Method and system for calculating power carbon emissions based on hybrid model prediction, and terminal

Abstract

The present application belongs to the field of electric power, and particularly relates to a power carbon emission calculation method, system and terminal based on a hybrid model prediction. According to the real-time output of each generator set and the weighted average of the fuel emission coefficient, the basic electric carbon factor of the day is calculated. Then, the power information in the power information collection system is obtained to construct the feature vector of each period, which is input into multiple time series models to predict multiple prediction factors. According to the residual error and the root mean square error of the model prediction, the different model predictions are fused according to the weight to obtain the fused carbon factor value. Then, the density peak clustering is applied to the historical daily load curve, and each period of the day is divided into different load clusters, and the carbon factor correction rate and the power factor correction coefficient are updated in combination with the load type. Finally, the total power carbon emission of the day is calculated by using the updated fused factor and the correction parameters. The present application solves the problems of systematic error and insufficient timeliness of the existing carbon emission accounting scheme.

Description

Technical Field

[0001] This invention belongs to the field of power, specifically relating to a method, system, and terminal for calculating electricity carbon emissions based on a hybrid model prediction. Background Technology

[0002] Accurate accounting and dynamic assessment of carbon emissions have become a crucial component of power system management and regulation. Traditional methods largely rely on fixed emission factors, estimating emissions by multiplying the annual or regional grid average carbon emission coefficient by user electricity consumption. However, in actual grid operation, various dynamic factors, such as changes in the proportion of renewable energy output, fluctuations in inter-regional electricity purchases, and differences in load distribution during different time periods, render fixed carbon factors ineffective in reflecting real-time emission intensity. This results in systematic errors and insufficient timeliness in estimating electricity carbon emissions.

[0003] Current research has introduced machine learning models to make short-term predictions of electricity carbon emission factors, but most of them ignore the influence of potential variables such as load patterns and power factors, and lack a unified fusion mechanism, making it difficult to balance accuracy and robustness. Other studies have attempted to build static feature libraries or use single clustering methods to analyze load, but their support for the generalization ability of models in practical applications is limited. Summary of the Invention

[0004] To address the systematic errors and timeliness issues in existing carbon emission accounting schemes, this invention provides a method, system, and terminal for calculating electricity carbon emissions based on a hybrid model.

[0005] This invention is achieved using the following technical solution: A method for calculating carbon emissions from electricity based on a hybrid model, comprising: The basic carbon factor for the day is calculated by weighting the real-time output and fuel emission coefficient of each generator set. CF base .

[0006] Historical data on electricity consumption information related to carbon emission calculations were obtained from the electricity consumption information collection system, and feature vectors for each time period were constructed. X ( i );Will X ( i Input to multiple pre-trained time series models X ( i Predictions are made for each time period of the day. i predictors ; j Indicates the timing model number. j =1…M, where M represents the total number of time series models.

[0007] according to and CF base The residuals and root mean square errors of the prediction results of each model are calculated, and the fused carbon factor values ​​for each time period are generated by combining the multi-model prediction results through a smooth soft update method based on the errors. CF rt ( i ).

[0008] Density peak clustering is applied to historical daily load curves to divide the day into different time periods. K In different load clusters, and add load cluster tags. k , k =1… K .

[0009] according to CF base The average carbon factor across all time periods within the same load cluster The carbon factor correction rate corresponding to each load cluster was calculated. .

[0010] Generate the power factor correction coefficient for each load cluster. ; and then combine CF rt ( i ), and The total carbon emissions E from electricity on that day are calculated using the following formula: ; In the above formula, Indicates that it belongs to the first of the day. k The first load cluster i Power generation during each time period; This indicates the duration of each time period.

[0011] As a further improvement to the present invention, the duration of each time period... This refers to the data sampling period setting for the electricity information collection system, namely: .

[0012] As a further improvement to the present invention, the basic electric carbon factor CF base The calculation formula is: ; In the above formula, Indicates the first i Power generation during each time period; EF ( i ) indicates the first i The fuel emission coefficient of the generator set during each time period.

[0013] As a further improvement of the present invention, the parameters in the feature vector include: wind power ratio, photovoltaic ratio, thermal power generation, imported electricity, exported electricity, self-consumed electricity, self-generated electricity, power, current, voltage, and historical carbon factor for historical periods.

[0014] Among them, the proportion of wind power, the proportion of photovoltaic power, thermal power generation, imported electricity, exported electricity, self-consumed electricity, self-generated electricity, power, current, and voltage values ​​have been normalized; the normalization formula used is: ; In the above formula, x Represents the original value of any parameter. express x The normalized value; and These represent the minimum and maximum values ​​of the corresponding parameters among all feature vectors for that day, respectively.

[0015] As a further improvement of the present invention, a time series model based on ARIMA, LSTM and Kalman filtering is used to generate predictors.

[0016] As a further improvement of the present invention, the carbon factor value is fused. CF rt ( i The calculation process is as follows: (1) Calculate the predictor of each model output for each time period. residual : ; (2) Calculate the mean square error of the prediction results output by each model for each time period. RSEM j : ; (3) Calculate the initial weights of each model inversely proportional to the mean square error. : , The weights are then smoothed using a soft update method to obtain the model weights. : ; In the above formula, Indicates the soft update coefficient; This represents the model weight of the j-th model in the previous time period.

[0017] (4) Combine the model weights of each model and predictors The fusion factor of the mixture model is calculated using the following formula. : .

[0018] (5) The basic carbon factor CF base and fusion factor The carbon factor value is obtained by fusing according to the smoothing coefficient. CF rt ( i ): .

[0019] As a further improvement of the present invention, the carbon factor correction rate corresponding to each load cluster The calculation formula is: ; In the above formula, The load cluster is marked as k Fusion carbon factor values ​​corresponding to each time period CF rt ( i M indicates that the load cluster in the historical load data for that day is marked as k The number of time periods.

[0020] As a further improvement of the present invention, the power factor correction coefficient corresponding to each load cluster is... The calculation formula is: ; In the above formula, Indicates the first i Power factor for each time period; This represents an empirical amplification factor related to the load type.

[0021] The present invention also includes a power carbon emission calculation system based on a hybrid model prediction, which includes: a data acquisition module, a fusion carbon factor generation module, a correction unit, and a carbon emission calculation module.

[0022] The data acquisition module is used to obtain historical data of all electricity information collected daily by the electricity information acquisition system; and to construct feature vectors and historical daily load curves for each time period.

[0023] The fusion carbon factor generation module deploys multiple pre-trained time series models. Each time series model can output the predicted factor for the next time period based on the feature vectors of each historical time period. The fusion carbon factor generation module is used to calculate the basic electric carbon factor for the day by weighting the real-time output and fuel emission coefficient of each generator unit. CF base It also utilizes multiple time-series models and feature vectors for prediction, and then combines them with... CF baseThe prediction results are analyzed and fused to generate fused carbon factor values ​​for each time period. CF rt ( i ).

[0024] The correction unit is used to cluster electricity information from different time periods in the historical daily load curve using density peak clustering, dividing the daily electricity information into multiple distinct load clusters; then, it updates the carbon factor correction rate by combining the load cluster labels of the data. and power factor correction coefficient .

[0025] The carbon emission calculation module is used to combine the updated CF rt ( i ), and The value is used to calculate the total carbon emissions E of electricity generated on that day based on the power generation information.

[0026] The present invention also includes a computer program product, which includes a computer program that, when executed by a processor, implements the aforementioned method for calculating electricity carbon emissions based on a hybrid model prediction, thereby completing the calculation of the carbon emissions generated on a given day.

[0027] The present invention also includes a storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for calculating electricity carbon emissions based on a hybrid model, thereby completing the calculation of the carbon emissions generated on that day.

[0028] The present invention also includes a carbon emission calculation terminal, which includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the aforementioned method for calculating the carbon emissions of electricity based on a hybrid model prediction, thereby completing the calculation of the carbon emissions of electricity generated on that day.

[0029] The technical solution provided by this invention has the following beneficial effects: This invention calculates carbon emission data based on real-time power changes by predicting carbon factor values ​​using a hybrid model, and then weights and softly updates the prediction results of the hybrid model by incorporating the root mean square error of the prediction results. This updated fused carbon factor can quickly respond to short-term fluctuations in grid structure and renewable energy, while maintaining long-term trend stability.

[0030] This invention incorporates real-time load changes and introduces bandwidth kernel functions, intercepts, and least-squares center fine-tuning into density peak clustering. This allows typical laboratory load characteristics to adaptively match diverse on-site load patterns, significantly improving the accuracy of power factor during intraday peaks, valleys, and intermediate periods. Then, based on load types at different times, an empirical amplification factor λ related to load conditions is introduced to the power factor deviation, quantifying and precisely compensating for network-added losses caused by reactive power, thus meeting diverse power factor requirements in industrial scenarios.

[0031] The present invention utilizes a fused carbon factor generated by a hybrid module, and the carbon emission data calculated based on the carbon factor correction rate and power factor correction coefficient updated according to the actual load has higher accuracy, is more suitable for short-term carbon emission analysis, and has higher practical value. Attached Figure Description

[0032] Figure 1 This is a flowchart of the steps in the method for calculating carbon emissions from electricity based on a hybrid model, as provided in Embodiment 1 of the present invention.

[0033] Figure 2 This is a flowchart of the steps in the carbon factor value update process in Embodiment 1 of the present invention.

[0034] Figure 3 This is an architecture diagram of the electricity carbon emission calculation system based on hybrid model prediction provided in Embodiment 2 of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0036] Example 1 This embodiment provides a method for calculating electricity carbon emissions based on a hybrid model. This method constructs a basic carbon factor based on the power generation structure, introduces a multi-model prediction mechanism, and combines load clustering analysis and power factor correction strategies to comprehensively integrate static structure and dynamic operating characteristics, achieving accurate dynamic prediction of carbon emission factors. Based on dynamically updated parameters, the solution in this embodiment can automatically calculate and display carbon emission data online, meeting users' needs for data accuracy and real-time performance.

[0037] Specifically, in this embodiment, the electricity consumption information collection system typically completes automated data collection and uploading every 15 minutes within a 24-hour period. Based on the sampling cycle of the electricity consumption information collection system, each period can be divided into 96 time slots. To achieve more refined calculation of carbon emission data, this embodiment generates a fused carbon emission factor based on the updated electricity information for each time slot. Furthermore, the state of the electrical load changes throughout the day. Based on these changes, this embodiment also dynamically adjusts the correction parameters of the power factor and carbon emission factor according to the load type of different time slots. After the electricity consumption information collection system completes the uploading of daily load and other electricity data, this embodiment can quickly calculate the daily carbon emission data by combining data from different time slots, demonstrating high real-time performance.

[0038] In detail, such as Figure 1 As shown, the method for calculating electricity carbon emissions based on a hybrid model prediction provided in this embodiment includes the following steps: S1: The basic carbon factor for the day is calculated by weighting the real-time output and fuel emission coefficient of each generator set. CF base .

[0039] The basic carbon factor is calculated using existing methods, primarily based on the real-time output (i.e., power generation) of the generator sets and the fuel emission coefficients corresponding to the various energy forms used by the generator sets. For example, the fuel emission coefficients per unit power differ for coal-fired generator sets, natural gas generator sets, hydropower, and solar power generation. Furthermore, when different types of power generation are connected to the grid at different times, the carbon emissions of each type of grid-connected power should be accumulated and weighted averaged. In practical applications, the basic carbon factor... CF base The calculation formula is: ; In the above formula, Indicates the first i Power generation during each time period; EF ( i ) indicates the first i The fuel emission coefficient of the generator set during each time period. This indicates the duration of each time period set according to the data sampling period of the electricity consumption information collection system; N represents the total number of time periods in a day. For example, when When set to 15 minutes, N=96. Of course, in practical applications, users can also flexibly adjust the data sampling frequency of the electricity information collection system according to their needs, and adaptively adjust the number of time periods for each data analysis.

[0040] S2: Obtain historical electricity data related to carbon emission calculation from the electricity information collection system, and construct feature vectors for each time period. X ( i );Will X ( i Input to multiple pre-trained time series models X ( i Predictions are made for each time period of the day. i Predictors.

[0041] In this embodiment, considering that actual carbon emission factors often exhibit time-series correlation, a time-series model is used to incrementally update the carbon factor value by combining various power information from the power grid at different times, thereby outputting the current prediction factor. To improve prediction accuracy, this embodiment employs a hybrid model containing multiple time-series models to generate the prediction factor, thus obtaining multiple prediction factors corresponding to each time-series model. ;in, j Indicates the timing model number. j =1…M.

[0042] In practical applications, time-series models based on ARIMA, LSTM, and Kalman filtering can be used for carbon factor prediction. These three models, after offline training on historical datasets, can be deployed and applied online. This allows for the prediction of carbon emission factors for the next time period based on the feature vectors representing the grid's operating status for each historical time period. In this embodiment, the parameters in the feature vectors input into the time-series model may include: wind power share, photovoltaic share, thermal power generation, imported electricity, transmitted electricity, self-consumed electricity, self-generated electricity, power, current, voltage, and historical carbon factors for historical time periods (e.g., past P-periods).

[0043] In this embodiment, considering that different parameters have different dimensions and significant differences in value, in order to balance the parameters, the parameters of wind power ratio, photovoltaic ratio, thermal power generation, imported electricity, exported electricity, self-consumed electricity, self-generated electricity, power, current, and voltage are preprocessed, and the influence of dimensions is eliminated through data normalization. The normalization formula used for each parameter is as follows: ; In the above formula, x Represents the original value of any parameter. express x The normalized value; and These represent the minimum and maximum values ​​of the corresponding parameters among all feature vectors for that day, respectively.

[0044] S3: According to andCF base The residuals and root mean square errors of the prediction results of each model are calculated, and the fused carbon factor values ​​for each time period are generated by combining the multi-model prediction results through a smooth soft update method based on the errors. CF rt ( i ).

[0045] In this embodiment, the fused carbon factor value is generated by fusing the predicted carbon factor values ​​from various time-series models. During the fusion process, it is necessary to analyze the error distribution of the predicted carbon factor values ​​output by each time-series model in conjunction with the basic carbon factor values, and then perform targeted smoothing before generation. Specifically, as... Figure 2 As shown, the fusion carbon factor value CF rt ( i The calculation process is as follows: (1) Calculate the predictor of each model output for each time period. residual : ; In the above formula, j In this embodiment, referring to each time series model, j =1…3.

[0046] (2) Based on each time period Calculate the mean squared error of the prediction results output by each model. RSEM j : ; (3) In this embodiment, if the mean square error of any time-series model prediction result is large, its weight in the final generated fused carbon factor is reduced. Furthermore, to ensure accuracy, the change in the fused carbon factor over consecutive time periods should avoid abrupt changes. Based on this principle, this embodiment calculates the initial weight of each model inversely proportional to the mean square error. : .

[0047] The weights are then smoothed using a soft update method to obtain the model weights. : ; In the above formula, Indicates the soft update coefficient; This represents the model weight of the j-th model in the previous time period.

[0048] (4) Combine the model weights of each model and predictors The fusion factor of the mixture model is calculated using the following formula. : .

[0049] (5) The basic carbon factor CF base and fusion factor By smoothing coefficient Fusion, to obtain the fusion carbon factor value CF rt ( i ): .

[0050] The smoothing coefficient can be generated based on the actual operating status of the power plant.

[0051] S4: Apply density peak clustering to the historical daily load curve to divide the various time periods of the day into... K In different load clusters, and add load cluster tags. k , k =1… K .

[0052] In this embodiment, to reduce the impact of load on carbon factors at different time periods, density peak clustering (supporting Euclidean distance and DTW) is applied to the historical daily load curves, thereby automatically dividing the daily load into K typical load clusters. The acquired historical load data includes the sequence values ​​of the total system load over the past several days, with a sampling period of 15 minutes. Density peak clustering automatically identifies K "typical electricity consumption patterns" in the load sequence, such as: base load, rising period, stable period, evening peak, and valley period. The load data under these different electricity consumption patterns form multiple different "clusters" after clustering. Each cluster has a cluster center C_k, which represents the most typical load trend under that pattern. This embodiment can divide different time periods of the day into different load clusters based on the clustering results, and add the corresponding load cluster category label to all data corresponding to each time period. k , k =1… K .For example. k =1 indicates that the corresponding time period belongs to the base load. k =2 indicates that the corresponding time period is in an upward phase, ... k =5 indicates that the corresponding time period is the trough period.

[0053] S5: According to CF base The average carbon factor across all time periods within the same load cluster The carbon factor correction rate corresponding to each load cluster was calculated. .

[0054] In this embodiment, the fused carbon factor values ​​generated for all time periods belonging to the same load cluster are... CF rt ( i The mean of ) is denoted as Then, by combining the deviation between the mean and the baseline carbon factor value, a carbon factor correction rate corresponding to each load cluster is calculated. The corresponding calculation formula is: ; In the above formula, This indicates the values ​​of each fusion carbon factor. CF rt ( i The load clusters selected in the process are labeled as follows: k The values ​​of the fused carbon factor for each time period; M represents the load cluster labeled as in the historical load data for that day. k The number of time periods.

[0055] S6: Combine the identified load clusters to generate the power factor correction coefficient for each load cluster. ; ; In the above formula, Indicates the first i Power factor for each time period; This represents an empirical amplification factor related to the load type.

[0056] Combined CF rt ( i ), and The total carbon emissions E from electricity on that day are calculated using the following formula: ; In the above formula, Indicates that it belongs to the first of the day. k The first load cluster i Power generation during each time period; This indicates the duration of each time period.

[0057] In summary, this embodiment uses ARIMA, LSTM, and Kalman filtering models in parallel when calculating carbon emission data based on real-time power changes, and applies RMSE inverse ratio and soft update weight fusion. This approach can quickly respond to short-term fluctuations in grid structure and renewable energy while maintaining long-term trend stability. Furthermore, considering real-time load changes, this embodiment introduces bandwidth kernel functions, cutoff radius, and least squares center fine-tuning in density peak clustering. This allows typical laboratory load characteristics to adaptively match diverse on-site load patterns, significantly improving the accuracy of factors during intraday peak and valley periods and intermediate times. For power factor deviations, an empirical amplification factor λ related to load conditions is introduced to quantify and accurately compensate for network-added losses caused by reactive power, meeting the diverse power factor requirements in industrial scenarios. Therefore, the final carbon emission calculation results generated by this embodiment are more accurate, more suitable for short-term carbon emission analysis, and have higher practical value.

[0058] Example 2 Based on the scheme in Example 1, this example further provides a system for calculating electricity carbon emissions based on a hybrid model. This system is used to dynamically update the carbon emission data and various correction parameters according to the data uploaded in real time by the electricity information collection system, using the scheme provided in Example 1, and to generate real-time carbon emission data online.

[0059] Specifically, such as Figure 3 As shown, the electricity carbon emission calculation system based on hybrid model prediction provided in this embodiment includes: a data acquisition module, a fusion carbon factor generation module, a correction unit, and a carbon emission calculation module.

[0060] The data acquisition module is used to obtain historical data of all electricity information collected daily by the electricity information collection system; and to construct feature vectors and historical daily load curves for each time period. Specifically, the data acquisition module acquires the values ​​of data such as wind power share, photovoltaic share, thermal power generation, imported electricity, exported electricity, self-consumption electricity, self-generated electricity, power, current, voltage, and historical carbon factor for historical periods by time period. Then, it uses a preset normalization formula to convert the values ​​of wind power share, photovoltaic share, thermal power generation, imported electricity, exported electricity, self-consumption electricity, self-generated electricity, power, current, and voltage into normalized values, and then encodes them into corresponding feature vectors. The historical daily load curve is automatically generated based on the load data of the day.

[0061] The fusion carbon factor generation module deploys multiple pre-trained time series models. Each time series model can output the predicted factor for the next time period based on the feature vectors of each historical time period. The fusion carbon factor generation module is used to calculate the basic electric carbon factor for the day by weighting the real-time output and fuel emission coefficient of each generator set. CF baseIt also utilizes multiple time-series models and feature vectors for prediction, and then combines them with... CF base The prediction results are analyzed and fused to generate fused carbon factor values ​​for each time period. CF rt ( i In this embodiment, the time series model deployed in the fusion carbon factor generation module can be trained using ARIMA, LSTM, and Kalman filtering as base models.

[0062] The correction unit is used to cluster electricity information from different time periods in the historical daily load curve using density peak clustering, dividing the daily electricity information into multiple distinct load clusters; then, it updates the carbon factor correction rate by combining the load cluster labels of the data. and power factor correction coefficient .

[0063] The carbon emission calculation module is used to combine the updated CF rt ( i ), and The value is used to calculate the total carbon emissions E from electricity generation on that day based on the power generation information; the calculation formula is: .

[0064] In practical applications, the system can also include a visualization module, which is used to visualize the calculated carbon emissions results by combining daily metering information, so as to display the results of automatic accounting online.

[0065] Example 3 In order to apply the scheme in Example 1 to the calculation method of carbon emissions of electricity, this example also provides a computer program product, storage medium and computing terminal for executing the corresponding scheme.

[0066] The computer program product provided in this embodiment includes a computer program. When the computer program is executed by a processor, it implements the method for calculating electricity carbon emissions based on a hybrid model prediction as described in Embodiment 1, thereby completing the calculation of the carbon emissions generated on that day.

[0067] The storage medium provided in this embodiment stores a computer program. When the computer program is executed by the processor, it implements the method for calculating the carbon emissions of electricity based on the hybrid model prediction as in Embodiment 1, thereby completing the calculation of the carbon emissions of electricity generated on that day.

[0068] The carbon emission calculation terminal provided in this embodiment includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the method for calculating electricity carbon emissions based on the hybrid model prediction as in Embodiment 1, thereby completing the calculation of the carbon emissions generated on that day.

[0069] The carbon emission calculation terminal for data-graph fusion provided in this embodiment is essentially a computer device. In practical applications, this computer device can be an embedded device; or it can be a separate computer device, such as a laptop, tablet, desktop computer, or a rack server, blade server, tower server, or cabinet server (including a standalone server or a server cluster composed of multiple servers) capable of executing computer programs.

[0070] The computer device in this embodiment includes, but is not limited to, a memory and a processor that can be interconnected via a system bus. In this embodiment, the memory (i.e., the readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory can be an internal storage unit of the computer device, such as the hard disk or RAM of the computer device. In other embodiments, the memory can also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Of course, the memory can also include both internal storage units and external storage devices of the computer device. In this embodiment, the memory is typically used to store the operating system and various application software installed on the computer device. Furthermore, the memory can also be used to temporarily store various types of data that have been output or will be output.

[0071] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is typically used to control the overall operation of a computer device.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of calculating power carbon emissions based on a mixture model prediction, characterized by, It includes: The basic electricity carbon factor of the day is calculated by weighting average of real-time output of each generator set and fuel emission coefficient CF base ; Historical data on electricity consumption information related to carbon emission calculations were obtained from the electricity consumption information collection system, and feature vectors for each time period were constructed. X ( i );Will X ( i The input is fed into various pre-trained time series models, and... X ( i Predictions are made for each time period of the day. i predictors ; j Indicates the timing model number; according to and CF base The residuals and root mean square errors of the prediction results of each model are calculated, and the fused carbon factor values ​​for each time period are generated by combining the multi-model prediction results through a smooth soft update method based on the errors. CF rt ( i ); Density peak clustering is applied to historical daily load curves to divide the day into different time periods. K In different load clusters, and add load cluster tags. k , k =1… K ; according to CF base The average carbon factor across all time periods within the same load cluster The carbon factor correction rate corresponding to each load cluster was calculated. ; Generate the power factor correction coefficient for each load cluster. ; and then combine CF rt ( i ), and The total carbon emissions E from electricity on that day are calculated using the following formula: ； In the above formula, Indicates that it belongs to the first of the day. k The first load cluster i Power generation during each time period; This indicates the duration of each time period.

2. The method for calculating electricity carbon emissions based on a hybrid model as described in claim 1, characterized in that: Duration of each time period This refers to the data sampling period setting for the electricity information collection system, namely: 。 3. The method for calculating electricity carbon emissions based on a hybrid model as described in claim 2, characterized in that, Basic electrocarbon factor CF base The calculation formula is: ； In the above formula, Indicates the first i Power generation during each time period; EF ( i ) indicates the first i The fuel emission coefficient of the generator set during each time period.

4. The method for calculating electricity carbon emissions based on a hybrid model as described in claim 1, characterized in that: The parameters in the feature vector include: wind power ratio, photovoltaic ratio, thermal power generation, imported electricity, exported electricity, self-consumed electricity, self-generated electricity, power, current, voltage, and historical carbon factor for historical periods. Among them, the proportion of wind power, the proportion of photovoltaic power, thermal power generation, imported electricity, exported electricity, self-consumed electricity, self-generated electricity, power, current, and voltage values ​​have been normalized; the normalization formula used is: ； In the above formula, x Represents the original value of any parameter. express x The normalized value; and These represent the minimum and maximum values ​​of the corresponding parameters in all feature vectors for that day, respectively. And / or, Predictors are generated using a time series model based on ARIMA, LSTM, and Kalman filtering.

5. The method for calculating electricity carbon emissions based on a hybrid model as described in claim 4, characterized in that: Fusion carbon factor value CF rt ( i The calculation process is as follows: (1) Calculate the predictor of each model output for each time period. residual : ； (2) Based on each time period Calculate the mean squared error of the prediction results output by each model. RSEM j : ； (3) Calculate the initial weights of each model inversely proportional to the mean square error. : ， The weights are then smoothed using a soft update method to obtain the model weights. : ； In the above formula, Indicates the soft update coefficient; This represents the model weight of the j-th model in the previous time period; (4) Combine the model weights of each model and predictors The fusion factor of the mixture model is calculated using the following formula. : ； (5) The basic carbon factor CF base and fusion factor By smoothing coefficient Fusion, to obtain the fusion carbon factor value CF rt ( i ): 。 6. The method for calculating electricity carbon emissions based on a hybrid model as described in claim 5, characterized in that: Carbon factor correction rate for each load cluster The calculation formula is: ； In the above formula, The load cluster is marked as k Fusion carbon factor values ​​corresponding to each time period CF rt ( i M indicates that the load cluster in the historical load data for that day is marked as k The number of time periods.

7. The method for calculating electricity carbon emissions based on a hybrid model as described in claim 1, characterized in that: Power factor correction coefficient for each load cluster The calculation formula is: ； In the above formula, Indicates the first i Power factor for each time period; This represents an empirical amplification factor related to the load type.

8. A system for calculating carbon emissions from electricity based on a hybrid model, characterized in that, It uses the method for calculating the carbon emissions of electricity generated on a given day based on a hybrid model, as described in any one of claims 1-7. It includes: The data acquisition module is used to acquire historical data of all electricity information collected daily by the electricity information acquisition system; and to construct feature vectors and historical daily load curves for each time period. The integrated carbon factor generation module deploys multiple pre-trained time series models. Each time series model is used to output the prediction factor for the next time period based on the feature vectors of each historical time period. The integrated carbon factor generation module is used to calculate the basic electric carbon factor for the day by weighted averaging the real-time output and fuel emission coefficient of each generator set. CF base It also utilizes multiple time-series models and feature vectors for prediction, and then combines them with... CF base The prediction results are analyzed and fused to generate fused carbon factor values ​​for each time period. CF rt ( i ); The correction unit is used to cluster electricity information from different time periods in the historical daily load curve using density peak clustering, dividing the daily electricity information into multiple distinct load clusters; then, it updates the carbon factor correction rate based on the load cluster labels of the data. and power factor correction coefficient ; The carbon emissions calculation module is used to combine the updated CF rt ( i ), and The value is used to calculate the total carbon emissions E of electricity generated on that day based on the power generation information.

9. A computer program product comprising a computer program, characterized in that: When the computer program is executed by the processor, it implements the method for calculating electricity carbon emissions based on hybrid model prediction as described in any one of claims 1-7, thereby completing the calculation of the carbon emissions generated on that day.

10. A carbon emission calculation terminal, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, it implements the method for calculating electricity carbon emissions based on hybrid model prediction as described in any one of claims 1-7, thereby completing the calculation of the carbon emissions generated on that day.

Images