A method, device and medium for accounting for carbon dioxide emissions based on a thermal power plant

Abstract

The application discloses a kind of based on carbon dioxide emission accounting method, equipment and medium of thermal power plant, it is related to intelligent emission accounting technical field, including, acquisition operating parameter data and coal quality test data, obtain observation data and fuel prior data;With observation data and fuel prior data assimilation calculation is carried out, obtains emission main value and uncertainty and constructs prediction simulation model, carries out discharge amount calculation to action set, obtains feedforward suggestion action and prediction emission curve;Through the residual set formed by emission main value and observation data and calculate the upper bound of uncertainty, compare prediction emission curve with the upper bound of uncertainty, to feedforward suggestion action is adjusted, obtains executable action and emission upper bound;Actual emission curve is formed with emission main value, compared with prediction emission curve, obtain prediction deviation.The application is analyzed by constructing prediction simulation model to discharge amount result, realizes that future emission trend can be predicted before action implementation.

Description

Technical Field

[0001] This invention relates to the field of intelligent emission accounting technology, and in particular to a method, equipment and medium for carbon dioxide emission accounting based on thermal power plants. Background Technology

[0002] With the adjustment of energy structure and the introduction of carbon emission reduction targets, the monitoring and accounting of carbon dioxide emissions from thermal power plants has gradually become a research focus in the power industry and environmental protection field. Existing accounting methods mainly rely on fuel consumption and coal quality test data to calculate emission factors, or obtain emission concentration and flow data through online flue gas monitoring. Some studies have begun to explore methods based on multi-source data fusion and model assimilation to improve the accuracy and real-time performance of emission accounting.

[0003] Existing emission accounting methods based on fuel statistics and laboratory results are insufficient to reflect the dynamic fluctuations in unit operating conditions, resulting in minute-level accuracy. On the other hand, relying solely on flue gas monitoring is easily affected by measurement errors and equipment calibration deviations, lacking robustness. Therefore, existing methods are inadequate in both accuracy and real-time performance, failing to achieve feedforward assessment and compliance gating control of predicted emission trends, and thus failing to meet the requirements of low-carbon operation and refined supervision. Summary of the Invention

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

[0005] Therefore, this invention provides a carbon dioxide emission accounting method based on thermal power plants to solve the problems of insufficient minute-level accounting accuracy and lack of feedforward optimization and gating control capabilities.

[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 calculating carbon dioxide emissions from thermal power plants, which includes collecting operating parameter data and coal quality test data, and obtaining observation data and prior fuel data.

[0008] The principal values ​​and uncertainties of emissions are obtained by assimilating and calculating the data from observation and prior fuel data.

[0009] A predictive simulation model is constructed by using a pre-defined set of executable actions, emission principal values, and uncertainties. Emissions are calculated for the pre-defined set of executable actions to obtain feedforward suggested actions and predicted emission curves.

[0010] The residual set is obtained based on the main emission values ​​and observation data. The upper bound of uncertainty is calculated through the residual set. The predicted emission curve is compared with the upper bound of uncertainty. The feedforward suggested actions are gating and adjusted to obtain the executable actions and the upper bound of emissions.

[0011] The actual emission curve is obtained by using the principal emission values. The actual emission curve is compared with the predicted emission curve to obtain the prediction deviation. The prediction deviation, emission upper limit and principal emission values ​​are then archived.

[0012] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the specific steps for collecting operating parameter data and coal quality test data, and obtaining observation data and prior fuel data are as follows.

[0013] Collect operational parameter data, perform time-consistent processing and caliber verification, and obtain the processed operational parameter data;

[0014] Collect coal quality test data, match the processed operating parameter data with the coal quality test data, and obtain observation data and fuel prior data.

[0015] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the steps for obtaining the principal emission values ​​and uncertainties through assimilation calculations using observed data and prior fuel data are as follows:

[0016] The observed data and fuel prior data are time-aligned, and combined with the assimilated state vector after process noise correction as the prior state. The assimilation operation is then performed in conjunction with carbon conservation constraints to obtain the assimilated emission parameters.

[0017] Uncertainty propagation analysis was performed on the assimilated emission parameters to obtain the principal values ​​and uncertainties of emissions.

[0018] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the specific steps for constructing a predictive simulation model using a preset set of executable actions, emission principal values, and uncertainties are as follows:

[0019] Sensitivity data is calculated by using a pre-defined set of executable actions, emission principal values, and uncertainties. The sensitivity data is then accumulated through windowing to obtain the cumulative response, which is used to build a predictive simulation model.

[0020] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the specific steps for calculating emissions from a preset set of executable actions to obtain feedforward suggested actions and predicted emission curves are as follows:

[0021] Emissions are calculated one by one using a predictive simulation model and a pre-set set of executable actions to obtain an emission data set.

[0022] By comparing and analyzing emission data sets, the optimal feedforward action for emission control can be obtained.

[0023] The predicted emission curve of the feedforward suggestion action is obtained by jointly calculating the predicted simulation model and the feedforward suggestion action with the optimal emission level.

[0024] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the steps for obtaining the residual set based on the principal emission values ​​and observation data, and calculating the upper bound of uncertainty using the residual set, are as follows:

[0025] The difference between the main emission values ​​and the observed data series is calculated to obtain the residual set. The residual set is then processed by a sliding time window to obtain the residual set after the sliding time window processing.

[0026] The residual set after being processed by a sliding time window is subjected to quantile calculation to obtain the quantiles. The quantiles are then superimposed with the principal emission values ​​to obtain the upper bound of the uncertainty.

[0027] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the steps of comparing the predicted emission curve with the upper bound of uncertainty, gating the feedforward suggested actions, and obtaining the executable actions and emission upper bounds are as follows:

[0028] The predicted emission curve is compared minute by minute with the upper bound of uncertainty to obtain comparative data.

[0029] By jointly analyzing the comparative data and the feedforward suggested actions, gating judgment data is obtained and adjusted to obtain the executable actions and emission upper limits.

[0030] As a preferred embodiment of the carbon dioxide emission accounting method based on thermal power plants described in this invention, the steps of obtaining the actual emission curve through the principal emission values, comparing the actual emission curve with the predicted emission curve to obtain the prediction deviation, and archiving the upper limit of emissions and the principal emission values ​​are as follows:

[0031] The main emission values ​​are continuously collected within a time window to obtain the actual emission curve. The actual emission curve is then compared with the predicted emission curve minute by minute to obtain the prediction deviation.

[0032] By correcting the sensitivity data for the next cycle through prediction bias, the assimilation calculation parameters under carbon conservation constraints are optimized, and the upper limit of emissions, principal values ​​of emissions, and prediction bias are stored accordingly to obtain archived records of prediction bias, upper limit of emissions, and principal values ​​of emissions.

[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 dioxide emission accounting method based on a thermal power plant 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 dioxide emission accounting method based on a thermal power plant as described in the first aspect of the present invention.

[0035] The beneficial effects of this invention are as follows: By aligning the observed data with the prior fuel data over time and combining it with carbon conservation constraints for assimilation calculations, the accuracy and consistency of the principal emission accounting results in a statistical sense are ensured, thereby improving the credibility of the total emissions; by constructing a predictive simulation model and combining it with a pre-set set of executable actions for emission result analysis, it is possible to predict future emission trends before the actions are implemented, thus providing feedforward decision support for unit operation optimization; by comparing the predicted emission curve with the upper bound of uncertainty minute by minute and triggering gating adjustments, it is ensured that the final output executable actions will not lead to excessive emissions, thus ensuring the compliance and traceability of the externally reported results. 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 carbon dioxide emission accounting method based on thermal power plants.

[0038] Figure 2 This is a flowchart for assimilating observational data with prior fuel data.

[0039] Figure 3 This is a flowchart for predicting simulation model derivations.

[0040] Figure 4 This is a flowchart for gating adjustment and evaluation. 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 This is one embodiment of the present invention, which provides a method for calculating carbon dioxide emissions from thermal power plants, including the following steps:

[0045] S1: Collect operational parameter data and coal quality test data, and obtain observation data and fuel prior data.

[0046] Specifically, the process involves collecting operational parameter data and coal quality test data, obtaining observational data and prior fuel data. The specific steps are as follows:

[0047] Collect operational parameter data, perform time-consistent processing and caliber verification, and obtain the processed operational parameter data;

[0048] Collect coal quality test data, match the processed operating parameter data with the coal quality test data, and obtain observation data and fuel prior data.

[0049] Furthermore, data on coal feed rate, unit output, and flue gas monitoring of thermal power units are collected at minute sampling intervals. The flue gas monitoring data includes carbon dioxide volume fraction, oxygen content, and flue gas volume flow rate. Operating parameter data is obtained, and the operating parameter data is aligned with a unified time reference and verified to obtain time-unified operating parameter data.

[0050] It should be noted that carbon content parameters obtained from daily coal sample testing are collected to acquire coal quality analysis data. Based on the carbon dioxide volume fraction and flue gas volume flow rate from flue gas monitoring data, the carbon dioxide volume flow rate under standard conditions is calculated. Then, the carbon dioxide molar mass is converted to the standard state molar volume to obtain the carbon dioxide mass emission rate, which is used as observation data. The theoretical emission rate is calculated by comparing the coal feed rate in the operating parameter data with the carbon content parameters in the coal quality analysis data. The conversion relationship follows the molar mass conversion of carbon and carbon dioxide and the principle of energy conservation. Specifically, the coal feed rate in the operating parameter data is first aligned to minute timestamps and unified to tons per hour. The carbon content parameters in the coal quality analysis data... The data is standardized as kilograms of carbon per kilogram of coal and is used as a constant within the same natural day, paired with each minute timestamp. At each minute timestamp, the product of the coal feed rate and the carbon content parameter is used as the carbon mass flow rate, which is converted to tons of carbon per hour based on the formula 1,000 kilograms equals 1 ton. Then, based on the fixed ratio between the molecular weight of carbon dioxide and the atomic weight of carbon, the carbon mass flow rate per ton of carbon per hour is converted to the carbon dioxide mass flow rate per ton of carbon per hour. When the coal feed rate or coal quality test data in the operating parameter data fails to collect valid values ​​at certain timestamps, it is determined to be a missing case. If the missing time is no more than two minutes, linear interpolation is used to fill the gap. If the missing time is more than two minutes, the previous valid value is used to fill the gap, thus obtaining the fuel prior data.

[0051] S2: By combining observational data and prior fuel data with carbon conservation constraints and prior states, assimilation calculations are performed to obtain the principal values ​​and uncertainties of emissions.

[0052] The observation data and prior fuel data are concatenated at the same timestamp to form a two-dimensional observation vector. To establish a carbon conservation constraint, the carbon input on the minute-level side is kept equal to the carbon content in the carbon dioxide emissions on the minute-level side. The carbon input on the minute-level side is obtained by reading the coal feed rate from the operating parameter data according to the minute timestamp, with the unit of coal feed rate being uniformly ton per hour. Then, the carbon content parameter from the coal quality test data is called according to the minute timestamp, with the unit of carbon content parameter being kilograms of carbon per kilogram of coal. At the same timestamp, the product of the coal feed rate and the carbon content parameter is taken as the total carbon input for that minute. The total carbon input is converted to tons of carbon per hour according to the conversion rule of 1,000 kilograms equaling 1 ton, thus obtaining the carbon input on the minute-level side. The product of unit output, combustion efficiency, effective carbon content, and coal consumption per unit of electricity is taken as the carbon content in the carbon dioxide emissions on the minute-level side. At each minute timestamp, the two-dimensional observation vector is assimilated with the prior state vector, and a carbon conservation constraint is applied during the assimilation process. The expression is:

[0053] ;

[0054] in, Represents the state vector. Represents a two-dimensional observation vector. This represents the emission mechanism function obtained by mapping state vectors. Represents the observation operator matrix, Represents the observation error covariance matrix. Represents the prior state vector. Denotes the prior covariance matrix. This represents the carbon conservation constraint weight parameter, with a value range of [10, 100]. This represents the residual due to the carbon conservation constraint. Represented by state vector The cost function is the input.

[0055] It should be noted that, The parameters, including effective carbon content, combustion efficiency, and coal consumption per unit of electricity, are initially set during the initialization phase using coal quality test data, unit design parameters, and historical operating data. During the operation phase, they are updated in real time through assimilation operations using the prior state vector from the previous minute combined with observed data and prior fuel data. The emission mechanism function is obtained by multiplying the effective carbon content, combustion efficiency, and coal consumption per unit of electricity in the state vector with the unit output in the operating parameter data. By analyzing the dimensional relationship between the observed data and the output of the mechanism function, the emission mechanism function output is copied twice to correspond to the observed data and the prior fuel data. It is obtained by performing variance statistics on the difference between observed data within the historical window and prior fuel data. During the initialization phase, the state vector is set based on the unit design parameters, coal quality test data, and historical operating data. During the operation phase, the state vector obtained from the assimilation of the previous minute is corrected for process noise and used as the new prior state vector. The estimation is based on the fluctuation range of historical operating data during the initialization phase, and updated during the operating phase using the posterior covariance obtained from the assimilation of the previous minute as a base, plus a small amount of process noise. The carbon content is obtained by calculating the input-side carbon content from operating parameter data and coal quality test data at the same minute timestamp, and the emission-side carbon content calculated from unit output, coal consumption per unit of electricity, effective carbon content, and combustion efficiency. It was determined by analyzing the differences between input-side and emission-side carbon content in historical data and combining assimilation experiments and stability tests. When the carbon balance is less than 10, the carbon conservation constraint is insufficient, and the difference between the carbon input and emission sides is too large. When the value is greater than 100, the carbon conservation constraint is too strong, resulting in insufficient response of the assimilation results to the observation data.

[0056] Using the prior state vector as the expansion point, a first-order Taylor expansion of the emission mechanism function is performed to obtain the linear approximation of the emission mechanism function at the expansion point. Weighted least squares is used to minimize the assimilation objective function and obtain the assimilated emission parameters.

[0057] Uncertainty propagation analysis was performed on the assimilated emission parameters to obtain the principal values ​​and uncertainties of emissions.

[0058] Furthermore, the assimilated emission parameters are output with minute timestamps and quality markers are recorded. The effective carbon content, combustion efficiency, and coal consumption per unit of electricity in the assimilated emission parameters are read, and the unit output is read at the same minute timestamp. The product of effective carbon content, coal consumption per unit of electricity, unit output, and combustion efficiency is obtained, and the molecular weight of carbon and carbon dioxide is converted to obtain the minute-level emission principal value. The minute-level sensitivity vector and the posterior covariance generated by the assimilation operation are subjected to first-order error propagation to obtain the variance of the minute-level emission principal value, and the square root is taken to obtain the uncertainty.

[0059] It should be noted that the method for obtaining the first-order sensitivity coefficients of minute-level emission principal values ​​for effective carbon content, combustion efficiency, and coal consumption per unit of electricity is as follows: read the effective carbon content, combustion efficiency, and coal consumption per unit of electricity from the assimilated emission parameters, and simultaneously read the unit output. Then, fix two of the parameters respectively, and only make a small perturbation to the other parameter. Calculate the difference between the emission principal values ​​before and after the perturbation. The ratio of the difference in emission principal values ​​to the amplitude of the parameter perturbation is used as the first-order sensitivity coefficient corresponding to the parameter. Combine the first-order sensitivity coefficients of effective carbon content, combustion efficiency, and coal consumption per unit of electricity to obtain the minute-level sensitivity vector.

[0060] S3: Construct a predictive simulation model using a preset set of executable actions, emission principal values, and uncertainties. Calculate emissions from the preset set of executable actions to obtain feedforward suggested actions and their predicted emission curves.

[0061] Sensitivity data is calculated by using a pre-defined set of executable actions, emission principal values, and uncertainties. The sensitivity data is then accumulated through windowing to obtain the cumulative response, which is used to build a predictive simulation model.

[0062] Furthermore, a preset set of executable actions is determined at the beginning. This set consists of discrete combinations of output fine-tuning and excess air coefficient fine-tuning (e.g., output fine-tuning values ​​are -2%, -1%, 0, +1%, +2%, and excess air coefficient fine-tuning values ​​are -0.02, -0.01, 0, +0.01, +0.02). At each minute timestamp, the principal emission value and uncertainty are read. The response strength to output fine-tuning and excess air coefficient fine-tuning is calculated using a univariate perturbation method while keeping other quantities constant. Specifically, the calculation caliber of the principal emission value is fixed, and only the output is subjected to very small positive and negative perturbations. The ratio of the difference between the two principal emission values ​​to the perturbation amplitude is used to obtain the sensitivity data of the unit output. The same method is used to obtain the sensitivity data of the excess air coefficient. In scenarios where output and excess air coefficient are coupled, cross-sensitivity data is obtained by simultaneously applying paired perturbations and removing univariate terms, thus obtaining a minute-level sensitivity data sequence.

[0063] Within a fixed evaluation window with a sampling granularity of minutes (e.g., a continuous 15 minutes), an uncertainty-weighted window accumulation calculation is performed on the minute-level sensitivity data sequence to obtain the cumulative response of any candidate action within that window, expressed as:

[0064] ;

[0065] ;

[0066] in, This refers to actions in a predefined set of executable actions. Cumulative response results within a window This indicates the number of minutes displayed in the window. Represents a timestamp. Indicates the timestamp Uncertainty weighting factor Indicates the timestamp Sensitivity data of primary emission values ​​to unit output. Indicates action Fine-tuning of unit output in the middle, Indicates the timestamp Sensitivity data of the main emission values ​​to the excess air coefficient. Indicates action The amount of excess air coefficient fine-tuning in the middle. timestamp The cross-sensitivity data of the main emission values ​​to power output and excess air coefficient. Indicates the timestamp Uncertainty of the principal values ​​of upper emissions Indicates the scale parameter.

[0067] It should be noted that, It obtains a representative scale parameter by statistically analyzing the uncertainty series within the current shift or historical window. .

[0068] Within a fixed time window, the principal emission values ​​for each minute within the window are accumulated minute by minute to obtain the total emission within the window. This total emission is used as the window baseline. Based on the window baseline and the accumulated response results, a predictive simulation model for a preset set of executable actions is constructed, expressed as:

[0069] ;

[0070] ;

[0071] in, This refers to actions in a predefined set of executable actions. Window-level counterfactual emissions estimates This refers to actions in a predefined set of executable actions. The minute-by-minute counterfactual trajectory, Indicates the timestamp The minute-level main emission values, Indicates the window baseline.

[0072] Actions in each set of executable actions The calculation results are output in a structured record format, yielding an emissions dataset where each record contains actions. Window-level counterfactual emissions estimates ,action minute-by-minute counterfactual trajectory With quality markings.

[0073] By comparing and analyzing the emission data sets, the optimal feedforward action for emission control can be obtained.

[0074] Furthermore, by using the emissions dataset and the corresponding window baseline, records lacking any field or with invalid quality markers are removed to obtain window-level counterfactual emissions estimates. Using the sole primary ranking criterion, the emission data set is sorted in ascending order to obtain a candidate sequence, and a resolution threshold is set. This is used to handle cases where values ​​are close to each other. Take 1% of the window baseline, when two adjacent candidate window-level counterfactual emission estimates The difference does not exceed the resolution threshold. Parallel elimination is performed using parallel elimination rule one. If parallel elimination rule one still cannot distinguish them, parallel elimination rule two is used for parallel elimination. Candidate sequences are judged from top to bottom and the first candidate that meets the conditions is selected to obtain the optimal feedforward suggestion action.

[0075] It should be noted that the digestion threshold The 1% threshold of the window baseline is chosen because the window baseline represents the cumulative emissions before any action is taken. Its value is typically in the range of tens to hundreds of tons of CO2, making it a relatively stable reference. If the difference is less than 1%, it means the emission reduction effects of the two actions are almost indistinguishable within the statistical fluctuation and uncertainty range. If the difference is greater than 1%, it means the difference in action effects has exceeded random fluctuations, and the choice can be made directly based on magnitude. Rule one for mitigation involves selecting the action with the smaller value using the action cost index. The action cost index is a linear weighted average of the relative change in output and the change in excess air coefficient (for example, the change in output is normalized to the unit's rated output, and the change in excess air coefficient is normalized in increments of 0.01, with a weighting ratio of 1:1). Rule two applies to actions with parallel candidate values. minute-by-minute counterfactual trajectory With timestamp Minute-level main emission values Comparison and statistics are performed within the window to make actions. minute-by-minute counterfactual trajectory ≤ at timestamp Minute-level main emission values Choose the option with the higher percentage of minutes.

[0076] The predicted emission curve of the feedforward suggestion action is obtained by jointly calculating the predicted simulation model and the feedforward suggestion action with the optimal emission level.

[0077] Furthermore, by using minute-level sensitive data sequences, the unit output fine-tuning and excess air coefficient fine-tuning are mapped to minute-level impacts on the main emission values. Specifically, the product of the unit output sensitivity data and the unit output fine-tuning is taken as the minute-level impact of the unit output; the product of the excess air coefficient sensitivity data and the excess air coefficient fine-tuning is taken as the minute-level impact of the excess air coefficient; the product of the cross-sensitivity data and the unit output fine-tuning is taken as the intermediate value; the product of the intermediate value and the excess air coefficient fine-tuning is taken as the minute-level impact resulting from their coupling; and the sum of the minute-level impacts of the unit output and the excess air coefficient is taken as the minute-level effect.

[0078] The minute effect size is scaled using an uncertainty weighting factor at the timestamp. When the uncertainty of the principal emission value at the minute level is greater than or equal to the scaling parameter, it is defined as a minute with greater uncertainty. Minutes with greater uncertainty will have a smaller scaling factor, thereby reducing the interference of the uncertainty peak on the predicted emission curve. The sum of the principal emission value and the minute effect quantity after uncertainty weighting is used as the predicted emission value for the minute. All the predicted emission values ​​for the minutes are connected in chronological order to form the predicted emission curve. When the predicted emission value is a non-physical negative value, the predicted emission value is set to zero and a limit quality mark is recorded. When the sensitivity data or uncertainty weighting factor is missing in individual minutes and the consecutive missing time does not exceed two minutes, linear interpolation is performed. When the consecutive missing time exceeds two minutes, the previous valid value is used and a SUB quality mark is added.

[0079] S4: By obtaining the residual set from the principal emission values ​​and observation data, the upper bound of uncertainty is calculated using the residual set. The predicted emission curve is compared with the upper bound of uncertainty, and the feedforward suggested actions are gating and adjusted to obtain the executable actions and the upper bound of emissions.

[0080] The difference between the main emission values ​​and the corresponding sequences of the observation data is calculated to obtain the residual set. The residual set is then processed by a sliding time window to obtain the residual set after the sliding time window processing.

[0081] Furthermore, a pair of records of emission principal values ​​and observed data are formed at each minute timestamp, denoted as the corresponding sequence of emission principal values ​​and observed data. The deviation between the two is obtained in the form of absolute difference, resulting in a set of residuals arranged in chronological order. Minutes with SUB quality labels are synchronously labeled, and a fixed-length sliding time window (e.g., 180 minutes) is determined. The residual set within the most recent 180 minutes is sliced ​​into a subset to obtain a residual set after being processed by the sliding time window, while retaining the quality label corresponding to each minute.

[0082] The residual set after being processed by a sliding time window is subjected to quantile calculation to obtain the quantiles. The quantiles are then superimposed with the principal emission values ​​to obtain the upper bound of the uncertainty.

[0083] Furthermore, in minute timestamps The residual set, after being processed by a sliding time window, is read. Weights are assigned to each minute sample in the residual set: samples labeled "normal" have a weight of 1, and samples labeled "SUB" have a weight of 0.5. The confidence level is calculated using a weighted empirical distribution. upper quantile (e.g.) =0.10 corresponds to the 90th percentile), by finding the smallest quantile. The quantiles are obtained by satisfying the following formula, expressed as:

[0084] ;

[0085] ;

[0086] in, The index representing the minute timestamp, starting from the current timestamp. Backtracking minute, Indicates the length of the sliding time window. Index representing the minute timestamp The residual values ​​between the principal emission values ​​and the observed data, Indicates the current timestamp Above, based on quantile probability The calculated quantiles, This is an indicator function that takes the value 1 when the residual value is less than or equal to the quantile, and 0 otherwise. Indicates the minute timestamp The weighting coefficients are set to 1 for normal samples and 0.5 for samples labeled SUB. This represents the quantile probability, with a value range of [0,1].

[0087] It should be noted that the weight of samples labeled as "normal" is 1 because they have the highest credibility, while the weight of samples labeled as "SUB" is 0.5, which is based on the empirical result of halving the information content of low-credibility samples after analyzing historical data.

[0088] At the same timestamp The upper bound of uncertainty is obtained by directly superimposing the quantiles with the principal emission values.

[0089] The predicted emission curves are compared minute by minute with the upper bound of uncertainty to obtain comparative data.

[0090] At a one-minute time granularity, the predicted emission curve and the upper bound of uncertainty are aligned one by one according to the timestamp, confirming that the units are consistent (e.g., carbon dioxide / hour). The difference between the predicted emission curve and the upper bound of uncertainty is calculated and a hard judgment label is given. When the difference between the predicted emission curve and the upper bound of uncertainty is greater than zero, it is recorded as exceeding the limit; when the difference is less than or equal to zero, it is recorded as not exceeding the limit. At the same time, the exceedance magnitude (the value of the difference when it is positive) and the safety margin (the absolute value of the difference between the upper bound of uncertainty and the predicted emission value when the difference is negative) are recorded. To avoid jitter caused by measurement noise, a soft-gated exceedance index is calculated at each minute timestamp to quantify the intensity of approaching or exceeding the upper bound of uncertainty. The expression is:

[0091] ;

[0092] in, Indicates the minute timestamp The soft gating over-limit index, Indicates the timestamp The predicted emission curve values ​​on the curve, Indicates the timestamp The upper bound of the uncertainty on, Indicates dead zone margin, This represents the scaling factor.

[0093] It should be noted that, The dead zone margin is calculated by multiplying the median of minute-level uncertainties within a statistical historical window by an empirical proportionality coefficient. This ensures that the dead zone margin matches the uncertainty level of the generator set itself and avoids false exceedances. The empirical proportionality coefficient is determined by comparing false exceedances with actual exceedances under different historical data values, balancing the reduction of noise interference with maintaining detection sensitivity. It is obtained by taking the interquartile range of the difference sequence between the statistically predicted emission curve and the upper bound of uncertainty, and then taking a certain proportion (0.25–0.5 times). Combined with historical data correction, it ensures that the soft-gated over-limit index can both smooth noise and maintain sensitivity. The 0.25 is taken as 0.25–0.5 times the interquartile range because this range can achieve the best balance between noise suppression and true over-limit sensitivity. Less than 0.25 times would be too sensitive, and more than 0.5 times would be too insensitive.

[0094] when When ≥0.5, it is consistent with the hard judgment of exceeding the limit. A threshold of 0.4–0.6 is used to indicate a critical state, 0.5 is a threshold naturally determined by the symmetry of the Sigmoid function, and 0.4–0.6 is defined by the input falling within ±0.25. The noise-sensitive interval derived within the range is confirmed as a critical interval through historical data verification. Based on hard decision markers and soft-gated over-limit indices, continuous over-limit segments are identified, and the start timestamp, end timestamp, maximum over-limit amplitude within the segment, and average soft-gated over-limit index of the segment are output for each segment. At the same time, the over-limit percentage (number of over-limit minutes / total number of minutes) and the maximum continuous over-limit duration (minutes) within the window are calculated, and minute-by-minute comparison data are output in chronological order, including timestamp, difference, hard decision marker, soft-gated over-limit index, over-limit amplitude / safety margin, and quality marker.

[0095] By jointly analyzing the comparative data and the feedforward suggested actions, gating judgment data is obtained and adjusted to obtain the executable actions and emission upper limits.

[0096] Furthermore, the comparison data is checked for instances where the predicted emission curve exceeds the upper uncertainty limit. If all minutes do not exceed the limit and the average soft-gating over-limit index is below 0.5, the feedforward suggestion action is directly deemed acceptable and requires no adjustment. Otherwise, adjustments are made. Specifically, when the predicted emission curve is detected to exceed the upper uncertainty limit for some minutes, the unit output fine-tuning and excess air coefficient fine-tuning of the feedforward suggestion action are reduced. The reduction process gradually decreases the fine-tuning amount proportionally from the original amount. After each reduction, the predicted emission curve is recalculated and compared with the upper uncertainty limit. When the predicted emission curve does not exceed the upper uncertainty limit for all minutes... When the uncertainty upper bound is exceeded, the fine-tuning amount under the reduction range is determined to be an executable action. If the requirement cannot be met even after reduction to zero, the action is determined to be unexecutable. During the reduction process, if the fine-tuning amount exceeds the allowable physical range or operating limit of the unit, such as the output change exceeding the rated value or the air coefficient change exceeding the safe range, truncation is performed to ensure that the final action is within the allowable range. Gating judgment data (pass or pass after adjustment) and the corresponding executable actions are obtained, including the adjusted unit output fine-tuning amount and the excessive air coefficient fine-tuning amount. The adjusted predicted emission curve and the uncertainty upper bound are output together as the emission upper bound.

[0097] S5: Obtain the actual emission curve through the principal emission value, compare the actual emission curve with the predicted emission curve, obtain the prediction deviation, and archive the upper limit of emissions and the principal emission value.

[0098] The main emission values ​​are continuously collected within a time window to obtain the actual emission curve. The actual emission curve is then compared with the predicted emission curve minute by minute to obtain the prediction deviation.

[0099] Furthermore, within a fixed time window (e.g., 60 minutes), emission master values ​​are continuously collected at one-minute sampling intervals. The minute-level emission master values ​​arranged in chronological order form the actual emission curve. The predicted emission curve within the same time window is read, and the values ​​of the actual emission curve are directly compared with the values ​​of the predicted emission curve. The difference between the two is calculated and the result is marked. A positive difference indicates that the predicted value is too low, a negative difference indicates that the predicted value is too high, and a zero difference indicates that the prediction is consistent with the actual value. The difference results obtained from the minute-by-minute comparison within the time window are arranged in chronological order to form the prediction bias, and a quality label is attached to the result of each minute.

[0100] By correcting the sensitivity data for the next period through prediction bias, the assimilation calculation parameters under the carbon conservation constraint are optimized, so that the subsequent predicted emission curves are more closely aligned with the actual emission curves. The prediction bias, emission upper limit, and emission principal values ​​are stored together, and corresponding emission accounting reports are generated.

[0101] This embodiment also provides a computer device applicable to the carbon dioxide emission accounting method based on thermal power plants, including: 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 dioxide emission accounting method based on thermal power plants as proposed in the above embodiment.

[0102] 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.

[0103] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the carbon dioxide emission accounting method based on a thermal power plant 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.

[0104] In summary, this invention achieves statistical accuracy and consistency in the principal emission calculation results by: aligning observed data with prior fuel data over time and performing assimilation calculations in conjunction with carbon conservation constraints, thereby improving the credibility of total emissions; by constructing a predictive simulation model and combining it with a pre-defined set of executable actions for emission result analysis, enabling the prediction of future emission trends before the actions are implemented, thus providing feedforward decision support for unit operation optimization; and by comparing the predicted emission curve with the upper bound of uncertainty minute by minute and triggering gating adjustments, ensuring that the final output executable actions will not lead to excessive emissions, guaranteeing the compliance and traceability of the reported results.

[0105] 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 calculating carbon dioxide emissions from thermal power plants, characterized in that: include, Collect operational parameter data and coal quality test data, and obtain observation data and fuel prior data; By combining observational data and prior fuel data with carbon conservation constraints and prior states, assimilation calculations are performed to obtain the principal values ​​and uncertainties of emissions. Sensitivity data is calculated using a preset set of executable actions, emission principal values, and uncertainties. This sensitivity data is then subjected to windowed cumulative calculation to obtain the cumulative response. Specifically, an uncertainty-weighted windowed cumulative calculation is performed on a minute-level sensitivity data sequence to obtain the cumulative response of any candidate action within that window. The expression is as follows: ； ； in, This refers to actions in a predefined set of executable actions. Cumulative response results within a window This indicates the number of minutes displayed in the window. Represents a timestamp. Indicates the timestamp Uncertainty weighting factor Indicates the timestamp Sensitivity data of primary emission values ​​to unit output. Indicates action Fine-tuning of unit output in the middle, Indicates the timestamp Sensitivity data of the main emission values ​​to the excess air coefficient. Indicates action The amount of excess air coefficient fine-tuning in the middle. In timestamp The cross-sensitivity data of the main emission values ​​to power output and excess air coefficient. Indicates the timestamp Uncertainty of the principal values ​​of upper emissions Indicates the scale parameter; A predictive simulation model is constructed using the cumulative response, expressed as follows: ； ； in, This refers to actions in a predefined set of executable actions. Window-level counterfactual emissions estimates This refers to actions in a predefined set of executable actions. The minute-by-minute counterfactual trajectory, Indicates the timestamp The minute-level main emission values, Indicates the window baseline; Emissions are calculated one by one using a predictive simulation model and a pre-set set of executable actions to obtain an emission data set. By comparing and analyzing emission data sets, the optimal feedforward action for emission control can be obtained. The predicted emission curve of the feedforward suggestion action is obtained by jointly calculating the predictive simulation model and the feedforward suggestion action with the optimal emission level. The difference between the main emission values ​​and the observed data series is calculated to obtain the residual set. The residual set is then processed by a sliding time window to obtain the residual set after the sliding time window processing. The residual set after being processed by the sliding time window is subjected to quantile calculation to obtain the quantiles. The quantiles are then superimposed with the principal emission values ​​to obtain the upper bound of the uncertainty. By comparing the predicted emission curve with the upper bound of uncertainty, gating adjustments are made to the feedforward suggested actions to obtain the executable actions and the upper bound of emissions. The actual emission curve is obtained by using the principal emission values. The actual emission curve is compared with the predicted emission curve to obtain the prediction deviation. The prediction deviation, emission upper limit and principal emission values ​​are then archived.

2. The carbon dioxide emission accounting method based on thermal power plants as described in claim 1, characterized in that: The specific steps for collecting operational parameter data and coal quality test data, and obtaining observation data and prior fuel data are as follows: Collect operational parameter data, perform time-consistent processing and caliber verification, and obtain the processed operational parameter data; Collect coal quality test data, match the processed operating parameter data with the coal quality test data, and obtain observation data and fuel prior data.

3. The carbon dioxide emission accounting method based on thermal power plants as described in claim 2, characterized in that: The process involves assimilation calculations using observational data and prior fuel data, combined with carbon conservation constraints and prior states, to obtain principal emission values ​​and uncertainties. The specific steps are as follows: The observation data and the fuel prior data are concatenated at the same timestamp to form a two-dimensional observation vector; The two-dimensional observation vector is combined with the prior state vector, and assimilation operation is performed in conjunction with the carbon conservation constraint to obtain the assimilated emission parameters. Uncertainty propagation analysis was performed on the assimilated emission parameters to obtain the principal values ​​and uncertainties of emissions.

4. The carbon dioxide emission accounting method based on thermal power plants as described in claim 3, characterized in that: The steps involve comparing the predicted emission curve with the upper bound of uncertainty, adjusting the gating of the feedforward suggested actions, and obtaining the executable actions and the upper bound of emissions. The predicted emission curve is compared minute by minute with the upper bound of uncertainty to obtain comparative data. By jointly analyzing the comparative data and the feedforward suggested actions, gating judgment data is obtained and adjusted to obtain the executable actions and emission upper limits.

5. The carbon dioxide emission accounting method based on thermal power plants as described in claim 4, characterized in that: The process involves obtaining the actual emission curve through principal emission values, comparing the actual emission curve with the predicted emission curve to obtain the prediction deviation, and archiving the prediction deviation, emission upper bound, and principal emission values. The specific steps are as follows: The main emission values ​​are continuously collected within a time window to obtain the actual emission curve. The actual emission curve is then compared with the predicted emission curve minute by minute to obtain the prediction deviation. By correcting the sensitivity data for the next cycle through prediction bias, the assimilation calculation parameters under carbon conservation constraints are optimized, and the upper limit of emissions, principal values ​​of emissions, and prediction bias are stored accordingly to obtain archived records of prediction bias, upper limit of emissions, and principal values ​​of emissions.

6. 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 dioxide emission accounting method based on thermal power plants as described in any one of claims 1 to 5.

7. 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 dioxide emission accounting method based on any one of claims 1 to 5.

Images