Carbon emission prediction method and device based on dynamic attribution, electronic equipment, storage medium and program product

Abstract

The application relates to the technical field of carbon emission prediction, and discloses a carbon emission prediction method and device based on dynamic attribution, electronic equipment, a storage medium and a program product. The method comprises the following steps: acquiring time sequence data of different carbon emission driving variables and different carbon emission response variables at a current time; constructing a state vector at the current time according to the time sequence data; calculating the distance between the state vector at the current time and state vectors at different historical times in a historical state library, and determining the weights of the different historical times according to the distance; calculating a local coefficient vector according to the response vectors at the different historical times, the state vectors and the weights; calculating a predicted value of a carbon emission response variable at a preset time step in the future according to the state vector at the current time and the local coefficient vector, and outputting carbon emission trend early warning information. Through causal screening, state similarity weighting and positive and negative disturbance effect quantity calculation, the application realizes carbon emission trend prediction and dynamic attribution of driving variables.

Description

Technical Field

[0001] This invention relates to the field of carbon emission prediction technology, and specifically to a carbon emission prediction method, apparatus, electronic device, storage medium, and program product based on dynamic attribution. Background Technology

[0002] Accurate prediction of carbon emissions and analysis of their driving mechanisms are the core foundation for achieving energy conservation and emission reduction. Whether it is the formulation of regional emission reduction pathways at the macro level or the optimization of energy efficiency in high-energy-consuming units at the micro level, both face common engineering challenges: it is necessary to accurately predict future carbon emission trends and to identify which key parameters can most effectively achieve emission reduction.

[0003] Currently, mainstream analytical techniques in engineering practice mainly rely on two paths: one is statistical models based on fixed empirical formulas, and the other is black-box prediction models based on deep learning. However, when facing real-world emission systems with complex nonlinear characteristics, both of these existing technologies have significant limitations in engineering applications. For example, fixed-parameter statistical models presuppose that the correlation coefficient between driving factors and carbon emissions is a constant value, making it impossible to change the fitting rules according to the dynamic changes in regional industrial and energy structures. When structural changes occur in the system's operating conditions, the carbon emission prediction results deviate significantly from the actual evolution trend. Deep learning prediction models rely on large-scale samples for parameter training, while regional carbon emissions are mostly short-term annual statistical data. Under small sample conditions, overfitting is very likely to occur, directly reducing the accuracy of carbon emission trend prediction.

[0004] Furthermore, existing carbon emission prediction technologies typically only output future emission values, making it difficult to identify the state-dependent effects, intensity, and regulatory classification attributes of driving variables within the same calculation process. Consequently, they cannot provide carbon emission management systems with readily accessible early warning or regulatory classification signals. Summary of the Invention

[0005] This invention provides a carbon emission prediction method, device, electronic device, storage medium, and program product based on dynamic attribution, to solve the problems of fixed parameters in existing models, difficulty in quantifying the dynamic influence of driving variables, insufficient prediction accuracy, and difficulty in outputting callable and controllable signals.

[0006] In a first aspect, the present invention provides a carbon emission prediction method based on dynamic attribution, the method comprising: Acquire time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment; construct a state vector at the current moment based on the time-series data of carbon emission driving variables and carbon emission response variables. The state vector includes the current term of the response variable, the historical lag term of the response variable, the current term of the driving variable, and the historical lag term of the driving variable; calculate the distance between the current moment's state vector and the state vectors of different historical moments in the historical state database, and determine the weight of different historical moments based on the distance. The historical state database contains state vectors and response vectors of different historical moments, and the response vectors of different historical moments are the true response values ​​of the carbon emission response variables at a future preset time step; calculate the local coefficient vector based on the response vectors, state vectors, and weights of different historical moments; calculate the predicted value of the carbon emission response variable at a future preset time step based on the current moment's state vector and local coefficient vector; output carbon emission trend warning information based on the predicted value of the carbon emission response variable.

[0007] The carbon emission prediction method based on dynamic attribution provided by this invention can gather basic raw information describing the changes in the operation of the emission system by acquiring time-series data of carbon emission driving and response variables. Based on the above data, a state vector containing the current and lagged terms of the response variable and the current and lagged terms of the driving variable is constructed. The discrete time-series information can be transformed into multi-dimensional spatial coordinates that can characterize the instantaneous comprehensive operating conditions of the system by relying on phase space reconstruction logic, thus restoring the actual operating characteristics of the system at present. Furthermore, by calculating the spatial distance between the current state and each historical state and assigning corresponding sample weights according to the distance, the historical data that is close to the current operating conditions has a higher reference contribution in the fitting calculation process. The local coefficient vector is solved by using the weighted historical sample data, which breaks away from the constraints of the traditional model that uses fixed regression coefficients throughout the process and adapts to the unique operating rules of the current moment. Finally, the future carbon emission prediction results are calculated by combining the current state vector and the specific local coefficients, which effectively improves the prediction accuracy of carbon emissions under different operating condition switching scenarios.

[0008] In an optional implementation, before the steps of acquiring time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment, the method further includes: Acquire historical time-series data sets and historical response variable data for different candidate driver variables; for each candidate driver variable, generate its own surrogate sequence using a surrogate data generation method that preserves spectral characteristics; use a convergent cross-mapping algorithm to calculate the prediction performance index of historical response variable data using historical time-series data of candidate driver variables and the distribution of surrogate sequence prediction performance index of historical response variable data using surrogate sequence; identify candidate driver variables whose prediction performance index is greater than the preset quantile of the surrogate sequence prediction performance index distribution and passes the verification correction as carbon emission driver variables.

[0009] The carbon emission prediction method based on dynamic attribution provided by this invention constructs a proxy sequence for each candidate driving variable that retains the original spectral characteristics but breaks the temporal coupling relationship. Then, it compares the cross-mapping prediction levels of the real variables and the proxy sequences to filter out pseudo-correlated indicators that only fluctuate synchronously with time and have no actual driving effect, and retains the variables that have a real driving effect on carbon emissions. This eliminates irrelevant and redundant indicators from the source and avoids pseudo-correlated variables interfering with the subsequent state vector construction and prediction calculation.

[0010] In one optional implementation, the local coefficient vector is calculated based on the response vector, state vector, and weights at different historical moments, including: A diagonal weight matrix is ​​constructed based on the weights at different historical moments; a locally weighted linear regression equation is constructed based on the response vector, state vector, and diagonal weight matrix at different historical moments; the locally weighted linear regression equation is solved using the singular value decomposition algorithm to obtain the local coefficient vector, which is used to characterize the local mapping relationship between each carbon emission driving variable and the carbon emission response variable in the current state vector.

[0011] The carbon emission prediction method based on dynamic attribution provided by this invention constructs a diagonal weight matrix using the weights corresponding to each historical moment. Then, it combines the historical response vector, state vector, and diagonal weight matrix to build a locally weighted linear regression equation. The differential weighting constraint of historical samples is completed in the equation construction stage, abandoning the construction form of equal weights for all samples. Furthermore, the equation is solved by the singular value decomposition algorithm, avoiding the problems of matrix singularity and incomputability that are easy to occur in conventional matrix inversion. The local coefficient vector is stably solved, and the coefficient vector can accurately describe the unique local mapping relationship between each driving variable and carbon emission at the current moment, providing a reliable coefficient basis for subsequent quantitative measurement of carbon emissions.

[0012] In one optional implementation, a state vector for the current moment is constructed based on time-series data of carbon emission driving variables and carbon emission response variables, including: The time delay step and optimal embedding dimension are determined based on the time series data of the carbon emission response variable. Taking the current moment as the baseline, lag time points are sequentially taken forward according to the time delay step. The total order of lag reconstruction is determined according to the optimal embedding dimension. The time series data of the carbon emission response variable are sampled at equal intervals to obtain the current term and historical lag term of the response variable. The current term, lag term, current term, and historical lag term of the driving variable are combined to form the candidate state vector at the current moment. The state vector at the current moment is selected from the candidate state vectors.

[0013] The carbon emission prediction method based on dynamic attribution provided by this invention determines the optimal embedding dimension and time delay step by traversing multiple combinations of embedding dimensions and time delays and optimizing parameters based on quantitative indicators, ensuring that the parameters match the time-series variation characteristics of carbon emissions. Lagged data is extracted in segments according to the selected delay interval and number of embeddings. Current and lagged data of the response and driving variables are summarized to generate candidate state vectors, fully covering various operational information. Furthermore, the current state vector is selected from the candidate state vectors to reduce interference from invalid data and accurately represent real-time operating conditions.

[0014] In one optional implementation, the carbon emission prediction method based on dynamic attribution further includes: According to the preset perturbation amplitude, positive and negative perturbations are applied to each carbon emission driving variable at the current time. The predicted values ​​of the response variables corresponding to the positive perturbation and the negative perturbation are calculated. The difference between the positive and negative perturbation response variable values ​​is divided by twice the preset perturbation amplitude to obtain the dynamic effect size of the carbon emission driving variable. The dynamic effect size curve of each carbon emission driving variable is constructed based on the dynamic effect size of the carbon emission driving variable. The dynamic effect size curve is used to characterize the evolution law of the continuous change of the dynamic effect size of the corresponding carbon emission driving variable over time.

[0015] The carbon emission prediction method based on dynamic attribution provided by this invention applies pre-set positive and negative bidirectional perturbations to each carbon emission driving variable, and performs difference normalization calculation on the carbon emission prediction results corresponding to the positive and negative perturbations. This accurately quantifies the true strength and direction of the influence of each driving variable on carbon emissions at each moment, achieving a quantitative and accurate measurement of the driving effect. Furthermore, it continuously solves the dynamic effect quantity time-by-time and constructs a time-series evolution curve, continuously presenting the dynamic fluctuation and real-time evolution characteristics of the carbon emission impact effect of each driving variable over time. This accurately captures the differences in the role of driving factors at different stages, providing a reliable quantitative basis for subsequent refined analysis of carbon emission driving mechanisms and the formulation of dynamic control strategies.

[0016] In one optional implementation, the dynamic effect size of the carbon emission driving variables includes the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity. The carbon emission prediction method based on dynamic attribution further includes: If the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity are both less than zero, then the carbon emission driving variable is determined to be a key regulatory driving variable; if the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity are both greater than zero, then the carbon emission driving variable is determined to be a deterioration driving variable; if the product of the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity is less than zero, then the carbon emission driving variable is determined to be a pending driving variable; based on the key regulatory driving variable, the deterioration driving variable, and the pending driving variable, carbon emission regulation early warning information is generated.

[0017] The carbon emission prediction method based on dynamic attribution provided by this invention quantifies the changing patterns of carbon emission indicators corresponding to fluctuations in various driving variables, classifies factors into three categories with different effects: key regulatory driving variables, deterioration driving variables, and undetermined driving variables. Based on the differences in the direction of the effects of various variables on carbon emissions, adjustment strategies are formulated in a differentiated manner. The carbon emission control scheme is optimized in a targeted manner based on the classification results, avoiding the blind control caused by the extensive formulation of control measures based on experience, and ensuring that emission reduction adjustment measures can match the actual impact characteristics of various variables.

[0018] Secondly, the present invention provides a carbon emission prediction device based on dynamic attribution, the device comprising: The data acquisition module is used to acquire time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment; The current state vector construction module is used to construct the current state vector based on the time series data of carbon emission driving variables and carbon emission response variables. The state vector includes the current term of the response variable, the historical lag term of the response variable, the current term of the driving variable, and the historical lag term of the driving variable. The state similarity weighting module is used to calculate the distance between the current state vector and the state vectors at different historical times in the historical state database, and to determine the weight of different historical times based on the distance. The historical state database contains state vectors and response vectors at different historical times. The response vectors at different historical times are the true response values ​​of the carbon emission response variable at a preset time step in the future. The local coefficient vector calculation module is used to calculate the local coefficient vector based on the response vector, state vector, and weights at different historical moments. The carbon emission prediction module is used to calculate the predicted value of the carbon emission response variable for a preset time step in the future based on the current state vector and local coefficient vector. The carbon emission early warning module is used to output early warning information on carbon emission trends based on the predicted values ​​of carbon emission response variables.

[0019] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the carbon emission prediction method based on dynamic attribution described in the first aspect or any corresponding embodiment thereof.

[0020] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the dynamic attribution-based carbon emission prediction method of the first aspect or any corresponding embodiment described above.

[0021] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the dynamic attribution-based carbon emission prediction method of the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the carbon emission prediction method based on dynamic attribution according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the carbon emission prediction results and the dynamic effect of the driving variables according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the dynamic attribution and regulation classification results of carbon emission driving variables according to an embodiment of the present invention. Figure 5 This is a structural block diagram of a carbon emission prediction device based on dynamic attribution according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] As an optional application scenario of this invention, the specific application environment architecture or specific hardware architecture on which the execution of the dynamic attribution-based carbon emission prediction method depends is described herein. Figure 1 As shown, the architecture system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0028] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0029] According to an embodiment of the present invention, a method for predicting carbon emissions based on dynamic attribution is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0030] This embodiment provides a carbon emission prediction method based on dynamic attribution, which can be executed by the aforementioned server and the results can be displayed via a mobile terminal. Figure 2 This is a flowchart of a carbon emission prediction method based on dynamic attribution according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain time series data of different carbon emission driving variables and different carbon emission response variables at the current moment.

[0031] In one optional embodiment, time-series data is a sequence of variable observations continuously sampled and recorded at fixed time nodes, with the time scale selectable as annual, quarterly, daily, etc. Time-series data for carbon emission response variables and carbon emission driving variables are collected separately according to a unified time indexing rule, with both types of data sharing the same time scale. Specifically, carbon emission response variables include total carbon emissions and carbon emission intensity, while carbon emission driving variables include multiple driving variables related to the carbon emission response variables.

[0032] Specifically, time-series data of carbon emission driving variables and carbon emission response variables can be obtained from statistical databases, energy metering systems, carbon emission monitoring systems, SCADA systems, DCS systems, or IoT sensors.

[0033] Step S202: Construct the state vector for the current moment based on the time series data of carbon emission driving variables and carbon emission response variables.

[0034] In one optional embodiment, the state vector includes the current response variable, the historical lag response variable, the current driving variable, and the historical lag driving variable. The current response variable refers to the measured data of the carbon emission response variable at the current moment; the historical lag response variable refers to historical carbon emission response variable data extracted at fixed intervals from the current moment; the current driving variable refers to the measured data of the carbon emission driving variable at the current moment; and the historical lag driving variable refers to historical carbon emission driving variable data extracted at fixed intervals from the current moment. The current response variable data and interval lag data are extracted based on the time-series data of the carbon emission driving variable and the carbon emission response variable, and the current and lag driving variable data are matched simultaneously. These multiple types of values ​​are then integrated and concatenated to generate the state vector for the current moment.

[0035] Step S203: Calculate the distance between the current state vector and the state vectors at different historical times in the historical state database, and determine the weights of different historical times based on the distance.

[0036] In one optional embodiment, the historical state database contains state vectors and response vectors at different historical moments, and the response vectors at different historical moments are the true response values ​​of the carbon emission response variables at a preset time step in the future.

[0037] The geometric distance between the current state vector and the state vectors at different historical moments in phase space is calculated using the Euclidean distance formula, as shown below:

[0038] In the formula, Let this be the state vector at the current moment. For a historic moment j The state vector, Indicates the current state With historical status The Euclidean distance between them. The smaller the geometric distance, the more similar the two states are in terms of emission levels, historical inertia, combinations of driving variables, and operating conditions.

[0039] Furthermore, the weights for different historical moments are calculated using the sequence mapping (S-map) weighting function:

[0040] In the formula, Representing historical state Regarding the current state The weight, Indicates the current state The average distance between the available samples in the historical state database and the historical state database. This represents the S-map localization parameter. When... When all historical samples have the same weight, the model approximates a global linear model; when In this case, historical samples that are closer to the current state receive higher weights, and the model behaves as a state-dependent local nonlinear model.

[0041] Step S204: Calculate the local coefficient vector based on the response vector, state vector, and weights at different historical moments.

[0042] In one optional embodiment, the local coefficient vector adapted to the current system operating condition is solved by combining the determined weights of each historical moment with all historical state vectors to form an independent variable matrix and the historical response vectors to form a dependent variable. Different current state vectors have different distances from the state vectors of each historical moment, thus resulting in different weights for each historical moment and ultimately different local coefficient vectors. For example, when the system is in a construction and expansion phase, historical moments closer to it with higher weights typically come from the construction period, and the model will obtain local evolution rules for the construction period. When the system enters a stable operation phase, historical moments closer to it will more often come from the low-carbon operation phase, and the model will obtain local evolution rules for the operation phase. Therefore, the same driving variable can exhibit different directions and intensities of influence in different states, thereby supporting the identification of operating condition switching and driving effect reversals.

[0043] Step S205: Calculate the predicted value of the carbon emission response variable for the future preset time step based on the current state vector and local coefficient vector.

[0044] In one optional embodiment, an augmented state vector is constructed by adding a constant term to the front of the original state vector. The carbon emission prediction value at the time corresponding to the preset step size is calculated by transposing the augmented state vector and performing an inner product operation with the local coefficient vector.

[0045] Specifically, to avoid confusion between state coordinates and intercept terms, the system uses the current state vector Expanded into augmented state vector :

[0046] In the formula, Indicates time The augmented state vector, first term Used to correspond to local intercept ,the remaining to The current state vector The coordinates of each state in the diagram.

[0047] The predicted carbon emissions for the next preset step size are calculated using the following formula:

[0048] In the formula, This represents the predicted value of the future response variable. This indicates the state coordinate number. If the target is total carbon emissions, then it is the predicted value of future total carbon emissions; if the target is carbon emission intensity, then it is the predicted value of future carbon emission intensity.

[0049] Step S206: Output carbon emission trend early warning information based on the predicted value of carbon emission response variable.

[0050] In one optional embodiment, the carbon emission trend early warning information includes predicted carbon emissions for future periods, determination results of emission change trends, multi-level early warning levels, and trend conclusions.

[0051] Specifically, based on the calculated predicted values ​​of carbon emission response variables, the system analyzes the future trends of carbon emission increases, decreases, and fluctuations to generate early warning information on carbon emission trends, and sends this information to the carbon emission management terminal or control interface.

[0052] The carbon emission prediction method based on dynamic attribution provided in this embodiment can gather basic raw information that characterizes the changes in the operation of the emission system by acquiring time-series data of carbon emission driving and response variables. Based on the above data, a state vector containing the current and lagged terms of the response variable and the current and lagged terms of the driving variable is constructed. The discrete time-series information can be transformed into multi-dimensional spatial coordinates that can represent the instantaneous comprehensive operating conditions of the system by relying on phase space reconstruction logic, thus restoring the actual operating characteristics of the system at present. Furthermore, by calculating the spatial distance between the current state and each historical state and assigning corresponding sample weights according to the distance, the historical data that is close to the current operating conditions has a higher reference contribution in the fitting calculation process. The local coefficient vector is solved by using the weighted historical sample data, which breaks away from the constraints of the traditional model that uses fixed regression coefficients throughout the process and adapts to the unique operating rules of the current moment. Finally, the future carbon emission prediction results are calculated by combining the current state vector and the specific local coefficients, which effectively improves the prediction accuracy of carbon emissions under different operating condition switching scenarios.

[0053] In some optional implementations, prior to the steps of acquiring time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment, the method further includes: Step a1: Obtain the historical time series data set of different candidate driving variables and the historical response variable data.

[0054] In one optional embodiment, candidate driving variables include macroeconomic driving variables, such as the proportion of added value of the secondary industry, the proportion of hydropower generation, and net electricity exports, as well as microeconomic driving variables, such as influent concentration, aeration rate, wind speed, and rotational speed.

[0055] Specifically, annual / quarterly statistical data are obtained through statistical databases, or high-frequency monitoring data, such as daily and hourly data, are obtained using SCADA systems, DCS systems, and IoT sensors. Time-series data of all initially selected candidate driving variables, total carbon emissions, and carbon emission intensity are then aggregated. Zero-mean standardization is performed on all raw time-series data, as shown in the following formula:

[0056] In the formula, Represents the standardized first i One candidate driver variable, This represents the original observed value of the variable. Indicates the first i The mean of each variable during the sample period This represents the standard deviation of the variable over the sample period. After standardization, each variable is transformed to a comparable scale, preventing variables with larger numerical dimensions from having an unreasonable advantage in distance calculation and model fitting.

[0057] Step a2: For each candidate driving variable, a proxy sequence is generated using a proxy data generation method that preserves spectral characteristics.

[0058] In an alternative embodiment, it is set Indicates the first b The agent sequence at time t The value of , ,in B This represents the number of surrogate sequences, for example, 500. The surrogate sequences can be generated using the Ebisuzaki phase randomization method, which preserves the autocorrelation and spectral characteristics of the original sequence but disrupts its temporal correspondence with the response variable.

[0059] Step a3: Using the convergent cross-mapping algorithm, calculate the prediction performance index for predicting historical response variable data using historical time series data of candidate driving variables and the second prediction skill index for predicting historical response variable data using surrogate sequences.

[0060] In an optional embodiment, the cross-mapping prediction performance of the real sequences is calculated using a convergent cross-mapping algorithm. Distribution of performance metrics for surrogate sequence prediction .in, Represents the actual observation sequence. This indicates a proxy sequence.

[0061] Step a4: Candidate driving variables whose predicted performance indicators are greater than the preset quantile of the distribution of predicted performance indicators of the proxy sequence and have passed the test correction are determined as carbon emission driving variables.

[0062] In an optional embodiment, if the true predictive ability is significantly higher than the null model distribution formed by the surrogate sequence, for example, higher than its 95th quantile, and passes Bonferroni correction or BH multiple test correction, then the variable is determined to pass the causal significance screening. The set of carbon emission driving variables obtained after the above processing is as follows:

[0063] In the formula, This represents the set of carbon emission driving variables retained after causal screening. q This indicates the number of variables to be retained, and q No more than the total number of candidate driver variables .

[0064] The carbon emission prediction method based on dynamic attribution provided in this embodiment constructs a proxy sequence for each candidate driving variable that retains the original spectral characteristics but breaks the temporal coupling relationship. Then, it compares the cross-mapping prediction levels of the real variables and the proxy sequences to filter out pseudo-correlated indicators that only fluctuate synchronously with time and have no actual driving effect, and retains the variables that have a real driving effect on carbon emissions. It eliminates irrelevant and redundant indicators from the source and avoids pseudo-correlated variables interfering with the subsequent state vector construction and prediction calculation.

[0065] In some optional implementations, a local coefficient vector is calculated based on the response vector, state vector, and weights at different historical moments, including: Step b1: Construct a diagonal weight matrix based on the weights at different historical moments.

[0066] Step b2: Construct a locally weighted linear regression equation based on the response vector, state vector, and diagonal weight matrix at different historical moments.

[0067] In an optional embodiment, a diagonal weight matrix is ​​constructed based on the weights at different historical moments. If each state vector contains p Each coordinate represents a historical moment. j The state vector can be written as By arranging all available historical samples row by row, the design matrix can be obtained. :

[0068] In the formula, Indicates the use of predicting the current state The local design matrix, the first column Represents the intercept term. This indicates the time when entering the current local fitting of historical samples. This indicates the number of available historical samples. The corresponding response vector. for:

[0069] In the formula, This indicates that the above historical samples will be in the future. The actual response value after the step.

[0070] Local coefficient vector for:

[0071] In the formula, This represents the local intercept in the current state. to These correspond to the local coefficients of each coordinate in the state vector. Since this coefficient vector changes with the current state... It changes with the environment, therefore it is not a globally fixed parameter in a traditional regression model.

[0072] The locally weighted linear regression equation is shown below:

[0073] In the formula, The weights of each historical sample The diagonal matrix formed, superscript This indicates the matrix transpose.

[0074] Step b3: Solve the locally weighted linear regression equation using the singular value decomposition algorithm to obtain the local coefficient vector.

[0075] In one optional embodiment, singular value decomposition (SVD) is used to solve the weighted least squares problem to obtain the local coefficient vector at the current time step. SVD is a numerical decomposition algorithm that breaks down an arbitrary matrix into the product of three smaller matrices. By decomposing the weighted design matrix using SVD, the matrix invertibility defect is avoided, and a unique local coefficient vector is stably calculated. This coefficient vector characterizes the local mapping relationship between each carbon emission driving variable and the carbon emission response variable in the current state vector.

[0076] The carbon emission prediction method based on dynamic attribution provided in this embodiment constructs a diagonal weight matrix using the weights corresponding to each historical moment. Then, it combines the historical response vector, state vector, and diagonal weight matrix to build a locally weighted linear regression equation. The differentiated weighting constraint of historical samples is completed in the equation construction stage, abandoning the construction form of equal weight for all samples. Furthermore, the equation is solved by the singular value decomposition algorithm, avoiding the matrix singularity and incomputability problems that are easy to occur in conventional matrix inversion. The local coefficient vector is stably solved, and the coefficient vector can accurately describe the unique local mapping relationship between each driving variable and carbon emission at the current moment, providing a reliable coefficient basis for subsequent quantitative measurement of carbon emissions.

[0077] In some optional implementations, a state vector for the current moment is constructed based on time-series data of carbon emission driving variables and carbon emission response variables, including: Step c1: Determine the time delay step and the optimal embedding dimension based on the time series data of the carbon emission response variable.

[0078] In one optional embodiment, multiple sets of embedding dimensions and time delay combinations are traversed within a preset parameter value range. Trial calculations are carried out based on the simplex projection algorithm. The prediction correlation coefficient, root mean square error, and mean absolute error are used as evaluation criteria to select the set of parameters with the best overall prediction accuracy. This set is then fixed as the optimal embedding dimension and time delay step size for subsequent reconstruction.

[0079] Step c2: Using the current time as a reference, take the lag time points sequentially forward according to the time delay step, determine the total order of lag reconstruction based on the optimal embedding dimension, and perform equal-interval lag sampling on the time series data of carbon emission response variables to obtain the current term of the response variable and the historical lag term of the response variable.

[0080] In one optional embodiment, starting from the current moment, historical carbon emission response variable time series data is extracted forward at every time delay step. The total number of data segments extracted is controlled according to the optimal embedding dimension. The current moment data is split as the current item of the response variable, and multiple segments of historical lagged data are used as multiple historical lagged items of the response variable.

[0081] Step c3: Combine the current term of the response variable, the lagged term of the response variable, the current term of the driving variable, and the historical lagged term of the driving variable to form the candidate state vector at the current moment.

[0082] In an alternative embodiment, it is set Indicates the first The lag period used for each driving variable, when "When" indicates the use of the current period driving variable. The time indicates the use of lagged driving variables. The candidate state vector is shown below:

[0083] In the formula, Indicates time A set of candidate state coordinates to choose from. This represents the current value of the response variable. to This represents the lagged term of the response variable. Indicates the first i The historical lags of each carbon emission driver variable at the corresponding lag period.

[0084] Step c4: Select the current state vector from the candidate state vectors.

[0085] In one alternative embodiment, a multivariate embedding or multi-view embedding method is used to select the coordinate combination with better prediction performance from the candidate state vectors as the state vector at the current moment.

[0086] Let the state coordinates be ,in P This indicates the number of coordinates contained in the current state vector. (Time) t The state vector is written as:

[0087] In the formula, Indicates time The state vector, each It can be the current term of the response variable, the lagged term of the response variable, the current term of the driving variable, or the lagged term of the driving variable. Indicates the state coordinate number.

[0088] For example, a typical state vector can be written as:

[0089] When the driving variable has a significant lag effect, it can be written as:

[0090] Repeat the above construction for all valid moments to form a historical state database. Let... j If we denote the historical sample time, then the historical state database can be represented as:

[0091] In the formula, This represents the historical state library used for EDM modeling. Representing historical moments The state vector, This indicates that the historical state will be in the future. The actual response value after the step. This indicates the number of valid historical samples.

[0092] The dynamic attribution-based carbon emission prediction method provided in this embodiment determines the optimal embedding dimension and time delay step by traversing multiple combinations of embedding dimensions and time delays and optimizing parameters based on quantitative indicators, ensuring that the parameters match the time-series variation characteristics of carbon emissions. Lagged data is extracted in segments according to the selected delay interval and number of embeddings. Current and lagged data of the response and driving variables are summarized to generate candidate state vectors, fully covering various operational information. Furthermore, the current state vector is selected from the candidate state vectors to reduce interference from invalid data and accurately represent real-time operating conditions.

[0093] In some alternative implementations, the dynamic attribution-based carbon emission prediction method further includes: Step d1: Apply positive and negative perturbations to each carbon emission driving variable at the current moment according to the preset perturbation amplitude.

[0094] In an alternative embodiment, for the first i A key driving variable, let... This indicates the preset disturbance magnitude for the variable. This disturbance magnitude can be set according to the actual management implications, such as increasing the secondary industry's share by 1 percentage point, increasing hydropower generation's share by 1 percentage point, increasing net electricity outflow by 10 TWh, or adding a fixed magnitude in the standardized space. The system at time... t Construct two perturbation states: one that will change the first state...i The state coordinates corresponding to each driving variable increase Another one reduces it Other state coordinates remain unchanged. If the model uses lagged terms for the driving variables, then a perturbation of the same magnitude is applied to the corresponding lagged coordinates. The two perturbation states are denoted as follows:

[0095]

[0096] In the formula, Indicates the first i The state vector after applying a positive perturbation to each carbon emission driving variable This represents the state vector after applying a negative perturbation; the + in parentheses i and- i It is only used to identify the direction of the disturbance and does not indicate a new type of variable.

[0097] Step d2: Calculate the predicted values ​​of the response variables corresponding to positive disturbances and negative disturbances.

[0098] In an optional embodiment, the S-map prediction method is used to calculate the predicted future responses for the two perturbation states, as shown below:

[0099]

[0100] In the formula, This represents the established S-map state dependency prediction process; and These represent the predicted future responses under positive and negative perturbations, respectively.

[0101] Step d3: Subtract the values ​​of the positive and negative disturbance response variables and divide by twice the preset disturbance amplitude to obtain the dynamic effect of the carbon emission driving variables.

[0102] In an optional embodiment, the first i The driving variables at time 1 t The dynamic effect size is denoted as The calculation formula is as follows:

[0103] In the formula, Indicates the first i The local marginal impact of each driving variable on the future response variable in the current state. If This indicates that increasing the driving variable near the current state will improve the future response variable; if This indicates that increasing the driving variable will decrease the future response variable; if A value close to 0 indicates that the marginal effect of the driving variable is weak in the current state.

[0104] Step d4: Construct dynamic effect curves for each carbon emission driving variable based on the dynamic effect of the carbon emission driving variables.

[0105] In an optional embodiment, when the response variables include total carbon emissions and carbon emission intensity, the above-described perturbation prediction is performed on both response variables separately. Let... Indicates the first i The dynamic effect of each driving variable on total carbon emissions. The dynamic effect of this effect on carbon emission intensity can be represented by a dynamic effect matrix:

[0106] In the formula, Indicates time t The dynamic effects matrix, also known as the dynamic Jacobian approximation matrix, is derived by each element from the finite difference results of the S-map positive and negative perturbation predictions. Each element represents the local effect of a driving variable on a specific response objective. Here, the Jacobian representation is a matrix representation in the sense of a local approximation, and is not equivalent to the globally fixed coefficients in traditional linear regression.

[0107] Furthermore, by repeating the above calculations for each time point, the dynamic effect size curves of each driving variable can be obtained. These dynamic effect size curves characterize the evolution of the dynamic effect size of the corresponding carbon emission driving variable over time.

[0108] In an alternative embodiment, such as Figure 3 As shown, the system generates historical fitted values ​​and future predicted values ​​for carbon emission response variables, and simultaneously outputs the dynamic effect size curves of each driving variable. The dynamic effect size curves are used to characterize the local marginal impact of the same driving variable on the carbon emission response variable at different historical stages or under different operating conditions, so as to identify the time intervals in which the driving effect is enhanced, weakened, or reversed.

[0109] The dynamic attribution-based carbon emission prediction method provided in this embodiment applies pre-set positive and negative bidirectional perturbations to each carbon emission driving variable. The carbon emission prediction results corresponding to these perturbations are then normalized using difference calculations. This accurately quantifies the true strength and direction of each driving variable's impact on carbon emissions at each moment, achieving a quantitative and precise measurement of the driving effect. Furthermore, the dynamic effect size is continuously solved moment-by-moment, and a time-series evolution curve is constructed. This continuously presents the dynamic fluctuations and real-time evolution characteristics of the carbon emission impact of each driving variable over time, accurately capturing the differences in the effects of driving factors at different stages. This provides a reliable quantitative basis for subsequent refined analysis of carbon emission driving mechanisms and the formulation of dynamic control strategies.

[0110] In some optional implementations, the dynamic effect size of carbon emission drivers includes the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity. The dynamic attribution-based carbon emission prediction method also includes: If the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity are both less than zero, then the carbon emission driving variable is determined to be a key regulatory driving variable.

[0111] In an alternative embodiment, for the first Each driving variable is analyzed, and its dynamic effect on total carbon emissions is read. and the dynamic effect on carbon emission intensity .like This indicates that increasing this driving variable can simultaneously reduce both total carbon emissions and carbon emission intensity, classifying the current state as a synergistic optimization zone, and the carbon emission driving variable as a key regulatory driving variable. In this state, this driving variable can be prioritized for support or expansion.

[0112] If the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity are both greater than zero, then the carbon emission driving variable is determined to be a deterioration driving variable.

[0113] In an alternative embodiment, if This indicates that increasing this driving variable will simultaneously increase both total carbon emissions and carbon emission intensity, classifying the current state as an inefficient loss zone, and the carbon emission driving variable as a deteriorating driving variable. In this state, further increases in this driving variable should be avoided, or alternative, reduction, and optimized operational measures should be prioritized.

[0114] If the product of the effect of total carbon emissions and the effect of carbon dioxide emission intensity is less than zero, then the carbon emission driving variable is determined to be an undetermined driving variable.

[0115] In an alternative embodiment, if If the condition is met, it means that increasing this driving variable will improve one objective but worsen another, classifying the current state as a trade-off zone, and the carbon emission driving variable as an undetermined driving variable. In this state, the magnitude of regulation needs to be determined by combining policy objectives, constraints, and multi-objective optimization results to avoid pursuing only a single indicator at the expense of another objective.

[0116] In engineering applications, to avoid misjudgments caused by minute numerical fluctuations, the system can set an effect size threshold. .when or In such cases, the corresponding effect can be classified as a weak or uncertain effect, and a recommendation is made to verify it further with more data or expert judgment. This represents the minimum acceptable effect threshold from a management perspective, which can be set based on data noise, carbon accounting accuracy, or management scale.

[0117] In an alternative embodiment, such as Figure 4 As shown, the system uses the dynamic effect of the driving variables on total carbon emissions and the dynamic effect on carbon emission intensity as two coordinate axes to map the driving variables to the control classification space. If the driving variable is located in the collaborative optimization region, the output is the priority control variable; if it is located in the inefficient loss region, the output is the constraint or reduction variable; if it is located in the trade-off region, the output is the undetermined variable that needs to be further optimized in conjunction with management objectives.

[0118] Carbon emission control early warning information is generated based on key control drivers, deterioration drivers, and undetermined drivers.

[0119] In one optional embodiment, the carbon emission control early warning information includes a control classification signal, which is used for display on the carbon emission management terminal, triggering early warnings, and invoking the control interface. Control schemes are formulated based on the attributes of the three types of driving variables. Expansion and enhancement measures are implemented for key control driving variables, reduction and substitution are carried out for deteriorating driving variables, and the adjustment range for undetermined driving variables is weighed in conjunction with control objectives, thus integrating these factors to form an optimized carbon emission control strategy.

[0120] The carbon emission prediction method based on dynamic attribution provided in this embodiment quantifies the changing patterns of carbon emission indicators corresponding to fluctuations in various driving variables, classifies factors into three categories with different effects: key regulatory driving variables, deterioration driving variables, and undetermined driving variables. Based on the different directions of the effects of various variables on carbon emissions, adjustment strategies are formulated in a differentiated manner. The carbon emission control scheme is optimized in a targeted manner based on the classification results, avoiding the blindness of control measures caused by extensive formulation of control measures based on experience, and ensuring that emission reduction adjustment measures can match the actual impact characteristics of various variables.

[0121] This embodiment also provides a carbon emission prediction device based on dynamic attribution, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0122] This embodiment provides a carbon emission prediction device based on dynamic attribution, such as... Figure 5 As shown, it includes: The data acquisition module 301 is used to acquire time series data of different carbon emission driving variables and different carbon emission response variables at the current moment.

[0123] The current state vector construction module 302 is used to construct the current state vector based on the time series data of carbon emission driving variables and carbon emission response variables. The state vector includes the current term of the response variable, the historical lag term of the response variable, the current term of the driving variable, and the historical lag term of the driving variable.

[0124] The state similarity weighting module 303 is used to calculate the distance between the current state vector and the state vectors at different historical times in the historical state database, and to determine the weight of different historical times based on the distance. The historical state database contains state vectors and response vectors at different historical times, and the response vectors at different historical times are the true response values ​​of the carbon emission response variable at a preset time step in the future.

[0125] The local coefficient vector calculation module 304 is used to calculate the local coefficient vector based on the response vector, state vector and weight at different historical moments.

[0126] The carbon emission prediction module 305 is used to calculate the predicted value of the carbon emission response variable for a preset time step in the future based on the current state vector and local coefficient vector. The carbon emission early warning module 306 is used to output carbon emission trend early warning information based on the predicted value of the carbon emission response variable.

[0127] The carbon emission prediction device based on dynamic attribution provided in this embodiment of the invention can execute the carbon emission prediction method based on dynamic attribution provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0128] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0129] The following is a detailed reference. Figure 6This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0130] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0131] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the dynamic attribution-based carbon emission prediction method of the embodiments of the present invention.

[0132] Figure 6 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0133] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the dynamic attribution-based carbon emission prediction method shown in the above embodiments is implemented.

[0134] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0135] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A carbon emission prediction method based on dynamic attribution, characterized in that, The method is executed by a carbon emission prediction server and includes: Obtain time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment; A state vector for the current moment is constructed based on the time-series data of the carbon emission driving variables and carbon emission response variables. The state vector includes the current term of the response variable, the historical lag term of the response variable, the current term of the driving variable, and the historical lag term of the driving variable. Calculate the distance between the current state vector and the state vectors at different historical times in the historical state database, and determine the weight of different historical times based on the distance. The historical state database contains state vectors and response vectors at different historical times, and the response vectors at different historical times are the true response values ​​of the carbon emission response variable at a future preset time step. Calculate the local coefficient vector based on the response vector, state vector, and weights at different historical moments; Calculate the predicted value of the carbon emission response variable for the future preset time step based on the current state vector and local coefficient vector; Based on the predicted values ​​of the carbon emission response variables, output carbon emission trend early warning information.

2. The method according to claim 1, characterized in that, Before the steps of acquiring time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment, the method further includes: Obtain historical time-series data sets of different candidate driver variables and historical response variable data; For each candidate driving variable, a surrogate sequence is generated using a surrogate data generation method that preserves spectral characteristics. Using the convergent cross-mapping algorithm, we calculate the prediction performance index of using historical time series data of candidate driver variables to predict historical response variable data and the distribution of surrogate sequence prediction performance index of using surrogate sequence to predict historical response variable data, respectively. Candidate driving variables whose predicted performance indicators are greater than the preset quantile of the distribution of predicted performance indicators of the proxy sequence and whose results have been verified are identified as carbon emission driving variables.

3. The method according to claim 1, characterized in that, The calculation of the local coefficient vector based on the response vector, state vector, and weights at different historical moments includes: Construct a diagonal weight matrix based on the weights at the different historical moments; A locally weighted linear regression equation is constructed based on the response vector, state vector, and diagonal weight matrix at different historical moments. The local weighted linear regression equation is solved using the singular value decomposition algorithm to obtain a local coefficient vector. This local coefficient vector is used to characterize the local mapping relationship between each carbon emission driving variable and the carbon emission response variable in the current state vector.

4. The method according to claim 1, characterized in that, The current state vector is constructed based on the time-series data of the carbon emission driving variables and carbon emission response variables, including: The time delay step and optimal embedding dimension are determined based on the time series data of carbon emission response variables; Based on the current time, the lag time points are sequentially taken forward according to the time delay step. The total order of lag reconstruction is determined according to the optimal embedding dimension. The time series data of the carbon emission response variable are sampled at equal intervals to obtain the current term of the response variable and the historical lag term of the response variable. The current term of the response variable, the lagged term of the response variable, the current term of the driving variable, and the historical lagged term of the driving variable are combined to form the candidate state vector at the current moment; The current state vector is selected from the candidate state vectors.

5. The method according to claim 1, characterized in that, The method also includes: According to the preset perturbation amplitude, positive and negative perturbations are applied to each carbon emission driving variable at the current moment; Calculate the predicted values ​​of the response variables for positive and negative disturbances; The dynamic effect of the carbon emission driving variable is obtained by subtracting the positive disturbance response variable value from the negative disturbance response variable value and dividing by twice the preset disturbance amplitude. Based on the dynamic effect size of the carbon emission driving variables, dynamic effect size curves are constructed for each carbon emission driving variable. The dynamic effect size curves are used to characterize the evolution of the dynamic effect size of the corresponding carbon emission driving variable over time.

6. The method according to claim 5, characterized in that, The dynamic effect size of the carbon emission driving variables includes the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity. The method also includes: If the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity are both less than zero, then the carbon emission driving variable is determined to be a key regulatory driving variable. If the effect size of total carbon emissions and the effect size of carbon dioxide emission intensity are both greater than zero, then the carbon emission driving variable is determined to be a deterioration driving variable. If the product of the effect of total carbon emissions and the effect of carbon dioxide emission intensity is less than zero, then the carbon emission driving variable is determined to be an undetermined driving variable. Carbon emission control early warning information is generated based on the key control driving variables, deterioration driving variables, and undetermined driving variables.

7. A carbon emission prediction device based on dynamic attribution, characterized in that, The device includes: The data acquisition module is used to acquire time-series data of different carbon emission driving variables and different carbon emission response variables at the current moment; The current state vector construction module is used to construct the current state vector based on the time series data of the carbon emission driving variables and carbon emission response variables. The state vector includes the current term of the response variable, the historical lag term of the response variable, the current term of the driving variable, and the historical lag term of the driving variable. The state similarity weighting module is used to calculate the distance between the current state vector and the state vectors at different historical times in the historical state database, and to determine the weight of different historical times based on the distance. The historical state database contains state vectors and response vectors at different historical times, and the response vectors at different historical times are the true response values ​​of the carbon emission response variable at a preset time step in the future. The local coefficient vector calculation module is used to calculate the local coefficient vector based on the response vector, state vector, and weights at different historical moments. The carbon emission prediction module is used to calculate the predicted value of the carbon emission response variable for a preset time step in the future based on the current state vector and local coefficient vector. The carbon emission early warning module is used to output carbon emission trend early warning information based on the predicted value of the carbon emission response variable.

8. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the carbon emission prediction method based on dynamic attribution as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the carbon emission prediction method based on dynamic attribution as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes computer instructions for causing a computer to execute the carbon emission prediction method based on dynamic attribution as described in any one of claims 1 to 6.

Images