Carbon emission comprehensive evaluation method and system for matrix type flue gas flow measurement
By combining matrix flow sensors with optical flow estimation algorithms and empirical mode decomposition algorithms, the problem of dynamic feature extraction and anomaly identification in flue gas flow and carbon emission monitoring is solved, achieving high-precision carbon emission assessment and real-time early warning.
Patent Information
- Application Number
- CN202510985699.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-17
- Publication Date
- 2025-10-17
AI Technical Summary
Existing flue gas flow and carbon emission monitoring technologies are unable to comprehensively and dynamically reflect flow change characteristics under complex working conditions. They lack the ability to extract multi-scale dynamic features and identify abnormal changes from multi-source time series data, resulting in large errors in carbon emission assessments and a lack of real-time early warning mechanisms.
A matrix flow sensor is used in conjunction with optical flow estimation algorithm and empirical mode decomposition algorithm to process multi-source time series data, reconstruct flue gas flow field, extract multi-scale flow characteristics, and generate carbon emission early warning when anomalies are detected.
It enables accurate identification of flue gas flow fields and high-precision calculation of carbon emissions, improves response sensitivity and real-time performance, and provides dynamic trend analysis and anomaly early warning capabilities for carbon emissions.
Smart Images
Figure CN120801624A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of environmental monitoring and industrial process control, and particularly relates to a carbon emission comprehensive evaluation method and system for matrix flue gas flow measurement. BACKGROUND
[0002] With the continuous advancement of industrialization, flue gas emission has become the core monitoring object of atmospheric environmental governance. The existing flue gas flow and carbon emission monitoring technology mainly relies on single-point or a small number of distributed flow sensors, combined with traditional volume flow and gas concentration measurement methods for total evaluation. This kind of method is often difficult to fully and dynamically reflect the flow change characteristics in the whole cross section under complex working conditions, multi-variable interference and uneven flow field distribution, resulting in a large error in the evaluation of carbon emission. In addition, the current flow and carbon emission analysis mostly uses static models or simple statistical methods, which lack the ability of multi-scale dynamic feature extraction and abnormal change identification for multi-source time series data, and cannot effectively respond to the dynamic working conditions and sudden abnormalities of industrial systems.
[0003] In recent years, although some researches have tried to introduce image processing algorithms or time series decomposition tools to assist in analysis, the global flow field dynamic reconstruction and in-depth multi-modal feature mining of matrix high-density flow data have not been realized. For complex flue gas flow fields, the existing methods are difficult to accurately identify local disturbances, spatial dynamic patterns and abnormal change processes in the flow field, and also lack real-time early warning mechanisms for carbon emission risks. Especially in the industrial flue environment with large amount of multi-source sensor data, complex structure and frequent changes, traditional algorithms face the problems of poor algorithm generalization ability, insufficient abnormality identification sensitivity and low carbon emission calculation precision.
[0004] Therefore, how to provide a carbon emission comprehensive evaluation method and system for matrix flue gas flow measurement is a problem that those skilled in the art need to solve. SUMMARY
[0005] One object of the present application is to provide a carbon emission comprehensive evaluation method and system for matrix flue gas flow measurement. The present application fully integrates matrix flow sensing, optical flow estimation algorithm and empirical mode decomposition algorithm, and describes in detail the flue gas flow field dynamic reconstruction, multi-scale flow feature extraction and carbon emission intelligent evaluation process based on multi-source time series data. The present application has the advantages of strong global flow field measurement, high dynamic response capability, timely abnormality early warning and high carbon emission calculation precision.
[0006] The carbon emission comprehensive evaluation method for matrix flue gas flow measurement according to the embodiment of the present application comprises:
[0007] A plurality of matrix flow sensors are arranged on an industrial flue cross section, and multi-source time series data of each matrix flow sensor is continuously collected, preprocessed, and matrix time series flow data is generated;
[0008] Based on the matrix time series flow data, an optical flow estimation algorithm is used for calculation to obtain flow velocity vector time series data of each matrix flow sensor at each time;
[0009] For the flow velocity vector time series data of each matrix flow sensor, an empirical mode decomposition algorithm is applied to obtain a plurality of intrinsic mode function components and residual components;
[0010] The intrinsic mode function components of all matrix flow sensors are spatially aligned and analyzed for correlation, multi-scale flow feature data is extracted, and spatial dynamic patterns and abnormal change characteristics of the flue gas flow field at different time scales are identified;
[0011] Combining the multi-scale flow feature data and the matrix time series flow data, the carbon emissions at each time are calculated according to a preset carbon emission calculation rule, and a continuous carbon emission comprehensive evaluation result is generated;
[0012] When any intrinsic mode function component or residual component of a matrix flow sensor is detected to have abnormal fluctuations, carbon emission abnormality warning information is automatically generated, and output for working condition tracing and response processing.
[0013] Optionally, the multi-source time series data specifically includes flue gas flow, flue gas temperature, flue gas pressure and flue gas carbon dioxide concentration collected by each sensing point.
[0014] Optionally, the pre-processing of the multi-source time series data specifically includes time synchronization, missing value filling, outlier removal and normalization processing of the multi-source time series data.
[0015] Optionally, based on the matrix time series flow data, an optical flow estimation algorithm is used for calculation to obtain flow velocity vector time series data of each matrix flow sensor at each time, including:
[0016] The matrix time series flow data is arranged into a three-dimensional space-time data matrix according to the spatial position of each matrix flow sensor and each sampling time;
[0017] For each matrix flow sensor in the three-dimensional space-time data matrix, based on the historical collection data of the matrix flow sensor, the mean and standard deviation are calculated respectively, the scaling factor and offset of the matrix flow sensor are set, and the scaling and offset correction of all collection data of the matrix flow sensor is performed to obtain a calibrated three-dimensional space-time data matrix;
[0018] On a time sequence, from the calibrated three-dimensional space-time data matrix, data frames of two adjacent time points are extracted in turn as input data pairs of an optical flow estimation algorithm;
[0019] For each group of input data pairs, for each matrix flow sensor, a number of adjacent matrix flow sensors in the spatial neighborhood are selected with the current matrix flow sensor as the center, the absolute values of the differences between the flow values of the matrix flow sensor and each adjacent matrix flow sensor at the current time are counted respectively, and the average of all absolute values is obtained. The average value is the flow change amplitude of the current matrix flow sensor and the spatial neighborhood. The spatial smoothing weight of the matrix flow sensor is set according to the flow change amplitude. The greater the flow change amplitude, the smaller the spatial smoothing weight. The smaller the change amplitude, the greater the spatial smoothing weight;
[0020] The spatial smoothing weights of all matrix flow sensors are substituted into the optical flow optimization objective function, and the horizontal direction flow velocity component and the vertical direction flow velocity component of each sensing point at the current time are solved by an iterative optimization method;
[0021] For the horizontal direction flow velocity component and the vertical direction flow velocity component of each matrix flow sensor at all time points, a sliding time window with a fixed length is set. The flow velocity time series data of each matrix flow sensor is traversed, the mean and standard deviation of the flow velocity in the window are counted, and it is judged whether the current flow velocity component deviates from the window mean by more than a preset threshold. The abnormal flow velocity point is marked;
[0022] The data marked as an abnormal flow velocity point is replaced or corrected by the mean of the flow velocity component at other time points in the window or according to the flow velocity trend at adjacent time points to obtain the corrected flow velocity component;
[0023] The corrected flow velocity components of all matrix flow sensors are arranged according to the spatial position and the sampling time to form flow velocity vector time series data.
[0024] Optionally, the flow velocity vector time series data of each matrix flow sensor is decomposed by an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode function components and residual components, including:
[0025] The flow velocity vector time series data of each matrix flow sensor is arranged as a flow velocity vector time series data sequence for each matrix flow sensor at all sampling times, including the horizontal direction and the vertical direction flow velocity components;
[0026] For each flow velocity component time series of each matrix flow sensor, before each round of decomposition, according to the trend of several points at the beginning and end of the flow velocity component time series, a specified number of virtual end points are generated at the beginning and end of the flow velocity component time series by using quadratic extrapolation, and the virtual end points are added to both ends of the flow velocity component time series;
[0027] For the flow velocity component time series with virtual end points, all maximum and minimum points are detected, the amplitude difference of each extreme point and adjacent data points is calculated, the extreme amplitude threshold is set according to the signal variance, and only the maximum and minimum points with an amplitude greater than the extreme amplitude threshold are retained;
[0028] The retained maximum points are used as nodes to construct the upper envelope line by using spline interpolation, and the retained minimum points are used as nodes to construct the lower envelope line, for each time point in the constructed upper envelope line and lower envelope line, the amplitude value of the upper envelope line and the amplitude value of the lower envelope line corresponding to the time point are read respectively, the two values are added and averaged to obtain the mean value of the current time point, and all time points are arranged in time sequence to form the mean value sequence of the upper and lower envelope lines corresponding to the flow velocity vector time series;
[0029] The current flow velocity component time series is subtracted from the mean value sequence to obtain a new candidate data sequence;
[0030] The current candidate data sequence is judged whether it meets the eigenmode function component condition, the eigenmode function component condition includes that the difference between the number of extreme points and the number of zero-crossing points in the candidate data sequence is not greater than 1, and the mean value of the upper envelope line and the lower envelope line of the candidate data sequence is close to zero as a whole in the entire time sequence; if the above two conditions are met, the candidate data sequence is output as an eigenmode function component, and the current flow velocity vector time series is subtracted from the eigenmode function component to obtain a new residual data sequence; if not, the current candidate data sequence is used as the basis to repeat the virtual extension of both ends of the data, the extreme point screening, the envelope line construction, the mean value calculation and the new candidate sequence construction until the candidate data sequence meets the above conditions;
[0031] After extracting an intrinsic mode function component, the proportion of the energy of the current remaining flow velocity vector time series data sequence to the total energy of the original flow velocity vector time series data sequence and the amplitude change between the current remaining flow velocity vector time series data sequence and the remaining flow velocity vector time series data sequence after the previous decomposition are calculated. If the energy of the current remaining flow velocity vector time series data sequence is lower than a set energy threshold or the change amplitude between the current remaining flow velocity vector time series data sequence and the remaining flow velocity vector time series data sequence after the previous decomposition is lower than a set change threshold, the decomposition is terminated, and the current remaining flow velocity vector time series data sequence is output as a residual component. Otherwise, the current remaining flow velocity vector time series data sequence is taken as new input data, and the decomposition is continued.
[0032] The intrinsic mode function components and the final residual component are output as the empirical mode decomposition result of the flow velocity component time series data of the matrix flow sensor.
[0033] Optionally, the intrinsic mode function components of all the matrix flow sensors are subjected to spatial alignment and correlation analysis, multi-scale flow feature data are extracted, and spatial dynamic modes and abnormal change features of the flue gas flow field at different time scales are identified, including:
[0034] The intrinsic mode function components of all the matrix flow sensors are spatially aligned according to the same component number, spatial position and sampling time, and the intrinsic mode function components of all the matrix flow sensors at the same order are formed into a spatially aligned intrinsic mode function component matrix.
[0035] For the intrinsic mode function component matrix, the mean and variance of the intrinsic mode function components corresponding to all spatial positions are calculated at each sampling time, and the spatiotemporal statistical feature data of the intrinsic mode function components of each order are obtained.
[0036] In the intrinsic mode function component matrix, the correlation coefficients of the intrinsic mode function components between all spatial positions are calculated, and the spatial correlation distribution matrix at each order and each sampling time is obtained.
[0037] The spatial correlation distribution matrix and the spatiotemporal statistical feature data are normalized to obtain a multi-scale, multi-spatial position and multi-sampling time flow feature data set.
[0038] The flow feature data set is subjected to multi-scale clustering analysis and mutation point detection, and feature intervals with significant correlation improvement, mutation or energy abnormal change in spatial distribution and time sequence are identified and divided into spatial dynamic dominant mode intervals and spatial abnormal change feature intervals.
[0039] The multi-scale flow feature data, the spatial dynamic dominant mode intervals and the spatial abnormal change feature intervals are sorted and output according to the spatial position, the sampling time and the decomposition order.
[0040] Optionally, the multi-scale flow characteristic data and the matrix time series flow data are combined to calculate the carbon emissions at each moment according to a preset carbon emissions calculation rule to generate a continuous carbon emissions comprehensive assessment result, including:
[0041] Extract multi-scale flow characteristic data and matrix time series flow data, and organize them into input data sets for carbon emission calculation;
[0042] At each sampling moment, the volume flow rate is calculated for the matrix flow sensor at each spatial position based on the current flow velocity component, horizontal flow velocity component, and vertical flow velocity component combined with the spatial distribution structure;
[0043] Acquire the flue gas carbon dioxide concentration data corresponding to each matrix flow sensor, perform spatial synchronization and time alignment, and form a gas component concentration data sequence that corresponds one-to-one to the volume flow rate;
[0044] For each sampling moment, the volume flow rate of all spatial positions is multiplied by the carbon dioxide concentration at the corresponding spatial position and sampling moment to obtain the carbon emissions per unit time at each spatial position and sampling moment. The carbon emissions per unit time of all spatial positions at the same sampling moment are accumulated to obtain the total carbon emissions of the entire section at that sampling moment, where the volume flow rate is the flue gas volume flow rate corresponding to the spatial position and sampling moment, and the carbon dioxide concentration is the concentration of carbon dioxide in the flue gas corresponding to the spatial position and sampling moment.
[0045] The total carbon emissions at each moment are arranged in chronological order to form continuous carbon emissions time series data. Combined with multi-scale flow characteristic data, trend analysis, periodicity analysis and anomaly detection are performed to generate comprehensive carbon emissions assessment results and output continuous comprehensive carbon emissions assessment results.
[0046] Optionally, when abnormal fluctuations are detected in the intrinsic mode function component or residual component of any matrix flow sensor, carbon emission abnormality warning information is automatically generated and output for operating condition tracing and response processing, including:
[0047] The intrinsic mode function components and residual components of all matrix flow sensors are continuously monitored in chronological order, and the amplitude change characteristic data of each component at each sampling moment are extracted;
[0048] Set the abnormal fluctuation judgment threshold, and judge whether the intrinsic mode function component or residual component has a drastic change exceeding the threshold at any sampling time based on the statistical characteristics of historical data, the component amplitude change rate and abnormal change experience;
[0049] Once the amplitude variation of the intrinsic mode function component or residual component of any matrix flow sensor exceeds the determination threshold, it is determined as abnormal fluctuation and the corresponding spatial position, sampling time and related component information are recorded;
[0050] Automatic generation of carbon emission anomaly warning information, including the spatial position, time, component type and amplitude variation value of abnormal fluctuation, and outputting the warning information to the monitoring device and related response terminal;
[0051] According to the warning information, the historical flow, concentration and running state data of the corresponding period are retrieved, the causes of the anomaly are traced back, and the traceability data report for operation and maintenance processing and abnormal response is output.
[0052] The carbon emission comprehensive evaluation system for matrix flue gas flow measurement according to the embodiment of the application comprises the following modules:
[0053] The sensing and collecting module is used for laying multiple matrix flow sensors and continuously collecting multi-source time series data;
[0054] The data preprocessing module is used for preprocessing the multi-source time series data to generate matrix time series flow data;
[0055] The optical flow estimation module is used for calculating the flow velocity vector time series data based on the matrix time series flow data by using the optical flow estimation algorithm;
[0056] The modal decomposition module is used for performing empirical mode decomposition on the flow velocity vector time series data to obtain the intrinsic mode function component and the residual component;
[0057] The feature analysis module is used for spatial alignment and correlation analysis of the intrinsic mode function component, extraction of multi-scale flow feature data, identification of dynamic mode and abnormal features;
[0058] The carbon emission evaluation module is used for combining the multi-scale flow feature and the matrix time series flow data, calculating the continuous carbon emission comprehensive evaluation result according to the carbon emission calculation rule;
[0059] The abnormal warning module is used for monitoring the abnormal fluctuation of the intrinsic mode function component or the residual component, automatically generating the abnormal warning information and outputting the working condition traceability and response processing.
[0060] The beneficial effects of the application are:
[0061] The present invention constructs a matrix flow sensor network, combines the optical flow estimation algorithm and the empirical mode decomposition algorithm, and performs dynamic analysis and multi-scale feature extraction on multi-source time-series flow data in the industrial flue, thereby realizing accurate identification of the spatial dynamic pattern and abnormal changes of the flue gas flow field. Compared with traditional single-point or a small amount of distributed measurement and static models, the present invention reconstructs the flue gas velocity vector field of the entire cross-section in real time, effectively captures the local disturbance and global evolution process of the flow field, and provides data support for the temporal and spatial distribution evaluation and dynamic trend analysis of carbon emissions. The present invention introduces multi-scale feature analysis and spatial correlation detection, which improves the response sensitivity to sudden abnormalities in carbon emissions and changes in operating conditions, realizes abnormal warning and operating condition tracing functions, improves the accuracy, real-time nature and intelligence level of industrial flue gas carbon emission monitoring, and meets the actual needs of modern industrial low-carbon management and intelligent operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0063] Figure 1 This is a flow chart of the carbon emission comprehensive assessment method for matrix flue gas flow measurement proposed in the present invention;
[0064] Figure 2 This is a structural diagram of the carbon emission comprehensive evaluation system for matrix flue gas flow measurement proposed in the present invention. DETAILED DESCRIPTION
[0065] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.
[0066] refer to Figure 1 , a comprehensive carbon emission assessment method for matrix flue gas flow measurement, including:
[0067] Multiple matrix flow sensors are deployed on the cross section of the industrial flue, and multi-source time series data of each matrix flow sensor is continuously collected. The multi-source time series data is pre-processed to generate matrix time series flow data.
[0068] Based on the matrix time series flow data, the optical flow estimation algorithm is used to calculate the flow velocity vector time series data of each matrix flow sensor at each moment;
[0069] The empirical mode decomposition algorithm is applied to the flow velocity vector time series data of each matrix flow sensor to decompose multiple intrinsic mode function components and residual components;
[0070] The intrinsic mode function components of all matrix flow sensors are spatially aligned and correlation analysis is performed, multi-scale flow feature data is extracted, and spatial dynamic patterns and abnormal change characteristics of the flue gas flow field at different time scales are identified;
[0071] In combination with the multi-scale flow feature data and the matrix time series flow data, the carbon emissions at each time are calculated according to a preset carbon emission calculation rule, and a continuous carbon emission comprehensive evaluation result is generated;
[0072] When it is detected that the intrinsic mode function component or the residual component of any matrix flow sensor abnormally fluctuates, carbon emission abnormal warning information is automatically generated, and output is used for working condition tracing and response processing.
[0073] In this embodiment, the multi-source time series data specifically includes flue gas flow, flue gas temperature, flue gas pressure and flue gas carbon dioxide concentration collected by each sensing point.
[0074] In this embodiment, the preprocessing of the multi-source time series data specifically includes time synchronization, missing value filling, outlier removal and normalization processing of the multi-source time series data.
[0075] In this embodiment, based on the matrix time series flow data, an optical flow estimation algorithm is used for calculation to obtain the flow velocity vector time series data of each matrix flow sensor at each time, including:
[0076] The matrix time series flow data is arranged into a three-dimensional space-time data matrix according to the spatial position of each matrix flow sensor and each sampling time;
[0077] For each matrix flow sensor in the three-dimensional space-time data matrix, the mean and standard deviation are calculated based on the historical collection data of the matrix flow sensor, the scaling coefficient and the offset of the matrix flow sensor are set, and the scaling and offset correction is performed on all collection data of the matrix flow sensor to obtain a calibrated three-dimensional space-time data matrix;
[0078] On the time series, the data frames of two adjacent time points are sequentially extracted from the calibrated three-dimensional space-time data matrix as a group of input data pairs of the optical flow estimation algorithm;
[0079] For each group of input data pairs, for each matrix flow sensor, a number of adjacent matrix flow sensors in the spatial neighborhood are selected with the current matrix flow sensor as the center, the absolute values of the differences between the flow values of the matrix flow sensor and each adjacent matrix flow sensor at the current time are counted respectively, and the average of all absolute values is obtained, which is the flow change amplitude of the current matrix flow sensor and the spatial neighborhood, and the spatial smoothing weight of the matrix flow sensor is set according to the flow change amplitude, the greater the flow change amplitude, the smaller the spatial smoothing weight, the smaller the flow change amplitude, the greater the spatial smoothing weight;
[0080] The spatial smoothing weights of all matrix flow sensors are substituted into the optical flow optimization objective function, and the horizontal direction flow velocity component and the vertical direction flow velocity component of each sensing point at the current time are solved respectively by an iterative optimization method:
[0081]
[0082] Wherein, E represents the optimization objective function in the optical flow estimation algorithm, I x represents the first-order partial derivative of the calibrated three-dimensional space-time data matrix in the spatial horizontal direction, u represents the horizontal direction flow velocity component of the current matrix flow sensor at the current time, I y represents the first-order partial derivative of the calibrated three-dimensional space-time data matrix in the spatial vertical direction, v represents the vertical direction flow velocity component of the current matrix flow sensor at the current time, I t represents the first-order partial derivative of the calibrated three-dimensional space-time data matrix in the time direction, and a(x, y) represents the spatial smoothing weight of the matrix flow sensor at the spatial position (x, y), represents the gradient of the horizontal direction flow velocity component of all matrix flow sensors in the spatial direction, represents the gradient of the vertical direction flow velocity component of all matrix flow sensors in the spatial direction.
[0083] For the horizontal direction flow velocity component and the vertical direction flow velocity component of each matrix flow sensor at all times, a fixed length sliding time window is set, the flow velocity time series data of each matrix flow sensor is traversed, the mean and standard deviation of the flow velocity in the window are counted, and it is judged whether the current flow velocity component differs from the window mean by more than a preset threshold, and marked as an abnormal flow velocity point;
[0084] For the data marked as an abnormal flow velocity point, the mean of the flow velocity component at other times in the window or according to the adjacent time flow velocity trend is replaced or corrected to obtain the corrected flow velocity component;
[0085] The corrected flow velocity components of all matrix flow sensors are arranged according to the spatial position and sampling time to form the flow velocity vector time series data.
[0086] In this embodiment, the flow velocity vector time series data of each matrix flow sensor is decomposed into multiple intrinsic mode function components and residual components by using the empirical mode decomposition algorithm, including:
[0087] The flow velocity vector time series data of each matrix flow sensor is sorted into flow velocity vector time series data sequences according to each matrix flow sensor and at all sampling times, including flow velocity components in the horizontal direction and the vertical direction.
[0088] For each flow velocity component time series data sequence of each matrix flow sensor, before each round of decomposition, according to the trend of the first and last points of the flow velocity component time series data sequence, a specified number of virtual end points are generated at the beginning and end of the flow velocity component time series data sequence by using quadratic extrapolation, and the virtual end points are added to both ends of the flow velocity component time series data sequence.
[0089] The flow velocity component time series data sequence with virtual end points is detected for all maximum and minimum points, the amplitude difference between each extreme point and the adjacent data point is calculated, the extreme amplitude threshold is set according to the signal variance, and only the maximum and minimum points with an amplitude greater than the extreme amplitude threshold are retained.
[0090] The retained maximum points are used as nodes to construct the upper envelope line by using spline interpolation, and the retained minimum points are used as nodes to construct the lower envelope line. For each time point in the constructed upper envelope line and lower envelope line, the upper envelope line amplitude value and the lower envelope line amplitude value corresponding to the time point are read respectively, and the average of the sum of the two is taken to obtain the mean value of the current time point. The mean values of all time points are arranged in time sequence to form the mean sequence of the upper and lower envelope lines corresponding to the flow velocity vector time series data sequence.
[0091] The current flow velocity component time series data sequence is subtracted from the mean sequence to obtain a new candidate data sequence.
[0092] It is judged whether the current candidate data sequence meets the intrinsic mode function component condition, which includes that the difference between the number of extreme points and the number of zero-crossing points in the candidate data sequence is not greater than 1, and the mean value of the upper envelope line and the lower envelope line of the candidate data sequence is close to zero as a whole in the entire time sequence. If the above two conditions are met, the candidate data sequence is output as an intrinsic mode function component, and the intrinsic mode function component is subtracted from the current flow velocity vector time series data sequence to obtain a new residual data sequence. If not, the current candidate data sequence is used as the basis to repeat the virtual extension of the data at both ends, the extreme point screening, the envelope line construction, the mean value calculation and the new candidate sequence construction until the candidate data sequence meets the above conditions.
[0093] After extracting an intrinsic mode function component, the proportion of the energy of the current remaining flow velocity vector time series data sequence to the total energy of the original flow velocity vector time series data sequence and the amplitude change between the current remaining flow velocity vector time series data sequence and the remaining flow velocity vector time series data sequence after the previous decomposition are calculated. If the energy of the current remaining flow velocity vector time series data sequence is lower than a set energy threshold or the change amplitude between the current remaining flow velocity vector time series data sequence and the remaining flow velocity vector time series data sequence after the previous decomposition is lower than a set change threshold, the decomposition is terminated, and the current remaining flow velocity vector time series data sequence is output as a residual component. Otherwise, the current remaining flow velocity vector time series data sequence is taken as new input data for continuous decomposition.
[0094] The intrinsic mode function components and the final residual component are output as the empirical mode decomposition result of the flow velocity component time series data of the matrix flow sensor.
[0095] In the embodiment, the spatial alignment and correlation analysis of the intrinsic mode function components of all matrix flow sensors are performed to extract multi-scale flow feature data, identify spatial dynamic patterns and abnormal change features of the flue gas flow field at different time scales, including:
[0096] The intrinsic mode function components of all matrix flow sensors are spatially aligned according to the same component number, spatial position and sampling time to form an intrinsic mode function component matrix of spatial alignment.
[0097] For the intrinsic mode function component matrix, the mean and variance of the intrinsic mode function components corresponding to all spatial positions are calculated at each sampling time to obtain the spatiotemporal statistical feature data of each order intrinsic mode function component.
[0098] In the intrinsic mode function component matrix, the correlation coefficients of the intrinsic mode function components between all spatial positions are calculated to obtain a spatial correlation distribution matrix at each order and each sampling time.
[0099] The spatial correlation distribution matrix and the spatiotemporal statistical feature data are normalized to obtain a multi-scale, multi-space position and multi-sampling time flow feature data set.
[0100] The multi-scale clustering analysis and mutation point detection are performed on the flow feature data set to identify feature intervals with significant correlation improvement, mutation or energy abnormal change in spatial distribution and time sequence, which are divided into spatial dynamic dominant mode intervals and spatial abnormal change feature intervals.
[0101] The multi-scale flow characteristic data, the spatial dynamic dominant mode interval and the spatial abnormal change characteristic interval are sorted and output according to the spatial position, the sampling time and the decomposition order.
[0102] In the embodiment, the multi-scale flow characteristic data and the matrix time sequence flow data are combined to calculate the carbon emission at each time according to a preset carbon emission calculation rule, and a continuous carbon emission comprehensive evaluation result is generated, including:
[0103] The multi-scale flow characteristic data and the matrix time sequence flow data are extracted and sorted into input data sets for carbon emission calculation;
[0104] At each sampling time, the volume flow of each spatial position is calculated according to the current flow velocity component, the horizontal direction flow velocity component and the vertical direction flow velocity component, combined with the spatial distribution structure;
[0105] The flue gas carbon dioxide concentration data corresponding to each matrix flow sensor are obtained and spatially synchronized and time-aligned to form a gas component concentration data sequence corresponding to the volume flow one by one;
[0106] For each sampling time, the volume flow of all spatial positions is multiplied by the corresponding spatial position and sampling time carbon dioxide concentration to obtain the carbon emission per unit time of each spatial position and sampling time, and the carbon emission per unit time of all spatial positions at the same sampling time is accumulated to obtain the total carbon emission of the entire cross section at the sampling time, wherein the volume flow is the flue gas volume flow rate corresponding to the spatial position and the sampling time, and the carbon dioxide concentration is the concentration of carbon dioxide in the flue gas corresponding to the spatial position and the sampling time;
[0107] The total carbon emission at each time is arranged in time sequence to form continuous carbon emission time sequence data, combined with the multi-scale flow characteristic data, for trend analysis, periodicity analysis and anomaly detection, to generate a carbon emission comprehensive evaluation result, and output the continuous carbon emission comprehensive evaluation result.
[0108] In the embodiment, when it is detected that the intrinsic mode function component or the residual component of any matrix flow sensor abnormally fluctuates, carbon emission abnormal warning information is automatically generated and output for working condition tracing and response processing, including:
[0109] The intrinsic mode function components and residual components of all matrix flow sensors are continuously monitored in time sequence, and the amplitude change characteristic data of each component at each sampling time are extracted;
[0110] An abnormal fluctuation judgment threshold is set, and whether the intrinsic mode function component or the residual component of the matrix flow sensor appears a sharp change exceeding the threshold at any sampling time is judged according to the statistical characteristics of the historical data, the component amplitude change rate and the abnormal change experience;
[0111] Once the amplitude change of the intrinsic mode function component or the residual component of any matrix flow sensor exceeds the judgment threshold, it is immediately judged as abnormal fluctuation and the corresponding spatial position, sampling time and related component information are recorded;
[0112] Abnormal fluctuation warning information is automatically generated, including the spatial position, time, component type and amplitude change value of the abnormal fluctuation, and the warning information is output to the monitoring device and the related response terminal;
[0113] According to the warning information, the historical flow, concentration and running state data of the corresponding period are searched, the causes of the abnormality are traced back, and the traceability data report for operation and maintenance processing and abnormal response is output.
[0114] Reference Figure 2 , a carbon emission comprehensive evaluation system for matrix flue gas flow measurement, comprising the following modules:
[0115] A sensing and collecting module is used to arrange multiple matrix flow sensors and continuously collect multi-source time series data;
[0116] A data preprocessing module is used to preprocess the multi-source time series data to generate matrix time series flow data;
[0117] An optical flow estimation module is used to calculate the flow velocity vector time series data based on the matrix time series flow data using an optical flow estimation algorithm;
[0118] A modal decomposition module is used to perform empirical mode decomposition on the flow velocity vector time series data to obtain intrinsic mode function components and residual components;
[0119] A feature analysis module is used to perform spatial alignment and correlation analysis on the intrinsic mode function components, extract multi-scale flow feature data, identify dynamic patterns and abnormal features;
[0120] A carbon emission evaluation module is used to combine the multi-scale flow features and the matrix time series flow data, calculate the continuous carbon emission comprehensive evaluation result according to the carbon emission calculation rules;
[0121] An abnormal warning module is used to monitor the abnormal fluctuation of the intrinsic mode function component or the residual component, automatically generate abnormal warning information and output working condition traceability and response processing.
[0122] Embodiment 1:
[0123] To verify the feasibility of the application in implementation, the application is applied to the sintering section of a certain large steel enterprise, which has long been faced with the problems of large fluctuation of flue gas emission flow, high error of total carbon emission assessment, and lag of abnormal monitoring. The traditional scheme adopts single-point flow meters and fixed-point gas analyzers, which can only cover partial areas of the flue, cannot reflect the overall cross-sectional flow field changes, leading to frequent high carbon emission monitoring, and often failing to timely alert abnormal conditions during process adjustment, affecting the implementation of precise carbon emission management and intelligent operation and maintenance of the enterprise.
[0124] Therefore, in May 2025, the enterprise arranged 24 matrix flow sensors at equal intervals on the cross section of the sintering main flue with a width of 3 meters and a height of 2.5 meters, forming a high-density flow measurement network. The system automatically collects the flow, temperature, pressure, carbon dioxide concentration and other multi-source data of each sensor every 10 seconds, synchronizes the time, fills in the missing data and removes the abnormal data through the server, and generates high-quality matrix time series flow data. Using the optical flow estimation algorithm proposed in the application, the enterprise can reconstruct the full cross-sectional flow velocity vector distribution at each moment, effectively capturing local flow field disturbances and flow direction mutations. Taking 9:00-12:00 on May 6, 2025 as an example, the sensor array data shows that the flow velocity in the B2, D3 and E4 regions fluctuates significantly during 9:50-10:10, and the flow velocity field reconstructed by the optical flow algorithm can truly reflect the transient disturbance caused by the sintering process adjustment.
[0125] The system automatically performs empirical mode decomposition on the flow velocity vector time series data of each sensor to separate the dominant flow change trend and residual signal at different time scales. On the morning of May 7, 2025, the amplitudes of the high-order intrinsic mode components of C1, D4 and F3 sensors have abnormal mutations of more than 0.6, and the spatial correlation analysis identifies the synchronous fluctuations in the C1-D4 interval, and the system generates an early warning and links to the process traceability. According to the multi-scale flow characteristics and matrix flow data, the system automatically completes the total carbon emission assessment of the full cross section, and the total carbon emission of the whole day on May 6, 2025 is calculated as 7700 kg, which is lower than the 8180 kg measured by the traditional single-point method.
[0126] Table 1 Comparison of key performance between the application scheme and the traditional scheme
[0127]
[0128]
[0129] The data in Table 1 clearly demonstrates the significant advantages of the present method over traditional approaches in industrial flue gas carbon emission monitoring and response. In calculating daily average carbon emissions, the present method achieves more accurate values than traditional approaches, with a measurement error of approximately ±2%, significantly lower than the ±8% of traditional approaches. This improvement in measurement accuracy is attributed to matrix-based flow acquisition and dynamic flow field reconstruction, which effectively reduces cross-sectional measurement blind spots and error accumulation.
[0130] In terms of abnormal response, the method of the present invention completes abnormal response within 4 to 5 minutes, which is significantly shorter than the response time of 15 to 17 minutes of traditional solutions. This demonstrates the sensitivity and efficiency of the present invention in real-time analysis of flow characteristics and abnormal warning mechanisms, ensuring that enterprises can identify abnormalities earlier, take measures, and reduce environmental risks. The warning accuracy of the method of the present invention has always remained above 95%, which is significantly higher than the 78% to 79% of traditional solutions, indicating that the system can more accurately identify actual process abnormalities and sudden changes in carbon emissions, avoiding false alarms and missed reports.
[0131] In terms of maximum intraday velocity fluctuation amplitude and spatial resolution, the proposed method, utilizing 24 measuring points covering the entire cross section, captures more detailed flow field fluctuations and operating disturbances. Traditional approaches, with only two measuring points, struggle to capture the true dynamics and spatial heterogeneity of the flow field. The proposed method achieves slightly higher maximum intraday fluctuation amplitude, demonstrating its ability to sensitively capture local changes and dynamic trends within the cross section, providing a more accurate data foundation for carbon emission assessment and production optimization.
[0132] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.
Claims
1. A comprehensive carbon emission assessment method for matrix flue gas flow measurement, characterized in that: include: Multiple matrix flow sensors are deployed on the cross section of the industrial flue, and multi-source time series data of each matrix flow sensor is continuously collected. The multi-source time series data is pre-processed to generate matrix time series flow data. Based on the matrix time series flow data, the optical flow estimation algorithm is used to calculate the flow velocity vector time series data of each matrix flow sensor at each moment; The empirical mode decomposition algorithm is applied to the flow velocity vector time series data of each matrix flow sensor to decompose multiple intrinsic mode function components and residual components; Perform spatial alignment and correlation analysis on the intrinsic mode function components of all matrix flow sensors to extract multi-scale flow characteristic data and identify the spatial dynamic patterns and abnormal change characteristics of the flue gas flow field at different time scales; Combining multi-scale flow characteristic data and matrix time series flow data, according to the preset carbon emission calculation rules, calculate the carbon emissions at each moment and generate continuous carbon emission comprehensive assessment results; When abnormal fluctuations are detected in the intrinsic mode function components or residual components of any matrix flow sensor, carbon emission abnormality warning information is automatically generated and output for operating condition tracing and response processing.
2. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: The multi-source time series data specifically includes the flue gas flow, flue gas temperature, flue gas pressure and flue gas carbon dioxide concentration collected by each sensor point.
3. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: The preprocessing of multi-source time series data specifically includes time synchronization, missing value filling, outlier removal and normalization processing of the multi-source time series data.
4. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: The optical flow estimation algorithm is used to calculate the matrix-based time series flow data to obtain the flow velocity vector time series data of each matrix flow sensor at each moment, including: The matrix time series flow data is organized into a three-dimensional spatiotemporal data matrix according to the spatial position of each matrix flow sensor and each sampling moment; For each matrix flow sensor in the three-dimensional spatiotemporal data matrix, the mean and standard deviation are calculated based on the historical data collected by the matrix flow sensor, the scaling factor and offset of the matrix flow sensor are set, and all the collected data of the matrix flow sensor are scaled and offset corrected to obtain a calibrated three-dimensional spatiotemporal data matrix; In the time series, the data frames of two adjacent moments are sequentially extracted from the calibrated three-dimensional spatiotemporal data matrix as a set of input data pairs for the optical flow estimation algorithm; For each set of input data pairs, for each matrix flow sensor, with the current matrix flow sensor as the center, several adjacent matrix flow sensors in the spatial neighborhood are selected, and the absolute value of the difference in flow value between the matrix flow sensor and each adjacent matrix flow sensor at the current moment is counted respectively, and all absolute values are averaged. The average value obtained is the flow change amplitude of the current matrix flow sensor and the spatial neighborhood. The spatial smoothing weight of the matrix flow sensor is set according to the flow change amplitude. The larger the flow change amplitude, the smaller the spatial smoothing weight, and the smaller the change amplitude, the larger the spatial smoothing weight. The spatial smoothing weights of all matrix flow sensors are substituted into the optical flow optimization objective function, and the horizontal and vertical flow velocity components of each sensor point at the current moment are solved respectively through the iterative optimization method. For the horizontal and vertical flow velocity components of each matrix flow sensor at all times, a sliding time window of fixed length is set. The flow velocity time series data of each matrix flow sensor is traversed, and the mean and standard deviation of the flow velocity within the window are calculated. It is determined whether the difference between the current flow velocity component and the window mean exceeds a preset threshold and is marked as an abnormal flow velocity point. For the data marked as abnormal velocity points, the average of the velocity components at other times in the window or the velocity trend at adjacent times are used to replace or correct them to obtain the corrected velocity components; The corrected velocity components of all matrix flow sensors are arranged according to spatial position and sampling time to form velocity vector time series data.
5. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: The empirical mode decomposition algorithm is applied to the flow velocity vector time series data of each matrix flow sensor to decompose the flow velocity vector time series data to obtain multiple intrinsic mode function components and residual components, including: The flow velocity vector time series data of each matrix flow sensor is sorted into a flow velocity vector time series data sequence at all sampling moments according to each matrix flow sensor, including flow velocity components in the horizontal direction and the vertical direction; For each velocity component time series data sequence of each matrix flow sensor, before each round of decomposition begins, a quadratic extrapolation method is used to generate a specified number of virtual endpoints at the beginning and end of the velocity component time series data sequence based on the trends of several points at the beginning and end of the velocity component time series data sequence, and then the virtual endpoints are added to both ends of the velocity component time series data sequence; For the velocity component time series data sequence with virtual endpoints, all maximum and minimum points are detected, the amplitude difference between each extreme point and the adjacent data points is calculated, the extreme amplitude threshold is set according to the signal variance, and only the maximum and minimum points with amplitudes greater than the extreme amplitude threshold are retained; The upper envelope is constructed using the retained maximum points as nodes using spline interpolation, and the lower envelope is constructed using the retained minimum points as nodes. For each time point in the constructed upper and lower envelopes, the amplitude values of the upper and lower envelopes corresponding to the time point are read respectively, and the two are added and averaged to obtain the mean value of the current time point. The mean values at all time points are arranged in chronological order to form the mean sequence of the upper and lower envelopes corresponding to the velocity vector time series data sequence; Subtract the mean sequence from the current velocity component time series data sequence to obtain a new candidate data sequence; Determine whether the current candidate data sequence meets the intrinsic mode function component conditions, wherein the intrinsic mode function component conditions include: the difference between the number of extreme points and the number of zero crossing points in the candidate data sequence is not greater than 1, and the mean of the upper envelope and the lower envelope of the candidate data sequence is close to zero as a whole in the entire time series; if the above two conditions are met, output the candidate data sequence as an intrinsic mode function component, and subtract the intrinsic mode function component from the current flow vector time series data sequence to obtain a new residual data sequence; if not, based on the current candidate data sequence, repeatedly perform virtual extension of the data at both ends, extreme point screening, envelope construction, mean calculation, and construct a new candidate sequence until the candidate data sequence meets the above conditions; After extracting an intrinsic mode function component, calculate the ratio of the energy of the current remaining velocity vector time series data sequence to the total energy of the original velocity vector time series data sequence, as well as the amplitude change between the velocity vector time series data sequence remaining after the previous round of decomposition. If the energy of the current remaining velocity vector time series data sequence is lower than the set energy threshold, or the amplitude change between the current remaining velocity vector time series data sequence and the velocity vector time series data sequence remaining in the previous round is lower than the set change threshold, then terminate the decomposition and output the current remaining velocity vector time series data sequence as the residual component; otherwise, use the current remaining velocity vector time series data sequence as new input data and continue decomposition. All intrinsic mode function components and the final residual components are output as the empirical mode decomposition results of the flow velocity component time series data of the matrix flow sensor.
6. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: The spatial alignment and correlation analysis of the intrinsic mode function components of all matrix flow sensors are performed to extract multi-scale flow characteristic data and identify the spatial dynamic patterns and abnormal change characteristics of the flue gas flow field at different time scales, including: The obtained intrinsic mode function components of each order of all matrix flow sensors are spatially aligned according to the same component sequence number, spatial position and sampling time to form a spatially aligned intrinsic mode function component matrix; For the intrinsic mode function component matrix, at each sampling moment, the mean and variance of the intrinsic mode function components corresponding to all spatial positions are calculated to obtain the spatiotemporal statistical characteristic data of the intrinsic mode function components of each order; In the intrinsic mode function component matrix, the correlation coefficients of the intrinsic mode function components between all spatial positions are calculated respectively to obtain the spatial correlation distribution matrix at each order and each sampling time; Normalize the spatial correlation distribution matrix and spatiotemporal statistical feature data to obtain a set of traffic feature data at multiple scales, multiple spatial locations, and multiple sampling times; Multi-scale cluster analysis and mutation point detection are performed on the traffic characteristic data set to identify characteristic intervals with significant correlation increase, mutation or abnormal energy change in spatial distribution and time series, and divide them into spatial dynamic dominant mode intervals and spatial abnormal change characteristic intervals; The multi-scale traffic characteristic data, spatial dynamic dominant mode intervals and spatial abnormal change characteristic intervals are sorted and output according to spatial position, sampling time and decomposition order.
7. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: The multi-scale flow characteristic data and matrix time series flow data are combined to calculate the carbon emissions at each moment according to the preset carbon emissions calculation rules, and generate continuous carbon emissions comprehensive assessment results, including: Extract multi-scale flow characteristic data and matrix time series flow data, and organize them into input data sets for carbon emission calculation; At each sampling moment, the volume flow rate is calculated for the matrix flow sensor at each spatial position based on the current flow velocity component, horizontal flow velocity component, and vertical flow velocity component combined with the spatial distribution structure; Acquire the flue gas carbon dioxide concentration data corresponding to each matrix flow sensor, perform spatial synchronization and time alignment, and form a gas component concentration data sequence that corresponds one-to-one to the volume flow rate; For each sampling moment, the volume flow rate of all spatial positions is multiplied by the carbon dioxide concentration at the corresponding spatial position and sampling moment to obtain the carbon emissions per unit time at each spatial position and sampling moment. The carbon emissions per unit time of all spatial positions at the same sampling moment are accumulated to obtain the total carbon emissions of the entire section at that sampling moment, where the volume flow rate is the flue gas volume flow rate corresponding to the spatial position and sampling moment, and the carbon dioxide concentration is the concentration of carbon dioxide in the flue gas corresponding to the spatial position and sampling moment. The total carbon emissions at each moment are arranged in chronological order to form continuous carbon emissions time series data. Combined with multi-scale flow characteristic data, trend analysis, periodicity analysis and anomaly detection are performed to generate comprehensive carbon emissions assessment results and output continuous comprehensive carbon emissions assessment results.
8. The carbon emission comprehensive assessment method for matrix flue gas flow measurement according to claim 1 is characterized in that: When abnormal fluctuations are detected in the intrinsic mode function components or residual components of any matrix flow sensor, carbon emission abnormality warning information is automatically generated and output for operating condition tracing and response processing, including: The intrinsic mode function components and residual components of all matrix flow sensors are continuously monitored in chronological order, and the amplitude change characteristic data of each component at each sampling moment are extracted; Set the abnormal fluctuation judgment threshold, and judge whether the intrinsic mode function component or residual component has a drastic change exceeding the threshold at any sampling time based on the statistical characteristics of historical data, the component amplitude change rate and abnormal change experience; Once the amplitude change of the intrinsic mode function component or residual component of any matrix flow sensor exceeds the judgment threshold, it is immediately determined to be an abnormal fluctuation and the corresponding spatial position, sampling time and related component information are recorded; Automatically generate carbon emission abnormality warning information, including the spatial location, time, component type and amplitude change value of the abnormal fluctuation, and output the warning information to the monitoring equipment and related response terminals; Based on the early warning information, the historical flow, concentration and operating status data of the corresponding period are retrieved, the cause of the abnormality is traced, and a traceability data report is output for operation and maintenance processing and abnormal response.
9. A carbon emission comprehensive assessment system for matrix flue gas flow measurement, which implements the carbon emission comprehensive assessment method for matrix flue gas flow measurement according to any one of claims 1 to 8, characterized in that: Includes the following modules: The sensor acquisition module is used to deploy multiple matrix flow sensors and continuously collect multi-source time series data; Data preprocessing module, used to preprocess multi-source time series data and generate matrix time series traffic data; The optical flow estimation module is used to calculate the flow velocity vector time series data based on the matrix time series flow data using the optical flow estimation algorithm; The modal decomposition module is used to perform empirical mode decomposition on the velocity vector time series data to obtain the intrinsic mode function components and residual components; Feature analysis module, used to perform spatial alignment and correlation analysis on intrinsic mode function components, extract multi-scale flow feature data, and identify dynamic patterns and abnormal features; The carbon emission assessment module is used to combine multi-scale flow characteristics and matrix time series flow data to calculate the continuous carbon emission comprehensive assessment results according to the carbon emission calculation rules; The abnormal warning module is used to monitor abnormal fluctuations of intrinsic mode function components or residual components, automatically generate abnormal warning information and output working condition tracing and response processing.