Search

CN-122022130-A - Method, system and medium for constructing carbon-to-carbon neutralization path evaluation model

CN122022130ACN 122022130 ACN122022130 ACN 122022130ACN-122022130-A

Abstract

The invention relates to the technical field of carbon emission management, in particular to a method, a system and a medium for constructing a carbon arrival peak carbon neutralization path evaluation model, comprising the following steps of S1, collecting multi-source time sequence data such as energy consumption, carbon emission, technical adoption, carbon sink change and the like in different periods, fusing multi-field dynamic characteristic embedding, and constructing a carbon path evolution memory map reflecting causal evolution relations among stages; S2, calculating path tensor similarity and tension indexes based on the atlas, identifying jump windows, judging specified execution strength and structural adjustment hysteresis, and outputting a control empty window area set, S3, constructing a carbon evolution driving boundary network, training a path evaluation model, and outputting a path sensitivity atlas and a regulation strategy suggestion. According to the invention, by constructing the carbon path evolution memory map and the carbon evolution driving boundary network, the accurate identification and path sensitivity analysis of the control empty window area are realized, so that the intelligent regulation strategy support is provided for the carbon-to-peak carbon neutralization path.

Inventors

  • LI QUANWEI
  • LIN XIAOHE
  • SUN YUNXIA

Assignees

  • 河北登望电力工程有限公司

Dates

Publication Date
20260512
Application Date
20251231

Claims (10)

  1. 1. The method for constructing the carbon-to-carbon neutralization path evaluation model is characterized by comprising the following steps of: S1, collecting multi-source time sequence data of energy consumption, carbon emission, specified requirements, technical adoption and carbon sink change in different periods, and constructing a carbon path evolution memory map for expressing causal evolution relations among path stages by combining a multi-field dynamic characteristic embedding mechanism; S2, calculating tensor similarity and influence tension indexes of adjacent stages in path evolution based on the carbon path evolution memory map, identifying a path jump window through a dynamic mutation detection method, judging regulation hysteresis of specified execution strength and structure, and outputting a control empty window area set with missing intervention mechanism or response delay; S3, tracking upstream influence factors and downstream path responses of the control empty window area set based on the control empty window area set, constructing a carbon evolution driving boundary network crossing the fields of energy, traffic and industry, training a path evaluation model based on the carbon evolution driving boundary network, and outputting path sensitivity maps and jump regulation strategy suggestions under different intervention combinations.
  2. 2. The method of constructing a carbon-to-carbon neutralization pathway evaluation model according to claim 1, wherein S1 comprises: S11, extracting multi-source time sequence data from an energy statistics annual survey, a carbon emission accounting list, an industry regulation file library, a green technology adoption report and a remote sensing carbon sink monitoring system, wherein the multi-source time sequence data comprises energy consumption Carbon emission Specifying a required index Technological adoption rate Carbon sink capacity And performing a timing alignment process on the multi-source timing data; S12, encoding multi-source time sequence data of different sources into a unified vector form to form a high-dimensional path state representation, and setting the total length of a time sequence as The total number of areas is Constructing a path state vector ; S13, embedding and modeling the path state vector by adopting GRU network, and outputting a carbon path memory embedding sequence ; S14, carbon path memory embedded sequence based on GRU network output Constructing path state embedded vectors by adopting attention mechanism Embedding vectors into path states Evolution side weight ; S15, fusing nodes and evolution side weights of the carbon path memory embedded sequence to construct a carbon path evolution memory map , wherein, For a set of nodes, Is a collection of edges.
  3. 3. The method of constructing a carbon-to-carbon neutralization pathway evaluation model according to claim 2, wherein S2 comprises: s21, based on the carbon path evolution memory map, comparing and analyzing path states of adjacent time phases, and when the state change amplitude between two continuous phases exceeds a state change threshold value and the change trend continuously exists, identifying the time zone as a path jump window; S22, comparing and analyzing the regulated execution condition corresponding to the stage with the regulation response of the industrial structure and the energy structure in the identified path jump window, judging whether the regulated execution is synchronous with the path change, and marking the time section as a control empty window area if the regulated execution intensity is not regulated correspondingly in more than two continuous time periods after the path jump window is started or the change of the structural variable is delayed from the path jump time point to exceed one time period and does not reach the response threshold value.
  4. 4. The method for constructing a carbon-to-carbon neutralization pathway evaluation model as set forth in claim 3, wherein said S21 comprises: s211, memorizing the embedded sequence of the carbon path in the carbon path evolution memory map The adjacent state vector in the path is used for calculating the path state change amplitude by adopting the cosine distance As a measure of path evolution intensity; s212, a change amplitude sequence according to the change amplitude of the path state Trend analysis is carried out to judge whether the continuous two or more time steps are satisfied And is also provided with , wherein, Is a state change threshold; S213 for satisfying And is also provided with Is a continuous period of time of (a) The time index set corresponding to the time zone is collected Marked as a path hop window.
  5. 5. The method of constructing a carbon-to-carbon neutralization pathway evaluation model as set forth in claim 4, wherein said S22 includes: S221, jumping the window in the identified path In, extract and path hopping window Corresponding time series of specified execution intensity, and setting the specified execution intensity index as Calculating the variation of the specified execution intensity before and after the starting point of the path jump window When the path jump window starts, the method satisfies the following conditions in more than two continuous time periods And determining that the specified execution intensity is not correspondingly adjusted in the path jump stage, recording the specified execution intensity as a specified response missing state, wherein, A starting time point of the path jump window; S222, extracting time sequences of structural variables of the industrial structure and the energy structure in the same path jump window, and setting the structural variables as Calculate the variation amplitude If the first effective change time of the structural variable Satisfy the following requirements Or at any point in time within the path hopping window A kind of electronic device It is determined that the structural adjustment response is insufficient, wherein, Is a structural response threshold; S223, combining the detection result of the specified execution intensity change and the analysis result of the structural variation response hysteresis, and when the specified execution intensity is not changed in more than two continuous time periods or the adjustment of the structural variable lags the path jump time point by more than one time period and the variation amplitude does not reach the structural response threshold value in the same path jump window, comparing the corresponding path jump window Marked as control empty window areas and outputting a set of control empty window areas.
  6. 6. The method of constructing a carbon-to-carbon neutralization pathway evaluation model as set forth in claim 5, wherein said S3 comprises: S31, based on the identified control empty window area set, extracting structural variables related to energy supply, transportation and industrial activities, and establishing a carbon evolution driving boundary network with the control empty window area as a node core and the structural variables as edges by analyzing a time sequence offset relation and response weight between the structural variables and path state changes; S32, constructing a path evaluation model based on the constructed carbon evolution driving boundary network, outputting path sensitivity maps under different intervention combinations, identifying key influence factors, sensitive path stages and high-risk regulation breakpoints, and generating jump regulation strategy suggestions.
  7. 7. The method of constructing a carbon-to-carbon neutralization pathway evaluation model as set forth in claim 6, wherein said S31 includes: s311, based on the identified control empty window area set Extracting structural variables of multiple departments of energy supply, transportation and industrial activities in a corresponding time period of each control empty window area, wherein, Representation department At the time of Is the first of (2) A structural feature value; S312, for each control empty window area Constructing path state embedded vectors And structural characteristic value Is identified by cross hysteresis window method Optimal response hysteresis with path state transitions And calculating a response weight of the structural variable to the path state change based on the correlation ; S313, controlling the time period of the empty window area For core nodes of the network, in response to the weights Is greater than the drive weight threshold and lags behind the optimal response And constructing a weighted directed graph by taking the structural variable with the corresponding relation with the path state as an edge connection source, and finally forming a carbon evolution driving boundary network.
  8. 8. The method of constructing a carbon-to-carbon neutralization pathway evaluation model as set forth in claim 7, wherein said S32 includes: S321, extracting a path state embedded vector corresponding to each control empty window area as a target output based on a carbon evolution drive boundary network, and outputting a structural characteristic value As input, constructing a path evaluation model by adopting a graph neural network, and training the path evaluation model by taking a path state prediction error as a loss function of a target; S322, calculating the sensitivity of each structural variable to path state change by adopting gradient back propagation according to the predicted path state embedded vector output by the trained path evaluation model Generating a path sensitivity map by counting and visualizing sensitivities of different time periods, different path stages and different structural variables; S323, focusing on a high-sensitivity structural variable and a high jump risk node corresponding to a control empty window area based on a path sensitivity map, and outputting jump regulation strategy suggestions by combining the current actual regulation state of the high-sensitivity structural variable and the high jump risk node, wherein the method specifically comprises the following steps: A priority intervention factor list, which lists structural variables needing to immediately improve the regulation and control force; recommending an optimal intervention time window; And the path jump early warning prompt is used for sending out jump early warning to the nodes with high risk trend in the path evaluation model prediction so as to assist decision makers in deployment in advance.
  9. 9. A system for constructing a carbon-to-carbon neutralization pathway evaluation model for implementing a method for constructing a carbon-to-carbon neutralization pathway evaluation model as defined in any one of claims 1-8, comprising the following modules: The data acquisition and fusion module acquires and fuses multi-source time sequence data of energy consumption, carbon emission, specified requirements, low-carbon technology adoption and carbon sink change in different periods, and performs multi-field dynamic characteristic embedding processing to generate a carbon path evolution memory map; The jump recognition and empty window diagnosis module is used for calculating tensor similarity and influencing tension indexes between adjacent sections of the path based on the carbon path evolution memory map, detecting and recognizing a path jump window through dynamic mutation, and outputting a control empty window region set with missing intervention mechanisms or response delay; the boundary network construction and model training module is used for constructing a carbon evolution driving boundary network covering the multiple fields of energy, traffic and industry based on the control air window area set and training a path evaluation model; and the output analysis module is used for outputting a path sensitivity map and jump regulation strategy suggestion based on the path evaluation model.
  10. 10. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the method of constructing a carbon-to-peak carbon neutralization path evaluation model of any of claims 1-8.

Description

Method, system and medium for constructing carbon-to-carbon neutralization path evaluation model Technical Field The invention relates to the technical field of carbon emission management, in particular to a method, a system and a medium for constructing a carbon-to-carbon neutralization path evaluation model. Background Along with the gradual advancement of global carbon peaks and carbon neutralization targets, transformation paths in the fields of high carbon emission such as energy, industry, traffic and the like become important attention objects for prescribing and making. For scientific planning and dynamic evaluation of the effects of various carbon emission reduction regulations, the time sequence tracking and path evolution modeling capability of the carbon emission process is continuously enhanced in each country, and the aim of identifying key intervention nodes influencing the carbon emission reduction progress is fulfilled through the causal relationship between the historical data mining regulation execution and the carbon emission, so that the regulation combination and execution strategy is optimized. Most of the existing carbon path evaluation methods are based on static data analysis or macroscopic statistical modeling, and are difficult to accurately capture the accumulated effect and the hysteresis response of variables such as specified execution, technical adoption, structural adjustment and the like in different time windows, and meanwhile, effective recognition means are not available for key dynamic events such as mutation, jump and the like in the path evolution process, so that the understanding of a driving mechanism behind carbon emission change is not deep enough. In addition, most of current path sensitivity analysis adopts univariate regression or principal component analysis, and interference strategy visualization and intelligent recommendation under multi-domain and multi-causal boundaries are difficult to support. Disclosure of Invention The invention provides a method, a system and a medium for constructing a carbon reaching peak carbon neutralization path evaluation model, wherein the system is used for describing causal and cumulative effects among carbon emission path stages by fusing multisource time sequence data, a dynamic characteristic embedding mechanism and a carbon path evolution memory map, identifying a path jump window by utilizing tensor similarity and tension indexes, accurately positioning response lag and defining a blank window area, further introducing a graph neural network and a boundary driving mechanism, training the path evaluation model, outputting a high-sensitivity variable and key regulation strategy map, and improving the accuracy of carbon neutralization path analysis and the scientificity of defining intervention. The method for constructing the carbon-to-carbon neutralization path evaluation model comprises the following steps of: S1, collecting multi-source time sequence data of energy consumption, carbon emission, regulation requirements, technical adoption and carbon sink change in different periods, and constructing a carbon path evolution memory map of causal evolution relation between expression path stages by combining a multi-field dynamic characteristic embedding mechanism, wherein the carbon path evolution memory map is used for retaining the accumulation effect of regulation and structure adjustment; S2, calculating tensor similarity and influence tension indexes of adjacent stages in path evolution based on the carbon path evolution memory map, identifying a path jump window through a dynamic mutation detection method, judging regulation hysteresis of specified execution strength and structure, and outputting a control empty window area set with missing intervention mechanism or response delay; S3, tracking upstream influence factors and downstream path responses of the control empty window area set based on the control empty window area set, constructing a carbon evolution driving boundary network crossing the fields of energy, traffic and industry, training a path evaluation model based on the carbon evolution driving boundary network, and outputting path sensitivity maps and jump regulation strategy suggestions under different intervention combinations. Optionally, the S1 includes: S11, extracting multi-source time sequence data from an energy statistics annual survey, a carbon emission accounting list, an industry regulation file library, a green technology adoption report and a remote sensing carbon sink monitoring system, wherein the multi-source time sequence data comprises energy consumption Carbon emissionSpecifying a required indexTechnological adoption rateCarbon sink capacityAnd performing a timing alignment process on the multi-source timing data; S12, encoding multi-source time sequence data of different sources into a unified vector form to form a high-dimensional path state representation, and setting the total