, such as a simple moving average, is used to reduce noise to ensure that the time series data is smoother when analyzed point by point. Through this step, the cleaned carbon emission long time series data set is output. After the cleaning is completed, the latest data node is selected from the cleaned data set as the starting point for segmentation. The segmentation starting point is the initial segmentation position of the data, which is used to start the subsequent near-expiry segmentation process. The latest data node is determined by date and timestamp sorting to ensure that the position of the segmentation starting point is correct, especially when there are multiple sources of data, and the time nodes are unified based on timestamp deduplication and alignment processing.

[0035] Next, with the segmentation starting point as a reference, the set data recognition time threshold is used to trace back in time and extract the time series data in segments. Here, the sliding window technology is used to set the window length to the data recognition time threshold, and the window is gradually traced back from the segmentation starting point. During the backtracking process, the data points in the window are extracted to form the comparison time series data. While extracting the comparison time series data, the data of the previous time period outside the backtracking window is set as the reference time series data. The data of the comparison and reference time periods are quickly extracted through the sequence slicing technology to ensure that each reference time period has the same step length as its corresponding comparison time period.

[0036] Then repeat the above segmentation operation on the cleaned carbon emission long time series data set, slide the window successively and trace back to the beginning of the data set, and generate a pair of comparison time series and reference time series each time. During the traversal process, the start and end time of each pair of data is automatically recorded so that each group of segmented data can be located in chronological order in subsequent analysis. The sliding window traversal based on loop iteration is adopted here to ensure that the segmented data pairs are evenly generated in the entire data set. After the traversal is completed, all the generated comparison time series data and reference time series data are output to form the interim segmentation result, which contains the segmented data in multiple time periods.

[0037] Step S400: Using the reference time series data as the benchmark data, a time series correlation analysis of the comparison time series data is performed in combination with the sliding window method to determine multiple time series correlation features of multiple carbon emission data sources, and based on the multiple time series correlation features, longitudinal quality fluctuation evaluation of multiple carbon emission data sources is performed to output the longitudinal quality fluctuation coefficient.

[0038] In an embodiment of the present application, the reference time series data is first used as the benchmark data, and the sliding window method is applied to compare and analyze the fluctuations in the comparison time series data. Based on the sliding window method, the window is slid in the comparison time series data in sequence to compare the characteristic values ​​of the data points in each window with the reference data segment by segment. The sliding window acts as a continuous sampling tool here. By gradually moving the window, it is possible to extract data in segments and capture subtle fluctuations. In the analysis of each sliding window, multiple time series correlation features are calculated between the comparison time series and the reference time series. Affine transformation analysis is used here to quantify the similarity between the comparison data and the reference data. The affine transformation matrix of the comparison data in the window relative to the reference data is calculated, and the linear and translation features therein are extracted to generate time series correlation features, which can reflect the numerical fluctuations, phase changes and trend differences of the carbon emission data.

[0039] After that, the entire comparison time series is traversed, the time series correlation features of all windows are extracted, and a complete set of correlation features is obtained after summarization. Next, based on these features, the longitudinal quality fluctuation evaluation of each carbon emission data source is performed. The longitudinal quality fluctuation evaluation is to analyze the fluctuation of the same data source in different time periods to determine the stability and consistency of the data source. Specifically, the linear transformation indicator set and the translation transformation indicator set are extracted from the feature set, and the numerical fluctuation coefficient, periodic fluctuation coefficient and phase fluctuation coefficient are calculated. These coefficients represent the fluctuation of carbon emission data in the comparison time period.

[0040] Finally, the numerical fluctuation coefficient, periodic fluctuation coefficient and phase fluctuation coefficient are weighted and processed to output the longitudinal quality fluctuation coefficient of each data source. The weights of the numerical fluctuation coefficient, periodic fluctuation coefficient and phase fluctuation coefficient are pre-set by technical experts, and the sum of the weights of the three coefficients is 1.

[0041] Furthermore, in the method provided in the embodiment of the application, the reference time series data is used as the benchmark data, and the time series correlation analysis of the comparison time series data is performed in combination with the sliding window method to determine multiple time series correlation features of multiple carbon emission data sources, and further includes:

[0042] Based on the data recognition time threshold, a sliding window is configured, wherein the window length of the sliding window is at least twice the length of the data recognition time threshold; the near-term segmentation results are traversed, the first comparison time series data and the first reference time series data are extracted, and the sliding window is applied to the first reference time series data for random sampling to obtain multiple first initial reference data sequences; the affine transformation matrices of the multiple first initial reference data sequences and the first comparison time series data are calculated and a first time series association feature is generated; the near-term segmentation results are traversed, and multiple affine transformation matrices of the multiple comparison time series data and the reference time series data are calculated until multiple time series association features are obtained.

[0043] In the embodiment of the present application, a sliding window is first configured based on a set data identification time threshold, and a technical expert sets the window length to twice the data identification time threshold to cover enough data points. The sliding window moves step by step on the reference time series data, one step at a time, to achieve segment-by-segment sampling.

[0044] Then, the results of the phase segmentation are traversed to extract the first comparison time series data and the corresponding first reference time series data. A sliding window is applied to the first reference time series data to generate multiple first initial reference data sequences by random sampling. In other words, multiple first initial reference data sequences correspond to data sequence segments at different time positions in the first reference time series data, and different initial reference sequences are generated accordingly.

[0045] Then, the affine transformation matrix of the first initial reference data sequence and the first comparison time series data is calculated, and the elements of the affine transformation matrix are extracted to generate the first time series correlation feature. Then, the results of the phase segmentation are traversed continuously, and the same process is repeated for each group of comparison time series data and the corresponding reference time series data, and multiple affine transformation matrices are generated one by one to generate the corresponding time series correlation features.

[0046] Finally, multiple time series correlation features are generated through traversal and sliding window sampling.

[0047] Furthermore, in the method provided in the embodiment of the application, the affine transformation matrix of the first initial reference data sequence and the first comparison time series data is calculated and the first time series correlation feature is generated, and the method further includes:

[0048] Based on the least squares method, taking the first comparison time series data as the affine transformation target, multiple affine transformation matrices of the first initial reference data sequences are calculated; the matrix elements and of the multiple affine transformation matrices are obtained, and the multiple affine transformation matrices are traversed to perform the minimum matrix element and selection; the minimum matrix element and the corresponding affine transformation matrix are set as the time series association feature.

[0049] In the embodiment of the present application, the first comparison time series data is first used as the target sequence of affine transformation, and the least square method is used to calculate the affine transformation matrix of multiple first initial reference data sequences. Through matrix operation optimization fitting, multiple affine transformation matrices are generated, each matrix describing the transformation relationship such as scaling, translation and rotation of a reference sequence relative to the comparison sequence.

[0050] After each affine transformation matrix is ​​generated, extract each element in the matrix and sum them to get the corresponding matrix element sum. The matrix element sum reflects the overall transformation amount of the transformation matrix. The smaller the value, the smaller the fitting error between the two sequences. Record the element sums of all affine transformation matrices for further screening.

[0051] Then, we traverse all generated affine transformation matrices and compare their matrix elements to find the minimum matrix element and the corresponding transformation matrix. By sorting, we lock the minimum matrix element and the associated transformation matrix, which represents the transformation relationship with the minimum error and the best fit.

[0052] Finally, the minimum matrix element and the corresponding affine transformation matrix are set as the temporal correlation feature.

[0053] Furthermore, in the method provided in the embodiment of the application, the longitudinal quality fluctuation evaluation of multiple carbon emission data sources is performed based on multiple time series correlation features, and the longitudinal quality fluctuation coefficient is output, which also includes:

[0054] Analyze the time series correlation features, extract the linear transformation index set and the translation transformation index set; based on the linear transformation index set, calculate the numerical fluctuation coefficient and the periodic fluctuation coefficient of the comparison time series data; based on the translation transformation index set, calculate the phase fluctuation coefficient of the comparison time series data; comprehensively consider the numerical fluctuation coefficient, the periodic fluctuation coefficient and the phase fluctuation coefficient, and weightedly calculate the longitudinal quality fluctuation coefficient.

[0055] In the embodiment of the present application, the time series correlation features are first analyzed to separate the linear transformation index set and the translation transformation index set. The linear transformation index set includes the scaling and rotation parameters in the affine transformation matrix, which are used to analyze the numerical fluctuation and periodic fluctuation characteristics of the data; the translation transformation index set includes the translation parameters, which are used to calculate the phase fluctuation of the data.

[0056] Then, the scaling coefficients in each affine transformation matrix are extracted from the linear transformation index set. The variance of these scaling coefficients is calculated by the variance calculation method to directly obtain the numerical fluctuation coefficient. The variance value reflects the magnitude of the change in the size or range of the comparison time series data. The larger the value, the stronger the fluctuation. The rotation parameters of each affine transformation matrix are further extracted from the linear transformation index set, and the standard deviation calculation method is used to obtain the standard deviation of these rotation parameters as the periodic fluctuation coefficient. A higher standard deviation indicates that the period length of the data is unstable, while a lower standard deviation indicates that the period is more consistent.

[0057] Then, the translation parameters in each affine transformation matrix are extracted from the translation transformation index set, and the mean absolute deviation calculation method is used to measure the differences of these translation parameters and generate the phase fluctuation coefficient. The mean absolute deviation is used to evaluate the amplitude of the phase change of the data on the time axis. A higher value indicates that the periodic displacement of the time series data in time is more significant.

[0058] After obtaining the numerical fluctuation coefficient, periodic fluctuation coefficient and phase fluctuation coefficient, a weighted average is performed to generate the final longitudinal mass fluctuation coefficient. Specifically, each coefficient is weighted according to a preset weight, so that the numerical fluctuation coefficient, periodic fluctuation coefficient and phase fluctuation coefficient respectively affect the final result according to the weight ratio, and the longitudinal mass fluctuation coefficient is obtained.

[0059] Step S500: align multiple time series correlation features to perform fluctuation consistency analysis, obtain multiple contemporaneous correlation features of multiple carbon emission data sources, and evaluate the lateral quality fluctuation of multiple carbon emission data sources, and output the lateral quality fluctuation coefficient.

[0060] In the embodiment of the present application, a plurality of time series correlation features are first statistically analyzed to determine the common fluctuation features under the current data recognition time threshold. Then, the common fluctuation features are combined with the time series correlation features of each carbon emission data source in pairs to generate multiple consistent discrimination groups. These consistent discrimination groups are traversed, and the matrix element variances of the common fluctuation features and time series correlation features in each group are calculated, and the output results are used as the contemporaneous correlation features of each carbon emission data source, thereby reflecting the fluctuation characteristics of each data source in the same time period.

[0061] After obtaining multiple concurrent correlation features, these features are normalized to standardize their data ranges for uniform comparison. Next, based on the normalized concurrent correlation features, a horizontal quality fluctuation evaluation of multiple carbon emission data sources is conducted. In the horizontal evaluation, the fluctuation consistency between the concurrent correlation features of each data source is comprehensively analyzed, and finally the horizontal quality fluctuation coefficient of each carbon emission data source is generated and output.

[0062] Furthermore, in the method provided in the embodiment of the application, a plurality of the time series correlation features are aligned to perform fluctuation consistency analysis, a plurality of concurrent correlation features of a plurality of carbon emission data sources are obtained, and a lateral quality fluctuation evaluation of the plurality of carbon emission data sources is performed, and a lateral quality fluctuation coefficient is output, and the method also includes:

[0063] Perform statistical analysis on multiple time series correlation features to determine common fluctuation features under the current data recognition time threshold; combine the common fluctuation features with multiple time series correlation features in pairs to obtain multiple consistent discrimination groups, and traverse the multiple consistent discrimination groups to calculate the variance of the matrix elements in the common fluctuation features and the time series correlation features, and output them as multiple contemporaneous correlation features; normalize the multiple contemporaneous correlation features to generate multiple lateral quality fluctuation coefficients.

[0064] In an embodiment of the present application, a plurality of time series correlation features are first statistically analyzed to determine the common fluctuation features under the current data recognition time threshold. Specifically, the mean of the time series correlation features of each data source is calculated to obtain a common fluctuation feature representing the overall trend, reflecting the average fluctuation of all data sources in the same time period. The common fluctuation feature refers to the common fluctuation trend or fluctuation standard between different carbon emission data sources in the same time period. This feature is used for horizontal consistency analysis, which aims to identify whether each data source exhibits similar fluctuation patterns in the same time period, thereby measuring their fluctuation consistency.

[0065] Next, the common fluctuation features are combined with the time series correlation features of each data source in pairs to generate multiple consistent discrimination groups. For each consistent discrimination group, the variance calculation method is used to calculate the difference between the common fluctuation feature matrix and the time series correlation feature matrix. The difference is calculated element by element, and the variance of these differences is calculated to generate the variance value of each consistent discrimination group as the contemporaneous correlation feature of the group, which is used to quantify the degree of deviation of each data source from the overall trend in the same time period.

[0066] After obtaining the contemporaneous correlation features of all data sources, these features are normalized to ensure that the features of different data sources are within a unified range for easy comparison. By using the maximum-minimum normalization method, the value of each contemporaneous correlation feature is scaled to between 0 and 1. Finally, by normalizing multiple contemporaneous correlation features, multiple horizontal mass fluctuation coefficients are generated.

[0067] Furthermore, in the method provided in the embodiment of the application, before aligning the plurality of time series correlation features to perform fluctuation consistency analysis and obtaining a plurality of concurrent correlation features of a plurality of carbon emission data sources, the method further includes:

[0068] Calculate the traces of multiple time series correlation features; use the multiple traces as a sorting basis, serialize the multiple time series correlation features in ascending order, and generate a time series correlation feature sequence; perform confidence screening on the time series correlation feature sequence according to a predefined confidence threshold, wherein the confidence threshold defines a selection ratio for confidence screening.

[0069] In the embodiment of the present application, the trace of each time series correlation feature is first calculated, that is, the sum of all elements on the diagonal of the matrix. When calculating the trace, the matrix summation method is used to extract and add only the diagonal elements to obtain the sum value of each feature. This sum value is used as the representative value of each time series correlation feature to compare the overall size of each feature.

[0070] Then, the trace values ​​of all time series correlation features are sorted in ascending order. The sorting process uses a quick sorting algorithm to sort the features from small to large according to the trace value, generating a time series correlation feature sequence with increasing magnitude, so that the features with smaller values ​​(i.e., data sources with smaller fluctuations) are at the front of the sequence, and the features with larger values ​​(data sources with larger fluctuations) are at the end of the sequence.

[0071] After the time-series correlation feature sequence is generated, confidence filtering is applied to filter outliers or extreme fluctuation data in the sequence to ensure the accuracy of common features. Specifically, first determine the screening range based on a predefined confidence threshold, such as the top 95% or exclude ±5%, and set the screening ratio to exclude extreme features at both ends of the sequence. Exemplarily, confidence filtering uses percentile calculations to locate the feature boundaries of the 5th and 95th percentiles of the sequence, and remove features outside these two boundaries, thereby retaining 90% of the main data. This screening eliminates data sources with large or small fluctuations in data quality, ensuring that the remaining features are more reliable.

[0072] Through the above steps, a set of time series correlation features that has been confidence-screened is finally obtained.

[0073] Step S600: Combining the longitudinal quality fluctuation coefficient and the transverse quality fluctuation coefficient, data quality fluctuations of multiple carbon emission data sources are identified.

[0074] In an embodiment of the present application, the longitudinal quality fluctuation coefficient and the transverse quality fluctuation coefficient of each carbon emission data source are comprehensively processed to identify data quality fluctuations. Specifically, the longitudinal and transverse quality fluctuation coefficients are first weighted averaged, and the two are weighted and summed by a preset weight parameter to generate a comprehensive data quality fluctuation index. A threshold judgment is then made on the index, and the data sources are classified according to the size of the fluctuation index to identify the fluctuation of data quality. Data above the threshold are marked as "high volatility" or "abnormal", and data below the threshold are marked as "stable" or "normal". Through this comprehensive analysis, the identification of data quality fluctuations of multiple carbon emission data sources is completed, and the data quality fluctuation status of each data source is obtained.

[0075] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0076] The present application interacts with multiple carbon emission data sources in the carbon emission trading market, conducts long-term data monitoring, and obtains a long-term carbon emission data set, wherein the long-term carbon emission data set includes multiple groups of time series data corresponding to multiple carbon emission data sources; defines a data identification time threshold based on the data quality fluctuation identification requirements of the target scenario; performs near-expiry segmentation on the long-term carbon emission data set with the data identification time threshold to obtain near-expiry segmentation results, wherein the near-expiry segmentation results include multiple groups of comparative time series data and reference time series data; uses the reference time series data as the benchmark data, combines the sliding window method to perform time series correlation analysis on the comparative time series data, determines multiple time series correlation features of multiple carbon emission data sources, and performs longitudinal quality fluctuation evaluation on multiple carbon emission data sources based on multiple time series correlation features, and outputs a longitudinal quality fluctuation coefficient; aligns multiple time series correlation features to perform fluctuation consistency analysis, obtains multiple concurrent correlation features of multiple carbon emission data sources, and performs transverse quality fluctuation evaluation on multiple carbon emission data sources, and outputs a transverse quality fluctuation coefficient; combines the longitudinal quality fluctuation coefficient with the transverse quality fluctuation coefficient to identify data quality fluctuations of multiple carbon emission data sources. The present invention solves the technical problem that the prior art is difficult to effectively identify the quality fluctuations of multi-source carbon emission data in time and space dimensions, resulting in insufficient data consistency and accuracy. By monitoring the long time series of carbon emission data, the data identification time threshold and sliding window method are used to perform time series correlation analysis, and the longitudinal and lateral quality fluctuation coefficients are output respectively to comprehensively evaluate the fluctuation consistency of the data, ultimately achieving accurate identification of quality fluctuations of multi-source carbon emission data, and achieving the technical effect of improving the reliability and consistency of carbon emission data.

[0077] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0079] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for identifying carbon emission data quality fluctuations based on time series correlation analysis, characterized in that: The method comprises: Interact with multiple carbon emission data sources in the carbon emission trading market, conduct long-term data monitoring, and obtain a carbon emission long-term data set, wherein the carbon emission long-term data set includes multiple groups of time series data corresponding to multiple carbon emission data sources; Define the data identification time threshold based on the data quality fluctuation identification requirements of the target scenario; Performing near-term segmentation on the carbon emission long time series data set with the data identification time threshold to obtain near-term segmentation results, wherein the near-term segmentation results include multiple groups of comparison time series data and reference time series data; Taking the reference time series data as the benchmark data, combining with the sliding window method, performing time series correlation analysis on the comparison time series data, determining multiple time series correlation features of multiple carbon emission data sources, and performing longitudinal quality fluctuation evaluation on multiple carbon emission data sources based on the multiple time series correlation features, and outputting a longitudinal quality fluctuation coefficient; Aligning multiple time series correlation features to perform fluctuation consistency analysis, obtaining multiple concurrent correlation features of multiple carbon emission data sources, and evaluating the horizontal quality fluctuation of multiple carbon emission data sources, and outputting a horizontal quality fluctuation coefficient; The longitudinal quality fluctuation coefficient and the transverse quality fluctuation coefficient are combined to identify data quality fluctuations of multiple carbon emission data sources.

2. The carbon emission data quality fluctuation identification method based on time series correlation analysis according to claim 1 is characterized in that: The carbon emission long time series data set is segmented according to the data identification time threshold to obtain the segmentation result, including: Performing data cleaning on the carbon emission long time series data set, and selecting the latest data node of the carbon emission long time series data set after data cleaning as the segmentation starting point; Based on the segmentation starting point, taking the data identification time threshold as the backtracking step, extracting multiple data nodes within the backtracking step as the comparison time series data, and outputting multiple data nodes outside the backtracking step as the reference time series data; The carbon emission long time series data set after data cleaning is traversed to obtain multiple groups of the comparison time series data and the reference time series data.

3. The method for identifying carbon emission data quality fluctuations based on time series correlation analysis according to claim 2, characterized in that: Taking the reference time series data as the benchmark data, the time series correlation analysis of the comparison time series data is performed in combination with the sliding window method to determine multiple time series correlation features of multiple carbon emission data sources, including: Based on the data recognition time threshold, configuring a sliding window, wherein the window length of the sliding window is at least twice the length of the data recognition time threshold; Traversing the proximal segmentation results, extracting first comparison time series data and first reference time series data, and applying the sliding window to the first reference time series data for random sampling to obtain a plurality of first initial reference data sequences; Calculating the affine transformation matrix of the first initial reference data sequences and the first comparison time series data and generating a first time series correlation feature; The near-term segmentation results are traversed, and multiple affine transformation matrices of the multiple comparison time series data and the reference time series data are calculated until multiple time series correlation features are obtained.

4. The method for identifying carbon emission data quality fluctuations based on time series correlation analysis according to claim 3 is characterized in that: Calculating the affine transformation matrix of the first initial reference data sequence and the first comparison time series data and generating a first time series association feature also includes: Based on the least square method, taking the first comparison time series data as an affine transformation target, calculating a plurality of affine transformation matrices of a plurality of the first initial reference data sequences; Get the matrix element sum of multiple affine transformation matrices, traverse multiple affine transformation matrices to select the minimum matrix element sum; The minimum matrix element and the corresponding affine transformation matrix are set as the temporal association feature.

5. The method for identifying carbon emission data quality fluctuations based on time series correlation analysis according to claim 4, characterized in that: Based on the multiple time series correlation features, longitudinal quality fluctuation evaluation of multiple carbon emission data sources is performed, and a longitudinal quality fluctuation coefficient is output, including: Analyze the time series correlation features and extract a linear transformation index set and a translation transformation index set; Based on the linear transformation indicator set, calculating the numerical fluctuation coefficient and the periodic fluctuation coefficient of the comparison time series data; Based on the translation transformation indicator set, calculating the phase fluctuation coefficient of the comparison time series data; The longitudinal quality fluctuation coefficient is weightedly calculated by comprehensively considering the numerical fluctuation coefficient, the periodic fluctuation coefficient and the phase fluctuation coefficient.

6. The method for identifying carbon emission data quality fluctuations based on time series correlation analysis according to claim 5, characterized in that: Align multiple time series correlation features to perform fluctuation consistency analysis, obtain multiple concurrent correlation features of multiple carbon emission data sources, and perform lateral quality fluctuation evaluation of multiple carbon emission data sources, and output lateral quality fluctuation coefficients including: Performing statistical analysis on a plurality of the time series correlation features to determine common fluctuation features under the current data recognition time threshold; Combining the common fluctuation features with the multiple time series correlation features in pairs to obtain multiple consistent discrimination groups, and traversing the multiple consistent discrimination groups to calculate the variance of the matrix elements in the common fluctuation features and the time series correlation features, and outputting them as multiple concurrent correlation features; Normalize the multiple concurrent correlation features to generate multiple lateral quality fluctuation coefficients.

7. The carbon emission data quality fluctuation identification method based on time series correlation analysis according to claim 1, aligning multiple time series correlation features to perform fluctuation consistency analysis, and obtaining multiple concurrent correlation features of multiple carbon emission data sources, before that, comprising: Calculating traces of a plurality of said time series correlation features; Based on the plurality of traces as sorting basis, the plurality of time series correlation features are serialized in ascending order to generate a time series correlation feature sequence; The time series correlation feature sequence is confidence screened according to a predefined confidence threshold, wherein the confidence threshold defines a selection ratio for confidence screening.

8. The method for identifying carbon emission data quality fluctuations based on time series correlation analysis according to claim 1, characterized in that: Based on the data quality fluctuation identification requirements of the target scenario, a data identification time threshold is defined, wherein the data identification time threshold includes multiple time thresholds corresponding to the identification requirements of multiple carbon emission data sources.

Images