A dynamic marginal cost-based carbon emission quota real-time regulation method and system

Abstract

The application discloses a kind of based on dynamic marginal cost's carbon emission quota real-time regulation method and system, it is related to carbon emission quota real-time regulation technical field, including the following steps, the real-time disturbance data of construction site is collected, current disturbance state is identified, and disturbance state is mapped to pre-defined state interval;Based on the state interval mapped, the corresponding interval model in pre-constructed carbon emission variable structure disturbance response model is called;Wherein, variable structure disturbance response model is composed of multiple interval models associated with different state intervals, each interval model includes interval variable structure control law and interval disturbance response function;Interval model is fitted, and real-time disturbance marginal elasticity is calculated, and marginal carbon emission cost function is constructed according to real-time disturbance marginal elasticity;Real-time regulation of carbon emission quota is carried out through interval model and marginal carbon emission cost function.To solve the problem of marginal carbon emission cost function distortion.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon emission quota real-time regulation method, more specifically, the present application relates to a carbon emission quota real-time regulation method and system based on dynamic marginal cost. BACKGROUND

[0002] Carbon emission intensity control and carbon quota allocation are important management methods for high-emission industries such as power, transportation and construction. In carbon-intensive construction scenarios such as power grid construction and major infrastructure projects, how to dynamically adjust the carbon emission quota of each stage according to the actual construction state to achieve optimal carbon emission efficiency under the condition of constant overall carbon emission quota is one of the core problems in current intelligent low-carbon construction. In the highly complex and frequently disturbed engineering scenarios, traditional linear modeling and fixed rule control methods have been difficult to meet the requirements. Therefore, there is an urgent need for a regulation mechanism that can dynamically identify the marginal influence of disturbances, flexibly respond to state switching, and realize real-time optimization and allocation of carbon emissions.

[0003] For example, the invention patent with the announcement number CN119065324A announced an industrial production negative carbon emission optimization control method and system, which includes obtaining a plurality of coordinate data and negative carbon emission data in a target area, then clustering the plurality of coordinate data into a first clustering cluster, and clustering the plurality of negative carbon emission data into a second clustering cluster, and then identifying a negative carbon emission abnormal area according to the difference between the first clustering cluster and the second clustering cluster for emission optimization control. The present application clusters the coordinate data and the corresponding negative carbon emission data respectively, analyzes the distribution characteristics of the coordinate data and the negative carbon emission data through clustering, and then utilizes the difference between the distribution characteristics to quickly locate the coordinate data and the negative carbon emission data with different distribution characteristics, thereby quickly identifying the position of the carbon emission abnormality, solving the problem of being unable to quickly locate the abnormal area of negative carbon emission in the prior art.

[0004] For example, the invention patent with the announcement number CN117742160B announced a carbon emission optimization control method and system based on artificial intelligence, which includes using a first fusion neural network and utilizing sample carbon emission monitoring data sequences to effectively learn and mine useful optimization control strategies from historical data. Secondly, by utilizing the dependency information between neural networks, the decision differences between different neural networks can be better understood and revealed, so that the network parameter learning and the adjustment of the fusion training error are more accurate. Finally, through simplification and optimization, the configuration amount of the neural network can be optimized, the computational complexity is reduced, the operation efficiency of the network is improved, and the consumption of computing resources is reduced. The target fusion neural network finally generated can make efficient and accurate carbon emission optimization control decisions for the target carbon emission source, realize the intelligentization and automation of the carbon emission optimization control decision process, and improve the accuracy and efficiency of the decisions.

[0005] The technical problems in the disclosed technical solutions are at least as follows.

[0006] Most of the existing carbon emission regulation methods use static carbon factor estimation or rule-based carbon quota allocation, which cannot fully reflect the dynamic marginal influence of disturbance variables on carbon emission changes under actual construction conditions, resulting in a regulation mechanism that cannot timely respond to the nonlinear jump of carbon emission behavior, and problems of low regulation accuracy and strong hysteresis.

[0007] Due to the large number of disturbance variables and complex influence mechanism, directly establishing a unified global response model often leads to high modeling dimension and poor stability, and is difficult to support online updating and real-time decision-making requirements in engineering practice.

[0008] To solve the above problems, the present application provides a solution. SUMMARY

[0009] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a carbon emission quota real-time regulation method and system based on dynamic marginal cost, which extracts the dominant mechanism of disturbance by interval division and sparse modeling, and solves the problems of strong carbon emission disturbance and regulation hysteresis in construction scenes by combining marginal cost optimization and sliding mode control.

[0010] To achieve the above-mentioned purposes, the present application provides the following technical solutions:

[0011] The carbon emission quota real-time regulation method based on dynamic marginal cost comprises the following steps: collecting real-time disturbance data of the construction site, identifying the current disturbance state, and mapping the disturbance state to a predefined state interval; based on the mapped state interval, calling the corresponding interval model in the pre-constructed carbon emission variable structure disturbance response model; wherein the variable structure disturbance response model is composed of a plurality of interval models associated with different state intervals, and each interval model includes an interval variable structure control law and an interval disturbance response function; fitting the interval model, calculating the real-time disturbance marginal elasticity, and constructing a marginal carbon emission cost function according to the real-time disturbance marginal elasticity; and performing real-time regulation of the carbon emission quota through the interval model and the marginal carbon emission cost function.

[0012] In a preferred embodiment, the collection of real-time disturbance data of the construction site, the identification of the current disturbance state, and the mapping of the disturbance state to the predefined state interval are as follows: collecting a historical data set of disturbance variables and carbon emission change rates of the construction site; constructing a plurality of disturbance state intervals based on density clustering; calculating the distance between the current disturbance state and the clustering centers of the predefined disturbance state intervals, and mapping it to the corresponding state interval according to the minimum distance principle.

[0013] In a preferred embodiment, the interval perturbation response function further comprises determining an optimal sparsity penalty coefficient, and the specific steps are as follows: taking the carbon emission change rate as the target response quantity, constructing a function base library based on the perturbation variable; setting a candidate set of sparsity penalty coefficients, and evaluating the fitting error under different sparsity parameters using a cross-validation strategy; selecting the sparsity penalty coefficient corresponding to the minimum fitting error as the optimal sparsity penalty coefficient.

[0014] In a preferred embodiment, the interval perturbation response function is obtained by a sparse regression algorithm, and the specific steps are as follows: based on the optimal sparsity penalty coefficient, solving the coefficients of the function base library by a sparse regression algorithm, and selecting the corresponding function base items according to the non-zero coefficients; generating the perturbation dominant response function expression of the current interval using the function base items; generating the corresponding interval perturbation dominant response function for a plurality of different perturbation state intervals; aggregating the perturbation dominant response functions of all intervals to construct an interval perturbation response function library.

[0015] In a preferred embodiment, the interval variable structure control law is specifically: obtaining a preset global carbon quota plan, and constructing an interval carbon emission state deviation function based on the plan and real-time carbon emission state data; based on the sliding mode control theory, defining a convergence target surface of the control system, wherein the convergence target surface aims to make the interval carbon emission state deviation function approach zero;

[0016] According to the convergence target surface, the interval variable structure control law for driving the carbon emission state to track the global carbon quota is derived and generated to obtain the interval variable structure control law.

[0017] In a preferred embodiment, the fitting of the interval model, the calculation of the real-time perturbation marginal elasticity, and the construction of the marginal carbon emission cost function based on the real-time perturbation marginal elasticity are specifically: obtaining the interval perturbation response function of the target system, and performing fitting operation at the current perturbation state point to obtain the perturbation marginal carbon emission elasticity; obtaining the control weight of the current perturbation variable through a machine learning model based on the historical data set;

[0018] Based on the perturbation marginal carbon emission elasticity and the control weight, a marginal carbon emission cost function for carbon emission optimization decision is constructed.

[0019] In a preferred embodiment, the real-time regulation of the carbon emission quota through the interval model and the marginal carbon emission cost function further comprises obtaining a preliminary optimization control amount of each perturbation variable, and the specific steps are as follows: based on the preset global carbon quota plan and the historical data set, generating the carbon emission control upper limit constraint and the perturbation variable boundary constraint through a machine learning model; taking the marginal carbon emission cost function as the target, constructing a real-time optimization model containing the emission control upper limit constraint and the perturbation variable boundary constraint, and solving the optimization model in real time to output the preliminary optimization control amount of each perturbation variable.

[0020] In a preferred embodiment, the real-time regulation of carbon emission quota by the interval model and the marginal carbon emission cost function is specifically: calling an interval variable structure control law corresponding to the current disturbance state interval, constructing a sliding mode control correction instruction based on a carbon emission state deviation function; linearly superimposing the sliding mode control correction instruction and an externally input preliminary optimization regulation amount; and outputting the superimposed result as a final regulation output for driving carbon emission regulation.

[0021] The application also discloses a real-time regulation system for carbon emission quota based on dynamic marginal cost, which comprises a state interval mapping module, an interval model acquisition module, a carbon cost function construction module and a real-time regulation module.

[0022] The state interval mapping module is used for collecting real-time disturbance data of a construction site, identifying a current disturbance state, and mapping the disturbance state to a predefined state interval.

[0023] The interval model acquisition module is used for calling a corresponding interval model in a pre-constructed carbon emission variable structure disturbance response model based on the mapped state interval; wherein the variable structure disturbance response model is composed of a plurality of interval models associated with different state intervals, and each interval model comprises an interval variable structure control law and an interval disturbance response function.

[0024] The carbon cost function construction module is used for fitting the interval model, calculating real-time disturbance marginal elasticity, and constructing a marginal carbon emission cost function according to the real-time disturbance marginal elasticity.

[0025] The real-time regulation module is used for real-time regulation of carbon emission quota by the interval model and the marginal carbon emission cost function.

[0026] As can be seen from the above scheme, the application constructs a disturbance dominant response function by a sparse identification dynamic modeling method, and forms a plurality of section interval response functions according to disturbance variable state division, effectively overcomes the problem of unstable response function estimation under high-dimensional disturbance sources, realizes accurate modeling of carbon emission change rate and state-driven segmented analysis, and improves modeling accuracy. The application introduces a variable structure control law and a sliding mode control strategy, finely divides the disturbance state interval and embeds a real-time feedback mechanism, can quickly respond in the case of severe nonlinear change or sudden mutation of interference disturbance of carbon emission, dynamically corrects a regulation amount, and realizes real-time tracking control of carbon emission quota. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 FIG. 1 is a flowchart of a real-time regulation method for carbon emission quota based on dynamic marginal cost according to the application;

[0028] Figure 2 FIG. 4 is a schematic diagram of DBSCAN clustering results of disturbance variables according to the application.

[0029] Figure 3 A schematic diagram of the fitting precision comparison results of the state interval disturbance response function of the present application;

[0030] Figure 4 A schematic diagram of the comparison results of the original carbon quota line and the carbon emission trend predicted by the LSTM of the present application;

[0031] Figure 5 A schematic diagram of the comparison results of the carbon emission trend curves of the three strategies of the present application;

[0032] Figure 6 A schematic diagram of the structure of the carbon emission quota real-time regulation system based on dynamic marginal cost of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0034] Embodiment 1, Figure 1 The present application provides a carbon emission quota real-time regulation method based on dynamic marginal cost, comprising the following steps:

[0035] S1, collecting real-time disturbance data of the construction site, identifying the current disturbance state, and mapping the disturbance state to a predefined state interval.

[0036] In the embodiment, the collection of real-time disturbance data of the construction site, the identification of the current disturbance state, and the mapping of the disturbance state to a predefined state interval are specifically as follows:

[0037] Collecting real-time disturbance data of the construction site, identifying the current disturbance state;

[0038] Calculating the distance between the current disturbance state and the clustering center of each predefined disturbance state interval, and mapping it to the corresponding state interval according to the minimum distance principle.

[0039] In the embodiment, the collection of real-time disturbance data of the construction site, the identification of the current disturbance state, and the mapping of the disturbance state to a predefined state interval are specifically as follows:

[0040] Collecting construction site data with a time span of several weeks to several months, including disturbance variables of material dimension, equipment dimension, environment dimension, and labor dimension, and carbon emission change rate corresponding to time, and processing the collected data into a structured data matrix

[0041] Each perturbation variable dimension is normalized by zero mean unit variance, abnormal perturbation behavior points are removed by Z-score method, and standardized samples are obtained to avoid variable scale difference from affecting clustering results.

[0042] The standardized samples of the perturbation variables are used as input vectors to perform DBSCAN clustering on the sample set to obtain a plurality of state intervals, and each state interval is represented by a sample mean vector as an interval center.

[0043] In the embodiment, the distance between the current perturbation state and the clustering center of each predefined perturbation state interval is calculated, and it is mapped to the corresponding state interval according to the minimum distance principle, specifically:

[0044] In actual operation, for the perturbation state obtained at any time, the Mahalanobis distance between the perturbation state and each state interval center is calculated, and the real-time perturbation state is mapped to the state interval with the smallest distance to complete real-time state interval identification.

[0045] The state interval mapping process provides prior positioning for subsequent interval model loading and control law calling, reduces the complexity of the state space, compresses the continuous high-dimensional perturbation state into a limited interval, and enhances the model calling efficiency. Each interval corresponds to a variable structure response model, which is convenient for real-time calling, supports dynamic updating, can periodically or conditionally trigger re-clustering, and adapts to construction phase switching.

[0046] S2, based on the mapped state interval, calling the corresponding interval model in the pre-constructed carbon emission variable structure perturbation response model; wherein the variable structure perturbation response model is composed of a plurality of interval models associated with different state intervals, and each interval model includes an interval variable structure control law and an interval perturbation response function.

[0047] In the embodiment, the interval perturbation response function further includes determining an optimal sparsity penalty coefficient, and the specific steps are as follows:

[0048] Taking the carbon emission change rate as the target response quantity, a function base library is constructed based on the perturbation variables;

[0049] A candidate set of sparsity penalty coefficients is set, and a cross-validation strategy is used to evaluate the fitting error under different sparsity parameters;

[0050] The sparsity penalty coefficient corresponding to the minimum fitting error is selected as the optimal sparsity penalty coefficient.

[0051] In the embodiment, the function base library is constructed based on the perturbation variables with the carbon emission change rate as the target response quantity, specifically:

[0052] A function base library including a first-order term, a second-order term, a nonlinear term and a constant term is constructed, and a target response variable of a carbon emission change rate is fitted as a weighted sum of the function base library;

[0053] A plurality of sparsity penalty coefficients are set, and a SINDy regression algorithm is used to solve sparse weights, each sparsity penalty coefficient corresponding to a group of sparse solutions, and retaining different numbers of base terms.

[0054] In the embodiment, a candidate set of sparsity penalty coefficients is set, and a cross-validation strategy is used to evaluate fitting errors under different sparsity parameters, specifically:

[0055] A data set composed of historical collected construction disturbance variables and carbon emission change rates is divided into a training set and a validation set, and a K-fold cross-validation method is used, that is, the data is divided into K parts, one part is taken as a validation set each time, and the remaining parts are taken as a training set, and the process is repeated K times;

[0056] A plurality of candidate sparsity penalty coefficients are set, and for each candidate sparsity penalty coefficient, a response function model is obtained on the training set using a sparse regression algorithm; then the model is tested on the validation set to evaluate its accuracy in predicting the carbon emission change rate. The evaluation standard is the mean square value of the prediction error, and the smaller the error, the better the modeling effect of the sparsity penalty coefficient.

[0057] The sparsity penalty coefficient corresponding to the minimum fitting error is selected as the optimal sparsity penalty coefficient, specifically: the errors of the models corresponding to all candidate sparsity penalty coefficients on the validation set are compared, and the sparsity penalty coefficient that minimizes the prediction error is selected as the optimal sparsity penalty coefficient for final modeling.

[0058] The purpose of the cross-validation process is to accurately identify the truly effective disturbance variable combination in the actual carbon emission scene with a large number of complex disturbance factors, to ensure the accuracy of the modeling result, to effectively eliminate redundant features, to reduce the complexity of the model, and to improve the real-time control efficiency.

[0059] The interval disturbance response function is obtained by a sparse regression algorithm, specifically:

[0060] Based on the optimal sparsity penalty coefficient, the coefficients of the function base library are solved by a sparse regression algorithm, and the corresponding function base terms are selected according to the non-zero coefficients;

[0061] The function base terms are used to generate a disturbance dominant response function expression for the current interval;

[0062] For a plurality of different disturbance state intervals, corresponding interval disturbance dominant response functions are generated respectively;

[0063] The disturbance dominant response functions of all intervals are aggregated to build an interval disturbance response function library.

[0064] The disturbance dominant response function is:

[0065]

[0066] The interval disturbance response function is:

[0067]

[0068] In the formula, is the carbon emission change rate, is the base term coefficient obtained by sparse regression, is the function base term, is the disturbance variable, is the effective function base term reserved in the i-th interval, is the disturbance response coefficient in the corresponding interval, is the i-th interval disturbance response function, and the value represents the i-th state interval and the current disturbance state , and the corresponding carbon emission model estimated value.

[0069] In this embodiment, the interval variable structure control law is specifically:

[0070] A preset global carbon quota plan is obtained, and an interval carbon emission state deviation function is built based on the plan and real-time carbon emission state data.

[0071] Based on the sliding mode control theory, a convergence target surface of the control system is defined, wherein the convergence target surface aims to make the interval carbon emission state deviation function approach zero.

[0072] According to the convergence target surface, an interval variable structure control law for driving the carbon emission state to track the global carbon quota is derived to obtain the interval variable structure control law.

[0073] In this embodiment, a preset global carbon quota plan is obtained, and an interval carbon emission state deviation function is built based on the plan and real-time carbon emission state data, specifically:

[0074] Through the carbon emission control plan provided by the construction management system or the competent department, a carbon emission quota planning table of the construction project in the whole cycle is obtained. The table usually includes carbon emission quota values of each construction stage (such as foundation excavation, main structure construction, equipment installation, etc.), construction time intervals, process types and carbon emission intensities, and carbon control constraint parameters such as monthly carbon quota or critical emission red line.

[0075] Real-time collection of current stage, construction progress (e.g. completion percentage of the stage where the current date is located), and current cumulative carbon emissions from data sources such as energy consumption monitoring systems, material management platforms, or BIM progress models deployed on site.

[0076] According to the state interval to which the current disturbance state belongs, the carbon quota allocation value corresponding to the state interval is extracted from the global carbon quota plan, and the target carbon emission baseline value at the current time is calculated according to the current progress percentage, i.e. the product of the state interval quota and the current progress percentage.

[0077] A carbon emission state deviation function is defined to quantify the difference between actual carbon emissions and planned carbon emissions, and the function value represents the carbon emission deviation under the current state. If it is greater than zero, it means that the carbon emissions exceed the current quota baseline, and timely inhibitory control needs to be performed; if it is less than zero, it means that the current carbon emissions are relatively generous, and flexible optimization can be considered; if it is equal to zero, it is in the balanced state of the control target. This deviation function will be used to define the carbon emission convergence target surface in the subsequent sliding mode control law construction, and is a key quantitative index for realizing closed-loop regulation of partitioned carbon emissions.

[0078] In this embodiment, based on the sliding mode control theory, a convergence target surface of the control system is defined, wherein the convergence target surface aims to make the interval carbon emission state deviation function approach zero; according to the convergence target surface, an interval variable structure control law is derived to drive the carbon emission state to track the global carbon quota, and the interval variable structure control law is obtained, specifically:

[0079] The carbon emission state deviation function constructed in the foregoing is taken as the main control state variable of the sliding mode control system, which is used to reflect the carbon emission error under the current disturbance state, a sliding surface function is defined, and relevant parameters are dynamically obtained through machine learning to balance the current carbon emission deviation and the trend of carbon emission deviation. The sliding surface constitutes a convergence target surface, which requires the actual regulation trajectory of the system to rapidly approach the surface and move along it, so as to design an interval variable structure control law that meets the sliding mode approach condition, and the form of the control law is constructed according to the following logic:

[0080] When the sliding surface function value > 0, the carbon emissions exceed, the control law should guide the regulation amount to increase in the negative direction to suppress the carbon emissions;

[0081] When the sliding surface function value < 0, the carbon emissions are insufficient, the control law should relax the regulation to release the regulation boundary;

[0082] When the sliding surface function value = 0, the control is on the sliding surface, and the regulation amount remains unchanged.

[0083] The corresponding sliding surface parameters, control gain coefficients and target functions of each block are stored as variable structure control law submodules for real-time stage calling and regulation.

[0084] wherein the state deviation function is:

[0085]

[0086] The variable structure control law is:

[0087]

[0088] In the formula, is a deviation function, is the upper limit of the carbon emission quota for the current period, is a variable structure control correction instruction, is a sign function, which ensures that the state deviation approaches zero, is a control gain coefficient.

[0089] S3, fitting the interval model, calculating the real-time disturbance marginal elasticity, and constructing a marginal carbon emission cost function according to the real-time disturbance marginal elasticity.

[0090] In this embodiment, the interval model is fitted, the real-time disturbance marginal elasticity is calculated, and the marginal carbon emission cost function is constructed according to the real-time disturbance marginal elasticity, specifically:

[0091] Obtain the interval disturbance response function of the target system, and perform fitting operation at the current disturbance state point to obtain the disturbance marginal carbon emission elasticity;

[0092] According to the historical data set, the control weight of the current corresponding disturbance variable is obtained through the machine learning model;

[0093] Based on the disturbance marginal carbon emission elasticity and the control weight, a marginal carbon emission cost function for carbon emission optimization decision is constructed.

[0094] In this embodiment, specifically: at the current disturbance state point, a small disturbance is added to each dimension of the disturbance variable respectively, and a group of micro-disturbance sample points are constructed;

[0095] For each disturbance direction, the corresponding interval disturbance response function is inputted to obtain the corresponding carbon emission change rate estimate, and the current disturbance marginal carbon emission elasticity is obtained in the form of difference quotient.

[0096] Using the historical carbon emission data set, a sample set of disturbance variables and actual control feedback is constructed; a weighted regression model is used to train the above samples, and the control weight of each disturbance variable under the current state interval is outputted;

[0097] Based on the disturbance marginal carbon emission elasticity and the control weight of the current disturbance state point, a marginal carbon emission cost function is defined.

[0098] The disturbance marginal carbon emission elasticity is specifically as follows:

[0099]

[0100] The marginal carbon emission cost function is specifically as follows:

[0101]

[0102] In the formula, is the disturbance marginal carbon emission elasticity, is the marginal carbon emission cost, is the regulation weight.

[0103] S4, real-time regulation of carbon emission quota is performed through the interval model and the marginal carbon emission cost function.

[0104] In the embodiment, the real-time regulation of carbon emission quota through the interval model and the marginal carbon emission cost function further comprises obtaining a preliminary optimized regulation amount of each disturbance variable, and specifically comprises:

[0105] Based on the preset global carbon quota plan and a historical data set, a carbon emission regulation upper limit constraint and a disturbance variable boundary constraint are generated through a machine learning model;

[0106] A real-time optimization model containing the emission regulation upper limit constraint and the disturbance variable boundary constraint is constructed with the marginal carbon emission cost function as a target, the optimization model is solved in real time, and the preliminary optimized regulation amount of each disturbance variable is output.

[0107] In the embodiment, the carbon emission regulation upper limit constraint and the disturbance variable boundary constraint are generated based on the preset global carbon quota plan and the historical data set through the machine learning model, and specifically comprises:

[0108] The theoretical upper limit control line under different time windows is built in combination with the project overall carbon quota target, construction cycle division, and stage carbon budget structure, and is cross-corrected with historical regulation execution values to obtain a fitted time distribution function, the influences of construction season, equipment start-stop cycle, and meteorological factors on the carbon emission base value are considered, a regression neural network is used to perform short-term prediction on the carbon emission trend, the regulation upper limit of the current period is corrected, and the real-time carbon emission regulation upper limit constraint is obtained.

[0109] With the historical regulation trajectory of the disturbance variable as a sample, the safe operation interval of each variable under different state intervals is counted by using cluster analysis in combination with the actual engineering log, a decision tree model is used to establish a variable boundary predictor, the current disturbance state is taken as input, and the minimum and maximum regulation amplitudes that can be accepted are output, and the real-time disturbance variable boundary constraint is obtained.

[0110] The real-time regulation of carbon emission quota through the interval model and the marginal carbon emission cost function is specifically as follows:

[0111] The interval variable structure control law corresponding to the current disturbance state interval is called, and a sliding mode control correction instruction is constructed based on a carbon emission state deviation function;

[0112] The sliding mode control correction instruction is linearly superimposed with the externally input preliminary optimization control amount;

[0113] The superposition result is output as the final control output for driving carbon emission control.

[0114] Through the above control law correction mechanism, the application realizes a dynamic control correction process based on disturbance state adaptive interval matching and sliding mode driven control compensation, effectively improves the stability and real-time performance of the carbon emission control strategy, and is particularly suitable for high dynamic construction scenes such as power grid construction with frequent carbon disturbance and significant nonlinear fluctuation.

[0115] The upper limit constraint of carbon emission control is:

[0116]

[0117] The disturbance variable boundary constraint is:

[0118]

[0119] Optimization model:

[0120]

[0121] In the formula, is a disturbance variable control vector, is the adjustment amount of the jth disturbance variable, is the upper limit of the carbon emission quota of the current period, is the minimum adjustment amplitude, is the maximum adjustment amplitude, is all disturbance variables.

[0122] Example 2, in order to verify the practicability and control precision of the application of “a real-time carbon emission quota control method based on dynamic marginal cost”, this embodiment is based on the typical project data of power grid construction engineering, constructs an example test environment, and compares and analyzes the performance of this method in disturbance identification, response modeling, control effect, etc. The specific is as follows:

[0123] 1) Test system and data setting

[0124] Taking a certain power transmission and transformation project construction stage as the background, the disturbance and carbon emission historical data of its construction site for three months (12 weeks) are selected as the modeling samples, the disturbance variables include four categories, a total of 8 dimensions, as shown in Table 1.

[0125] Table 1

[0126]

[0127] Corresponding to each record, the carbon emission change rate (kgCO2 / h) is monitored as the modeling target.

[0128] After removing abnormal disturbance values by Z-score method, DBSCAN clustering analysis is performed on the standardized samples to divide them into 4 disturbance state intervals, and the disturbance center mean vector (denoted as state 1~4) is constructed, as shown in Figure 2 The horizontal and vertical coordinates are the first two principal component dimensions extracted after principal component analysis (PCA) of the original disturbance variable data.

[0129] 2) Interval response model construction and marginal elasticity extraction

[0130] Based on the sample set of each state interval, sparse response models are constructed, and SINDy method is used to select effective items (including linear, quadratic and cross nonlinear terms) from the basic function library to determine the interval disturbance response function, as shown in Table 2.

[0131] Table 2

[0132]

[0133] The fitting accuracy of the disturbance response function of each state interval is compared as shown in Figure 3 , where the X-axis is the state interval and the Y-axis is the relative mean square error (RMSE) of the validation set. It can be seen from the figure that the fitting effect of state 4 is the best.

[0134] Under the current disturbance state (construction date is the 65th day, classified into state interval 3), the marginal carbon emission elasticity of disturbance is calculated by perturbation method, and the results are shown in Table 3:

[0135] Table 3

[0136]

[0137] 3) Design of carbon emission control upper limit prediction module

[0138] To improve the timeliness and robustness of the control model, this embodiment introduces a machine learning model to predict the future short-term carbon emission trend, and dynamically adjusts the control upper limit based on the carbon quota baseline as one of the input parameters of the optimization model, the specific steps are as follows:

[0139] Step 1: Data source and preprocessing

[0140] Collect daily carbon emission total (kgCO2 / day) and disturbance variables (after normalization) from day 1 to day 65; use sliding window to build time series sample set, every 10 days as an input sequence, predict the next day's emission value.

[0141] Step two, prediction model and parameter setting:

[0142] Model structure: Long Short-Term Memory (LSTM) neural network and a fully connected layer;

[0143] Loss function: Mean Squared Error (MSE);

[0144] Optimizer: Adam;

[0145] Training rounds: 200 rounds.

[0146] Step three, prediction results and regulation correction strategy:

[0147] The comparison results of LSTM prediction of carbon emission trend and original carbon quota line are shown in Figure 4 The predicted value and the corrected upper limit value are shown in Table 4, and the regulation correction strategy is as follows: compare the predicted value with the stage carbon quota baseline provided by the construction plan; if the predicted value is greater than the threshold value of 80%, reduce the upper limit of the current period (tend to be conservative); if the predicted value is less than the threshold value of 60%, relax the carbon emission space (improve flexibility); finally generate the dynamic upper limit value at the current time.

[0148] Table 4

[0149]

[0150] 4) Disturbance variable boundary constraint modeling

[0151] Based on historical regulation trajectory and engineering log, for state interval 3, train decision tree boundary predictor, output disturbance variable boundary under current state, as shown in Table 5.

[0152] Table 5

[0153]

[0154] 5) Real-time regulation optimization model solution

[0155] According to the carbon emission regulation upper limit of the 67th day predicted and corrected by the LSTM model and the disturbance variable boundary under state interval 3, the following real-time optimization model is constructed:

[0156] Objective function:

[0157]

[0158] Constraint conditions:

[0159] Carbon emission does not exceed the quota upper limit =2.10 kgCO2 / h;

[0160] Each disturbance variable adjustment boundary ;

[0161] The solution results are shown in Table 6:

[0162] Table 6

[0163]

[0164] 6) Sliding mode control law correction mechanism and final regulation fusion

[0165] The carbon emission target at the current state is = 2.10, the model prediction value under state interval 3 is = 2.22, and thus ;

[0166] The final control correction result based on the sliding mode control law is shown in Table 7.

[0167] Table 7

[0168]

[0169] The final simulation result shows that the carbon emission adjustment value is 1.98 kgCO2 / h, which is successfully controlled below the upper limit of the quota, and the total regulation cost is minimized.

[0170] 7) Comparative analysis

[0171] This embodiment compares the regulation performance of the three strategies under the same disturbance state, and the results are shown in Table 8, Figure 5 is the carbon emission trend curve of the three strategies, which shows that the strategy of the present application quickly approaches the quota line and has higher stability.

[0172] Table 8

[0173]

[0174] In summary, the carbon emission quota real-time regulation method based on dynamic marginal cost proposed by the present application can realize fine regulation of disturbance variables and dynamic optimization of carbon emission allocation under the premise of ensuring global carbon quota constraint. By introducing disturbance state interval modeling, marginal cost function construction and sliding mode control law correction mechanism, not only the accuracy of carbon emission response and the real-time performance of regulation strategy are significantly improved, but also the adaptability of the system to nonlinear disturbance is enhanced.

[0175] Embodiment 3, Figure 6 is a carbon emission quota real-time regulation system based on dynamic marginal cost, which includes a state interval mapping module, an interval model acquisition module, a carbon cost function construction module and a real-time regulation module;

[0176] The state interval mapping module is configured to collect real-time disturbance data of the construction site, identify a current disturbance state, and map the disturbance state to a predefined state interval.

[0177] The interval model acquisition module is configured to call a corresponding interval model in a pre-constructed carbon emission variable structure disturbance response model based on the mapped state interval, wherein the variable structure disturbance response model is composed of a plurality of interval models associated with different state intervals, and each interval model includes an interval variable structure control law and an interval disturbance response function.

[0178] The carbon cost function construction module is configured to fit the interval model, calculate real-time disturbance marginal elasticity, and construct a marginal carbon emission cost function according to the real-time disturbance marginal elasticity.

[0179] The real-time regulation module is configured to perform real-time regulation of carbon emission quotas through the interval model and the marginal carbon emission cost function.

[0180] The above formulas are all dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0181] The above embodiments can be realized wholly or partially by software, hardware, firmware, or any combination thereof. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product.

[0182] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solutions. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0183] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically, or two or more modules can be integrated in one module.

[0184] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0185] Finally: the above only for the preferred embodiments of the present application, and not for limiting the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the scope of protection of the present application.

Claims

1. A dynamic marginal cost-based carbon emission quota real-time regulation method, characterized in that, The method comprises the following steps: Collecting real-time disturbance data of the construction site, identifying the current disturbance state, and mapping the disturbance state to a predefined state interval; Based on the mapped state interval, the corresponding interval model in the pre-constructed carbon emission variable structure disturbance response model is called; wherein the variable structure disturbance response model is composed of multiple interval models associated with different state intervals, and each interval model includes an interval variable structure control law and an interval disturbance response function; The interval disturbance response function includes determining the optimal sparsity penalty coefficient, specifically: Taking the carbon emission change rate as the target response quantity, a function base library is constructed based on the disturbance variable; Set the sparsity penalty coefficient candidate set, and evaluate the fitting error under different sparsity parameters using the cross-validation strategy; Select the sparsity penalty coefficient corresponding to the minimum fitting error as the optimal sparsity penalty coefficient; The interval variable structure control law is specifically: Obtain the preset global carbon quota plan, and based on the plan and real-time carbon emission state data, construct an interval carbon emission state deviation function; Based on the sliding mode control theory, define the convergence target surface of the control system, wherein the convergence target surface aims to make the interval carbon emission state deviation function approach zero; According to the convergence target surface, the interval variable structure control law driving the carbon emission state to track the global carbon quota is derived to obtain the interval variable structure control law; Fitting the interval model, calculating the real-time disturbance marginal elasticity, and constructing the marginal carbon emission cost function according to the real-time disturbance marginal elasticity; Real-time regulation and control of carbon emission quota through interval model and marginal carbon emission cost function: calling the interval variable structure control law corresponding to the current disturbance state interval, constructing the sliding mode control correction instruction based on the carbon emission state deviation function; linearly superimpose the sliding mode control correction instruction and the externally input preliminary optimization control quantity; output the superposition result as the final control output driving carbon emission regulation and control.

2. The method of claim 1, wherein, The collection of real-time disturbance data of the construction site, the identification of the current disturbance state, and the mapping of the disturbance state to the predefined state interval are specifically: Collecting real-time disturbance data of the construction site, identifying the current disturbance state; Calculate the distance between the current disturbance state and the clustering center of each predefined disturbance state interval, and map it to the corresponding state interval according to the minimum distance principle.

3. The method of claim 1, wherein, The interval disturbance response function is obtained by sparse regression algorithm, specifically: Based on the optimal sparsity penalty coefficient, the coefficient of the function base library is solved by sparse regression algorithm, and the corresponding function base item is selected according to the non-zero coefficient; Use the function base item to generate the disturbance dominant response function expression of the current interval; For multiple different disturbance state intervals, the corresponding interval disturbance dominant response function is generated; Aggregate the disturbance dominant response functions of all intervals to construct the interval disturbance response function library.

4. The method of claim 1, wherein, The fitting of the interval model, the calculation of the real-time disturbance marginal elasticity, and the construction of the marginal carbon emission cost function according to the real-time disturbance marginal elasticity are specifically: Obtain the interval disturbance response function of the target system, and perform fitting operation at the current disturbance state point to obtain the disturbance marginal carbon emission elasticity; According to the historical data set, the control weight of the current corresponding disturbance variable is obtained through the machine learning model; Based on the disturbance marginal carbon emission elasticity and the regulation weight, a marginal carbon emission cost function for carbon emission optimization decision is constructed.

5. The method of claim 4, wherein, The real-time regulation of the carbon emission quota through the interval model and the marginal carbon emission cost function further includes obtaining a preliminary optimization regulation amount of each disturbance variable, specifically: Based on the preset global carbon quota planning and historical data set, a carbon emission regulation upper limit constraint and a disturbance variable boundary constraint are generated through a machine learning model; With the marginal carbon emission cost function as the target, a real-time optimization model containing the emission regulation upper limit constraint and the disturbance variable boundary constraint is constructed, the optimization model is solved in real time, and the preliminary optimization regulation amount of each disturbance variable is output.

6. A system for real-time regulation using the dynamic marginal cost based carbon emission quota of any one of claims 1-5, characterized in that, It includes a state interval mapping module, an interval model acquisition module, a carbon cost function construction module, and a real-time regulation module. The state interval mapping module is used to collect real-time disturbance data of the construction site, identify the current disturbance state, and map the disturbance state to a predefined state interval. The interval model acquisition module is used to call the corresponding interval model in the pre-constructed carbon emission variable structure disturbance response model based on the mapped state interval; wherein the variable structure disturbance response model is composed of multiple interval models associated with different state intervals, and each interval model includes an interval variable structure control law and an interval disturbance response function. The carbon cost function construction module is used to fit the interval model, calculate the real-time disturbance marginal elasticity, and construct the marginal carbon emission cost function according to the real-time disturbance marginal elasticity. The real-time regulation module is used to regulate the carbon emission quota in real time through the interval model and the marginal carbon emission cost function.

Images