A method, equipment, and medium for predicting carbon emissions from coal-fired power units.

Abstract

This invention discloses a method, equipment, and medium for predicting carbon emissions from coal-fired power units, relating to the field of intelligent energy data analysis technology. The method includes: calculating predicted carbon emission intensity and confidence interval boundary values ​​based on a real-time coal quality fusion operating parameter set and equipment health index, generating a dynamic carbon emission prediction package; comparing the dynamic carbon emission prediction package with CEMS real-time monitoring data, calculating the prediction error rate, and extracting multi-dimensional error features based on the prediction error rate to generate a multi-dimensional error feature vector; dynamically adjusting the equipment health index correlation factor and coal quality confidence weight parameters based on the multi-dimensional error feature vector to generate a dynamic correction instruction set, and combining it with the dynamic carbon emission prediction package to generate a carbon emission prediction report. This invention achieves dynamic noise reduction, enhancement, and feature quantization of multispectral coal flow images, directly generating high-fidelity coal quality feature vectors, solving the problem of coal quality data lag, and improving the response capability to sudden changes in coal quality.

Description

Technical Field

[0001] This invention relates to the field of intelligent energy data analysis technology, and in particular to a method, equipment and medium for predicting carbon emissions from coal-fired power units. Background Technology

[0002] In the field of carbon emission monitoring for coal-fired power units, two main technical approaches are typically employed: static calculation methods based on fuel element analysis and operating parameters, which use standard formulas combined with coal quality data and unit efficiency parameters to estimate theoretical emissions; and methods relying on real-time monitoring of flue gas components to directly obtain gas concentration and velocity data to generate emission results. In recent years, data-driven methods have been gradually applied, using time-series analysis techniques to integrate operating status parameters for trend prediction, and related mechanism research and practical applications have established a certain technical foundation.

[0003] However, existing methods have certain limitations at the dynamic prediction level. Conventional coal quality data has a long update cycle, making it difficult to capture minute-level fluctuations in the characteristics of fuel entering the furnace, which leads to biases in the prediction model during coal type switching scenarios; and fixed efficiency coefficients cannot characterize the continuous impact of equipment performance degradation on carbon emissions, resulting in a gradual decrease in prediction accuracy for long-term operation scenarios. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a carbon emission prediction method for coal-fired power units to address the problems of lagging dynamic response of coal quality and lack of equipment condition compensation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for predicting carbon emissions from coal-fired power units, comprising: acquiring raw coal flow image data and performing preprocessing and real-time feature extraction to obtain a real-time coal quality dataset; acquiring equipment operating parameters and performing data cleaning in conjunction with the real-time coal quality dataset to generate a real-time coal quality fusion operating parameter set, and simultaneously obtaining an equipment health index in conjunction with an equipment health assessment rule base; calculating the predicted carbon emission intensity and confidence interval boundary values ​​based on the real-time coal quality fusion operating parameter set and the equipment health index to generate a dynamic carbon emission prediction package; comparing the dynamic carbon emission prediction package with CEMS real-time monitoring data, calculating the prediction error rate, and extracting multi-dimensional error features based on the prediction error rate to generate a multi-dimensional error feature vector; and dynamically adjusting the equipment health index correlation factor and coal quality confidence weight parameters based on the multi-dimensional error feature vector to generate a dynamic correction instruction set, and combining the dynamic carbon emission prediction package to generate a carbon emission prediction report.

[0008] As a preferred embodiment of the carbon emission prediction method for coal-fired power units described in this invention, the steps for acquiring real-time coal quality data are as follows:

[0009] The raw coal flow image data refers to multispectral imaging data;

[0010] The preprocessing includes noise reduction, enhancement, and standardization.

[0011] Based on the preprocessed raw coal flow image data, color distribution features, texture roughness features, and particle size distribution features are extracted and integrated to generate a coal quality multimodal feature vector;

[0012] Based on the multimodal feature vector of coal quality, linear interpolation calculation is performed through a preset feature value-coal composition mapping table to obtain a real-time coal quality dataset.

[0013] As a preferred embodiment of the carbon emission prediction method for coal-fired power units described in this invention, the generation of the real-time coal quality fusion operation parameter set refers to performing timestamp alignment, missing value filling, and outlier removal processing on the real-time coal quality dataset and equipment operation parameters to generate the real-time coal quality fusion operation parameter set.

[0014] In a preferred embodiment of the carbon emission prediction method for coal-fired power units according to the present invention, the steps for obtaining the equipment health index are as follows:

[0015] Based on the real-time coal quality fusion operation parameter set, the equipment operation parameters are extracted, and the equipment operation parameters are compared with the corresponding equipment operation parameter benchmark values ​​to calculate the normalized offset and generate a parameter normalized offset set.

[0016] Based on the parameter normalization offset set, a predefined equipment health assessment rule library is called, and a weighted fusion algorithm is used to calculate the comprehensive equipment performance score. The comprehensive equipment performance score is then mapped to generate an equipment health index.

[0017] As a preferred embodiment of the carbon emission prediction method for coal-fired power units according to the present invention, the steps for generating the dynamic carbon emission prediction package are as follows:

[0018] Based on the real-time coal quality fusion operation parameter set, the theoretical carbon emission baseline value is calculated using the carbon conservation formula;

[0019] Based on the equipment health index, query the preset index-correction factor mapping rule to obtain the carbon emission correction factor, and dynamically correct the theoretical carbon emission baseline value based on the carbon emission correction factor to obtain the actual carbon emission value.

[0020] By coupling actual carbon emissions with real-time power generation data, the predicted carbon emission intensity is calculated, generating a dynamic carbon emission intensity prediction set.

[0021] Real-time data collection of coal consumption fluctuation range, combined with the changing trend of equipment health index, and calculation of confidence interval boundary values ​​through linear error synthesis algorithm to generate dynamic confidence interval set;

[0022] The dynamic carbon emission intensity prediction set and the dynamic confidence interval set are integrated to generate a dynamic carbon emission prediction package.

[0023] As a preferred embodiment of the carbon emission prediction method for coal-fired power units described in this invention, the steps for generating the multidimensional error feature vector are as follows:

[0024] The CO2 concentration, flow rate and temperature of flue gas are collected in real time by the equipment flue gas emission monitoring equipment to generate a CEMS real-time monitoring dataset.

[0025] The dynamic carbon emission intensity prediction set is matched with the CEMS real-time monitoring dataset on the time axis, and the clock deviation is compensated by interpolation to generate a spatiotemporal synchronization verification dataset.

[0026] Based on the spatiotemporal synchronous verification dataset, the prediction error rate is calculated point by point, and the average error rate is statistically analyzed. At the same time, high error intervals are marked according to the average error rate, and a high error persistence interval marking sequence is generated.

[0027] Based on the high-error persistent interval labeled sequence, three-dimensional features are extracted to generate the original error feature vector;

[0028] The original error feature vector is mapped to the equipment health index correlation factor adjustment coefficient and coal quality characteristic confidence weight parameter through a predefined engineering rule base, thereby generating a multidimensional error feature vector.

[0029] As a preferred embodiment of the carbon emission prediction method for coal-fired power units according to the present invention, the steps for generating the dynamic correction instruction set are as follows:

[0030] Analyze the equipment attenuation weight component and coal quality mutation weight component in the multidimensional error feature vector to generate a set of error dominant factors;

[0031] Based on the error-dominant factor set, a predefined parameter adjustment mapping table is called to dynamically adjust the equipment health index correlation factors and coal quality confidence weight parameters, generating a dynamic correction instruction set.

[0032] As a preferred embodiment of the carbon emission prediction method for coal-fired power units described in this invention, the carbon emission prediction report is generated by real-time correction of the dynamic carbon emission prediction package using a dynamic correction instruction set.

[0033] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the carbon emission prediction method for coal-fired power unit carbon emissions as described in the first aspect of the present invention.

[0034] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the carbon emission prediction method for coal-fired power unit carbon emissions as described in the first aspect of the present invention.

[0035] The beneficial effects of this invention are as follows: by collecting raw coal flow image data and performing preprocessing and real-time feature extraction, dynamic noise reduction, enhancement, and feature quantization of multispectral coal flow images are achieved, directly generating high-fidelity coal quality feature vectors, solving the problem of coal quality data lag, and improving the coal quality sudden change response capability; by dynamically adjusting the equipment health index correlation factor and coal quality confidence weight parameters based on multidimensional error feature vectors, a rule mapping from error source features to parameter correction instructions is realized, enhancing long-term prediction stability. Attached Figure Description

[0036] 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of a method for predicting carbon emissions from coal-fired power units.

[0038] Figure 2 A flowchart for generating the equipment health index.

[0039] Figure 3 A flowchart for generating dynamic carbon emission prediction packages.

[0040] Figure 4 A flowchart for generating carbon emission forecast reports. Detailed Implementation

[0041] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0042] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0044] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for predicting carbon emissions from coal-fired power units, comprising the following steps:

[0045] S1. Collect raw coal flow image data and perform preprocessing and real-time feature extraction to obtain a real-time coal quality dataset;

[0046] Raw coal flow image data refers to multispectral imaging data;

[0047] It should be noted that the multispectral imaging data is generated by acquiring the visible light (RGB) and near-infrared (NIR) band reflectance of coal flow through a multispectral industrial camera, and generating a multi-channel digital array containing spatial and spectral dimensions.

[0048] Preprocessing includes noise reduction, enhancement, and normalization.

[0049] It should be noted that noise reduction refers to spatial domain filtering and purification of pixel-level noise in multispectral imaging data caused by industrial environmental interference, replacing abnormal pixel values, and eliminating snowflake-like spots and banded interference patterns; enhancement refers to local contrast reconstruction of details lost in multispectral imaging data under complex lighting conditions, improving the visibility of coal cracks and mineral reflective bands; standardization refers to geometric and radiometric consistency correction of multispectral imaging data: unifying pixel size in the spatial dimension, converting to HSV space to separate luminance components in the color dimension, and normalizing pixel values ​​to a fixed floating-point range in the numerical dimension;

[0050] Based on the preprocessed raw coal flow image data, color distribution features, texture roughness features, and particle size distribution features are extracted and integrated to generate a coal quality multimodal feature vector;

[0051] Furthermore, based on the preprocessed raw coal flow image data, the component images of the H channel in the HSV color space are extracted, and the hue value of each pixel in the H channel component image is statistically analyzed in terms of full hue quantization interval. The occurrence frequency of pixels within each quantization level is obtained, generating a color distribution feature vector characterizing the color characteristics of the coal. The preprocessed raw coal flow image data is converted into a grayscale image. The grayscale spatial relationship of pixel pairs is analyzed along the horizontal, vertical, and two diagonal directions. For each direction, the distribution of grayscale co-occurrence of paired pixels with specific spatial intervals and directions in the grayscale image is statistically analyzed, forming a grayscale co-occurrence matrix for each direction. Four texture feature parameters—contrast, correlation, energy, and entropy—are extracted from the co-occurrence matrix for each direction to form a texture roughness feature vector characterizing surface roughness. Based on the preprocessed raw coal flow image data... The grayscale image of the original coal flow image data is used to obtain the coal particle contour through edge detection using the Canny edge detection algorithm, and noise interference is removed. Connected region marking is performed on the processed grayscale image to identify and distinguish each independent coal particle region. All marked coal particle regions are traversed, and the total number of pixels contained in each coal particle region is obtained to represent the actual physical area of ​​each coal particle region. According to the preset area interval division rules, the distribution frequency of the actual physical area of ​​each coal particle region is statistically analyzed to generate a particle size distribution histogram feature vector reflecting the particle size composition characteristics. The color distribution feature vector, texture roughness feature vector, and particle size distribution histogram feature vector are subjected to Min-Max normalization to make the values ​​of each feature vector fall within a unified fixed interval. They are then spliced ​​and fused in the order of color, texture, and particle size to generate a coal quality multimodal feature vector.

[0052] It should be noted that the full hue quantization range is a discretized hue grading system defined by linearly compressing the hue circle (0°-360°) of the HSV color space to a single-byte storage range (0-179); the preset area interval division rule is a multi-level interval division system set by analyzing the particle size distribution characteristics of historical coal particle samples and adopting a dynamic binning strategy, including three levels: basic particle size interval, transitional particle size interval, and abnormal particle size interval.

[0053] Based on the multimodal feature vector of coal quality, linear interpolation calculation is performed through a preset feature value-coal composition mapping table to obtain a real-time coal quality dataset;

[0054] Furthermore, the multimodal feature vectors of coal quality are analyzed to extract feature values ​​representing color, texture, and particle size characteristics, generating a feature set to be matched. A pre-defined feature value-coal composition mapping table is invoked to match each feature value in the feature set to the table, obtaining a partially matched set of coal components. Simultaneously, coal component data points that did not correspond precisely after matching are extracted, generating an unmatched set of coal component data points. Based on the adjacent known data points in the feature value-coal composition mapping table for each coal component data point in the unmatched set, the relative position of each coal component data point within the numerical range of the feature values ​​is determined. The coal component values ​​of each data point are estimated proportionally using the coal component values ​​of adjacent known data points, obtaining a coal component interpolation dataset. Finally, the coal component interpolation dataset and the partially matched set of coal components are combined to generate a real-time coal quality dataset.

[0055] It should be noted that the preset feature value-coal composition mapping table is a data comparison table established by regression analysis of historical coal quality test data and corresponding coal flow image features. It includes the corresponding relationship data between coal quality feature values ​​(such as color, texture, and particle size characteristics) and coal composition (such as carbon content, moisture, and ash content).

[0056] S2. Collect equipment operating parameters and perform data cleaning in conjunction with real-time coal quality datasets to generate a real-time coal quality fusion operating parameter set. At the same time, obtain the equipment health index by combining the equipment health assessment rule base.

[0057] The real-time coal quality dataset and equipment operating parameters are processed by timestamp alignment, missing value filling and outlier removal to generate a real-time coal quality fusion operating parameter set.

[0058] Furthermore, the unit's distributed sensors collect equipment operating parameters in real time, and compare the timestamps of the real-time coal quality dataset with those of the equipment operating parameters. Data points in the equipment operating parameters that do not match the timestamps of the real-time coal quality dataset are shifted and matched along the time axis to give all data points a unified time identifier, generating a timestamp-aligned coal quality and operating parameter dataset. The unit identifies gaps in the coal quality and operating parameter dataset caused by data acquisition interruptions, and infers reasonable values ​​for the gaps based on the numerical change trends of adjacent data before and after the gaps, generating a complete time-series dataset of coal quality and operating parameters. Finally, the unit identifies and removes abnormal data points in the time-series dataset that deviate from the normal range, generating a real-time coal quality fusion operating parameter set.

[0059] It should be noted that equipment operating parameters refer to the set of physical quantities reflecting the operating status of the coal-fired unit, which are collected in real time by the unit's distributed sensors. These include boiler load, flue gas oxygen content, fan speed, and air preheater pressure differential. The normal range is a threshold range determined by statistical analysis of historical operating data (such as mean ± 3 times standard deviation) or equipment design parameters (rated fluctuation range provided by the manufacturer), used to identify abnormal data points (for example, coal consumption suddenly drops to 0 or surges to 3 times the mean).

[0060] Based on the real-time coal quality fusion operation parameter set, the equipment operation parameters are extracted, and the equipment operation parameters are compared with the corresponding equipment operation parameter benchmark values ​​to calculate the normalized offset and generate a parameter normalized offset set.

[0061] Furthermore, based on the real-time coal quality fusion operation parameter set, equipment operation parameters directly related to the equipment operating status are extracted to generate an equipment operation parameter set. Corresponding equipment operation parameter benchmark values ​​are retrieved from a pre-set equipment parameter database and historical best operation data records to generate an equipment operation parameter benchmark value set. Each parameter value in the equipment operation parameter set is compared one by one with the corresponding parameter benchmark value in the equipment operation parameter benchmark value set to obtain the equipment operation parameter offset. Each equipment operation parameter offset is then normalized to obtain a normalized offset value for each equipment operation parameter. Finally, the normalized offset values ​​of each equipment operation parameter are integrated to generate a parameter normalized offset set.

[0062] It should be noted that the preset equipment parameter database is a structured data set constructed by collecting rated parameter values ​​(such as boiler rated efficiency, fan design air volume, and air preheater design differential pressure) from the design specifications and installation and commissioning reports provided by the equipment manufacturer; the historical best operating data record is a benchmark dataset generated by statistically analyzing the operating values ​​of various parameters when the equipment is in a healthy operating state within a specific period.

[0063] Based on the parameter normalization offset set, a predefined equipment health assessment rule library is called, a weighted fusion algorithm is used to calculate the comprehensive equipment performance score, and the comprehensive equipment performance score is mapped to generate the equipment health index.

[0064] Furthermore, based on a predefined equipment health assessment rule base, the weight coefficients of each equipment operating parameter and the corresponding parameter scoring calculation rules are extracted to generate an equipment health assessment rule set. According to the weight coefficients specified in the equipment health assessment rule set, each parameter offset in the parameter normalization offset set is weighted to obtain a weighted score for each operating parameter. The weighted scores of all parameters are then integrated to generate a comprehensive equipment performance score. Based on the scoring mapping rules set in the equipment health assessment rule set, the comprehensive equipment performance score is mapped to a fixed range to generate a standardized comprehensive equipment performance score. Based on the equipment health assessment rule set, predefined slope coefficients and intercept coefficients are extracted to generate a linear transformation parameter set. The standardized comprehensive equipment performance score and the linear transformation parameter set are substituted into the linear transformation formula to obtain an initial equipment health index value. This initial equipment health index value is then constrained within a standard range to generate the final equipment health index.

[0065] The comprehensive performance rating expression for computing devices is:

[0066] ;

[0067] in, It is a comprehensive score of equipment performance; This is the total number of equipment operating parameters; It is the index number of the equipment's operating parameters; It is the first A scoring function for the operating parameters of a device; It is the first parameter in the normalized offset set. Normalized offset of each device's operating parameters; It is the first in the set of equipment health assessment rules. Weighting coefficients for the operating parameters of each device;

[0068] It should be noted that the predefined equipment health assessment rule base is a set of rules constructed by analyzing a large amount of historical equipment operation data and combining it with the technical specifications provided by the equipment manufacturer. It includes the weight allocation coefficients of each equipment operation parameter, parameter scoring calculation rules, and the mapping relationship between the score and the health index. The slope coefficient is the optimal proportional factor obtained by analyzing the statistical relationship between historical equipment performance data and health status, with an exemplary value range of 80 to 120. The intercept coefficient is the baseline adjustment amount calculated by an optimization algorithm that minimizes the prediction error of historical equipment health status, with an exemplary value range of -10 to 10.

[0069] It should be noted that the fixed range is set based on the equipment manufacturer's design parameters and historical health operation data statistics, for example [0,1]; the standard range is strictly limited by boundary constraint functions to ensure industrial interpretability, for example [0,100].

[0070] S3. Based on the real-time coal quality fusion operation parameter set and equipment health index, calculate the predicted value of carbon emission intensity and the boundary value of the confidence interval, and generate a dynamic carbon emission prediction package.

[0071] Based on the real-time coal quality fusion operation parameter set, the theoretical carbon emission baseline value is calculated using the carbon conservation formula;

[0072] Furthermore, based on the real-time coal quality fusion operation parameter set, the real-time coal consumption, the basic carbon content of the boiler feed, and the boiler efficiency benchmark value are extracted to obtain real-time coal consumption and coal carbon content data; based on the coal consumption and coal carbon content data, the theoretical carbon emissions are calculated through the carbon conservation formula, and the theoretical carbon emissions are converted into standard units to generate the theoretical carbon emission baseline value.

[0073] The expression for calculating the theoretical baseline value of carbon emissions is:

[0074] ;

[0075] in, This is the theoretical baseline value for carbon emissions; This is the real-time coal consumption; It is the received basic carbon content; This is the benchmark value for boiler efficiency; It is the conversion coefficient of carbon mass to carbon dioxide (a fixed constant strictly set according to the molecular weight ratio of carbon to carbon dioxide, with an exemplary value of 3.667). It is a unit conversion factor used to convert carbon emissions from kilograms to tons;

[0076] Based on the equipment health index, query the preset index-correction factor mapping rule to obtain the carbon emission correction factor, and dynamically correct the theoretical carbon emission baseline value based on the carbon emission correction factor to obtain the actual carbon emission value.

[0077] Furthermore, based on the equipment health index, the preset index-correction factor mapping rule is queried to obtain the corresponding carbon emission correction factor value (for example, if the equipment health index falls into the range of [0,60], the corresponding correction factor value is 1.5). The carbon emission correction factor value is then used to scale the theoretical carbon emission baseline value proportionally to achieve dynamic scaling correction of the theoretical carbon emission baseline value and obtain the actual carbon emission value.

[0078] It should be noted that the preset index-correction factor mapping rule is a quantitative correspondence established by analyzing the correlation between historical equipment performance degradation data and carbon emission deviation. It includes equipment health index segment intervals and corresponding correction factor values, which are used to quantify the equipment health status into compensation coefficients for carbon emission calculation.

[0079] By coupling actual carbon emissions with real-time power generation data, the predicted carbon emission intensity is calculated, generating a dynamic carbon emission intensity prediction set.

[0080] Furthermore, by deploying smart meters at the generator outlet to collect real-time power generation data, a real-time power generation dataset with millisecond-level timestamps is generated. The actual carbon emission values ​​are then matched and aligned with the real-time power generation dataset to generate a time-synchronized carbon emission and power generation dataset. Based on the time-synchronized carbon emission and power generation dataset, the carbon emission intensity is obtained by comparing the actual carbon emission values ​​with the real-time power generation data, generating a carbon emission intensity prediction value sequence. The carbon emission intensity prediction value sequence is then combined and encapsulated with the corresponding timestamps to generate a dynamic carbon emission intensity prediction set containing time-intensity key-value pairs.

[0081] Real-time data collection of coal consumption fluctuation range, combined with the changing trend of equipment health index, and calculation of confidence interval boundary values ​​through linear error synthesis algorithm to generate dynamic confidence interval set;

[0082] Furthermore, coal consumption data is collected in real time by weighing sensors deployed at the coal conveyor belt, and the standard deviation of coal consumption fluctuation per minute is obtained based on the real-time coal consumption data to generate a coal consumption fluctuation range dataset. Based on historical data of equipment health index, the slope of equipment health index change is calculated using a linear regression method to generate an equipment health index change trend dataset. The coal consumption fluctuation range dataset and the equipment health index change trend dataset are combined, and the confidence interval boundary values ​​are calculated using a linear error synthesis algorithm. The confidence interval boundary values ​​are then combined with the corresponding timestamps to generate a dynamic confidence interval set.

[0083] The expression for calculating the boundary values ​​of the confidence interval is:

[0084] ;

[0085] in, These are the boundary values ​​of the confidence interval (including the upper and lower bounds of the confidence interval). This is a predicted value for carbon emission intensity; It is the confidence level quantile; It is the change in the predicted value of carbon emission intensity; It is the change in real-time coal consumption; It is the change in the equipment health index; It is the sensitivity coefficient of carbon emission intensity prediction to real-time coal consumption; It is the sensitivity coefficient of carbon emission intensity prediction to equipment health index; It is the standard deviation of the fluctuation in real-time coal consumption; It is the slope of the change in the equipment health index;

[0086] The dynamic carbon emission intensity prediction set and the dynamic confidence interval set are integrated to generate a dynamic carbon emission prediction package;

[0087] Furthermore, the timestamp sequence of the dynamic carbon emission intensity prediction set is precisely matched with the timestamp sequence of the dynamic confidence interval set to generate a dynamic carbon emission intensity prediction set and a dynamic confidence interval set with perfectly aligned timestamps. The carbon emission intensity prediction values ​​in the dynamic carbon emission intensity prediction set with perfectly aligned timestamps and the confidence interval boundary values ​​in the dynamic confidence interval set with perfectly aligned timestamps are combined and bound by time points to generate a dynamic carbon emission prediction package.

[0088] S4. Compare the dynamic carbon emission prediction package with the CEMS real-time monitoring data, calculate the prediction error rate, and extract multi-dimensional error features based on the prediction error rate to generate a multi-dimensional error feature vector.

[0089] The CO2 concentration, flow rate and temperature of flue gas are collected in real time by the equipment flue gas emission monitoring equipment to generate a CEMS real-time monitoring dataset.

[0090] Furthermore, by simultaneously collecting CO2 concentration, flow rate, and temperature parameters in the flue gas using non-dispersive infrared sensors, Pitot tube flow meters, and thermocouple temperature probes deployed on the flue gas cross-section, a raw flue gas data set containing multi-parameter readings is generated. The raw flue gas data set is then processed to unify the units, generating a standardized flue gas monitoring data set. The standardized flue gas monitoring data set is then aligned with a high-precision clock source using millisecond-level timestamps and encapsulated to generate a CEMS real-time monitoring dataset.

[0091] The dynamic carbon emission intensity prediction set is matched with the CEMS real-time monitoring dataset on the time axis, and the clock deviation is compensated by interpolation to generate a spatiotemporal synchronization verification dataset.

[0092] Furthermore, the timestamp sequence of the dynamic carbon emission intensity prediction set is compared with the timestamp sequence of the CEMS real-time monitoring dataset to obtain the average offset and offset direction between the two sets of timestamps, generating timestamp deviation analysis results. Based on the timestamp deviation analysis results, the timestamp of each data point in the dynamic carbon emission intensity prediction set is shifted forward or backward, and the two adjacent data points before and after the adjusted timestamp in the dynamic carbon emission intensity prediction set are located. The weight ratio is obtained according to the relative position relationship between the adjusted timestamp and the adjacent timestamps before and after. At the same time, the values ​​of the two data points before and after are merged according to the weight ratio to obtain the compensated data value, completing the numerical compensation of the data value at the adjusted timestamp position, generating a time-compensated dynamic carbon emission intensity prediction set. The time-compensated dynamic carbon emission intensity prediction set is matched and aligned one by one with the data points of the corresponding timestamps in the CEMS real-time monitoring dataset to generate a spatiotemporal synchronization verification dataset.

[0093] Based on the spatiotemporal synchronous verification dataset, the prediction error rate is calculated point by point, and the average error rate is statistically analyzed. At the same time, high error intervals are marked according to the average error rate, and a high error persistence interval marking sequence is generated.

[0094] Furthermore, the spatiotemporal synchronization verification dataset is analyzed to extract the predicted and measured carbon emission intensity values ​​corresponding to each timestamp. The absolute difference between the predicted and measured carbon emission intensity values ​​for each timestamp is calculated. Simultaneously, the prediction error rate is obtained based on the ratio of the absolute difference to the measured carbon emission intensity value, generating a prediction error rate sequence. Based on the prediction error rate sequence, the arithmetic mean of all prediction error rates within a fixed-length time window is calculated to obtain the average error rate, generating an average error rate sequence. Based on the average error rate sequence and a preset error rate threshold, time intervals in which the average error rate continuously exceeds the preset error rate threshold are identified and marked, generating a high-error-persistence interval marker sequence containing the start time, end time, and average error rate.

[0095] The expression for calculating the prediction error rate is:

[0096] ;

[0097] in, It is the first Prediction error rate at each time point; It is the first Predicted carbon emission intensity values ​​at each point in time; It is the first Measured values ​​of carbon emission intensity at each time point; It is a timestamp index;

[0098] It should be noted that the preset error rate threshold is a percentage value set based on historical operating data statistical analysis and equipment performance requirements. It is used to dynamically determine whether the prediction error exceeds the acceptable range. An exemplary value range is 3% to 10%.

[0099] Based on the high-error persistent interval labeled sequence, three-dimensional features are extracted to generate the original error feature vector;

[0100] Furthermore, based on the high-error duration interval labeling sequence, the start and end times of each labeling interval are analyzed, and the duration is obtained based on the difference between the start and end times. The arithmetic mean of all prediction error rates within each labeling interval is extracted as the average error amplitude. Simultaneously, the standard deviation of all prediction error rates within each labeling interval is extracted as the error volatility. The duration, average error amplitude, and error volatility are integrated to generate a three-dimensional feature set. Each feature dimension in the three-dimensional feature set is subjected to Min-Max normalization and linearly transformed to the range [0,1] to generate a standardized three-dimensional feature set. The three standardized features (standardized duration, standardized average error amplitude, and standardized error volatility) corresponding to each labeling interval in the standardized three-dimensional feature set are combined into a multi-dimensional vector in a fixed order to generate the original error feature vector.

[0101] The original error feature vector is mapped to the equipment health index correlation factor adjustment coefficient and coal quality characteristic confidence weight parameter through a predefined engineering rule base, thereby generating a multidimensional error feature vector.

[0102] Furthermore, the original error feature vector is analyzed to extract three feature values: standardized duration, standardized average error amplitude, and standardized error volatility, generating a set of error feature values. Based on the set of error feature values, a matching query is performed in the mapping relationship table in the predefined engineering rule base to obtain the corresponding equipment health index correlation factor adjustment coefficient and coal quality feature confidence weight parameter, generating a set of parameter adjustment coefficients. Based on the set of parameter adjustment coefficients, the equipment health index correlation factor adjustment coefficient and coal quality feature confidence weight parameter are combined in a fixed order (e.g., equipment health index correlation factor adjustment coefficient first, coal quality feature confidence weight parameter second) to generate a multidimensional error correction feature vector.

[0103] It should be noted that the predefined engineering rule base is a set of rules generated by performing machine learning cluster analysis on historical equipment operation data and coal quality data. It consists of a mapping table of equipment health index correlation factor adjustment coefficients and a mapping table of coal quality characteristic confidence weight parameters.

[0104] S5. Based on the multidimensional error feature vector, dynamically adjust the equipment health index correlation factor and coal quality confidence weight parameters to generate a dynamic correction instruction set, and combine it with the dynamic carbon emission prediction package to generate a carbon emission prediction report.

[0105] Analyze the equipment attenuation weight component and coal quality mutation weight component in the multidimensional error feature vector to generate a set of error dominant factors;

[0106] Furthermore, the multidimensional error feature vector is analyzed to extract the values ​​of the equipment attenuation weight component and the coal quality mutation weight component, generating a set of weight components. The equipment attenuation weight component and the coal quality mutation weight component in the set of weight components are normalized to ensure that the sum of the normalized equipment attenuation weight component and the coal quality mutation weight component reaches a unit sum, generating a set of normalized weight components. Based on the set of normalized weight components, the equipment attenuation weight component and the coal quality mutation weight component are encapsulated in a fixed key-value pair format to generate a set of error dominant factors.

[0107] It should be noted that the unit sum refers to the sum of all weighted components after normalization being set to a standard benchmark value, determined by a preset proportional relationship in the weight allocation rule base, to ensure that the relative contribution of each component weight remains balanced.

[0108] Based on the error-dominant factor set, a predefined parameter adjustment mapping table is called to dynamically adjust the equipment health index correlation factors and coal quality confidence weight parameters, generating a dynamic correction instruction set.

[0109] Furthermore, the error-dominant factor set is analyzed, and the values ​​of equipment attenuation weight components and coal quality mutation weight components are extracted to generate weight value pairs. Based on the weight value pairs, a predefined parameter adjustment mapping table is queried to obtain the corresponding equipment health index correlation factor adjustment coefficient and coal quality confidence weight parameter adjustment coefficient, generating adjustment coefficient pairs. Based on the adjustment coefficient pairs, the equipment health index correlation factor adjustment coefficient is multiplied with the current equipment health index correlation factor to obtain the optimized equipment health index correlation factor. At the same time, the coal quality confidence weight parameter adjustment coefficient is multiplied with the current coal quality confidence weight parameter to obtain the optimized coal quality confidence weight parameter. The optimized equipment health index correlation factor and the optimized coal quality confidence weight parameter are combined into key-value pairs and encapsulated to generate a dynamic correction instruction set.

[0110] It should be noted that the predefined parameter adjustment mapping table is a correspondence table established by analyzing the correlation between the equipment attenuation weight component, the coal quality mutation weight component and the optimal parameter adjustment coefficient in historical data. It includes the equipment health index correlation factor adjustment coefficient and the coal quality confidence weight parameter adjustment coefficient corresponding to different weight component intervals.

[0111] The dynamic carbon emission prediction package is corrected in real time using a dynamic correction instruction set to generate a carbon emission prediction report.

[0112] Furthermore, the dynamic correction instruction set is analyzed to extract the correlation factors of the optimized equipment health index and the confidence weight parameters of the optimized coal quality. The correlation factors of the optimized equipment health index are used as proportional coefficients and multiplied point by point with the predicted carbon emission intensity values ​​in the dynamic carbon emission prediction package to achieve proportional scaling correction of the predicted carbon emission intensity values, generating a sequence of corrected carbon emission intensity prediction values. At the same time, the confidence weight parameters of the optimized coal quality are used as confidence coefficients and multiplied point by point with the upper and lower boundary values ​​of the confidence interval in the dynamic carbon emission prediction package to achieve synchronous adjustment of the confidence interval, generating a confidence interval correction sequence. The carbon emission intensity prediction correction sequence and the confidence interval correction sequence are combined and packaged to generate a dynamic carbon emission prediction correction package. The dynamic carbon emission prediction correction package is integrated with the trend data of the equipment health index and the real-time coal quality fusion operation parameter set to generate a carbon emission prediction report.

[0113] This embodiment also provides a computer device applicable to a carbon emission prediction method for coal-fired power units, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the carbon emission prediction method for coal-fired power units as proposed in the above embodiment.

[0114] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0115] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the carbon prediction method for carbon emissions from coal-fired power units as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0116] In summary, this invention achieves dynamic noise reduction, enhancement, and feature quantization of multispectral coal flow images by acquiring raw coal flow image data, performing preprocessing, and real-time feature extraction. This directly generates high-fidelity coal quality feature vectors, solving the problem of coal quality data lag and improving the response capability to sudden changes in coal quality. Furthermore, by dynamically adjusting the equipment health index correlation factor and coal quality confidence weight parameters based on multidimensional error feature vectors, a rule-based mapping from error source features to parameter correction instructions is achieved, enhancing long-term prediction stability.

[0117] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting carbon emissions from coal-fired power units, characterized in that: include, Collect raw coal flow image data and perform preprocessing and real-time feature extraction to obtain a real-time coal quality dataset; Collect equipment operating parameters and perform data cleaning by combining them with real-time coal quality datasets to generate a real-time coal quality fusion operating parameter set. At the same time, combine the equipment health assessment rule base to obtain the equipment health index. Based on the real-time coal quality fusion operation parameter set and equipment health index, the predicted value of carbon emission intensity and the boundary value of confidence interval are calculated to generate a dynamic carbon emission prediction package. The dynamic carbon emission prediction package is compared with the CEMS real-time monitoring data to calculate the prediction error rate. Based on the prediction error rate, multi-dimensional error features are extracted to generate a multi-dimensional error feature vector. Based on multidimensional error feature vectors, the correlation factors of equipment health index and coal quality confidence weight parameters are dynamically adjusted to generate a dynamic correction instruction set. Combined with a dynamic carbon emission prediction package, a carbon emission prediction report is generated.

2. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The steps for obtaining real-time coal quality data are as follows: The raw coal flow image data refers to multispectral imaging data; The preprocessing includes noise reduction, enhancement, and standardization. Based on the preprocessed raw coal flow image data, color distribution features, texture roughness features, and particle size distribution features are extracted and integrated to generate a coal quality multimodal feature vector; Based on the multimodal feature vector of coal quality, linear interpolation calculation is performed through a preset feature value-coal composition mapping table to obtain a real-time coal quality dataset.

3. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The process of generating a real-time coal quality fusion operation parameter set refers to performing timestamp alignment, missing value filling, and outlier removal on the real-time coal quality dataset and equipment operation parameters to generate the real-time coal quality fusion operation parameter set.

4. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The steps for obtaining the device health index are as follows: Based on the real-time coal quality fusion operation parameter set, the equipment operation parameters are extracted, and the equipment operation parameters are compared with the corresponding equipment operation parameter benchmark values ​​to calculate the normalized offset and generate a parameter normalized offset set. Based on the parameter normalization offset set, a predefined equipment health assessment rule library is called, and a weighted fusion algorithm is used to calculate the comprehensive equipment performance score. The comprehensive equipment performance score is then mapped to generate an equipment health index.

5. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The steps for generating the dynamic carbon emission prediction package are as follows: Based on the real-time coal quality fusion operation parameter set, the theoretical carbon emission baseline value is calculated using the carbon conservation formula; Based on the equipment health index, query the preset index-correction factor mapping rule to obtain the carbon emission correction factor, and dynamically correct the theoretical carbon emission baseline value based on the carbon emission correction factor to obtain the actual carbon emission value. By coupling actual carbon emissions with real-time power generation data, the predicted carbon emission intensity is calculated, generating a dynamic carbon emission intensity prediction set. Real-time data collection of coal consumption fluctuation range, combined with the changing trend of equipment health index, and calculation of confidence interval boundary values ​​through linear error synthesis algorithm to generate dynamic confidence interval set; The dynamic carbon emission intensity prediction set and the dynamic confidence interval set are integrated to generate a dynamic carbon emission prediction package.

6. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The steps for generating the multidimensional error feature vector are as follows: The CO2 concentration, flow rate and temperature of flue gas are collected in real time by the equipment flue gas emission monitoring equipment to generate a CEMS real-time monitoring dataset. The dynamic carbon emission intensity prediction set is matched with the CEMS real-time monitoring dataset on the time axis, and the clock deviation is compensated by interpolation to generate a spatiotemporal synchronization verification dataset. Based on the spatiotemporal synchronous verification dataset, the prediction error rate is calculated point by point, and the average error rate is statistically analyzed. At the same time, high error intervals are marked according to the average error rate, and a high error persistence interval marking sequence is generated. Based on the high-error persistent interval labeled sequence, three-dimensional features are extracted to generate the original error feature vector; The original error feature vector is mapped to the equipment health index correlation factor adjustment coefficient and coal quality characteristic confidence weight parameter through a predefined engineering rule base, thereby generating a multidimensional error feature vector.

7. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The steps for generating the dynamic correction instruction set are as follows: Analyze the equipment attenuation weight component and coal quality mutation weight component in the multidimensional error feature vector to generate a set of error dominant factors; Based on the error-dominant factor set, a predefined parameter adjustment mapping table is called to dynamically adjust the equipment health index correlation factors and coal quality confidence weight parameters, generating a dynamic correction instruction set.

8. The carbon emission prediction method for coal-fired power units as described in claim 1, characterized in that: The carbon emission prediction report is generated by using a dynamic correction instruction set to correct the dynamic carbon emission prediction package in real time.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the carbon emission prediction method for coal-fired power units as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the carbon emission prediction method for coal-fired power units as described in any one of claims 1 to 8.

Images