Carbon footprint factor calculation method and system

Abstract

The application discloses a kind of carbon footprint factor calculation method and system, method includes: obtaining carbon emission source data, based on waveform form and energy distribution characteristics carry out multistage clustering and obtain carbon emission characteristic set;Extract target time sequence fragment in set to meet integrity condition, adopt multi-scale decomposition and fragment assembly to extract carbon fingerprint gene fragment and characteristic parameter, construct carbon fingerprint gene library;According to product production track, match carbon fingerprint gene sequence from gene library, carry out time sequence consistency check and conflict resolution;Accumulate the carbon emission contribution of each gene fragment, and obtain product carbon footprint factor using double-layer weighting correction strategy based on historical distribution and gene proportion.The application realizes the accurate separation and explainable accounting of carbon emission source by decomposing carbon emission waveform into pulse type, oscillation type and ladder type gene fragments, which significantly improves the calculation efficiency and accuracy of large-scale product carbon footprint.

Description

Technical Field

[0001] This invention belongs to the field of carbon emission accounting technology, and in particular relates to a method and system for calculating carbon footprint factors. Background Technology

[0002] The carbon footprint factor is the basic data for calculating the carbon emissions of a product throughout its entire life cycle. Its accuracy and calculation efficiency directly affect the credibility of carbon emission accounting.

[0003] Currently, carbon footprint factor calculation mainly employs the following technical solutions: First, the process-based carbon emission factor accumulation method, which tracks the energy and material consumption of a product at each process node, multiplies them by the corresponding carbon emission factor, and then accumulates them to obtain the product's carbon footprint. This method requires individual calculation for each product, resulting in a large computational load and inability to handle emission allocation issues when multiple products are produced on the same production line. Second, the system-wide total emission inversion allocation method, which solves for the carbon footprint factor of each product by constructing a system of linear equations. This method has high data quality requirements, lacks physical constraints in the solution process, and yields poor stability.

[0004] The existing technology has the following technical defects: First, it is difficult to separate the mixed emissions of carbon emission sources, and it is impossible to accurately identify the contribution of each product to the total emissions of the system; second, the carbon footprint factor calculation process lacks interpretability, and the results are difficult to trace and verify; third, it is inefficient in calculating the carbon footprint of massive products in large-scale production scenarios. Summary of the Invention

[0005] This invention provides a method and system for calculating carbon footprint factors, which solves the technical problem of difficulty in separating mixed emissions from carbon emission sources in the prior art.

[0006] In a first aspect, the present invention provides a method for calculating carbon footprint factors, comprising:

[0007] Acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features.

[0008] A target time sequence segment that meets the gene integrity condition is extracted from a certain carbon emission feature set. The target time sequence segment is then decomposed and assembled at multiple scales according to a preset carbon fingerprint gene sequencing strategy. At least one carbon fingerprint gene segment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene segment are extracted. The carbon fingerprint gene segment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time.

[0009] A carbon fingerprint gene library is constructed based on the at least one carbon fingerprint gene fragment and the gene characteristic parameters corresponding to the at least one carbon fingerprint gene fragment;

[0010] The production trajectory data of the target product during the monitoring period is obtained, and based on the production trajectory data, the carbon fingerprint gene sequence corresponding to the production trajectory data is matched from the carbon fingerprint gene library. The carbon fingerprint gene sequence is then subjected to time consistency verification and conflict resolution to obtain the target carbon fingerprint gene sequence.

[0011] The carbon emission contributions corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence are summed to obtain the summed result. The summed result is then corrected using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

[0012] Secondly, the present invention provides a carbon footprint factor calculation system, comprising:

[0013] The clustering module is configured to acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features.

[0014] The extraction module is configured to extract a target time sequence segment that meets the gene integrity condition from a certain carbon emission feature set, and to perform multi-scale decomposition and fragment assembly on the target time sequence segment according to a preset carbon fingerprint gene sequencing strategy, thereby extracting at least one carbon fingerprint gene fragment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene fragment, wherein the carbon fingerprint gene fragment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time.

[0015] The construction module is configured to construct a carbon fingerprint gene library based on the at least one carbon fingerprint gene fragment and the gene feature parameters corresponding to the at least one carbon fingerprint gene fragment;

[0016] The matching module is configured to acquire production trajectory data of the target product during the monitoring period, and match the carbon fingerprint gene sequence corresponding to the production trajectory data from the carbon fingerprint gene library based on the production trajectory data, and perform time-series consistency verification and conflict resolution on the carbon fingerprint gene sequence to obtain the target carbon fingerprint gene sequence.

[0017] The correction module is configured to accumulate the carbon emission contribution corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence to obtain the accumulation result, and to correct the accumulation result using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

[0018] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the carbon footprint factor calculation method of any embodiment of the present invention.

[0019] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the carbon footprint factor calculation method of any embodiment of the present invention.

[0020] The carbon footprint factor calculation method and system of this application, Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart of a carbon footprint factor calculation method provided in an embodiment of the present invention;

[0023] Figure 2 This is a structural block diagram of a carbon footprint factor calculation system provided in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 The flowchart shown here illustrates a method for calculating the carbon footprint factor according to this application.

[0027] like Figure 1 As shown, the carbon footprint factor calculation method specifically includes the following steps:

[0028] Step S101: Obtain carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features.

[0029] In this step, the carbon emission source data is segmented using a preset sliding window to obtain multiple time-series segments;

[0030] Multiple sliding sub-windows of different time lengths are used to extract the morphological features of a certain time segment, resulting in a morphological feature vector of a certain time segment at multiple time scales. The morphological feature vector includes a rising edge slope, a falling edge slope, a peak duration, and a peak-to-valley ratio.

[0031] Wavelet packet transform is used to decompose a time segment into multiple frequency band components. The energy proportion of each frequency band component is calculated as the first energy feature, the energy centroid frequency of all frequency band components is calculated as the second energy feature, and the variance of the energy distribution of each frequency band component is calculated as the third energy feature. The first energy feature, the second energy feature and the third energy feature are combined into a certain energy distribution feature vector of a certain time segment.

[0032] By concatenating the morphological feature vector with the energy distribution feature vector, a multidimensional feature vector of the time series segment is constructed.

[0033] Based on the multidimensional feature vectors of each time series segment, density clustering algorithm is used for the first-level clustering, dividing each time series segment into multiple initial clusters;

[0034] For each time-series segment within an initial cluster, a self-organizing map network is used for second-level clustering to further divide the initial cluster into multiple sub-clusters, resulting in at least one set of carbon emission features.

[0035] In one specific embodiment, carbon emission source data is acquired through a Continuous Emission Monitoring System (CEMS) deployed at each emission point (such as chimneys and exhaust ducts) of the target production system, with a sampling frequency of 1 Hz and a monitoring period of 24 consecutive hours. The collected raw data is a time series, denoted as... ,in, express Carbon emission rate at any given time (unit: kg CO2 / h).

[0036] The carbon emission source data X is segmented using a preset sliding window. The window length L is set to 300 sampling points (i.e., 5 minutes), and the sliding step size S is set to 60 sampling points (i.e., 1 minute). Multiple time series segments are obtained by traversing the entire time series through the sliding window. Each segment contains L consecutive data points. The window length and sliding step size can be adaptively adjusted according to the actual fluctuation characteristics of the carbon emission source; typical values ​​are selected in this embodiment.

[0037] For each time segment Three sliding sub-windows of different time lengths are used to extract morphological features to capture waveform variation patterns at different time scales:

[0038] The first sub-window, with a length of 10 sampling points (10 seconds), is used to extract local transient features. Within this sub-window, the rising edge slope is calculated using the maximum rate of change from the trough to the crest within the sub-window; the falling edge slope is calculated using the maximum rate of change from the crest to the trough.

[0039] The second sub-window, with a length of 30 sampling points (30 seconds), is used to extract short- to medium-term fluctuation characteristics. Within this sub-window, consecutive peaks are identified and the duration between adjacent peaks is calculated. The average of all durations is taken as the peak duration.

[0040] The third sub-window, with a length of 60 sampling points (1 minute), is used to extract long-term trend features. Within this sub-window, the peak-to-valley ratio is calculated: the ratio of the average of all peak intensities to the average of all valley intensities within the sub-window.

[0041] The feature values ​​extracted from the three sub-windows are concatenated in sequence to obtain the time series segment. Multi-timescale morphological feature vectors =[Rising edge slope, Falling edge slope, Peak duration, Peak-to-valley ratio].

[0042] Wavelet packet transform for time segments A three-level decomposition was performed, using the db4 wavelet basis function. After the three-level decomposition, eight frequency band components were obtained. Each frequency band component corresponds to a different frequency range.

[0043] Energy percentage: For the j-th frequency band component, calculate its energy. ,in For the k-th coefficient of the j-th frequency band component, Let be the number of frequency band coefficients. Then the energy percentage of the j-th frequency band component is: The energy percentage of all 8 frequency bands is taken as the first energy feature, resulting in a total of 8 dimensions.

[0044] Energy centroid frequency: The weighted average frequency for calculating all frequency band components. ,in Let be the center frequency of the j-th frequency band. This value serves as the second energy feature, and has one dimension.

[0045] Energy distribution variance: Calculate the variance of the energy proportion of each frequency band. ,in This is the average of the energy proportions across the eight frequency bands. This value serves as the third energy characteristic, comprising one dimension.

[0046] Combining the first energy feature (8 dimensions), the second energy feature (1 dimension), and the third energy feature (1 dimension) yields a time series segment. Energy distribution characteristic vector It has 10 dimensions.

[0047] morphological feature vectors (4-dimensional) and energy distribution eigenvector (10-dimensional) concatenation of the first and last parts yields a 14-dimensional multidimensional feature vector for each time segment. .

[0048] Multidimensional feature vectors of all time segments As input, the DBSCAN density clustering algorithm is used for the first-level clustering. Algorithm parameters are set as follows: the neighborhood radius ε is the 20th quantile of the pairwise Euclidean distance between all feature vectors, and the minimum number of neighborhood points is 5. The clustering results divide the time segment into multiple initial clusters. Fragments within each cluster are similar in morphology and energy distribution. Noise points that do not belong to any cluster are either marked as independent clusters or discarded.

[0049] For each initial cluster The system uses the temporal segment feature vectors within the network as input to train a self-organizing map (SOM) network. The SOM network output layer employs a 5×5 two-dimensional grid structure, and the training iterations are set to 200. After training, the SOM network maps the input vectors to different neurons in the output layer, with segments having similar features mapped to adjacent neurons. Based on the aggregation of neurons in the output layer, each initial cluster is further divided into multiple sub-clusters, each sub-cluster corresponding to a carbon emission feature set.

[0050] Finally, the set of all subclusters is at least one carbon emission feature set output in step S101. The time segments in the same carbon emission feature set not only have similar waveform morphology (such as steep rising edges and short peak duration), but also have similar energy distribution (such as high proportion of high frequency components), and physically correspond to the same type of carbon emission source operation mode (such as pulse emission generated by equipment start-up and shutdown, oscillating emission generated by process fluctuations, and stepped emission generated by material batch changes).

[0051] In summary, the following technical effects were achieved through the above multi-level clustering steps:

[0052] Simultaneously employing morphological features (time domain) and energy distribution features (frequency domain), the intrinsic patterns of carbon emission waveforms are characterized from different dimensions. Multi-timescale morphological feature extraction can capture complete waveform changes from second-level transients to minute-level trends; while wavelet packet transform multi-band energy analysis reveals the energy distribution characteristics of carbon emission signals in different frequency ranges. The two complement each other, significantly improving the distinguishability of feature expression.

[0053] The first-level density clustering does not require a preset number of categories, can automatically identify the natural cluster structure in the data, and is robust to noise points. The second-level self-organizing map network performs finer-grained division within each initial cluster, solving the problem that density clustering cannot further distinguish when there are substructures within the cluster. The two-level clustering refines the data layer by layer, avoiding the defects of single clustering methods being too coarse or too fine, and achieving adaptive, multi-level identification of carbon emission patterns.

[0054] Each carbon emission feature set obtained by clustering corresponds to a time series segment with the same carbon emission source action mode, providing high-quality and highly consistent input data for the subsequent steps of "extracting target time series segments" and "extracting carbon fingerprint gene segments". This data-driven grouping method, compared with traditional manual division or fixed threshold segmentation, can more objectively and accurately reflect the dynamic change pattern of carbon emissions in the actual production process.

[0055] Step S102: Extract a target time sequence segment that meets the gene integrity condition from a certain carbon emission feature set, and perform multi-scale decomposition and fragment assembly on the target time sequence segment according to a preset carbon fingerprint gene sequencing strategy to extract at least one carbon fingerprint gene fragment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene fragment. The carbon fingerprint gene fragment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time.

[0056] In this step, the time segments in a certain carbon emission feature set are arranged in chronological order to obtain a time segment sequence;

[0057] The earliest time segment is selected as a candidate segment, and each candidate segment is sequentially judged to meet a preset gene integrity condition, wherein the gene integrity condition includes:

[0058] Data integrity condition: The number of data points in the candidate segment is not less than the preset minimum number of points threshold;

[0059] Waveform purity condition: The signal-to-noise ratio of the candidate segment is greater than the preset signal-to-noise ratio threshold, where the signal-to-noise ratio is calculated by the ratio of signal power to noise power;

[0060] Boundary smoothness condition: The absolute values ​​of the waveform slopes at the start and end points of the candidate segment are both less than a preset slope threshold;

[0061] Internal consistency condition: The waveform similarity of each sub-segment within a candidate segment is not lower than a preset similarity threshold, wherein the waveform similarity is calculated by dynamic time warping distance;

[0062] If a candidate fragment meets the preset gene integrity conditions, then the candidate fragment is determined as the target time sequence fragment;

[0063] If a candidate fragment does not meet the preset gene integrity conditions, the next time sequence fragment is selected as a new candidate fragment in chronological order, and the gene integrity conditions are re-evaluated until a target time sequence fragment that meets the gene integrity conditions is determined.

[0064] Furthermore, wavelet packet decomposition is performed on the target time segment to obtain low-frequency trend components, mid-frequency fluctuation components, and high-frequency transient components;

[0065] Change points are detected on the low-frequency trend component, and all time points when the slope changes on the low-frequency trend component exceed a preset threshold are taken as change points. All change points are arranged in chronological order, and the time interval between each two adjacent change points is determined as a ladder-type candidate gene fragment.

[0066] Sliding window analysis is performed on the mid-frequency fluctuation components to calculate the local standard deviation and local autocorrelation coefficient of the mid-frequency fluctuation components within each window. Continuous window intervals with local standard deviations less than a first preset threshold and local autocorrelation coefficients greater than a second preset threshold are identified as oscillating candidate gene fragments.

[0067] Pulse detection is performed on the high-frequency transient components, and the peak intervals in the high-frequency transient components whose amplitude exceeds the third preset threshold and whose duration is less than the fourth preset duration are identified as pulse-type candidate gene fragments.

[0068] The ladder-type candidate gene fragments, oscillating candidate gene fragments, and pulse-type candidate gene fragments are arranged in chronological order to obtain candidate gene fragment sequences;

[0069] Traverse the candidate gene fragment sequence, and when the time interval between two adjacent candidate gene fragments is less than a preset merging threshold, merge the two adjacent candidate gene fragments into one candidate gene fragment;

[0070] For each merged candidate gene fragment, the maximum value of all data points within the merged candidate gene fragment is extracted as the peak intensity, the integral value of the merged candidate gene fragment on the time axis is extracted as the integral area, and the difference between the end time and the start time of the merged candidate gene fragment is extracted as the waveform duration, which are used as the gene feature parameters of the carbon fingerprint gene fragment corresponding to the merged candidate gene fragment.

[0071] In one specific embodiment, suppose that a certain carbon emission feature set obtained from step S101 contains multiple time-series segments. Arranged in chronological order (time nodes are determined by the start time of the segment). Select the time segment with the earliest time node. As candidate fragments, we sequentially determine whether they meet the following four gene integrity conditions:

[0072] Data integrity condition: Check the number of data points in the candidate segment. Is it not less than the preset minimum number of points threshold? In this embodiment, .like If so, the condition is not met.

[0073] Waveform purity condition: Calculate the signal-to-noise ratio of the candidate segment waveform. Signal power Take the mean of the squares of all data points within the candidate segment; noise power The variance of the candidate segments is obtained after first-order differencing (removing the trend). In this embodiment, the preset signal-to-noise ratio threshold is 20dB. If If dB, then the condition is not met.

[0074] Boundary smoothness condition: Calculate the average absolute value of the waveform slope at the starting points (first 5 points) of the candidate segment. The average absolute value of the waveform slope at the end point (the last 5 points) The waveform slope is calculated using the difference between adjacent points. The preset slope threshold is 0.5 (dimensionless, after normalization). or If so, the condition is not met.

[0075] Internal consistency condition: Divide the candidate fragment into 4 equal sub-segments (each segment length is...). The dynamic time warping (DTW) distance between each adjacent segment is calculated. The maximum DTW distance of all adjacent segments is taken as the internal inconsistency. The upper limit of the DTW distance corresponding to the preset similarity threshold is 0.15 (after feature normalization). If the maximum DTW distance > 0.15, the condition is not met.

[0076] If a candidate segment meets all four conditions mentioned above, it is identified as a "target time series segment"; otherwise, the next time series segment is selected in chronological order. As a new candidate segment, the above judgment is repeated until a segment that meets all conditions is found. In practical applications, if all segments in a certain carbon emission feature set do not meet the integrity condition, the set is preprocessed (such as smoothing filtering, boundary clipping, and resampling) and then judged again, which will not be elaborated here.

[0077] For the target time segment determined above (denoted as...) Perform the following gene sequencing operations.

[0078] Wavelet packet decomposition: using db4 wavelet basis functions to... Three-level wavelet packet decomposition is performed, resulting in eight frequency band components. Based on the center frequency range of each band, the frequency band components are recombined into three components:

[0079] Low-frequency trend component: Combines the two frequency bands with the lowest center frequency (corresponding to 0~0.125π rad / sample), reflecting long-term slow changes.

[0080] Mid-frequency fluctuation component: Combines the four middle frequency bands (corresponding to 0.125π~0.5π rad / sample) to reflect periodic fluctuations.

[0081] High-frequency transient components: combining the two highest frequency bands (corresponding to 0.5π~π rad / sample) to reflect peak transients.

[0082] Step-type candidate gene fragment detection: Change points are detected for low-frequency trend components. The PELT (Pruned Exact Linear Time) algorithm is used to find slope change points, with a penalty parameter set to 0.5. All time points where the slope change exceeds a preset threshold (threshold = 0.2 in this embodiment) are identified as change points. All change points are arranged in chronological order, and the time interval between any two adjacent change points (including the start and end points) is determined as a step-type candidate gene fragment. Each fragment corresponds to a carbon emission step caused by a change in material batch.

[0083] Detection of oscillating candidate gene fragments: Sliding window analysis was performed on the mid-frequency fluctuation component. The window length was set to 30 sampling points (corresponding to 2-3 times the typical oscillation period), and the sliding step size was 5 sampling points. Local standard deviations were calculated within each window. and local autocorrelation coefficient (Lag by 1 cycle). In this embodiment, the first preset threshold... The second preset threshold . Will continuously satisfy and The window intervals were merged to serve as oscillating candidate gene fragments. Each fragment corresponds to a periodic change in carbon emissions caused by process fluctuations.

[0084] Pulse-type candidate gene fragment detection: Pulse detection is performed on high-frequency transient components. A threshold-based method is used to identify all regions where the amplitude (absolute value) exceeds a third preset threshold (three times the component standard deviation in this embodiment), and the duration of this region is less than a fourth preset duration (20 sampling points in this embodiment). These regions are identified as pulse-type candidate gene fragments. Each fragment corresponds to a transient impact on carbon emissions caused by a single equipment start-up and shutdown.

[0085] Candidate gene fragment merging: All detected ladder-like, oscillating, and pulse-like candidate gene fragments are sorted according to their start time to obtain candidate gene fragment sequences. Traverse the candidate gene fragment sequences; if two adjacent fragments... and The time interval between ( Start time - If the end time is less than the preset merging threshold (10 sampling points in this embodiment), then the two will be merged into a new segment: the start time is... The start time and end time are taken as... The end time is marked as "composite". After merging, continue checking backwards.

[0086] Gene feature parameter extraction: For each candidate gene fragment obtained after merging (hereinafter referred to as "current fragment"), the following feature extraction operation is performed. The extraction result is used as the gene feature parameters of the carbon fingerprint gene fragment corresponding to the current fragment. Here, "the carbon fingerprint gene fragment corresponding to the current fragment" refers to the data structure formed after the current fragment itself has been parameterized. This data structure contains the waveform data of the fragment and its feature parameters, which are used for subsequent storage in the carbon fingerprint gene library or for matching calculations. The specific extraction steps are as follows:

[0087] Peak intensity: The maximum value among all data points in the current segment, denoted as . For pulse-type segments, this value reflects the maximum intensity of the instantaneous impact of carbon emissions; for oscillating segments, this value is the maximum value of the upper envelope of the oscillating waveform; for step-type segments, this value is the steady-state value after the step.

[0088] Area of ​​integration: Calculate the definite integral of the current segment on the time axis, denoted as Numerical integration is performed using the trapezoidal rule: = ,in The number of data points contained in the current segment. For the first The magnitude of each data point The sampling time interval (in this embodiment) (seconds). The integral area reflects the total carbon emissions corresponding to this segment.

[0089] Waveform duration: Calculate the difference between the end time and the start time of the current segment, denoted as . Let the start timestamp of the current segment be... (Time corresponding to the first data point within the segment), end timestamp is (The time corresponding to the last data point in the segment), then The unit is seconds. This parameter reflects the duration of a carbon emission event.

[0090] Ultimately, each candidate gene fragment is converted into a carbon fingerprint gene fragment, along with its gene type (pulse / oscillatory / stepwise / complex) and the aforementioned characteristic parameters.

[0091] In summary, by merging adjacent gene segments with too close an interval, we can avoid over-segmentation caused by short pauses, ensuring that each gene segment fully reflects an independent carbon emission event. The physical meaning of the merged gene segments is more complete, and the subsequent cumulative results of carbon emission contributions are more accurate.

[0092] The extracted peak intensity, integral area, and waveform duration parameters can not only fully describe the carbon emission characteristics of carbon fingerprint gene fragments, but also facilitate subsequent clustering, library construction, and matching calculations, achieving a balance between information preservation and computational efficiency.

[0093] Step S103: Construct a carbon fingerprint gene library based on the at least one carbon fingerprint gene fragment and the gene characteristic parameters corresponding to the at least one carbon fingerprint gene fragment.

[0094] In this step, the gene type identifier of at least one carbon fingerprint gene fragment is obtained. Carbon fingerprint gene fragments with gene type identifier of pulse type are classified into the pulse type gene fragment set, carbon fingerprint gene fragments with gene type identifier of oscillation type are classified into the oscillation type gene fragment set, and carbon fingerprint gene fragments with gene type identifier of ladder type are classified into the ladder type gene fragment set.

[0095] For a set of pulsed gene fragments, equal-width binning is performed using peak intensity and waveform duration as classification features to obtain multiple pulsed gene subclasses. The dynamic time warping center of all carbon fingerprint gene fragments within a certain pulsed gene subclass is calculated on the waveform data as the first standard waveform morphology. The arithmetic mean of the peak intensity, waveform duration, and integral area of ​​all carbon fingerprint gene fragments within a certain pulsed gene subclass is calculated as the standard peak intensity, standard duration, and standard integral area. The first standard waveform morphology, the standard peak intensity, the standard duration, and the standard integral area are combined to generate a certain pulsed standard gene template, and the certain pulsed standard gene template is stored in the carbon fingerprint gene library.

[0096] For a set of oscillating gene fragments, equal-width binning is performed using oscillation period and oscillation amplitude as classification features to obtain multiple oscillating gene subclasses. The dynamic time warping center of all carbon fingerprint gene fragments within a certain oscillating gene subclass is calculated on the waveform data as the second standard waveform morphology. The arithmetic mean of the oscillation period and oscillation amplitude of all carbon fingerprint gene fragments within a certain oscillating gene subclass is calculated as the standard oscillation period and standard oscillation amplitude. The standard waveform morphology, the standard oscillation period, and the standard oscillation amplitude are combined to generate a certain oscillating standard gene template, and the certain oscillating standard gene template is stored in the carbon fingerprint gene library.

[0097] For a set of step-type gene fragments, equal-width binning is performed using step amplitude and response time as classification features to obtain multiple step-type gene subclasses. The dynamic time warping center of all carbon fingerprint gene fragments in a certain step-type gene subclass is calculated on the waveform data as the third standard waveform morphology. The arithmetic mean of the step amplitude and response time of all carbon fingerprint gene fragments in a certain step-type gene subclass is calculated as the standard step amplitude and standard response time. The third standard waveform morphology, the standard step amplitude, and the standard response time are combined to generate a certain step-type standard gene template, and the certain step-type standard gene template is stored in the carbon fingerprint gene library.

[0098] In one specific embodiment, the gene type identifier (i.e., pulse, oscillatory, or ladder-like) of each carbon fingerprint gene fragment output in step S102 is obtained. Three empty sets are then established:

[0099] Pulse-type gene fragment set Stores carbon fingerprint gene fragments of all gene types identified as "pulse type".

[0100] Oscillating gene fragment set Stores carbon fingerprint gene fragments of all gene types identified as "oscillating".

[0101] Stepped gene fragment set Stores carbon fingerprint gene fragments that identify all gene types as "ladder type".

[0102] If a segment is marked as "composite" in step S102, it is classified into the closest single type according to the main characteristics of its waveform (such as peak intensity ratio, duration ratio).

[0103] The pulsed gene fragment processing specifically involves:

[0104] For pulse-type gene fragment sets Each pulsed gene fragment has two key characteristics: peak intensity. and waveform duration .Will All pulsed gene fragments are mapped to x-axis In the two-dimensional feature space with the ordinate as the vertical axis, set the parameters for equal-width binning:

[0105] Peak intensity number of bins Width of each compartment ,in , They are respectively The maximum and minimum peak intensities of all pulsed gene fragments.

[0106] Waveform duration bin number Width of each compartment ,in , They are respectively The maximum and minimum durations of waveforms for all pulsed gene fragments.

[0107] Each pulse-type gene fragment is arranged according to its Falling grid cell (i.e., the first) The peak intensity sub-box and the first The intersection of duration bins is assigned to the same pulse-type gene subclass. All fragments falling into the same grid cell constitute a subclass. If a grid cell contains no fragments, that subclass is skipped.

[0108] For the kth pulse-type gene subclass Suppose it contains Each carbon fingerprint gene fragment has a waveform data sequence. Each waveform data sequence is a time series (the length may vary). The Dynamic Time Warping (DTW) centering algorithm is used to calculate the standard waveform shape. This algorithm solves the following optimization problem:

[0109] ,

[0110] in, To obtain the optimal central waveform, i.e., the standard waveform morphology of this pulsed gene subclass, it is a time series whose waveform length may differ from that of other segments within the subclass. Typically, the length is taken as the median or mean of the lengths of all segments within the subclass. Each element represents the standardized amplitude at the corresponding time.

[0111] The mathematical symbol represents finding the independent variable that minimizes the following expression. , The candidate center waveform to be solved;

[0112] Candidate center waveform The first in the pulse-type gene subclass Waveform of a carbon fingerprint gene fragment The dynamic time-normalized distance between them;

[0113] In practical algorithm implementations, this optimization problem is typically solved approximatingly using the iterative averaging method, rather than directly solving the aforementioned minimization expression. The iterative steps are as follows:

[0114] Randomly select a waveform from the subclasses as the initial center. ;

[0115] For the In the nth iteration, calculate the... The center of the next iteration With each DTW alignment path;

[0116] Based on the alignment path, all The new center is obtained by averaging (or taking the median) the corresponding points. ;

[0117] Repeat the above steps until the change in the center waveform is less than the preset threshold.

[0118] The final convergence obtained That is, as the kth pulse-type gene subclass Standard waveform This method is computationally stable and can handle sequences of unequal length, and has been widely used in time series clustering and centroid computation.

[0119] For the kth pulse-type gene subclass Calculate the arithmetic mean of the following three characteristic parameters:

[0120] Standard peak intensity ,in For the first Peak intensity of each segment;

[0121] Standard duration ,in For the first The duration of the waveform in each segment;

[0122] Standard integral area ,in For the first The integral area of ​​each segment.

[0123] The above calculation results are combined into a pulse-type standard gene template. :

[0124] ={Type=Pulse,Standard Waveform= Standard peak intensity = Standard duration = Standard integral area = };

[0125] The pulse-type standard gene template was stored in the carbon fingerprint gene bank. .

[0126] The specific treatment of oscillating gene fragments is as follows:

[0127] For oscillating gene fragment sets Each oscillating gene segment has an oscillation period. and oscillation amplitude Two key features: The oscillation period is obtained by calculating the average interval between adjacent peaks in the waveform of the oscillating gene segment; the oscillation amplitude is taken as the average difference between the peaks and troughs within the oscillating gene segment. An equal-width binning method is used, and the number of bins for the oscillation period is... oscillation amplitude in bins Mapping oscillating gene fragments to Multiple oscillating gene subclasses are obtained in the two-dimensional grid.

[0128] For the k-th oscillating gene subclass The standard waveform shape is calculated using the same DTW center algorithm as the pulse type. .

[0129] Calculate the k-th oscillating gene subclass The arithmetic mean of the oscillation periods of all oscillating gene segments is used as the standard oscillation period. The arithmetic mean of the oscillation amplitudes is calculated as the standard oscillation amplitude. .

[0130] Oscillating standard gene template for:

[0131] ={Type=Oscillating, Standard Waveform Shape= Standard oscillation period = Standard oscillation amplitude = };

[0132] The oscillating standard gene template was stored in the carbon fingerprint gene library. .

[0133] The specific steps of ladder-type gene fragment processing are as follows:

[0134] For ladder-type gene fragment sets Each ladder-type gene segment has a step amplitude. and response time Two key characteristics: The step amplitude is the absolute value of the difference between the steady-state values ​​at the beginning and end of the step; the response time is the time elapsed from the start of the step to reaching 90% of the new steady state. Equal-width bins are used, and the number of bins for the step amplitude is... Response time for the number of boxes This yielded multiple ladder-type gene subclasses.

[0135] For the k-th ladder-type gene subclass The standard waveform morphology is calculated using the DTW center algorithm. .

[0136] Calculate the arithmetic mean of the step amplitudes of all segments within the k-th step-type gene subclass as the standard step amplitude. The arithmetic mean of the response times is calculated as the standard response time. .

[0137] Step-type standard gene template for:

[0138] ={Type=Stepped,Standard Waveform= Standard step amplitude = Standard response time = };

[0139] The ladder-type standard gene template is stored in the carbon fingerprint gene bank. .

[0140] Carbon fingerprint gene bank Each template in the database represents a typical carbon emission waveform pattern, and the subsequent step S104 will match based on this carbon fingerprint gene library.

[0141] In summary, equal-width binning is performed based on the key features of the three gene types (peak intensity + duration, oscillation period + amplitude, step amplitude + response time), which avoids the information loss caused by using a uniform clustering feature. The equal-width binning method is simple and efficient, and the number of bins can be dynamically adjusted according to the amount of data, ensuring that each subclass has a sufficient sample size for statistical calculation.

[0142] Since the waveform lengths of different carbon fingerprint gene fragments may vary (pulse-type fragments may be short or long, and oscillatory fragments may have different number of cycles), the Dynamic Time Warping Center (DTW) algorithm is used to calculate the standard waveform shape. This algorithm can naturally handle sequences with inconsistent lengths and find typical waveforms that represent the subclass center trend. Compared with simple truncation or interpolation alignment, DTW centers retain the temporal scaling invariance of the waveform, making the standard template more representative.

[0143] Step S104: Obtain production trajectory data of the target product during the monitoring period, and match the carbon fingerprint gene sequence corresponding to the production trajectory data from the carbon fingerprint gene library according to the production trajectory data, and perform time consistency verification and conflict resolution on the carbon fingerprint gene sequence to obtain the target carbon fingerprint gene sequence.

[0144] In this step, the production trajectory data is converted into a time-series event sequence consisting of equipment identifier, operating status, duration, process parameters, and the correlation between preceding and following events. The correlation between preceding and following events includes the equipment identifier of the preceding event segment and the equipment identifier of the following event segment for each event segment.

[0145] For each event segment in the time-series event sequence, a set of corresponding candidate standard gene templates is selected from the carbon fingerprint gene library based on the device identifier and operating status of the event segment.

[0146] Based on the duration of the event fragment, the Gaussian kernel function is used to calculate the duration matching degree between the event fragment and each candidate standard gene template, where the duration matching degree = exp(-(actual duration - standard duration)² / (2σ²)), and σ is the preset duration matching tolerance parameter;

[0147] Based on the process parameters of the event fragment, Mahalanobis distance was used to calculate the deviation of the process parameters between the event fragment and each candidate standard gene template.

[0148] Based on the correlation between events before and after the event fragment, the transfer probability between different gene types in the carbon fingerprint gene library is calculated. The gene type of the standard gene template matched by the preceding event fragment is taken as the preorder type, and the gene type of the standard gene template matched by the following event fragment is taken as the postorder type. The probability value of transferring from the preorder type to the postorder type is queried from the transfer probability and used as the context matching degree.

[0149] The duration matching degree, process parameter deviation degree and context matching degree are weighted and fused to obtain the comprehensive matching degree of each candidate standard gene template. The standard gene template with the highest comprehensive matching degree is selected as the carbon fingerprint gene fragment corresponding to the event fragment.

[0150] The carbon fingerprint gene fragments corresponding to each event fragment are combined in chronological order of the events to obtain the initial carbon fingerprint gene sequence.

[0151] For any two adjacent carbon fingerprint gene segments in the initial carbon fingerprint gene sequence, the last data point of the previous carbon fingerprint gene segment and the first data point of the next carbon fingerprint gene segment are extracted. The absolute difference of the waveform amplitude between the last data point and the first data point is calculated as the amplitude jump value. The average first waveform slope of the last three data points of the previous carbon fingerprint gene segment and the average second waveform slope of the first three data points of the next carbon fingerprint gene segment are extracted. The absolute difference between the average first waveform slope and the average second waveform slope is calculated as the slope difference value.

[0152] When the slope difference value is greater than the third preset threshold or the amplitude jump value is greater than the fourth preset threshold, the waveform morphology of the last three sampling points of the end segment of the previous carbon fingerprint gene fragment and the first three sampling points of the beginning segment of the next carbon fingerprint gene fragment are used as control points. A continuous and smooth transition waveform is generated by Bézier curve fitting. The transition waveform is inserted as a transition gene fragment between the previous carbon fingerprint gene fragment and the next carbon fingerprint gene fragment to obtain the target carbon fingerprint gene sequence.

[0153] In one specific embodiment, the production trajectory data of the target product is extracted from the Manufacturing Execution System (MES). Each production trajectory record contains the following information: event sequence number, equipment identifier (e.g., "kiln A", "cooler B", etc.), operating status (e.g., "heating", "heat holding", "cooling", etc.), start time (formatted as hours:minutes:seconds), end time, and process parameters (e.g., for the kiln heating section, process parameters include starting temperature, target temperature, heating rate, etc.).

[0154] Each of the above records is converted into a time-series event fragment E_iEi. Each event fragment contains the following fields:

[0155] device_id: Device identifier (string);

[0156] state: running status (string);

[0157] duration: The actual duration, in seconds, is calculated by subtracting the start time from the end time;

[0158] process_params: A vector of process parameters. For example, for the heating section of a kiln, the vector can be represented as [starting temperature, target temperature, heating rate], for the cooling section it can be represented as [starting temperature, target temperature, cooling rate], and for the heat preservation section it can be represented as [heat preservation temperature, temperature fluctuation range].

[0159] prev_device: The device identifier of the previous event segment (this field is empty if the current segment is the first event);

[0160] next_device: The device identifier for the next event segment (this field is empty if the current segment is the last event);

[0161] Arrange all event fragments in chronological order to obtain a time-series event sequence. ,in This represents the total number of event segments experienced by the target product during the production process.

[0162] For the i-th event segment Based on its device_id and state, the carbon fingerprint gene library constructed in step S103 is used to... The corresponding candidate standard gene template set was selected from the data.

[0163] The screening rules are predefined using a mapping table stored in the system. Each row of the mapping table defines a correspondence between a device-state pair and a gene type. For example:

[0164] For equipment "Kiln A" and its operating status is "heating", the corresponding gene type is "ladder type", and the candidate templates are all ladder type standard gene templates in the carbon fingerprint gene library.

[0165] For equipment "Kiln A" and its operating status is "heat preservation", the corresponding gene type is "oscillating", and the candidate template is all oscillating standard gene templates in the carbon fingerprint gene library.

[0166] For device "cooler B" and its operating status is "cooling", the corresponding gene type is "pulse type", and the candidate template is all pulse type standard gene templates in the carbon fingerprint gene library.

[0167] In practice, an index list is maintained for each device-state pair, which directly points to the relevant templates in the gene bank to improve screening efficiency.

[0168] For the i-th event segment and each candidate standard gene template The duration matching degree, process parameter deviation degree, and context matching degree are calculated respectively. Specifically:

[0169] The duration matching degree is calculated using the Gaussian kernel function, as shown in the following formula:

[0170] ,

[0171] in, For the i-th event fragment The actual duration, Standard gene template The standard duration is specifically taken as follows: for pulse-type templates, the "standard duration" in the template is taken; for oscillating templates, the "standard oscillation period" is taken by multiplying the number of periods implied in the template (the number of periods can be determined by the number of complete periods in the standard waveform); for stepped templates, the "standard response time" is taken.

[0172] A tolerance parameter is matched to the preset duration. In this embodiment, This is 20% of the standard duration. This value ensures that when the relative deviation between the actual duration and the standard duration is within 20%, the duration matching degree is greater than 0.8.

[0173] Example: The standard duration of a certain pulse template is 10 seconds, and the actual event duration is 12 seconds. Then the duration matching degree = exp(-(12-10)² / (2×(2)²)) = exp(-4 / 8) = exp(-0.5) ≈ 0.6065.

[0174] Mahalanobis distance is used to calculate the deviation of process parameters. Let the i-th event segment be... The process parameter vector is Its dimensions are When constructing the carbon fingerprint gene library, for each candidate standard gene template... The mean vector and covariance matrix of the process parameters for all corresponding original carbon fingerprint gene fragments were pre-calculated. Mahalanobis distance is defined as:

[0175] ,

[0176] in, For the i-th event fragment With candidate standard gene template Mahalanobis distance between them;

[0177] as candidate standard gene templates The corresponding process parameter mean vector, dimension and Similarly, when constructing a carbon fingerprint gene library, for each standard gene template, the process parameters of all corresponding original carbon fingerprint gene fragments are statistically analyzed, and the arithmetic mean of each parameter is calculated to form a mean vector. For example, for a given pulse-type template, its corresponding original segment may come from multiple device start-stop events. Each event records parameters such as temperature and pressure, and the average is taken to obtain the result. ;

[0178] as candidate standard gene templates The corresponding process parameter covariance matrix is ​​a The process parameter covariance matrix is ​​a symmetric positive definite matrix, where the element in the i-th row and j-th column represents the covariance between the i-th process parameter and the other process parameters. This process parameter covariance matrix reflects the correlation and scale differences among the process parameters. When constructing the gene library, it is also calculated based on process parameter samples of all original fragments corresponding to the template.

[0179] It is the inverse matrix of the process parameter covariance matrix;

[0180] For vector difference transpose;

[0181] Mahalanobis distance takes into account the correlation between various process parameters and is dimensionless. The smaller the distance, the closer the process parameters of the event segment are to the standard process parameters of the template.

[0182] Then, the Mahalanobis distance is converted into process parameter deviation. The conversion formula is:

[0183] ,

[0184] In the formula, For the i-th event fragment With candidate standard gene template Deviation of process parameters between;

[0185] Context matching utilizes the transfer probability between different gene types during the production process. First, the transfer probability between gene types (pulse, oscillatory, and stepwise) of all standard gene templates in the carbon fingerprint gene library is statistically analyzed. The statistical method is as follows: from the carbon fingerprint gene fragment sequences obtained in step S102 (in chronological order), all adjacent fragments are traversed, and the frequency of gene type transfer from the previous fragment to the gene type of the next fragment is recorded. A 3×3 transfer probability matrix is ​​then constructed. ,in Indicates from type Transfer to type probability The sum of the probabilities of each row is 1.

[0186] For the i-th event segment Given the fragment of the previous event Matched gene types (denoted as) ), and the next event fragment The gene type to be matched (denoted as) It's important to note that when matching the current segment, the matching result for the next segment is not yet determined. Therefore, the actual implementation uses an iterative approach: first, a preliminary matching is performed, ignoring the context matching degree, to obtain the preliminary matching types of all segments. Then, the context matching degree is calculated using these preliminary types, and a second round of matching is performed. This embodiment uses a one-pass method: Gene types to be replaced with candidate standard gene templates This can be a prior estimate, or simply using the forward probability. A simplified approach is to use only the forward probability:

[0187] ,

[0188] in, For the i-th event fragment With candidate standard gene template The context matching degree between the two, with a value between 0 and 1, reflects the likelihood that the gene type matched by the previous event fragment will be transferred to the gene type of the current candidate template in the context of the current production sequence.

[0189] To be based on gene type Transfer to gene type The probability, which is derived from a pre-statistical transition probability matrix. The query yielded the result; in the matrix transition probability matrix, the nth... line, number The corresponding element in the column; for example, if the type of the previous event fragment is "pulse type" and the type of the current candidate standard gene template is "oscillation type", then the query value is... ;

[0190] like If it is the first event fragment, the context matching degree is 1 (i.e., no constraint).

[0191] The matching degree after converting duration matching degree, process parameter deviation matching degree, and context matching degree are weighted and fused to obtain the comprehensive matching degree:

[0192] ,

[0193] In the formula, For the i-th event fragment With candidate standard gene template Overall matching degree between them , , These are the weights for duration matching, process parameter deviation, and context matching, respectively. The weights can be adjusted based on the actual application scenario through grid search or experience.

[0194] The candidate standard gene template with the highest overall matching degree is selected as the i-th event fragment. The corresponding carbon fingerprint gene fragment. Simultaneously, the matching score (i.e., overall matching score) of this fragment is recorded for subsequent correction in step S105.

[0195] The carbon fingerprint gene fragments corresponding to each event fragment are arranged in chronological order of the events (i.e., By combining these sequences, the initial carbon fingerprint gene sequence is obtained. Each carbon fingerprint gene fragment contains its standard waveform morphology (i.e., the standard waveform morphology extracted from the carbon fingerprint gene library), but the duration of this standard waveform morphology is a standard duration. The duration of the actual event segment Possibly with They are different. Therefore, it is necessary to stretch or compress the time axis of the standard waveform: multiply the horizontal axis (time) of the standard waveform by a coefficient. This yields waveform data adapted to the actual duration.

[0196] Because adjacent gene fragments in the initial sequence come from different standard templates, and after time-axis scaling, waveform discontinuities (amplitude jumps or slope abrupt changes) may occur at the splicing points. To simulate the continuous changes in carbon emissions during real physical processes, verification and repair are necessary.

[0197] For any two adjacent carbon fingerprint gene fragments and ( ), perform the following operations:

[0198] extract The last data point is denoted as The corresponding time is recorded as ;

[0199] extract The first data point is denoted as The corresponding time is recorded as (Notice > );

[0200] Calculate the amplitude jump value: ;

[0201] extract The last three data points have the following coordinates: , , Calculate the average slope of the first waveform. Linear regression or simply using the endpoint slopes can be used. ;

[0202] extract The first three data points, with coordinates as , , Calculate the average slope of the second waveform. ;

[0203] Calculate the slope difference value: ;

[0204] Preset third preset threshold (Dimensionless), Fourth Preset Threshold ,in For the range of carbon emission data (e.g., 10% of the maximum possible emission rate), if or If the splice is discontinuous, a transition gene fragment needs to be inserted.

[0205] Conflict resolution uses cubic Bézier curve fitting to generate the transition waveform. Specific steps:

[0206] Pick The last three sampling points of the final segment are used as part of the control points: , , These serve as the first three control points of the Bézier curve;

[0207] Pick The first three sampling points of the initial segment: , , .

[0208] Construct a cubic Bézier curve, starting from... The destination is The control point is and The parametric equation of the curve is:

[0209] ,

[0210] in, For parameters, , ;

[0211] In the interval Uniform sampling One point (in this embodiment, we take) For each sampling point , For the first transition waveform Time coordinates of each sampling point This is the index number of the sampling point, with a value of [value]. Calculate the corresponding Then calculate The ordinate is used to obtain the data point sequence of the transition waveform;

[0212] The generated transition waveform sequence is used as a new transition gene fragment. Insert into and Between these points, the gene type of the transition gene fragment is marked as "transitional," and its waveform data corresponds to the sampling points on the aforementioned Bézier curve, with a time range of [missing information]. Transition gene fragments are not included in the calculation of carbon emission contribution (their integral area is considered zero), and are only used to maintain waveform continuity.

[0213] If the time interval between adjacent segments If the interval is less than the preset minimum interval threshold (e.g., 1 second), linear interpolation is used directly, without generating complex Bézier curves.

[0214] After performing the above verification and repair on all adjacent fragment pairs, the final target carbon fingerprint gene sequence is obtained. The waveforms between all adjacent segments in this sequence are continuous and smooth.

[0215] when or When there are fewer than 3 data points, the average slope cannot be calculated. In this case, only the amplitude jump value is used. Make a judgment. If Then, linear interpolation is used to generate the transition waveform: In and linear interpolation between One point.

[0216] If the time interval between adjacent segments The interval is greater than the preset long interval threshold (e.g., 10 seconds), and and If the values ​​are all very small, then a transition segment can be omitted, and the interval can be retained directly. However, when accumulating subsequent carbon emission contributions, the carbon emissions within that interval should be considered zero (or estimated based on historical data). This embodiment inserts a transition segment by default to ensure continuity.

[0217] In summary, the duration matching degree (Gaussian kernel function) considers the tolerance of deviation between the event duration and the standard template; the process parameter deviation degree (Madara distance) considers the correlation between multidimensional parameters; and the context matching degree (transition probability) utilizes the statistical regularity of gene type changes during the production process. The weighted fusion of these three factors significantly improves the matching accuracy and avoids misjudgment due to single feature matching. By calculating the amplitude jump and slope difference at the splicing points of adjacent gene segments, the waveform continuity is quantitatively evaluated, and a smooth transition waveform is generated by fitting a Bézier curve. This ensures that the final carbon fingerprint gene sequence is physically continuous and differentiable, conforming to the gradual characteristics of the actual carbon emission process. This lays the foundation for the accurate accumulation of carbon emission contributions in the subsequent step S105 and avoids unreasonable mutations caused by discontinuity.

[0218] Step S105: The carbon emission contribution corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence is accumulated to obtain the accumulated result. The accumulated result is then corrected using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

[0219] In this step, the integral area in the standard gene template matching a certain carbon fingerprint gene fragment in the target carbon fingerprint gene sequence is obtained as a certain baseline contribution.

[0220] Obtain the actual duration and standard duration of the event fragment corresponding to a certain carbon fingerprint gene fragment, and calculate a certain time ratio coefficient, where the time ratio coefficient = actual duration / standard duration;

[0221] The actual running intensity and standard intensity of the event fragment corresponding to a certain carbon fingerprint gene fragment are obtained, and a certain intensity ratio coefficient is calculated, where the intensity ratio coefficient = actual running intensity / standard intensity;

[0222] The carbon emission contribution corresponding to a certain carbon fingerprint gene fragment is calculated based on a certain baseline contribution, a certain time ratio coefficient, and a certain intensity ratio coefficient. The carbon emission contribution is calculated as: a certain baseline contribution × a certain time ratio coefficient × a certain intensity ratio coefficient.

[0223] The carbon emission contributions of all carbon fingerprint gene fragments in the target carbon fingerprint gene sequence are summed to obtain the summation result.

[0224] Furthermore, the historical carbon footprint factor distribution of the target product is obtained, and the probability density function of the historical carbon footprint factor distribution is fitted using the kernel density estimation method;

[0225] Based on the probability density function, the quantile position of the accumulated result in the historical carbon footprint factor distribution is calculated, and the first-level correction coefficient is determined.

[0226] Obtain the gene type distribution of all standard gene templates in the carbon fingerprint gene library, calculate the proportion of the number of standard gene templates corresponding to each gene type to the total number of standard gene templates in the carbon fingerprint gene library, and use it as the standard proportion of various gene fragments under the standard production mode. The standard gene templates include pulse-type standard gene templates, ladder-type standard gene templates and oscillating standard gene templates.

[0227] The proportion of each type of carbon fingerprint gene fragment in the target carbon fingerprint gene sequence to the total number of carbon fingerprint gene fragments in the target carbon fingerprint gene sequence is calculated as the actual proportion;

[0228] Calculate the absolute value of the difference between the actual proportion and the standard proportion to obtain the proportion difference value of each type of gene fragment. Sum the proportion difference values ​​of all categories to obtain the total difference value. Determine the second-level correction coefficient based on the total difference value.

[0229] The accumulated result is multiplied by the first-level correction coefficient and the second-level correction coefficient in sequence to obtain the corrected total carbon emissions. The corrected total carbon emissions are then divided by the output of the target product to obtain the carbon footprint factor of the target product.

[0230] In one specific embodiment, for the target carbon fingerprint gene sequence (in The total number of gene fragments in the sequence, including any transition gene fragments that may be inserted in step S104 (but transition gene fragments do not participate in carbon emission calculations and their contribution is zero), is used to perform the following operations for each non-transition gene fragment:

[0231] Obtaining baseline contribution: Standard gene templates matching non-transitional gene fragments read from the carbon fingerprint gene library. The integral area is denoted as The unit of the integral area is kg CO2 (carbon emission mass). This value is obtained by taking the arithmetic mean of the integral areas of all original fragments within the subclass when constructing the gene library, and represents the carbon emission of this type of event under typical conditions.

[0232] Calculate the time scaling factor: Obtain the event fragment corresponding to the non-transition gene fragment. actual duration (Unit: seconds), and standard gene template Standard duration (Unit: seconds). Time scaling factor. Defined as:

[0233] ,

[0234] This time proportionality factor reflects the deviation between the actual duration and the standard duration. If the actual duration is longer, carbon emissions should increase proportionally; conversely, they should decrease.

[0235] Calculate the intensity ratio coefficient: Obtain the event fragment corresponding to the non-transition gene fragment. Actual operating intensity (Units depend on process parameters; for example, for the kiln heating section, intensity can be taken as heating rate ℃ / s; for motor operation, intensity can be taken as power kW), and standard gene templates. Standard strength Strength proportionality factor Defined as:

[0236] ,

[0237] This intensity proportionality coefficient reflects the deviation between the actual operating intensity and the standard intensity. For example, a faster heating rate may result in higher carbon emissions per unit time, but the duration may be shorter, and the product of the two reflects the total carbon emissions.

[0238] Calculate the carbon emission contribution of this non-transition gene fragment:

[0239] ,

[0240] For transition gene fragments (of type "transitional"), their carbon emission contribution is 0 because the transition waveform is only used to maintain continuity and does not represent actual carbon emission events.

[0241] The carbon emission contributions of all non-transition gene segments in the target carbon fingerprint gene sequence are summed to obtain the preliminary total carbon emissions. :

[0242] ,

[0243] in, The contribution of the i-th gene segment (the contribution of the transition segment is 0).

[0244] Obtain historical carbon footprint factor distribution: Collect historical carbon footprint factor data for all batches of the target product produced within a past period (e.g., the past year), denoted as set. ,in, This refers to the number of historical batches. For the first Historical data on carbon footprint factors for each batch.

[0245] Kernel density estimation fits the probability density function: The probability density function of historical data is fitted using the Gaussian kernel density estimation method. :

[0246] ,

[0247] in, For bandwidth parameters, it is usually taken as... , The standard deviation of historical data, For the first Historical data on carbon footprint factors for each batch Pi is the mathematical constant of a circle.

[0248] Calculating quantile positions: This involves calculating the initial total carbon emissions. Divide by the production volume of this batch of products (Quantity per unit product) to obtain preliminary carbon footprint factor Then calculate. quantile position in historical distribution :

[0249] ,

[0250] Right now The corresponding cumulative distribution function value. The value ranges from [0,1]. The closer it is to 0.5, the closer it is to the median. The closer it is to 0 or 1, the further it deviates from the mainstream.

[0251] Determine the first-level correction factor: Define the correction factor for:

[0252] ,

[0253] in, The preset correction strength parameter is used in this embodiment. This formula makes when When it is below the historical median, Correct upwards; when When it is higher than the historical median, Correct downwards; when When it is exactly the median, The correction magnitude increases linearly with the degree of deviation.

[0254] The correction based on gene type proportions is used to compensate for deviations caused by the relative frequency of various carbon emission events during production deviating from the standard pattern. Specifically:

[0255] Calculation of standard proportion: from carbon fingerprint gene library The distribution of gene types across all standard gene templates is statistically analyzed. Assume there are a total of [number missing] gene types in the carbon fingerprint gene database. There are 1 standard gene templates, of which the number of pulse-type templates is 1. The number of oscillating templates is The number of stepped templates is The standard percentage is:

[0256] , , ,

[0257] In the formula, This represents the standard proportion of pulse-type standard gene templates. This represents the standard proportion of oscillating standard gene templates. The standard proportion of the ladder-type standard gene template;

[0258] Calculate the actual proportion: In the target carbon fingerprint gene sequence, count the number of non-transition gene fragments. Let the number of pulse-type fragments in the sequence be... The number of oscillating segments is The number of stepped segments is The total number of non-transitional segments is The actual percentage is:

[0259] , , ,

[0260] In the formula, This represents the actual proportion of pulse-type carbon fingerprint gene fragments. This represents the actual proportion of oscillating carbon fingerprint gene fragments. This represents the actual proportion of ladder-type carbon fingerprint gene fragments.

[0261] Calculate the total difference value: the difference value of the proportion of each type of gene fragment is... Similarly , Total difference value The sum of the three:

[0262] ,

[0263] The value range is [0,2] (because the sum of the three differences is at most 2, for example, changing from all pulse types to all non-pulse types). The larger the value, the greater the deviation between the current batch's gene type composition and the standard pattern.

[0264] Determine the second-level correction factor: Define the correction factor ,in The preset correction strength parameter is used in this embodiment. =0.3. This formula means that the larger the difference between the actual proportion and the standard proportion, the larger the correction factor (upward correction). This is because if the gene type distribution of the current batch deviates from the standard pattern, it may mean abnormal production conditions, resulting in an underestimation of carbon emissions, which needs to be adjusted upward.

[0265] Preliminary total carbon emissions Multiply by the first-level correction factor in sequence Second-level correction coefficient The corrected total carbon emissions are as follows:

[0266] ,

[0267] Finally, the revised total carbon emissions are divided by the production volume of that batch of target products. The final carbon footprint factor is obtained. .

[0268] In summary, the carbon footprint factor calculation method and system of this application, by decomposing carbon emission waveforms into carbon fingerprint gene fragments with clear physical meaning (pulse type reflects equipment start-up and shutdown, oscillation type reflects process fluctuations, and step type reflects material batch changes), and constructing a reusable carbon fingerprint gene library, realizes a paradigm shift in carbon footprint factors from the traditional "positive accumulation" to "gene matching and assembly," achieving the following significant technical effects:

[0269] Traditional methods struggle to handle the mixing of carbon emission sources during multi-product co-production, leading to inaccurate carbon footprint accounting results. This application separates the carbon emission time-series signal into three components—low-frequency trend, mid-frequency fluctuation, and high-frequency transient—using wavelet packet decomposition. Step-type, oscillating, and pulse-type gene fragments are extracted using change point detection, sliding window statistics, and pulse detection techniques, respectively. This achieves physical decoupling of three types of carbon emission sources: material batch changes, process fluctuations, and equipment start-up and shutdown. Each carbon fingerprint gene fragment has an independent waveform morphology and characteristic parameters (peak intensity, integral area, waveform duration, etc.), making the source composition of carbon emissions clear and traceable. Users can intuitively understand the physical causes of each part of the carbon emissions, significantly improving the interpretability of the carbon footprint accounting results.

[0270] Traditional methods require tracking the process path of each product individually and accumulating carbon emissions. The computational complexity is linearly related to the number of products, making it difficult to cope with large-scale production scenarios. This application constructs a carbon fingerprint gene library and pre-standardizes and stores typical carbon emission patterns. For a new batch of products, it is only necessary to match its production trajectory with the standard template in the gene library to quickly generate a carbon fingerprint gene sequence, without having to repeat complex signal decomposition and feature extraction for each product.

[0271] The carbon emission contribution of each carbon fingerprint gene fragment is determined by the product of the baseline integral area, the time proportionality coefficient, and the intensity proportionality coefficient. Among them, the time proportionality coefficient (actual duration / standard duration) and the intensity proportionality coefficient (actual operating intensity / standard intensity) directly reflect the deviation between the actual production conditions and the standard template. This allows the carbon emission calculation to dynamically adapt to actual production factors such as equipment aging, operating condition fluctuations, and raw material changes, avoiding the systematic errors caused by the use of fixed carbon emission factors in traditional methods. At the same time, the transition gene fragments generated by time consistency verification and Bézier curve fitting ensure the waveform continuity between adjacent carbon emission events, simulate the gradual characteristics of real physical processes, and further improve the accuracy of the cumulative results.

[0272] The first-level correction (historical distribution quantile correction) uses kernel density estimation to fit the probability distribution of historical carbon footprint factors. By calculating the quantile position of the current batch's preliminary results in the historical distribution, it automatically compensates for systematic deviations caused by long-term changes in raw materials, climate, equipment aging, and other factors. The second-level correction (gene type proportion correction) compares the proportion of various gene fragments in the actual carbon fingerprint gene sequence with the standard proportion in the gene library to identify abnormal changes in the composition of carbon emission events during the production process (such as an abnormal increase in equipment start-up and shutdown) and makes corresponding adjustments to the total carbon emissions. The synergistic effect of the two-level corrections enables this method to maintain high-precision accounting results under different production conditions and time scales, and has strong robustness and adaptability.

[0273] Please see Figure 2 The diagram shows a structural block diagram of a carbon footprint factor calculation system according to this application.

[0274] like Figure 2 As shown, the carbon footprint factor calculation system 200 includes a clustering module 210, an extraction module 220, a construction module 230, a matching module 240, and a correction module 250.

[0275] The clustering module 210 is configured to acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering of the carbon emission source data based on the morphological and energy distribution characteristics of the carbon emission waveform to obtain at least one carbon emission feature set; the extraction module 220 is configured to extract a target time series segment that meets the gene integrity condition from a certain carbon emission feature set, and perform multi-scale decomposition and fragment assembly on the target time series segment according to a preset carbon fingerprint gene sequencing strategy to extract at least one carbon fingerprint gene fragment contained in the target time series segment and the gene feature parameters corresponding to each carbon fingerprint gene fragment, wherein the carbon fingerprint gene fragment is an indivisible waveform unit reflecting the independent action of a single carbon emission source per unit time; the construction module 230 is configured to... A carbon fingerprint gene library is constructed based on the at least one carbon fingerprint gene fragment and the gene feature parameters corresponding to the at least one carbon fingerprint gene fragment; the matching module 240 is configured to acquire the production trajectory data of the target product during the monitoring period, and match the carbon fingerprint gene sequence corresponding to the production trajectory data from the carbon fingerprint gene library based on the production trajectory data, and perform time-series consistency verification and conflict resolution on the carbon fingerprint gene sequence to obtain the target carbon fingerprint gene sequence; the correction module 250 is configured to accumulate the carbon emission contribution corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence to obtain the accumulation result, and use a weighted correction strategy based on the distribution of historical carbon emission source data to correct the accumulation result to obtain the carbon footprint factor of the target product.

[0276] It should be understood that Figure 2 The modules and references described in the document Figure 1 The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0277] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the carbon footprint factor calculation method in any of the above method embodiments.

[0278] In one embodiment, the computer-readable storage medium of the present invention stores computer-executable instructions, which are configured as follows:

[0279] Acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features.

[0280] A target time sequence segment that meets the gene integrity condition is extracted from a certain carbon emission feature set. The target time sequence segment is then decomposed and assembled at multiple scales according to a preset carbon fingerprint gene sequencing strategy. At least one carbon fingerprint gene segment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene segment are extracted. The carbon fingerprint gene segment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time.

[0281] A carbon fingerprint gene library is constructed based on the at least one carbon fingerprint gene fragment and the gene characteristic parameters corresponding to the at least one carbon fingerprint gene fragment;

[0282] The production trajectory data of the target product during the monitoring period is obtained, and based on the production trajectory data, the carbon fingerprint gene sequence corresponding to the production trajectory data is matched from the carbon fingerprint gene library. The carbon fingerprint gene sequence is then subjected to time consistency verification and conflict resolution to obtain the target carbon fingerprint gene sequence.

[0283] The carbon emission contributions corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence are summed to obtain the summed result. The summed result is then corrected using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

[0284] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the carbon footprint factor calculation system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely located relative to the processor, and this remote memory may be connected to the carbon footprint factor calculation system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0285] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the carbon footprint factor calculation method described in the above embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the carbon footprint factor calculation system. The output device 340 may include a display screen or other display device.

[0286] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0287] In one implementation, the above-described electronic device is applied to a carbon footprint factor calculation system for a client, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0288] Acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features.

[0289] A target time sequence segment that meets the gene integrity condition is extracted from a certain carbon emission feature set. The target time sequence segment is then decomposed and assembled at multiple scales according to a preset carbon fingerprint gene sequencing strategy. At least one carbon fingerprint gene segment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene segment are extracted. The carbon fingerprint gene segment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time.

[0290] A carbon fingerprint gene library is constructed based on the at least one carbon fingerprint gene fragment and the gene characteristic parameters corresponding to the at least one carbon fingerprint gene fragment;

[0291] The production trajectory data of the target product during the monitoring period is obtained, and based on the production trajectory data, the carbon fingerprint gene sequence corresponding to the production trajectory data is matched from the carbon fingerprint gene library. The carbon fingerprint gene sequence is then subjected to time consistency verification and conflict resolution to obtain the target carbon fingerprint gene sequence.

[0292] The carbon emission contributions corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence are summed to obtain the summed result. The summed result is then corrected using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

[0293] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0294] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for calculating carbon footprint factor, characterized in that, include: Acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features. A target time sequence segment that meets the gene integrity condition is extracted from a certain carbon emission feature set. The target time sequence segment is then decomposed and assembled at multiple scales according to a preset carbon fingerprint gene sequencing strategy. At least one carbon fingerprint gene segment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene segment are extracted. The carbon fingerprint gene segment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time. A carbon fingerprint gene library is constructed based on the at least one carbon fingerprint gene fragment and the gene characteristic parameters corresponding to the at least one carbon fingerprint gene fragment; The production trajectory data of the target product during the monitoring period is obtained, and based on the production trajectory data, the carbon fingerprint gene sequence corresponding to the production trajectory data is matched from the carbon fingerprint gene library. The carbon fingerprint gene sequence is then subjected to time consistency verification and conflict resolution to obtain the target carbon fingerprint gene sequence. The carbon emission contributions corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence are summed to obtain the summed result. The summed result is then corrected using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

2. The carbon footprint factor calculation method according to claim 1, characterized in that, The carbon emission source data is clustered at multiple levels based on the morphological and energy distribution characteristics of carbon emission waveforms to obtain at least one set of carbon emission features, including: The carbon emission source data is segmented using a preset sliding window to obtain multiple time-series segments; Multiple sliding sub-windows of different time lengths are used to extract the morphological features of a certain time segment, resulting in a morphological feature vector of a certain time segment at multiple time scales. The morphological feature vector includes a rising edge slope, a falling edge slope, a peak duration, and a peak-to-valley ratio. Wavelet packet transform is used to decompose a time segment into multiple frequency band components. The energy proportion of each frequency band component is calculated as the first energy feature, the energy centroid frequency of all frequency band components is calculated as the second energy feature, and the variance of the energy distribution of each frequency band component is calculated as the third energy feature. The first energy feature, the second energy feature and the third energy feature are combined into a certain energy distribution feature vector of a certain time segment. By concatenating the morphological feature vector with the energy distribution feature vector, a multidimensional feature vector of the time series segment is constructed. Based on the multidimensional feature vectors of each time series segment, density clustering algorithm is used for the first-level clustering, dividing each time series segment into multiple initial clusters; For each time-series segment within an initial cluster, a self-organizing map network is used for second-level clustering to further divide the initial cluster into multiple sub-clusters, resulting in at least one set of carbon emission features.

3. The carbon footprint factor calculation method according to claim 1, characterized in that, The extraction of a target time sequence segment that satisfies the gene integrity condition from a certain carbon emission feature set includes: Arrange the time-series segments in the set of carbon emission characteristics in chronological order to obtain a time-series segment sequence; The earliest time segment is selected as a candidate segment, and each candidate segment is sequentially judged to meet a preset gene integrity condition, wherein the gene integrity condition includes: Data integrity condition: The number of data points in the candidate segment is not less than the preset minimum number of points threshold; Waveform purity condition: The signal-to-noise ratio of the candidate segment is greater than the preset signal-to-noise ratio threshold, where the signal-to-noise ratio is calculated by the ratio of signal power to noise power; Boundary smoothness condition: The absolute values ​​of the waveform slopes at the start and end points of the candidate segment are both less than a preset slope threshold; Internal consistency condition: The waveform similarity of each sub-segment within a candidate segment is not lower than a preset similarity threshold, wherein the waveform similarity is calculated by dynamic time warping distance; If a candidate fragment meets the preset gene integrity conditions, then the candidate fragment is determined as the target time sequence fragment; If a candidate fragment does not meet the preset gene integrity conditions, the next time sequence fragment is selected as a new candidate fragment in chronological order, and the gene integrity conditions are re-evaluated until a target time sequence fragment that meets the gene integrity conditions is determined.

4. The carbon footprint factor calculation method according to claim 1, characterized in that, The step of performing multi-scale decomposition and fragment assembly on a target time segment according to a preset carbon fingerprint gene sequencing strategy, and extracting at least one carbon fingerprint gene fragment contained in the target time segment and the gene feature parameters corresponding to each carbon fingerprint gene fragment, includes: Wavelet packet decomposition is performed on a target time segment to obtain low-frequency trend components, mid-frequency fluctuation components and high-frequency transient components; Change point detection is performed on the low-frequency trend component to obtain the time points when the slope change of the low-frequency trend component exceeds a preset threshold as change points. All change points are arranged in chronological order, and the time interval between each two adjacent change points is determined as a ladder-type candidate gene fragment. A sliding window analysis is performed on the mid-frequency fluctuation components to calculate the local standard deviation and local autocorrelation coefficient of the mid-frequency fluctuation components within each window. Continuous window intervals with local standard deviations less than a first preset threshold and local autocorrelation coefficients greater than a second preset threshold are identified as oscillating candidate gene fragments. Pulse detection is performed on the high-frequency transient components, and the peak intervals in the high-frequency transient components whose amplitude exceeds the third preset threshold and whose duration is less than the fourth preset duration are identified as pulse-type candidate gene fragments. The ladder-type candidate gene fragments, oscillating candidate gene fragments, and pulse-type candidate gene fragments are arranged in chronological order to obtain candidate gene fragment sequences; Traverse the candidate gene fragment sequence, and when the time interval between two adjacent candidate gene fragments is less than a preset merging threshold, merge the two adjacent candidate gene fragments into one candidate gene fragment; For each merged candidate gene fragment, the maximum value of all data points within the merged candidate gene fragment is extracted as the peak intensity, the integral value of the merged candidate gene fragment on the time axis is extracted as the integral area, and the difference between the end time and the start time of the merged candidate gene fragment is extracted as the waveform duration, which are used as the gene feature parameters of the carbon fingerprint gene fragment corresponding to the merged candidate gene fragment.

5. The carbon footprint factor calculation method according to claim 1, characterized in that, The construction of the carbon fingerprint gene library based on the at least one carbon fingerprint gene fragment and the gene characteristic parameters corresponding to the at least one carbon fingerprint gene fragment includes: Obtain the gene type identifier of at least one carbon fingerprint gene fragment, and classify carbon fingerprint gene fragments with gene type identifier of pulse type into pulse type gene fragment set, carbon fingerprint gene fragments with gene type identifier of oscillation type into oscillation type gene fragment set, and carbon fingerprint gene fragments with gene type identifier of ladder type into ladder type gene fragment set. For a set of pulsed gene fragments, equal-width binning is performed using peak intensity and waveform duration as classification features to obtain multiple pulsed gene subclasses. The dynamic time warping center of all carbon fingerprint gene fragments within a certain pulsed gene subclass is calculated on the waveform data as the first standard waveform morphology. The arithmetic mean of the peak intensity, waveform duration, and integral area of ​​all carbon fingerprint gene fragments within a certain pulsed gene subclass is calculated as the standard peak intensity, standard duration, and standard integral area. The first standard waveform morphology, the standard peak intensity, the standard duration, and the standard integral area are combined to generate a certain pulsed standard gene template, and the certain pulsed standard gene template is stored in the carbon fingerprint gene library. For a set of oscillating gene fragments, equal-width binning is performed using oscillation period and oscillation amplitude as classification features to obtain multiple oscillating gene subclasses. The dynamic time warping center of all carbon fingerprint gene fragments within a certain oscillating gene subclass is calculated on the waveform data as the second standard waveform morphology. The arithmetic mean of the oscillation period and oscillation amplitude of all carbon fingerprint gene fragments within a certain oscillating gene subclass is calculated as the standard oscillation period and standard oscillation amplitude. The standard waveform morphology, the standard oscillation period, and the standard oscillation amplitude are combined to generate a certain oscillating standard gene template, and the certain oscillating standard gene template is stored in the carbon fingerprint gene library. For a set of step-type gene fragments, equal-width binning is performed using step amplitude and response time as classification features to obtain multiple step-type gene subclasses. The dynamic time warping center of all carbon fingerprint gene fragments in a certain step-type gene subclass is calculated on the waveform data as the third standard waveform morphology. The arithmetic mean of the step amplitude and response time of all carbon fingerprint gene fragments in a certain step-type gene subclass is calculated as the standard step amplitude and standard response time. The third standard waveform morphology, the standard step amplitude, and the standard response time are combined to generate a certain step-type standard gene template, and the certain step-type standard gene template is stored in the carbon fingerprint gene library.

6. The carbon footprint factor calculation method according to claim 1, characterized in that, The step of matching carbon fingerprint gene sequences corresponding to the production trajectory data from the carbon fingerprint gene library based on the production trajectory data, and performing temporal consistency verification and conflict resolution on the carbon fingerprint gene sequences to obtain the target carbon fingerprint gene sequence includes: The production trajectory data is converted into a time-series event sequence consisting of equipment identifier, operating status, duration, process parameters, and the correlation between preceding and following events. The correlation between preceding and following events includes the equipment identifier of the preceding event segment and the equipment identifier of the following event segment for each event segment. For each event segment in the time-series event sequence, a set of corresponding candidate standard gene templates is selected from the carbon fingerprint gene library based on the device identifier and operating status of the event segment. Based on the duration of the event fragment, the Gaussian kernel function is used to calculate the duration matching degree between the event fragment and each candidate standard gene template, where the duration matching degree = exp(-(actual duration - standard duration)² / (2σ²)), and σ is the preset duration matching tolerance parameter; Based on the process parameters of the event fragment, Mahalanobis distance was used to calculate the deviation of the process parameters between the event fragment and each candidate standard gene template. Based on the correlation between events before and after the event fragment, the transfer probability between different gene types in the carbon fingerprint gene library is calculated. The gene type of the standard gene template matched by the preceding event fragment is taken as the preorder type, and the gene type of the standard gene template matched by the following event fragment is taken as the postorder type. The probability value of transferring from the preorder type to the postorder type is queried from the transfer probability and used as the context matching degree. The duration matching degree, process parameter deviation degree and context matching degree are weighted and fused to obtain the comprehensive matching degree of each candidate standard gene template. The standard gene template with the highest comprehensive matching degree is selected as the carbon fingerprint gene fragment corresponding to the event fragment. The carbon fingerprint gene fragments corresponding to each event fragment are combined in chronological order of the events to obtain the initial carbon fingerprint gene sequence. For any two adjacent carbon fingerprint gene segments in the initial carbon fingerprint gene sequence, the last data point of the previous carbon fingerprint gene segment and the first data point of the next carbon fingerprint gene segment are extracted. The absolute difference of the waveform amplitude between the last data point and the first data point is calculated as the amplitude jump value. The average first waveform slope of the last three data points of the previous carbon fingerprint gene segment and the average second waveform slope of the first three data points of the next carbon fingerprint gene segment are extracted. The absolute difference between the average first waveform slope and the average second waveform slope is calculated as the slope difference value. When the slope difference value is greater than the third preset threshold or the amplitude jump value is greater than the fourth preset threshold, the waveform morphology of the last three sampling points of the end segment of the previous carbon fingerprint gene fragment and the first three sampling points of the beginning segment of the next carbon fingerprint gene fragment are used as control points. A continuous and smooth transition waveform is generated by Bézier curve fitting. The transition waveform is inserted as a transition gene fragment between the previous carbon fingerprint gene fragment and the next carbon fingerprint gene fragment to obtain the target carbon fingerprint gene sequence.

7. The carbon footprint factor calculation method according to claim 1, characterized in that, The step of summing up the carbon emission contributions corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence to obtain the summation result includes: The integral area in the standard gene template matching a carbon fingerprint gene fragment in the target carbon fingerprint gene sequence is obtained as a baseline contribution. Obtain the actual duration and standard duration of the event fragment corresponding to a certain carbon fingerprint gene fragment, and calculate a certain time ratio coefficient, where the time ratio coefficient = actual duration / standard duration; The actual running intensity and standard intensity of the event fragment corresponding to a certain carbon fingerprint gene fragment are obtained, and a certain intensity ratio coefficient is calculated, where the intensity ratio coefficient = actual running intensity / standard intensity; The carbon emission contribution corresponding to a certain carbon fingerprint gene fragment is calculated based on a certain baseline contribution, a certain time ratio coefficient, and a certain intensity ratio coefficient. The carbon emission contribution is calculated as: a certain baseline contribution × a certain time ratio coefficient × a certain intensity ratio coefficient. The carbon emission contributions of all carbon fingerprint gene fragments in the target carbon fingerprint gene sequence are summed to obtain the summation result.

8. The carbon footprint factor calculation method according to claim 1, characterized in that, The step of correcting the accumulated result using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product includes: Obtain the historical carbon footprint factor distribution of the target product, and fit the probability density function of the historical carbon footprint factor distribution using the kernel density estimation method; Based on the probability density function, the quantile position of the accumulated result in the historical carbon footprint factor distribution is calculated, and the first-level correction coefficient is determined. Obtain the gene type distribution of all standard gene templates in the carbon fingerprint gene library, calculate the proportion of the number of standard gene templates corresponding to each gene type to the total number of standard gene templates in the carbon fingerprint gene library, and use it as the standard proportion of various gene fragments under the standard production mode. The standard gene templates include pulse-type standard gene templates, ladder-type standard gene templates and oscillating standard gene templates. The proportion of each type of carbon fingerprint gene fragment in the target carbon fingerprint gene sequence to the total number of carbon fingerprint gene fragments in the target carbon fingerprint gene sequence is calculated as the actual proportion; Calculate the absolute value of the difference between the actual proportion and the standard proportion to obtain the proportion difference value of each type of gene fragment. Sum the proportion difference values ​​of all categories to obtain the total difference value. Determine the second-level correction coefficient based on the total difference value. The accumulated result is multiplied by the first-level correction coefficient and the second-level correction coefficient in sequence to obtain the corrected total carbon emissions. The corrected total carbon emissions are then divided by the output of the target product to obtain the carbon footprint factor of the target product.

9. A carbon footprint factor calculation system, characterized in that, include: The clustering module is configured to acquire carbon emission source data of the target production system during the monitoring period, and perform multi-level clustering on the carbon emission source data based on the morphological characteristics and energy distribution characteristics of the carbon emission waveform to obtain at least one set of carbon emission features. The extraction module is configured to extract a target time sequence segment that meets the gene integrity condition from a certain carbon emission feature set, and to perform multi-scale decomposition and fragment assembly on the target time sequence segment according to a preset carbon fingerprint gene sequencing strategy, thereby extracting at least one carbon fingerprint gene fragment contained in the target time sequence segment and the gene feature parameters corresponding to each carbon fingerprint gene fragment, wherein the carbon fingerprint gene fragment is an indivisible waveform unit that reflects the independent action of a single carbon emission source per unit time. The construction module is configured to construct a carbon fingerprint gene library based on the at least one carbon fingerprint gene fragment and the gene feature parameters corresponding to the at least one carbon fingerprint gene fragment; The matching module is configured to acquire production trajectory data of the target product during the monitoring period, and match the carbon fingerprint gene sequence corresponding to the production trajectory data from the carbon fingerprint gene library based on the production trajectory data, and perform time-series consistency verification and conflict resolution on the carbon fingerprint gene sequence to obtain the target carbon fingerprint gene sequence. The correction module is configured to accumulate the carbon emission contribution corresponding to each carbon fingerprint gene fragment in the target carbon fingerprint gene sequence to obtain the accumulation result, and to correct the accumulation result using a weighted correction strategy based on the distribution of historical carbon emission source data to obtain the carbon footprint factor of the target product.

Images