Regional carbon emission flow analysis and management method based on deep reinforcement learning

Abstract

The invention discloses a regional carbon emission flow analysis and management method based on deep reinforcement learning, and relates to the technical field of carbon emission flow management, and the method comprises the steps: collecting carbon emission related data corresponding to carbon emission sources in a plurality of carbon emission regions, and carrying out the feature extraction of the carbon emission related data of each carbon emission region, the method comprises the following steps: acquiring a carbon emission characteristic data set of a corresponding carbon emission region, establishing a regional carbon emission flow analysis model through a deep reinforcement learning algorithm, inputting the carbon emission characteristic data set of each carbon emission region into the regional carbon emission flow analysis model, and monitoring respective carbon emission flows of a plurality of carbon emission regions in real time. And constructing a regional carbon emission flow scheduling model, and carrying out carbon emission flow scheduling optimization on the carbon emission region with abnormal carbon emission flow.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission flow management, and specifically to a regional carbon emission flow analysis and management method based on deep reinforcement learning. Background Art

[0002] As the global climate change problem becomes increasingly serious, carbon emissions have become the focus of attention of all countries. However, the existing carbon emission management methods have the following problems: first, the monitoring and analysis capabilities of carbon emission-related data are insufficient, and abnormal carbon emission flows are often not discovered; second, for carbon emission areas with abnormal carbon emission flows, there is a lack of effective carbon emission flow optimization and scheduling methods, resulting in regional carbon emissions not meeting standards. Therefore, it is of great significance to study an efficient regional carbon emission flow analysis and management method. Summary of the invention

[0003] In order to solve the above problems, the purpose of the present invention is to provide a regional carbon emission flow analysis and management method based on deep reinforcement learning.

[0004] The purpose of the present invention can be achieved by the following technical solution: A regional carbon emission flow analysis and management method based on deep reinforcement learning, comprising the following steps:

[0005] Step S1: collecting carbon emission related data corresponding to carbon emission sources in several carbon emission areas, and performing feature extraction on the carbon emission related data of each carbon emission area, thereby obtaining a carbon emission feature data set of the corresponding carbon emission area;

[0006] Step S2: Establish a regional carbon emission flow analysis model through a deep reinforcement learning algorithm, input the carbon emission characteristic data set of each carbon emission area into the regional carbon emission flow analysis model, and then monitor the carbon emission flows of several carbon emission areas in real time;

[0007] Step S3: Construct a regional carbon emission flow scheduling model, and optimize the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows.

[0008] Furthermore, the process of collecting carbon emission related data corresponding to carbon emission sources in several carbon emission areas includes:

[0009] Several carbon emission areas are labeled and recorded as i, i=1, 2, 3, ..., n, where n is a natural number greater than 0, and carbon emission monitoring equipment is arranged in each carbon emission area, and the carbon emission monitoring equipment collects carbon emission related data of the carbon emission sources in the corresponding carbon emission area;

[0010] The carbon emission monitoring equipment includes a gas analyzer, a gas flow meter and a remote sensing detection device;

[0011] The carbon emission related data include carbon emission concentration, carbon emission amount, carbon emission flow rate and gas composition;

[0012] The gas analyzer is used to collect the carbon emission concentration in each carbon emission area;

[0013] The gas flow meter is used to collect the carbon emission amount and carbon emission flow rate in each carbon emission area;

[0014] The remote sensing detection equipment is used to collect gas composition in each carbon emission area;

[0015] The carbon emission related data corresponding to the carbon emission area labeled i is recorded as T-pf[i], T-pf[i] = {C[i], D[i], V[i], Ω[i]};

[0016] Among them, C[i] represents the carbon emission concentration corresponding to the carbon emission area labeled i, D[i] represents the carbon emission amount corresponding to the carbon emission area labeled i, V[i] represents the carbon emission flow rate corresponding to the carbon emission area labeled i, and Ω[i] is used to represent the gas composition corresponding to the carbon emission area labeled i.

[0017] Furthermore, the process of extracting features from the carbon emission related data of each carbon emission area and obtaining the carbon emission feature data set of the corresponding carbon emission area includes:

[0018] Set the feature extraction threshold corresponding to the carbon emission related data of each carbon emission area;

[0019] The feature extraction threshold includes the lower limit threshold of carbon emission concentration and the upper limit threshold of carbon emission concentration corresponding to the carbon emission area labeled i, denoted as C[i] 下限 and C[i] 上限 , and also includes the carbon emission lower limit threshold and carbon emission upper limit threshold corresponding to the carbon emission area labeled i, denoted as D[i] 下限 and D[i] 上限 , and also includes the lower flow threshold and the upper flow threshold corresponding to the carbon emission area labeled i, denoted as V[i] 下限 and V[i] 上限 ;

[0020] For the carbon emission region labeled i, C[i] 下限 ≤C[i]≤C[i] 上限 The carbon emission concentration of the corresponding part, D[i] 下限 ≤D[i]≤D[i] 上限 The carbon emissions of the corresponding part, V[i] 下限 ≤V[i]≤V[i] 上限The carbon emission flow rate of the corresponding part and Ω[i] are integrated as the carbon emission characteristic data set of the corresponding carbon emission area, and other data in the carbon emission area except the currently integrated data are marked as invalid data and filtered out.

[0021] Furthermore, the process of establishing a regional carbon emission flow analysis model through a deep reinforcement learning algorithm includes:

[0022] Select a model architecture, select a deep reinforcement learning algorithm as the execution algorithm of the selected model architecture, obtain carbon emission-related data of several carbon emission regions at historical time nodes, set the state space, and set model input nodes, model output nodes, and iterative learning nodes for state space association, and then build an initial regional carbon emission flow analysis model;

[0023] The carbon emission-related data of each carbon emission region at a historical time node are associated with the carbon emission flow status of the corresponding carbon emission region, and the carbon emission flow status includes a normal carbon emission flow status and an abnormal carbon emission flow status;

[0024] The model output node of the initial regional carbon emission flow analysis model is used to output the carbon emission flow state of each carbon emission region's carbon emission related data at the historical time node, and determine whether the output result is consistent with the actual carbon emission flow state of each carbon emission region;

[0025] The state determination rate of the initial regional carbon emission flow analysis model is calculated, and the model iteration threshold is set. According to the numerical relationship between the state determination rate and the model iteration threshold, it is determined whether it is necessary to iterate the initial regional carbon emission flow analysis model until the initial regional carbon emission flow analysis model is constructed as the final regional carbon emission flow analysis model.

[0026] Furthermore, the process of determining whether it is necessary to iterate the initial regional carbon emission flow analysis model includes:

[0027] The state determination rate is recorded as τ, and the expression formula of τ is:

[0028]

[0029] Among them, N 成功 Represents the initial regional carbon emission flow analysis model. According to the carbon emission related data of the carbon emission region at the historical time nodes, the number of times the carbon emission flow state outputted by the result is consistent with the actual carbon emission flow state of the corresponding carbon emission region; N 总 Indicates the total number of times the initial regional carbon emission flow analysis model outputs results;

[0030] The model iteration threshold is denoted as Dd;

[0031] If τ ≥ Dd, no model iteration is performed, indicating that the initial regional carbon emission flow analysis model has met the expected prediction accuracy, and the initial regional carbon emission flow analysis model is used as the final regional carbon emission flow analysis model;

[0032] If τ<Dd, it means that the initial regional carbon emission flow analysis model does not meet the expected prediction accuracy. The initial regional carbon emission flow analysis model is iterated until the final regional carbon emission flow analysis model that meets the expected prediction accuracy is constructed.

[0033] Furthermore, the model iteration process includes:

[0034] Data sampling is performed on carbon emission-related data at all historical time nodes to generate corresponding sample data, and the sample data division ratio is set to divide the sample data into a training set and a test set. The training set is input into the initial regional carbon emission flow analysis model through the iterative learning node, and the number of iterations is set. After each iteration, the test set is input into the initial regional carbon emission flow analysis model, and the model fitting degree, model normalization rate and model prediction accuracy after each iteration are recorded;

[0035] Set the corresponding compliance numerical ranges for the model fit, model normalization rate and model prediction accuracy of the initial regional carbon emission flow analysis model. If, after a certain iteration, the model fit, model normalization rate and model prediction accuracy of the initial regional carbon emission flow analysis model are all within their respective compliance numerical ranges, stop the model iteration and convert the initial regional carbon emission flow analysis model into the final regional carbon emission flow analysis model; otherwise, continue to iterate the initial regional carbon emission flow analysis model.

[0036] Furthermore, the process of inputting the carbon emission characteristic data set of each carbon emission region into the regional carbon emission flow analysis model, and then monitoring the carbon emission flows of several carbon emission regions in real time includes:

[0037] A data audit node is constructed for each carbon emission area, and the carbon emission characteristic data set of each carbon emission area is preferentially entered into the data audit node. The data audit node audits the carbon emission characteristic data set, screens out the alienated data information in the entered carbon emission characteristic data set, and removes all alienated data information;

[0038] Input the carbon emission characteristic data set that has been audited by the data audit node corresponding to each carbon emission area into the regional carbon emission flow analysis model, and use the regional carbon emission flow analysis model to analyze the carbon emission characteristic data set, and then monitor the carbon emission flow of each carbon emission area in real time;

[0039] and obtaining the carbon emission flow state of each carbon emission region corresponding to the carbon emission flow;

[0040] If the carbon emission characteristic data set is input into the regional carbon emission flow analysis model, and the output result is that the carbon emission flow corresponding to the current carbon emission area is in a normal carbon emission flow state, no operation is performed;

[0041] If the carbon emission characteristic data set is input into the regional carbon emission flow analysis model, and the output result is that the carbon emission flow corresponding to the current carbon emission area is in an abnormal carbon emission flow state, then a regional carbon emission flow scheduling model is constructed to optimize the carbon emission flow scheduling for the corresponding carbon emission area.

[0042] Furthermore, a regional carbon emission flow scheduling model is constructed, and the process of optimizing the carbon emission flow scheduling for carbon emission regions with abnormal carbon emission flows includes:

[0043] For each carbon emission region with abnormal carbon emission flow, a corresponding regional carbon emission flow scheduling model is constructed. The regional carbon emission flow scheduling model consists of a model decision layer, a model constraint layer and a model feedback layer.

[0044] The model decision layer generates a carbon emission scheduling process for the carbon emission area where the current carbon emission flow is abnormal. The carbon emission scheduling process consists of several sub-execution processes under the scheduling time nodes, and each sub-execution process corresponds to a scheduling stage of carbon emission scheduling;

[0045] When the model constraint layer sets the execution of the carbon emission scheduling process, each sub-execution process has its own corresponding carbon emission constraint target, which includes the total limit of regional carbon emissions, the peak of regional carbon emission concentration, and the regional carbon emission gas distribution regulations;

[0046] The model feedback layer evaluates whether there is any abnormality in the corresponding carbon emission flow in each sub-execution process of the carbon emission scheduling process in the corresponding scheduling stage. If so, the carbon emission constraint target of the corresponding scheduling stage is reset until there is no abnormality in the carbon emission flow in the current carbon emission area, and the carbon emission flow scheduling optimization is stopped. If not, no operation is performed.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows: by collecting carbon emission related data of carbon emission sources in several carbon emission areas and performing feature extraction, a carbon emission characteristic data set of the corresponding carbon emission areas is obtained, a regional carbon emission flow analysis model is established through a deep reinforcement learning algorithm, and the carbon emission characteristic data set of each carbon emission area is input into the regional carbon emission flow analysis model to monitor the carbon emission flows of several carbon emission areas in real time, build a regional carbon emission flow scheduling model, and optimize the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows. On the one hand, feature extraction of carbon emission related data obtains a carbon emission characteristic data set, and the carbon emission characteristic data set is processed and analyzed through the regional carbon emission flow analysis model, which effectively improves the monitoring and analysis capabilities of carbon emission related data and achieves timely discovery of carbon emission flow anomalies. On the other hand, a regional carbon emission flow scheduling model is built to optimize the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows, thereby ensuring that carbon emissions in carbon emission areas meet the standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0049] like Figure 1 As shown, a regional carbon emission flow analysis and management method based on deep reinforcement learning includes the following steps:

[0050] Step S1: collecting carbon emission related data corresponding to carbon emission sources in several carbon emission areas, and performing feature extraction on the carbon emission related data of each carbon emission area, thereby obtaining a carbon emission feature data set of the corresponding carbon emission area;

[0051] Step S2: Establish a regional carbon emission flow analysis model through a deep reinforcement learning algorithm, input the carbon emission characteristic data set of each carbon emission area into the regional carbon emission flow analysis model, and then monitor the carbon emission flows of several carbon emission areas in real time;

[0052] Step S3: Construct a regional carbon emission flow scheduling model, and optimize the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows.

[0053] It should be further explained that, in the specific implementation process, the process of collecting carbon emission related data corresponding to carbon emission sources in several carbon emission areas includes:

[0054] Several carbon emission areas are numbered and the number is recorded as i, then i=1, 2, 3, ..., n, where n is a natural number greater than 0, and a carbon emission monitoring device is arranged in each carbon emission area, and the carbon emission monitoring device collects carbon emission related data corresponding to the carbon emission source in the corresponding carbon emission area;

[0055] The carbon emission monitoring equipment includes a gas analyzer, a gas flow meter and a remote sensing detection device;

[0056] The carbon emission related data include carbon emission concentration, carbon emission amount, carbon emission flow rate and gas composition;

[0057] The gas analyzer is used to collect the carbon emission concentration in each carbon emission area;

[0058] The gas flow meter is used to collect the carbon emission amount and carbon emission flow rate in each carbon emission area;

[0059] The remote sensing detection equipment is used to collect gas composition in each carbon emission area;

[0060] The carbon emission related data corresponding to the carbon emission area labeled i is recorded as T-pf[i];

[0061] T-pf[i]={C[i], D[i], V[i], Ω[i]};

[0062] Among them, C[i] represents the carbon emission concentration corresponding to the carbon emission area labeled i, D[i] represents the carbon emission amount corresponding to the carbon emission area labeled i, V[i] represents the carbon emission flow rate corresponding to the carbon emission area labeled i, and Ω[i] is used to represent the gas composition corresponding to the carbon emission area labeled i.

[0063] It should be further explained that, in the specific implementation process, the process of extracting features from the carbon emission related data of each carbon emission area and then obtaining the carbon emission feature data set of the corresponding carbon emission area includes:

[0064] Set a feature extraction threshold corresponding to the carbon emission related data of each carbon emission area, analyze the relationship between the carbon emission related data of each carbon emission area and the feature extraction threshold, and then obtain the carbon emission feature data set of the corresponding carbon emission area through feature extraction;

[0065] The feature extraction threshold includes a carbon emission concentration lower limit threshold and a carbon emission concentration upper limit threshold corresponding to the carbon emission concentration. The carbon emission concentration lower limit threshold and the carbon emission concentration upper limit threshold corresponding to the carbon emission area labeled i are respectively denoted as C[i] 下限 and C[i] 上限 ;

[0066] The feature extraction threshold also includes the carbon emission lower limit threshold and the carbon emission upper limit threshold corresponding to the carbon emission. The carbon emission lower limit threshold and the carbon emission upper limit threshold corresponding to the carbon emission area labeled i are respectively denoted as D[i] 下限 and D[i] 上限 ;

[0067] The feature extraction threshold also includes the flow lower limit threshold and flow upper limit threshold corresponding to the carbon emission flow rate. The flow lower limit threshold and flow upper limit threshold corresponding to the carbon emission area labeled i are respectively denoted as V[i] 下限 and V[i] 上限 ;

[0068] For the carbon emission region labeled i, C[i] 下限 ≤C[i]≤C[i] 上限 The carbon emission concentration of the corresponding part, D[i] 下限 ≤D[i]≤D[i] 上限 The carbon emissions of the corresponding part, V[i] 下限 ≤V[i]≤V[i] 上限 The carbon emission flow rate of the corresponding part and Ω[i] are integrated as the carbon emission characteristic data set of the corresponding carbon emission area, and other data in the carbon emission area except the currently integrated data are marked as invalid data, and the invalid data are filtered out.

[0069] It should be further explained that in the specific implementation process, the process of establishing a regional carbon emission flow analysis model through a deep reinforcement learning algorithm includes:

[0070] Select a model architecture, select a deep reinforcement learning algorithm as the execution algorithm of the selected model architecture, obtain carbon emission-related data of several carbon emission regions at historical time nodes, set the state space, and set model input nodes, model output nodes, and iterative learning nodes for state space association, and then build an initial regional carbon emission flow analysis model;

[0071] The carbon emission related data of each carbon emission area at a historical time node are associated with the carbon emission flow state of the corresponding carbon emission area, wherein the carbon emission flow state includes a normal carbon emission flow state and an abnormal carbon emission flow state;

[0072] For the carbon emission area in the normal state of carbon emission flow, the carbon emission related data corresponding to the carbon emission area is used as the input data of the model input node, and the legal value range corresponding to the carbon emission concentration, carbon emission amount and carbon emission flow rate at this time is used as the first judgment basis for the normal state, and the gas composition of the carbon emission area at this time is used as the second judgment basis for the normal state;

[0073] For the carbon emission area in the abnormal state of carbon emission flow, the carbon emission related data corresponding to the carbon emission area is used as the input data of the model input node, and the abnormal numerical range corresponding to the carbon emission concentration, carbon emission amount and carbon emission flow rate at this time is used as the first judgment basis for the abnormal state, and the gas composition of the carbon emission area at this time is used as the second judgment basis for the abnormal state;

[0074] The model output node of the initial regional carbon emission flow analysis model is used to output the carbon emission flow state of each carbon emission region's carbon emission related data at the historical time node, and determine whether the output result is consistent with the actual carbon emission flow state of each carbon emission region;

[0075] The state determination rate of the initial regional carbon emission flow analysis model is calculated and recorded as τ;

[0076] Then the expression formula of the state judgment rate τ is as follows:

[0077]

[0078] Among them, N 成功 Represents the initial regional carbon emission flow analysis model. According to the carbon emission related data of the carbon emission region at the historical time nodes, the number of times the carbon emission flow state outputted by the result is consistent with the actual carbon emission flow state of the corresponding carbon emission region; N 总 Indicates the total number of times the initial regional carbon emission flow analysis model outputs results;

[0079] Set the model iteration threshold, and record the model iteration threshold as Dd;

[0080] According to the numerical relationship between the state determination rate of the initial regional carbon emission flow analysis model and the model iteration threshold, it is determined whether it is necessary to perform model iteration on the initial regional carbon emission flow analysis model;

[0081] If τ ≥ Dd, no model iteration is performed, indicating that the initial regional carbon emission flow analysis model has met the expected prediction accuracy, and the initial regional carbon emission flow analysis model is used as the final regional carbon emission flow analysis model;

[0082] If τ<Dd, it means that the initial regional carbon emission flow analysis model does not meet the expected prediction accuracy. The initial regional carbon emission flow analysis model is iterated until the final regional carbon emission flow analysis model that meets the expected prediction accuracy is constructed.

[0083] The content of the model iteration is as follows:

[0084] Data sampling is performed on carbon emission-related data at all historical time nodes to generate corresponding sample data, and the division ratio of the sample data is set to divide the sample data into a training set and a test set. The training set is input into the initial regional carbon emission flow analysis model through an iterative learning node;

[0085] Set the number of iterations. After each iteration, input the test set into the initial regional carbon emission flow analysis model, and record the model fit, model normalization rate, and model prediction accuracy after each iteration.

[0086] Set the corresponding compliance numerical ranges for the model fit, model normalization rate and model prediction accuracy of the initial regional carbon emission flow analysis model. If, after a certain iteration, the model fit, model normalization rate and model prediction accuracy of the initial regional carbon emission flow analysis model are all within their respective compliance numerical ranges, stop the model iteration and convert the initial regional carbon emission flow analysis model into the final regional carbon emission flow analysis model; otherwise, continue to iterate the initial regional carbon emission flow analysis model.

[0087] It should be further explained that, in the specific implementation process, the process of inputting the carbon emission characteristic data set of each carbon emission area into the regional carbon emission flow analysis model, and then real-time monitoring of the carbon emission flows of several carbon emission areas includes:

[0088] A data audit node is constructed for each carbon emission area, and the carbon emission characteristic data set of each carbon emission area is preferentially entered into the data audit node. The data audit node audits the carbon emission characteristic data set, screens out the alienated data information in the entered carbon emission characteristic data set, and removes all alienated data information;

[0089] The alienated data information refers to part of the data information after feature extraction that does not conform to the corresponding carbon emission characteristic data set, that is, the carbon emission concentration that is not between the lower limit threshold of carbon emission concentration and the upper limit threshold of carbon emission concentration, the carbon emission that is not between the lower limit threshold of carbon emission amount and the upper limit threshold of carbon emission, and also includes the carbon emission flow rate that is not between the lower limit threshold of flow and the upper limit threshold of flow;

[0090] Input the carbon emission characteristic data set that has been audited by the data audit node corresponding to each carbon emission area into the regional carbon emission flow analysis model, and use the regional carbon emission flow analysis model to analyze the carbon emission characteristic data set, and then monitor the carbon emission flow of each carbon emission area in real time;

[0091] and obtaining the carbon emission flow state of each carbon emission region corresponding to the carbon emission flow;

[0092] If the carbon emission characteristic data set is input into the regional carbon emission flow analysis model, and the output result is that the carbon emission flow corresponding to the current carbon emission area is in a normal carbon emission flow state, no operation is performed;

[0093] If the carbon emission characteristic data set is input into the regional carbon emission flow analysis model, and the output result is that the carbon emission flow corresponding to the current carbon emission area is in an abnormal carbon emission flow state, then a regional carbon emission flow scheduling model is constructed to optimize the carbon emission flow scheduling for the corresponding carbon emission area.

[0094] It should be further explained that, in the specific implementation process, the process of constructing a regional carbon emission flow scheduling model and optimizing the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows includes:

[0095] For each carbon emission region with abnormal carbon emission flow, a corresponding regional carbon emission flow scheduling model is constructed. The regional carbon emission flow scheduling model consists of a model decision layer, a model constraint layer and a model feedback layer.

[0096] The model decision layer generates a carbon emission scheduling process for the carbon emission area where the current carbon emission flow is abnormal. The carbon emission scheduling process consists of several sub-execution processes under the scheduling time nodes, and each sub-execution process corresponds to a scheduling stage of carbon emission scheduling;

[0097] When the model constraint layer sets the execution of the carbon emission scheduling process, each sub-execution process has its own corresponding carbon emission constraint target, which includes the total limit of regional carbon emissions, the peak of regional carbon emission concentration, and the regional carbon emission gas distribution regulations;

[0098] The model feedback layer evaluates whether there is any abnormality in the corresponding carbon emission flow in each sub-execution process of the carbon emission scheduling process in the corresponding scheduling stage. If so, the carbon emission constraint target of the corresponding scheduling stage is reset until there is no abnormality in the carbon emission flow in the current carbon emission area, and the carbon emission flow scheduling optimization is stopped. If not, no operation is performed.

[0099] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A regional carbon emission flow analysis and management method based on deep reinforcement learning, characterized in that: The following steps are involved: Step S1: collecting carbon emission related data corresponding to carbon emission sources in several carbon emission areas, and performing feature extraction on the carbon emission related data of each carbon emission area, thereby obtaining a carbon emission feature data set of the corresponding carbon emission area; Step S2: Establish a regional carbon emission flow analysis model through a deep reinforcement learning algorithm, input the carbon emission characteristic data set of each carbon emission area into the regional carbon emission flow analysis model, and then monitor the carbon emission flows of several carbon emission areas in real time; Step S3: Construct a regional carbon emission flow scheduling model, and optimize the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows.

2. According to claim 1, a regional carbon emission flow analysis and management method based on deep reinforcement learning is characterized in that: The process of collecting carbon emission related data corresponding to carbon emission sources in several carbon emission areas includes: Several carbon emission areas are labeled and recorded as i, i=1, 2, 3, ..., n, where n is a natural number greater than 0, and carbon emission monitoring equipment is arranged in each carbon emission area, and the carbon emission monitoring equipment collects carbon emission related data of the carbon emission sources in the corresponding carbon emission area; The carbon emission monitoring equipment includes a gas analyzer, a gas flow meter and a remote sensing detection device; The carbon emission related data include carbon emission concentration, carbon emission amount, carbon emission flow rate and gas composition; The gas analyzer is used to collect the carbon emission concentration in each carbon emission area; The gas flow meter is used to collect the carbon emission amount and carbon emission flow rate in each carbon emission area; The remote sensing detection equipment is used to collect gas composition in each carbon emission area; The carbon emission related data corresponding to the carbon emission area labeled i is recorded as T-pf[i], T-pf[i] = {C[i], D[i], V[i], Ω[i]}; Among them, C[i] represents the carbon emission concentration corresponding to the carbon emission area labeled i, D[i] represents the carbon emission amount corresponding to the carbon emission area labeled i, V[i] represents the carbon emission flow rate corresponding to the carbon emission area labeled i, and Ω[i] is used to represent the gas composition corresponding to the carbon emission area labeled i.

3. According to claim 2, a regional carbon emission flow analysis and management method based on deep reinforcement learning is characterized in that: The process of extracting features from the carbon emission related data of each carbon emission area and obtaining the carbon emission feature data set of the corresponding carbon emission area includes: Set the feature extraction threshold corresponding to the carbon emission related data of each carbon emission area; The feature extraction threshold includes the lower limit threshold of carbon emission concentration and the upper limit threshold of carbon emission concentration corresponding to the carbon emission area labeled i, denoted as C[i] 下限 and C[i] 上限 , and also includes the carbon emission lower limit threshold and carbon emission upper limit threshold corresponding to the carbon emission area labeled i, denoted as D[i] 下限 and D[i] 上限 , and also includes the lower flow threshold and the upper flow threshold corresponding to the carbon emission area labeled i, denoted as V[i] 下限 and V[i] 上限 ; For the carbon emission region labeled i, C[i] 下限 ≤C[i]≤C[i] 上限 The carbon emission concentration of the corresponding part, D[i] 下限 ≤D[i]≤D[i] 上限 The carbon emissions of the corresponding part, V[i] 下限 ≤V[i]≤V[i] 上限 The carbon emission flow rate of the corresponding part and Ω[i] are integrated as the carbon emission characteristic data set of the corresponding carbon emission area, and other data in the carbon emission area except the currently integrated data are marked as invalid data and filtered out.

4. According to claim 3, a regional carbon emission flow analysis and management method based on deep reinforcement learning is characterized in that: The process of establishing a regional carbon emission flow analysis model through deep reinforcement learning algorithm includes: Select a model architecture, select a deep reinforcement learning algorithm as the execution algorithm of the selected model architecture, obtain carbon emission-related data of several carbon emission regions at historical time nodes, set the state space, and set model input nodes, model output nodes, and iterative learning nodes for state space association, and then build an initial regional carbon emission flow analysis model; The carbon emission-related data of each carbon emission region at a historical time node are associated with the carbon emission flow status of the corresponding carbon emission region, and the carbon emission flow status includes a normal carbon emission flow status and an abnormal carbon emission flow status; The model output node of the initial regional carbon emission flow analysis model is used to output the carbon emission flow state of each carbon emission region's carbon emission related data at the historical time node, and determine whether the output result is consistent with the actual carbon emission flow state of each carbon emission region; The state determination rate of the initial regional carbon emission flow analysis model is calculated, and the model iteration threshold is set. According to the numerical relationship between the state determination rate and the model iteration threshold, it is determined whether it is necessary to iterate the initial regional carbon emission flow analysis model until the initial regional carbon emission flow analysis model is constructed as the final regional carbon emission flow analysis model.

5. According to claim 4, a regional carbon emission flow analysis and management method based on deep reinforcement learning is characterized in that: The process of determining whether it is necessary to iterate the initial regional carbon emission flow analysis model includes: The state determination rate is recorded as τ, and the expression formula of τ is: Among them, N 成功 Represents the initial regional carbon emission flow analysis model. According to the carbon emission related data of the carbon emission region at the historical time nodes, the number of times the carbon emission flow state outputted by the result is consistent with the actual carbon emission flow state of the corresponding carbon emission region; N 总 Indicates the total number of times the initial regional carbon emission flow analysis model outputs results; The model iteration threshold is denoted as Dd; If τ ≥ Dd, no model iteration is performed, indicating that the initial regional carbon emission flow analysis model has met the expected prediction accuracy, and the initial regional carbon emission flow analysis model is used as the final regional carbon emission flow analysis model; If τ<Dd, it means that the initial regional carbon emission flow analysis model does not meet the expected prediction accuracy. The initial regional carbon emission flow analysis model is iterated until the final regional carbon emission flow analysis model that meets the expected prediction accuracy is constructed.

6. According to claim 5, a regional carbon emission flow analysis and management method based on deep reinforcement learning is characterized in that: The model iteration process includes: Data sampling is performed on carbon emission-related data at all historical time nodes to generate corresponding sample data, and the sample data division ratio is set to divide the sample data into a training set and a test set. The training set is input into the initial regional carbon emission flow analysis model through the iterative learning node, and the number of iterations is set. After each iteration, the test set is input into the initial regional carbon emission flow analysis model, and the model fitting degree, model normalization rate and model prediction accuracy after each iteration are recorded; Set the corresponding compliance numerical ranges for the model fit, model normalization rate and model prediction accuracy of the initial regional carbon emission flow analysis model. If, after a certain iteration, the model fit, model normalization rate and model prediction accuracy of the initial regional carbon emission flow analysis model are all within their respective compliance numerical ranges, stop the model iteration and convert the initial regional carbon emission flow analysis model into the final regional carbon emission flow analysis model; otherwise, continue to iterate the initial regional carbon emission flow analysis model.

7. A regional carbon emission flow analysis and management method based on deep reinforcement learning according to claim 6, characterized in that: The process of inputting the carbon emission characteristic data set of each carbon emission region into the regional carbon emission flow analysis model and then monitoring the carbon emission flows of several carbon emission regions in real time includes: A data audit node is constructed for each carbon emission area, and the carbon emission characteristic data set of each carbon emission area is preferentially entered into the data audit node. The data audit node audits the carbon emission characteristic data set, screens out the alienated data information in the entered carbon emission characteristic data set, and removes all alienated data information; Input the carbon emission characteristic data set that has been audited by the data audit node corresponding to each carbon emission area into the regional carbon emission flow analysis model, and use the regional carbon emission flow analysis model to analyze the carbon emission characteristic data set, and then monitor the carbon emission flow of each carbon emission area in real time; and obtaining the carbon emission flow state of each carbon emission region corresponding to the carbon emission flow; If the carbon emission characteristic data set is input into the regional carbon emission flow analysis model, and the output result is that the carbon emission flow corresponding to the current carbon emission area is in a normal carbon emission flow state, no operation is performed; If the carbon emission characteristic data set is input into the regional carbon emission flow analysis model, and the output result is that the carbon emission flow corresponding to the current carbon emission area is in an abnormal carbon emission flow state, then a regional carbon emission flow scheduling model is constructed to optimize the carbon emission flow scheduling for the corresponding carbon emission area.

8. The regional carbon emission flow analysis and management method based on deep reinforcement learning according to claim 7 is characterized in that: The process of constructing a regional carbon emission flow scheduling model and optimizing the carbon emission flow scheduling for carbon emission areas with abnormal carbon emission flows includes: For each carbon emission region with abnormal carbon emission flow, a corresponding regional carbon emission flow scheduling model is constructed. The regional carbon emission flow scheduling model consists of a model decision layer, a model constraint layer and a model feedback layer. The model decision layer generates a carbon emission scheduling process for the carbon emission area where the current carbon emission flow is abnormal. The carbon emission scheduling process consists of several sub-execution processes under the scheduling time nodes, and each sub-execution process corresponds to a scheduling stage of carbon emission scheduling; When the model constraint layer sets the execution of the carbon emission scheduling process, each sub-execution process has its own corresponding carbon emission constraint target, which includes the total limit of regional carbon emissions, the peak of regional carbon emission concentration, and the regional carbon emission gas distribution regulations; The model feedback layer evaluates whether there is any abnormality in the corresponding carbon emission flow in each sub-execution process of the carbon emission scheduling process in the corresponding scheduling stage. If so, the carbon emission constraint target of the corresponding scheduling stage is reset until there is no abnormality in the carbon emission flow in the current carbon emission area, and the carbon emission flow scheduling optimization is stopped. If not, no operation is performed.

Images