Method and system for optimizing carbon management of a public building area

Abstract

The application relates to a public building area energy carbon management optimization method and system, which comprises the following steps: step S1: a virtual energy carbon node network of a public building area is constructed, the virtual energy carbon node network is composed of multiple virtual energy carbon nodes, each virtual energy carbon node corresponds to a physical entity or a functional space in the public building area, and is configured with a unique identifier, an energy carbon attribute set and an association rule set; step S2: real-time operation data and static characteristic data of each physical entity and functional space in the public building area are acquired, the real-time operation data and the static characteristic data are mapped to corresponding virtual energy carbon nodes based on the virtual energy carbon node network, and an initialized energy carbon digital twin is generated; and step S3: in the energy carbon digital twin, multi-source data fusion and dynamic baseline construction are performed based on the association rule set of each virtual energy carbon node; the application can realize better public building area energy carbon management optimization.

Description

Technical Field

[0001] This invention relates to the field of carbon emission management, and specifically to a method and system for optimizing energy and carbon management in public building areas. Background Technology

[0002] The current public building areas are characterized by complex business formats and close energy and carbon node connections between various physical entities and functional spaces. Energy and carbon consumption is dynamically affected by multiple dimensions of factors such as the external environment, personnel flow, and energy supply. This necessitates increasingly sophisticated, intelligent, and collaborative energy and carbon management. However, traditional energy and carbon management methods for public buildings have many technical shortcomings and are no longer suitable for actual control needs. Specifically: traditional energy and carbon benchmark values ​​are mostly statically set, failing to incorporate dynamic adjustments based on spatial / functional connections between nodes, real-time environmental parameters, and personnel flow, thus becoming disconnected from actual operating conditions and resulting in a lack of accurate numerical basis for energy and carbon anomaly detection; energy and carbon situation prediction often uses single-timescale models, making it difficult to consider ultra-short-term trends. The current system lacks comprehensive forecasting capabilities across short-term and medium-term timeframes. Relying solely on manual investigation or simple data comparisons for energy and carbon anomaly tracing fails to accurately analyze the transmission paths and core sources of deviations between related nodes, resulting in insufficient accuracy and efficiency in situation prediction and anomaly location. Energy and carbon optimization often employs single-objective models, failing to balance multiple objectives such as minimizing carbon emissions, minimizing operating costs, and maximizing overall energy efficiency. Optimization constraints are not aligned with the actual energy and carbon situation and anomaly rectification practices. Traditional optimization algorithms also fail to incorporate anomaly tracing characteristics, leading to blind optimization and neglecting the interests of multiple stakeholders, including owners, property management, and energy suppliers. This makes it difficult to implement control measures due to conflicting demands. Furthermore, energy and carbon management is inadequate. The inclusion of dynamic constraints on renewable energy and energy storage nodes neglects regional clean energy absorption capacity and energy storage dispatch potential, resulting in a mismatch between the energy supply and demand balance logic and the actual energy structure, which is detrimental to carbon emission reduction and energy efficiency improvement. Furthermore, the virtual carbon node's energy and carbon attribute representation is singular, failing to account for the entire lifecycle carbon emissions of energy, materials, and water resources, focusing only on direct carbon emissions during operation, leading to a lack of comprehensive control dimensions and hindering full-dimensional energy and carbon management. The closed-loop control mechanism lacks precise execution feedback and response deviation verification mechanisms for control commands; model parameters are statically fixed, lacking targeted self-correction and iterative optimization methods, and the energy and carbon digital twin cannot accurately reflect the operational status of the physical entity. The current system lacks real-time updates, making it difficult to maintain accurate mapping, analysis, and prediction capabilities. Energy and carbon management results are presented only in fragmented data formats, lacking intuitive visualization methods for anomaly location, quantitative analysis of energy and carbon reduction potential, and traceable reviews of energy and carbon efficiency improvements. This fails to provide comprehensive and scientific decision support for short-term rectification, medium-term planning, and long-term optimization of energy and carbon management. Furthermore, the overall energy and carbon management system lacks a systematic closed-loop design, with poor data interoperability and coordination between different stages. This makes it difficult to achieve refined and comprehensive control of energy and carbon in public building areas, and also fails to effectively balance the multiple demands of carbon reduction, operating cost control, and energy comfort. Therefore, a method for optimizing energy and carbon management in public building areas is proposed. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for optimizing energy and carbon management in public building areas, comprising the following steps: Step S1: Construct a virtual energy and carbon node network for the public building area. The virtual energy and carbon node network consists of multiple virtual energy and carbon nodes. Each virtual energy and carbon node corresponds to a physical entity or a functional space within the public building area and is configured with a unique identifier, a set of energy and carbon attributes, and a set of association rules. Step S2: Obtain real-time operation data and static feature data of each physical entity and functional space within the public building area. Based on the virtual energy carbon node network, map the real-time operation data and static feature data to the corresponding virtual energy carbon nodes to generate an initialized energy carbon digital twin. Step S3: In the energy carbon digital twin, based on the association rule set of each virtual energy carbon node, perform multi-source data fusion and dynamic baseline construction to generate dynamic energy carbon baseline values ​​for each virtual energy carbon node. Step S4: Based on the dynamic energy and carbon baseline values, real-time operation data and static characteristic data of each virtual energy and carbon node, perform multi-timescale prediction and spatial source tracing of energy and carbon status by coupling the time series prediction model and the spatial transmission model, and generate energy and carbon status prediction results and energy and carbon anomaly source tracing paths. Step S5: Based on the energy and carbon situation prediction results and the energy and carbon anomaly tracing path, call the preset multi-objective collaborative optimization model to generate a set of control instructions for at least one virtual energy and carbon node in the virtual energy and carbon node network. Step S6: Send the control command set to the corresponding physical entity actuator, receive execution feedback, update the energy carbon digital twin according to the execution feedback, and form a closed-loop control.

[0004] Furthermore, step S3 generates dynamic energy-carbon baseline values ​​for each virtual energy-carbon node, specifically including: For any virtual energy carbon node i, obtain its historical concurrent operating data and the real-time operating data of its associated nodes that have spatial or functional connections with it. The energy and carbon efficiency feature vector is extracted from historical data of the same period through a sliding window mechanism. The energy and carbon efficiency feature vector includes energy consumption intensity per unit area, carbon emission intensity per person-time, and energy efficiency fluctuation rate. Then, by combining the real-time operating data of the associated nodes, the current external environmental parameters, and the personnel flow density data, a weighted regression algorithm is used to calculate the dynamic correction factor; Based on the energy-carbon efficiency eigenvector, dynamic correction factor, and static characteristic data of the node, a dynamic energy-carbon baseline value is generated. The calculation process is as follows: ; in, Let i be the dynamic energy-carbon baseline value of the virtual energy-carbon node i at time t. Let M be the m-th type of energy-carbon efficiency feature value of node i at historical time k, where M is the total number of feature types, W is the sliding window width, and N is the number of sampling points within the window. The adaptive weight coefficient for the m-th feature is dynamically adjusted by minimizing the deviation between historical data and measured data. The spatial correlation correction factor is calculated based on real-time operational data of associated nodes. The calculation process is as follows: ; in Let be the spatial coupling coefficient between node i and its associated node j. The real-time carbon intensity of the associated node j. The static energy efficiency benchmark value is obtained from the statistical database of similar public building areas for the category to which the associated node j belongs; This is an environmental dynamic correction factor calculated based on external environmental parameters and population flow density. The calculation process is as follows: ; in and These represent the temperature, humidity, and population density at time t, respectively. and For the corresponding reference base value, and This represents the environmental impact weighting coefficient.

[0005] Furthermore, the specific process of performing multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4 is as follows: First, a time-series prediction model is constructed. The time-series prediction model adopts a combination architecture of temporal convolutional network and gated recurrent unit. The dynamic energy and carbon baseline value, real-time operation data and static feature data of each virtual energy and carbon node are used as input to generate multi-timescale energy and carbon situation prediction results for the ultra-short term, short term and medium term. Secondly, a spatial transmission model is constructed. The spatial transmission model is based on the topology and spatial coupling coefficient of the virtual energy carbon node network. When the deviation between the real-time energy carbon data of any virtual energy carbon node and the dynamic energy carbon benchmark value exceeds a preset threshold, the spatial source tracing mechanism is triggered. The spatial transmission model is used to analyze the transmission path and contribution of the deviation between related nodes and generate an energy carbon anomaly source tracing path. The spatial propagation model performs spatial source tracing calculations, including: Step 1: Calculate the instantaneous deviation of each virtual energy carbon node and define the deviation function. ,in Let be the real-time carbon energy intensity of node i at time t. The dynamic energy carbon baseline value is used, and the deviation function values ​​of all nodes are used to construct the deviation vector. where n is the total number of nodes; Step 2: Construct a spatial propagation model of the bias in the node network, assuming the observed bias vector From the external disturbance vector Generated through coupling and propagation between nodes, the propagation relationship satisfies: Where A is the adjacency matrix. The spatial coupling coefficient matrix, For conduction attenuation factor, It represents the Hadamardi (or Hadama) stack; Step 3: Rewrite the conduction relation in Step 2 as follows: Furthermore, by solving this equation, the contribution of each potential source node can be analytically determined. ; Step 4: Assign contribution Nodes exceeding a preset threshold are marked as sources of anomalies, and a path for tracing the source of energy and carbon anomalies is formed by tracing back along the topological relationships of the adjacency matrix.

[0006] Furthermore, the multi-objective collaborative optimization model preset in step S5 includes objective functions, constraints, and solution algorithms. The objective functions include at least minimizing total regional carbon emissions, minimizing total regional operating costs, and maximizing regional comprehensive energy efficiency. The constraints are constructed in the following ways: based on the energy and carbon situation prediction results, dynamic energy and carbon quota boundaries are set for key nodes in the virtual energy and carbon node network; based on the energy and carbon anomaly tracing path, energy and carbon efficiency recovery targets and penalty coefficients are set for abnormal nodes. The multi-objective collaborative optimization model is solved using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm introduces a topological distance weight based on the energy carbon anomaly tracing path into the velocity update formula. The specific process of velocity update is as follows: ; in, Let be the velocity of particle i in the d-th dimension during the (k+1)th iteration. Where w is the contraction factor and w is the inertia weight, and As a learning factor, and It is a random number. This represents the optimal position for an individual particle. To be the globally optimal position The contribution of the anomaly source node j. Let be the shortest path distance between the control scheme represented by particle i and the abnormal node j in the topology. This represents the local optimal control position corresponding to the abnormal node j; The source set is a collection of anomaly source nodes whose contribution exceeds a preset threshold in the energy and carbon anomaly tracing path.

[0007] Furthermore, the energy and carbon attribute set of the virtual energy and carbon node in step S1 also includes the real-time energy and carbon intensity, dynamic energy and carbon benchmark value, energy and carbon efficiency rating, and life-cycle carbon emission factor of the virtual energy and carbon node. The life-cycle carbon emission factor is obtained by calculating the carbon emissions of energy, materials, and water resources consumed by the physical entity or functional space corresponding to the virtual energy carbon node during the upstream production, transportation, and downstream disposal stages. The specific calculation process is as follows: ; in, The life-cycle carbon emission factor of node i. Let be the consumption of node i for the e-th type of energy. The life-cycle carbon emission coefficient for energy type e. Let be the consumption of the m-th type of material at node i. Let m be the carbon emission coefficient for the entire life cycle of the m-th material. Let be the consumption of the w-th type of water resource by node i. Let w be the carbon emission coefficient for the entire life cycle of water resources. The life-cycle carbon emission factor is stored in the energy and carbon attribute set and used for multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4.

[0008] Furthermore, after generating the control instruction set in step S5, the following steps are also included: Based on game theory, a multi-party coordination mechanism is constructed, defining different owners, property management companies, and energy suppliers within the public building area as game participants, and constructing a game utility function based on each participant's energy and carbon quotas, energy comfort, and economic costs. Energy comfort level is obtained by weighted calculation based on a combination of indoor temperature, humidity, illuminance, and air quality index. By solving for the Nash equilibrium, the control instruction set is negotiated and corrected to generate an optimized control instruction set that is agreed upon by multiple parties. The process of constructing the game utility function is as follows: ; in, Let be the utility function value of participant s. and For participant s, the weighting coefficients represent their preferences for carbon emissions, comfort, and economic costs. This represents the actual carbon emissions of participant s. The energy and carbon allowances allocated to it, For actual comfort indicators, The target comfort index is obtained by weighting a preset comfort target range with the personalized preferences of participant s. For actual economic costs, The benchmark economic cost is used; the solution process of Nash equilibrium adopts a distributed iterative algorithm. In each round of iteration, each participant updates its own policy by maximizing its own utility function based on the current policy of other participants, until the rate of change of the policies of all participants is lower than the preset convergence threshold.

[0009] Furthermore, the construction of the dynamic baseline in step S3 also includes: Identify renewable energy nodes and energy storage nodes within public building areas, and generate output confidence intervals for renewable energy nodes and dispatchable capacity intervals for energy storage nodes based on meteorological forecast data and historical output data. The output confidence interval and dispatchable capacity interval are used as dynamic constraints and embedded in the generation process of dynamic energy carbon benchmark values ​​to correct the rule logic of energy supply and demand balance in the association rule set. The process of generating the output confidence interval is as follows: ; in, The predicted power output of renewable energy nodes at time t is obtained through a weather-power output mapping model based on a long short-term memory network. The standard deviation of the prediction error is obtained through statistical analysis of historical prediction errors, and z is the coefficient corresponding to the confidence level. The schedulable capacity range is determined by both the state of charge and charge / discharge energy constraints, and its calculation process is as follows: ; in and The minimum and maximum dispatchable capacity of the energy storage node at time t (expressed as a percentage of state of charge). , These are the minimum and maximum allowable states of charge for the energy storage system. Let be the current state of charge at time t. , This represents the maximum discharge and charge power. For scheduling time intervals, This refers to the rated capacity of the energy storage system.

[0010] Furthermore, the specific process of updating the energy carbon digital twin based on execution feedback in step S6 includes: Calculate the response deviation between the execution feedback and the control instruction set. If the response deviation exceeds the preset deviation tolerance range, trigger the model self-correction process. The model self-correction process uses an incremental learning algorithm to fine-tune the internal parameters of the time series prediction model and the multi-objective collaborative optimization model online using response bias, and stores the response bias and its corresponding scene features in an anomaly case library for optimizing the subsequent construction of dynamic baselines. The parameter update process of the incremental learning algorithm is as follows: ; in, For the updated model parameter set, Here is the current model parameter set, and t is the number of exception cases that have been processed. The gradient of the loss function with respect to the model parameters is given. The loss function is constructed based on the weighted mean square error of response bias and scene features, and the weight coefficients are dynamically adjusted according to the similarity between scene features and historical cases in the abnormal case library.

[0011] Furthermore, the method also includes step S7: generating and outputting an energy and carbon management optimization report based on the energy and carbon digital twin. The energy and carbon management optimization report includes: an energy and carbon leakage heat map generated based on the energy and carbon anomaly tracing path, an energy and carbon emission reduction potential analysis generated based on a multi-objective collaborative optimization model, and an energy and carbon efficiency improvement curve generated based on historical control command set and execution feedback. The carbon leakage heatmap is generated by visually rendering it on a building information model by associating the degree of carbon anomaly of each virtual carbon node with spatial coordinates. Its rendering parameters are obtained through the following process: ; in, Let i be the thermal rendering intensity of node i. Let i be the energy-carbon anomaly sensitivity coefficient. This is a normalization function that maps the degree of anomaly to the visual rendering range. The analysis of carbon emission reduction potential is generated by sensitivity analysis of the Pareto front of the multi-objective synergistic optimization model, identifying the combination of control measures that contributes the most to each objective function and their corresponding marginal emission reduction costs; the energy and carbon efficiency improvement curve is generated by fitting the cumulative change rate of unit energy consumption output or unit carbon emission output in each cycle after sorting the historical control command set by time series, where unit energy consumption output is the ratio of regional total output to total energy consumption, and unit carbon emission output is the ratio of regional total output to total carbon emissions.

[0012] A public building area energy and carbon management optimization system, the system comprising: The virtual node network construction module is used to construct a virtual energy and carbon node network in a public building area. The virtual energy and carbon node network consists of multiple virtual energy and carbon nodes. Each virtual energy and carbon node corresponds to a physical entity or a functional space in the public building area and is configured with a unique identifier, a set of energy and carbon attributes, and a set of association rules. The data mapping and twin generation module is connected to the virtual node network construction module. It is used to obtain real-time operation data and static feature data of each physical entity and functional space within the public building area. Based on the virtual energy and carbon node network, the real-time operation data and static feature data are mapped to the corresponding virtual energy and carbon nodes to generate an initialized energy and carbon digital twin. The dynamic baseline construction module, connected to the data mapping and twin generation module, is used to perform multi-source data fusion and dynamic baseline construction in the energy and carbon digital twin based on the association rule set of each virtual energy and carbon node, and generate dynamic energy and carbon baseline values ​​for each virtual energy and carbon node. The energy and carbon situation prediction and source tracing module is connected to the dynamic baseline construction module. Based on the dynamic energy and carbon baseline values, real-time operation data and static characteristic data of each virtual energy and carbon node, it performs multi-timescale prediction and spatial source tracing of energy and carbon situation by coupling the time series prediction model and the spatial transmission model, and generates energy and carbon situation prediction results and energy and carbon anomaly source tracing paths. The multi-objective collaborative optimization module is connected to the energy and carbon situation prediction and source tracing module. It is used to call the preset multi-objective collaborative optimization model based on the energy and carbon situation prediction results and the energy and carbon anomaly source tracing path to generate a set of control instructions for at least one virtual energy and carbon node in the virtual energy and carbon node network. The closed-loop control and twin update module, connected to the multi-objective collaborative optimization module, is used to send the control command set to the corresponding physical entity actuator, receive execution feedback, update the energy carbon digital twin according to the execution feedback, and form closed-loop control. The report generation and output module is connected to the closed-loop control and twin update module. Based on the energy and carbon digital twin, it generates and outputs an energy and carbon management optimization report. The energy and carbon management optimization report includes: an energy and carbon leakage heat map generated based on the energy and carbon anomaly tracing path, an energy and carbon emission reduction potential analysis generated based on the multi-objective collaborative optimization model, and an energy and carbon efficiency improvement curve generated based on the historical control command set and execution feedback.

[0013] The present invention has the following advantages over the prior art: By employing a sliding window mechanism and a multi-factor correction-based dynamic energy carbon baseline generation method, the energy carbon baseline is made more closely aligned with the real-time operational scenarios of public building areas. A temporal prediction model coupling a temporal convolutional network and a gated recurrent unit, along with a topology-based spatial transmission model, achieves accurate prediction of energy carbon status across ultra-short-term, short-term, and medium-term timescales, as well as rapid source tracing and transmission path analysis of energy carbon anomalies. A pre-defined multi-objective collaborative optimization model, combined with an improved particle swarm optimization algorithm, minimizes total regional carbon emissions, minimizes total regional operating costs, and maximizes overall regional energy efficiency while enabling targeted regulation and optimization of anomalous nodes. A multi-stakeholder coordination mechanism based on game theory incorporates owners, property management companies, and energy suppliers into the regulation considerations. By solving for Nash equilibrium and correcting the regulation instruction set, the regulation instructions take into account the energy carbon quotas, energy comfort, and economic cost requirements of all parties, ensuring the executability of the regulation instructions. The dynamic baseline construction incorporates the output confidence intervals of renewable energy nodes and the dispatchable capacity intervals of energy storage nodes. Dynamic constraints optimize the rules and logic of regional energy supply and demand balance, improve the utilization efficiency of renewable energy, and realize online fine-tuning of model parameters and dynamic updating of energy and carbon digital twins through incremental learning algorithms in closed-loop control. Deviations and scenario characteristics are also stored in an anomaly case library to continuously optimize the construction of dynamic baselines, realizing self-learning and self-optimization of energy and carbon management. The management optimization report generated based on the energy and carbon digital twin includes energy and carbon leakage heat maps, emission reduction potential analysis, and energy and carbon efficiency improvement curves, realizing the visualization of energy and carbon management results. This makes it easy for stakeholders to intuitively understand the regional energy and carbon status and formulate targeted strategies. Overall, it realizes refined, intelligent, and collaborative closed-loop management of energy and carbon in public building areas, effectively reducing the total carbon emissions and operating costs of public building areas, significantly improving the overall energy efficiency of the area, and ensuring the comfort of energy use in the area. It enhances the scientific, accurate, and effective nature of energy and carbon management in public building areas, and can also achieve full-dimensional control of energy and carbon emissions in public building areas through the calculation of carbon emission factors throughout the entire life cycle. Attached Figure Description

[0014] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0015] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0016] like Figure 1 As shown, a method and system for optimizing energy and carbon management in public building areas includes the following steps: Step S1: Construct a virtual energy and carbon node network for the public building area. The virtual energy and carbon node network consists of multiple virtual energy and carbon nodes. Each virtual energy and carbon node corresponds to a physical entity or a functional space within the public building area and is configured with a unique identifier, a set of energy and carbon attributes, and a set of association rules. Step S2: Obtain real-time operation data and static feature data of each physical entity and functional space within the public building area. Based on the virtual energy carbon node network, map the real-time operation data and static feature data to the corresponding virtual energy carbon nodes to generate an initialized energy carbon digital twin. Step S3: In the energy carbon digital twin, based on the association rule set of each virtual energy carbon node, perform multi-source data fusion and dynamic baseline construction to generate dynamic energy carbon baseline values ​​for each virtual energy carbon node. Step S4: Based on the dynamic energy and carbon baseline values, real-time operation data and static characteristic data of each virtual energy and carbon node, perform multi-timescale prediction and spatial source tracing of energy and carbon status by coupling the time series prediction model and the spatial transmission model, and generate energy and carbon status prediction results and energy and carbon anomaly source tracing paths. Step S5: Based on the energy and carbon situation prediction results and the energy and carbon anomaly tracing path, call the preset multi-objective collaborative optimization model to generate a set of control instructions for at least one virtual energy and carbon node in the virtual energy and carbon node network. Step S6: Send the control command set to the corresponding physical entity actuator, receive execution feedback, update the energy carbon digital twin according to the execution feedback, and form a closed-loop control.

[0017] Step S3 generates the dynamic energy-carbon baseline value for each virtual energy-carbon node, specifically including: For any virtual energy carbon node i, obtain its historical concurrent operating data and the real-time operating data of its associated nodes that have spatial or functional connections with it. The energy and carbon efficiency feature vector is extracted from historical data of the same period through a sliding window mechanism. The energy and carbon efficiency feature vector includes energy consumption intensity per unit area, carbon emission intensity per person-time, and energy efficiency fluctuation rate. Then, by combining the real-time operating data of the associated nodes, the current external environmental parameters, and the personnel flow density data, a weighted regression algorithm is used to calculate the dynamic correction factor; Based on the energy-carbon efficiency eigenvector, dynamic correction factor, and static characteristic data of the node, a dynamic energy-carbon baseline value is generated. The calculation process is as follows: ; in, Let i be the dynamic energy-carbon baseline value of the virtual energy-carbon node i at time t. Let M be the m-th type of energy-carbon efficiency feature value of node i at historical time k, where M is the total number of feature types, W is the sliding window width, and N is the number of sampling points within the window. The adaptive weight coefficient for the m-th feature is dynamically adjusted by minimizing the deviation between historical data and measured data. The spatial correlation correction factor is calculated based on real-time operational data of associated nodes. The calculation process is as follows: ; in Let be the spatial coupling coefficient between node i and its associated node j. The real-time carbon intensity of the associated node j. The static energy efficiency benchmark value is obtained from the statistical database of similar public building areas for the category to which the associated node j belongs; This is an environmental dynamic correction factor calculated based on external environmental parameters and population flow density. The calculation process is as follows: ; in and These represent the temperature, humidity, and population density at time t, respectively. and For the corresponding reference base value, and Environmental impact weighting coefficient; By combining historical concurrent operating data of virtual energy and carbon nodes with real-time operating data of associated nodes, and incorporating multi-dimensional influencing factors such as external environmental parameters and personnel flow density, this method utilizes a sliding window mechanism to accurately extract energy and carbon efficiency feature vectors. Combined with a weighted regression algorithm to calculate multi-dimensional dynamic correction factors, the generated dynamic energy and carbon benchmark value fully reflects the real-time operating conditions of each node in the public building area. This overcomes the technical limitations of traditional static energy efficiency benchmark values, which cannot adapt to changes in inter-node relationships, environmental fluctuations, and dynamic changes in personnel flow, achieving dynamic adaptive adjustment of the energy and carbon benchmark value. The method configures adaptive weight coefficients for the energy and carbon efficiency feature vectors, dynamically adjusted to minimize the deviation between historical and measured data. This optimizes the weight ratio of various features based on changes in actual node operating data, improving the accuracy of energy and carbon efficiency feature extraction and making the benchmark value calculation more closely reflect the core energy and carbon consumption characteristics of the nodes. Simultaneously, separate spatial correlation correction factors and environmental dynamic correction factors are constructed to quantify the linkage effects of spatial / functional relationships between nodes and the real-time impact of external temperature, humidity, and personnel density, respectively. This makes the calculation of the dynamic energy and carbon benchmark value more comprehensive and scientifically sound. It can accurately reflect the reasonable energy and carbon consumption benchmark of each virtual energy and carbon node at a specific time, providing accurate and realistic numerical reference for subsequent energy and carbon situation prediction and energy and carbon anomaly tracing, thereby improving the accuracy and scientific nature of energy and carbon management optimization in the entire public building area from the source.

[0018] Let the lobby of an office building on the first floor of a public building area be designated as virtual energy and carbon node i. This node is a high-traffic functional space with no fixed equipment energy consumption fluctuation pattern. A dynamic energy and carbon baseline value needs to be calculated by considering multiple factors. The dynamic energy and carbon baseline value of this node at time t (9:00 AM) during the morning peak on a weekday is now calculated. The specific parameters and calculation process are as follows: Set basic parameters: Total number of energy and carbon efficiency characteristic types M=3, namely energy consumption intensity per unit area m1, carbon emission intensity per person m2, and energy efficiency fluctuation rate m3; The sliding window width W = 7 days (taking data from the same period of the last 7 working days), and sampling is performed once per hour within the window, with a sampling number of N = 168; The adaptive weighting coefficients are obtained by dynamically adjusting the data to minimize the deviation between historical and measured data. ; The association set of node i includes adjacent foreground nodes. Elevator lobby node Spatial coupling coefficient ; Foreground Node Category Static Energy Efficiency Benchmark Value Elevator lobby node Category Static Energy Efficiency Benchmark Value ; time t Real-time carbon intensity , Real-time carbon intensity ; External environment reference baseline value Persons / hour, actual environmental and personnel data at time t: T(t) = 27℃, H(t) = 55%, P(t) = 150 persons / hour; Environmental impact weighting coefficient .

[0019] The historical carbon efficiency characteristic mean part is calculated by extracting the historical data of node i over the past 7 days using a sliding window. Calculate according to the formula: .

[0020] Calculate the spatial correlation correction factor Substituting the data into the formula yields... .

[0021] Computational Environment Dynamic Correction Factor Substituting the data into the formula, we get: .

[0022] Calculate the dynamic energy carbon baseline value of node i at time t: Substitute the data into the formula to obtain... The dynamic energy and carbon baseline value of the office building lobby node i at time t is 10.631866 (unit of comprehensive energy and carbon intensity). This value fully combines the historical operating characteristics of the node, the linkage effect of adjacent nodes, and the actual situation of the morning peak environment and dense personnel. Compared with the traditional static baseline value, it is more in line with the reasonable level of actual energy and carbon consumption of the node at this time.

[0023] The specific process of performing multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4 includes: The specific process of performing multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4 is as follows: First, a time-series prediction model is constructed. The time-series prediction model adopts a combination architecture of temporal convolutional network and gated recurrent unit. The dynamic energy and carbon baseline value, real-time operation data and static feature data of each virtual energy and carbon node are used as input to generate multi-timescale energy and carbon situation prediction results for the ultra-short term, short term and medium term. Secondly, a spatial transmission model is constructed. The spatial transmission model is based on the topology and spatial coupling coefficient of the virtual energy carbon node network. When the deviation between the real-time energy carbon data of any virtual energy carbon node and the dynamic energy carbon benchmark value exceeds a preset threshold, the spatial source tracing mechanism is triggered. The spatial transmission model is used to analyze the transmission path and contribution of the deviation between related nodes and generate an energy carbon anomaly source tracing path. The spatial propagation model performs spatial source tracing calculations, including: Step 1: Calculate the instantaneous deviation of each virtual energy carbon node and define the deviation function. ,in Let be the real-time carbon energy intensity of node i at time t. The dynamic energy carbon baseline value is used, and the deviation function values ​​of all nodes are used to construct the deviation vector. where n is the total number of nodes; Step 2: Construct a spatial propagation model of the bias in the node network, assuming the observed bias vector From the external disturbance vector Generated through coupling and propagation between nodes, the propagation relationship satisfies: Where A is the adjacency matrix. The spatial coupling coefficient matrix, For conduction attenuation factor, It represents the Hadamardi (or Hadama) stack; Step 3: Rewrite the conduction relation in Step 2 as follows: Furthermore, by solving this equation, the contribution of each potential source node can be analytically determined. ; Step 4: Assign contribution Nodes exceeding a preset threshold are marked as sources of anomalies, and a path for tracing the source of energy and carbon anomalies is formed by tracing back along the topological relationships of the adjacency matrix. The above process constructs a time-series prediction model based on a combination of temporal convolutional networks and gated recurrent units. This model, combined with dynamic energy and carbon baseline values, real-time operational data, and external environmental data, enables multi-timescale predictions of energy and carbon status at ultra-short-term, short-term, and medium-term scales. It leverages both the local feature extraction advantages of temporal convolutional networks and the long-sequence dependency capture capabilities of gated recurrent units, resulting in predictions that are both accurate and forward-looking. This overcomes the limitations of traditional single-timescale predictions, which cannot meet the needs of different management stages in public building areas. Simultaneously, a deviation threshold is set to trigger a spatial source tracing mechanism. A standardized deviation propagation function is defined to quantify the degree of energy and carbon anomalies at nodes. A deviation propagation equation is constructed by combining the Hadamard product of the adjacency matrix and spatial coupling coefficient matrix of the virtual energy and carbon node network, and a propagation attenuation factor is introduced to consider the propagation of deviations between nodes. The loss effect of the guide makes the calculation of deviation transmission more consistent with the actual relationship between nodes in the public building area. Then, by analyzing the deviation contribution of each node through the backpropagation algorithm, the source node of energy and carbon anomalies can be accurately located and the deviation transmission path can be traced. This changes the current situation of vague and unquantifiable anomaly investigation. The energy and carbon situation prediction results of multiple time scales can provide data support for the forward-looking regulation of energy and carbon management in public building areas. The accurate energy and carbon anomaly tracing path can clarify the root cause and transmission link of the anomaly, making subsequent energy and carbon regulation instructions more targeted and precise. From the two core dimensions of situation prediction and anomaly location, the forward-looking, accurate and efficient energy and carbon management in public building areas has been improved. This lays a solid data analysis foundation for the subsequent generation of scientific regulation instruction sets by multi-objective collaborative optimization models.

[0024] Continuing with the public building area scenario, the lobby on the first floor of the office building is taken as virtual energy and carbon node i, and the reception desk is taken as node i. Elevator lobby as a node The dynamic energy carbon baseline value of node i at time t (9:00) during the morning rush hour on a weekday. (Comprehensive energy and carbon intensity unit), preset energy and carbon data deviation threshold of 10%, conduction attenuation factor θ=0.8, virtual energy and carbon node network only includes The three nodes are now performing multi-timescale prediction and spatial source tracing of the energy and carbon situation. The specific process and calculations are as follows: Multi-timescale energy and carbon situation prediction: A time-series prediction model combining a temporal convolutional network (TCN) and a gated recurrent unit (GRU) is constructed. The dynamic energy and carbon baseline values ​​of each node, real-time energy and carbon intensity, external environmental data such as temperature at time t (27℃), humidity (55%), and population density (150 people / hour), as well as historical energy and carbon operation data of the nodes are used as model inputs. After model training and inference, the energy and carbon situation prediction results are generated for ultra-short-term (9:00-10:00 within 1 hour), short-term (9:00 to 9:00 the next day within 24 hours), and medium-term (9:00 on the current day to 9:00 on the 7th day within 7 days). The mean value of the ultra-short-term predicted energy and carbon intensity of node i is 10.85, the mean value of the short-term predicted energy and carbon intensity is 10.72, and the mean value of the medium-term predicted energy and carbon intensity is 10.68. This result provides a forward-looking numerical reference for the real-time regulation, daily scheduling, and weekly planning of regional energy and carbon.

[0025] Spatial source tracing of energy and carbon anomalies: Real-time energy and carbon intensity of node i at time t. (Comprehensive energy carbon intensity unit), first calculate the deviation propagation function, substitute the data to obtain That is, 14.81%, which exceeds the preset deviation threshold of 10%, triggering the spatial tracing mechanism; a 3×3 adjacency matrix A is defined (where spatial / functional associations between nodes are denoted as 1, and no association is denoted as 0), because i and Directly related and There is no direct correlation, therefore The spatial coupling coefficient matrix Λ (where elements are the spatial coupling coefficients between nodes, with 0 representing no correlation) is: Calculate the Hadamard product of A and Λ. ; calculate The identity matrix is: Therefore ; Find the inverse matrix of this matrix. (Calculated by matrix inversion); Let the deviation vector of each node be... External disturbance vector According to the deviation transmission equation Substituting the inverse matrix and the deviation vector into the calculation, we can obtain... That is, the deviation contribution of each node is respectively ; The preset contribution threshold is 10%, then the node As the source node of the anomaly, and combining the topological relationships of the adjacency matrix, the backtracking path for the energy and carbon anomaly is as follows: This means that the energy and carbon anomalies at the front-end nodes are transmitted to the lobby nodes through spatial correlation, causing the energy and carbon data at the lobby nodes to exceed the threshold deviation. This tracing path clarifies the root cause of the anomaly and the transmission link, providing a precise basis for subsequent targeted regulation.

[0026] The multi-objective collaborative optimization model preset in step S5 includes objective functions, constraints, and solution algorithms. The objective functions include at least minimizing total regional carbon emissions, minimizing total regional operating costs, and maximizing regional comprehensive energy efficiency. The constraints are constructed in the following ways: based on the energy and carbon situation prediction results, dynamic energy and carbon quota boundaries are set for key nodes in the virtual energy and carbon node network; based on the energy and carbon anomaly tracing path, energy and carbon efficiency recovery targets and penalty coefficients are set for abnormal nodes. The multi-objective collaborative optimization model is solved using an improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm introduces a topological distance weight based on the energy carbon anomaly tracing path into the velocity update formula. The specific process of velocity update is as follows: ; in, Let be the velocity of particle i in the d-th dimension during the (k+1)th iteration. Where w is the contraction factor and w is the inertia weight, and As a learning factor, and It is a random number. This represents the optimal position for an individual particle. To be the globally optimal position The contribution of the anomaly source node j. Let be the shortest path distance between the control scheme represented by particle i and the abnormal node j in the topology. This represents the local optimal control position corresponding to the abnormal node j; The source set is a collection of anomaly source nodes whose contribution exceeds a preset threshold in the energy and carbon anomaly source tracing path; The constructed multi-objective collaborative optimization model incorporates minimizing regional total carbon emissions, minimizing regional total operating costs, and maximizing regional comprehensive energy efficiency into a unified objective system. This overcomes the technical limitations of traditional single-objective optimization, which often suffers from unintended consequences, and achieves a Pareto optimal balance of multi-dimensional benefits. Simultaneously, based on energy and carbon situation prediction results, dynamic energy and carbon quota boundaries are set for key nodes. Combined with energy and carbon anomaly tracing paths, energy and carbon efficiency recovery targets and penalty coefficients are configured for anomalous nodes, ensuring that optimization constraints align with the real-time operational status of regional energy and carbon emissions and the needs for anomaly rectification, making the optimization results more practically feasible. The model employs an improved particle swarm optimization algorithm, innovatively introducing topological distance weights based on energy and carbon anomaly tracing paths into the velocity update formula, integrating the anomaly source node... The contribution value and local optimal control position allow the algorithm iteration process to target the optimal control strategy for abnormal nodes, avoiding the blind optimization of traditional particle swarm optimization and improving the convergence speed and targeting of the control scheme. In addition, the introduction of the source set allows the algorithm to focus only on abnormal source nodes whose contribution value exceeds the threshold, reducing the amount of invalid computation and further improving optimization efficiency. The final generated control instruction set not only meets the global optimal requirements of energy and carbon management in public building areas, but also accurately targets abnormal nodes and related nodes for control, making energy and carbon control measures more scientific, precise and efficient. This provides a practical execution basis for subsequent closed-loop control and effectively ensures the implementation of multiple objectives of energy and carbon emission reduction, cost control and energy efficiency improvement in public building areas.

[0027] Continuing with the aforementioned public building area scenario, let's take the virtual energy carbon node i in the lobby on the first floor of the office building and the abnormal node at the front desk as examples. Elevator lobby node The source set has been determined as the research object. ( Contribution (exceeding the 10% threshold), node i and Topological shortest path distance Corresponding local optimal control position (Right now The static energy efficiency benchmark value is used. Based on the multi-objective collaborative optimization model, an improved particle swarm optimization algorithm is adopted to solve the control instruction set, focusing on the energy and carbon intensity control dimension (d=1) of node i. The specific parameter setting and calculation process are as follows: Setting constraints and core parameters for the multi-objective collaborative optimization model: Based on the energy and carbon situation prediction results, a dynamic energy and carbon quota boundary is set for node i as [10.0, 11.0] (in units of comprehensive energy and carbon intensity), which is considered an abnormal node. The target for energy and carbon efficiency recovery is set at a regression static baseline value of 8.00, with a penalty coefficient of 1.2. Improved Particle Swarm Optimization Algorithm Parameters as Contraction Factors Inertia weight w=0.9, learning factor random numbers (Using fixed values ​​simplifies the calculation; in reality, it's a random number between 0 and 1). The velocity of particle i in dimension d=1 at the k-th iteration. (Combined energy carbon intensity units / iteration), current particle position (i.e., real-time carbon intensity at node it), the optimal position of the individual particle. (i.e., the dynamic energy carbon baseline value at node it), the globally optimal position. (Regional overall optimal energy carbon intensity).

[0028] Calculate the summation terms related to the source set: since the source set contains only Therefore Substituting the data yields .

[0029] Calculate the velocity update value of the improved particle swarm optimization algorithm: Calculate each item in parentheses according to the formula in sequence: ; ; ; ; Summing all the terms, we get 0.18 - 1.6144 - 1.7496 - 0.1660 = -3.35; Multiply by the shrinkage factor to get .

[0030] Generate a set of control instructions: Based on the velocity update value and particle position update logic, the position of particle i in the (k+1)th iteration will move towards the global optimum, the individual optimum, and the local optimum of the abnormal node. Combining the model's multi-objective constraints and iterative convergence judgment, a set of control instructions for this region is finally generated. For abnormal nodes (The front-end) issued an energy carbon intensity control command, reducing its real-time energy carbon intensity from 8.8. Adjust to static benchmark value The rectification period is 1 hour; An energy and carbon intensity control command is issued to node i (lobby). Based on the optimization results, its energy and carbon intensity is controlled from 12.206946 to 10.631866 (comprehensive energy and carbon intensity unit) within the dynamic quota range. This is achieved by adjusting the power of ventilation and lighting equipment. For nodes (Elevator lobby) Maintain the existing energy and carbon emission control status and conduct real-time monitoring. This set of control instructions takes into account the multiple objectives of reducing regional total carbon emissions, controlling operating costs, and improving overall energy efficiency, while also precisely targeting the sources of abnormal emissions. The rectification measures, while aligning with the dynamic energy and carbon quota constraints of node i, possess strong practical enforceability and targeted regulation effects.

[0031] The energy and carbon attribute set of the virtual energy and carbon node in step S1 also includes the real-time energy and carbon intensity, dynamic energy and carbon benchmark value, energy and carbon efficiency rating and life cycle carbon emission factor of the virtual energy and carbon node. The life-cycle carbon emission factor is obtained by calculating the carbon emissions of energy, materials, and water resources consumed by the physical entity or functional space corresponding to the virtual energy carbon node during the upstream production, transportation, and downstream disposal stages. The specific calculation process is as follows: ; in, The life-cycle carbon emission factor of node i. Let be the consumption of node i for the e-th type of energy. The life-cycle carbon emission coefficient for energy type e. Let be the consumption of the m-th type of material at node i. Let m be the carbon emission coefficient for the entire life cycle of the m-th material. Let be the consumption of the w-th type of water resource by node i. Let w be the carbon emission coefficient for the entire life cycle of water resources. The life cycle carbon emission factor is stored in the energy and carbon attribute set and used for multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4. The addition of real-time energy carbon intensity, dynamic energy carbon benchmark value, energy carbon efficiency rating, and full life cycle carbon emission factor to the energy carbon attribute set of virtual energy carbon nodes enriches the representation dimensions of node energy carbon attributes. This allows for a more comprehensive and accurate digital mapping of the energy carbon characteristics of physical entities / functional spaces by virtual energy carbon nodes, breaking through the limitations of traditional energy carbon attributes that only focus on basic operational data. The full life cycle carbon emission factor systematically calculates the carbon emissions of energy, materials, and water resources consumed by the node throughout the entire process of upstream production, transportation, and downstream disposal. This extends energy carbon management from a single operational stage to the entire life cycle, making up for the data shortcomings of traditional methods that only calculate direct carbon emissions. Incorporating the full life cycle carbon emission factor into the energy carbon attribute set and using it for energy... The multi-timescale prediction and spatial source tracing of carbon status allow the prediction process to be revised by incorporating the full-cycle characteristics of carbon emissions, and the source tracing process to analyze the potential causes of energy and carbon anomalies from the source, thus improving the comprehensiveness of status prediction and the depth and scientific nature of anomaly source tracing. The addition of energy and carbon efficiency rating provides a standardized evaluation dimension for the energy and carbon performance of each virtual energy and carbon node, facilitating horizontal comparison, hierarchical control, and targeted optimization of energy and carbon management at each node within the public building area. The new attributes, together with the existing attributes, form a complete energy and carbon attribute system, providing richer and more scientific basic data support for all aspects of the energy and carbon management optimization system in the entire public building area, thereby improving the refinement, comprehensiveness, and scientific nature of energy and carbon management from the data source level.

[0032] Continuing with the aforementioned public building area scenario, let's take the abnormal node at the office building reception as an example. For the accounting object, calculate its life cycle carbon emission factor. The new content of its energy and carbon attribute set is clarified, and the application of this factor in energy and carbon situation prediction and spatial source tracing is explained. The calculation cycle is 1 day, and the specific parameter settings and calculation process are as follows: Define accounting dimensions and basic parameters: Set nodes Energy type E=2 (electricity, natural gas), material type M=2 (office furniture, decorative panels, daily consumption calculated on an annual amortization basis), water resource type W=1 (municipal tap water); the consumption of each category and the corresponding life-cycle carbon emission coefficient are as follows: electricity , ; natural gas (No natural gas consumption at the front desk) ; Daily amortization of office furniture , ; Daily amortization of decorative panels , ; Municipal tap water , .

[0033] Calculating carbon emissions in stages: Carbon emissions in the energy sector sky; Carbon emissions in the materials stage sky; Carbon emissions during the water resources phase sky.

[0034] Calculate the life-cycle carbon emission factor: sum the results of each dimension to obtain... Day, this value is the node. The full life cycle carbon emission factor is stored in the node energy carbon attribute set.

[0035] Improving the carbon attribute set and rating of nodes: Nodes The new addition to the energy-carbon attribute set is: real-time energy-carbon intensity. The dynamic energy carbon baseline value is combined with the calculation. Life cycle carbon emission factor sky; According to the regional energy and carbon efficiency rating standard (AE level, A level is the best, with a deviation of ≤5% and a low level of carbon emission factor throughout the life cycle in the region), due to node... The real-time energy carbon intensity deviates from the dynamic benchmark by 7.32%, and the life-cycle carbon emission factor is at a medium level in the region. Its energy carbon efficiency is rated as C.

[0036] Practical application of the life cycle carbon emission factor: Incorporate energy and carbon situation prediction model inputs and correct node values. The ultra-short-term energy carbon situation forecast value, from the original Revised to This improved the accuracy of predictions; During the process of tracing the source of carbon anomalies, combined with Analysis revealed that the nodes The carbon emission coefficient of the upstream power generation stage of electricity consumption is relatively high. Higher than the regional average This is a potential cause of its carbon intensity exceeding the benchmark value, allowing the source analysis to be extended from the simple operation stage to the entire life cycle stage, supplementing the analysis dimension of the root cause of the anomaly, and adding an optimization direction of "changing the green electricity procurement channel" for subsequent regulation, making the regulation strategy more comprehensive.

[0037] After generating the control instruction set in step S5, the following steps are also included: Based on game theory, a multi-party coordination mechanism is constructed, defining different owners, property management companies, and energy suppliers within the public building area as game participants, and constructing a game utility function based on each participant's energy and carbon quotas, energy comfort, and economic costs. Energy comfort level is obtained by weighted calculation based on a combination of indoor temperature, humidity, illuminance, and air quality index. By solving for the Nash equilibrium, the control instruction set is negotiated and corrected to generate an optimized control instruction set that is agreed upon by multiple parties. The process of constructing the game utility function is as follows: ; in, Let be the utility function value of participant s. and For participant s, the weighting coefficients represent their preferences for carbon emissions, comfort, and economic costs. This represents the actual carbon emissions of participant s. The energy and carbon allowances allocated to it, For actual comfort indicators, The target comfort index is obtained by weighting a preset comfort target range with the personalized preferences of participant s. For actual economic costs, The benchmark economic cost is used as the basis for solving the Nash equilibrium. The solution process adopts a distributed iterative algorithm. In each round of iteration, each participant updates its own policy by maximizing its own utility function based on the current policy of other participants, until the rate of change of all participants' policies is lower than the preset convergence threshold. After generating the initial control instruction set, a multi-stakeholder coordination mechanism is constructed based on game theory. Owners, property management companies, and energy suppliers within the public building area are defined as game participants. A game utility function is constructed using the core demands of each participant—energy and carbon emission quotas, energy comfort, and economic costs. This overcomes the limitations of traditional energy and carbon emission control instructions that focus solely on global technical optimization while neglecting the interests of multiple stakeholders, leading to difficulties in implementation. Energy comfort is objectively quantified through weighted calculations of indoor temperature, humidity, illuminance, and air quality index, providing a unified numerical reference for multi-party negotiations. The game utility function accurately quantifies the degree of loss due to deviations in carbon emissions from quotas and comfort levels from targets using quadratic terms. It also adapts to the individualized preference weights of each participant, aligning with their specific interests and making utility calculations more aligned with their actual operational needs through distributed iteration. The algorithm solves for Nash equilibrium, allowing each participant to maximize their own utility and update their own strategy in each iteration based on the current strategies of other participants, until the rate of change of the strategy falls below the convergence threshold. This achieves autonomous game and dynamic balance of interests among multiple parties, rather than a single forced regulation. The negotiated correction of the initial regulation instruction set based on the Nash equilibrium result retains the global optimization goals of carbon emission reduction, cost control, and energy efficiency improvement in public building areas, while also taking into account the energy comfort of owners, the operation and management costs of properties, and the energy supply revenue demands of energy suppliers. This significantly improves the acceptability and actual enforceability of regulation instructions. At the same time, the convergence of Nash equilibrium ensures the efficiency of multi-party negotiation and the stability of the results, upgrading the carbon emission control of public building areas from a single technical optimization to a comprehensive optimization that combines technology and interests, effectively ensuring the implementation and long-term sustainability of carbon emission control measures.

[0038] Continuing with the aforementioned scenario of office building public areas, an initial set of control instructions has been generated based on a multi-objective collaborative optimization model: abnormal nodes in the front end Energy carbon intensity from Adjustment to ; The carbon intensity of the lobby node i was adjusted from 12.206946 units to 10.631866 units. Elevator lobby node To maintain the current operational status, a multi-party coordination mechanism will be constructed based on game theory, defining office building owners. Property management Urban energy suppliers For the game participants, the game utility function is constructed and the Nash equilibrium is solved. The initial control instruction set is then modified. The specific parameter setting, calculation, and modification process is as follows: Determine the core basic parameters: Energy comfort level D is calculated by weighting indoor temperature (weight 0.4), humidity (weight 0.2), illuminance (weight 0.3), and air quality index (weight 0.1), with a maximum score of 10 points. This represents the target comfort level for all three participating parties. All are set at 8.5 points; Carbon allowances for each participating party Benchmark economic cost Based on the regional energy and carbon management plan and actual operational data, the owner... Undertake regional core carbon quotas Day, benchmark cost Yuan / day; Property management Bearing carbon quotas in the operation process Day, benchmark cost Yuan / day; Energy suppliers No direct carbon emission allowances ( The revenue is reflected in the energy supply cost, with the benchmark cost as the basis. Yuan / day; Each participating party sets its preference weighting coefficients for carbon emissions, energy comfort, and economic costs based on its own interests. (Owner) Focusing on energy-saving comfort, ; Property A balanced approach should be taken between carbon allowances and operating costs. ; Energy companies Focus on energy supply benefits (economic costs). ; The convergence threshold for the Nash equilibrium distributed iteration is preset to be the policy change rate < 0.01.

[0039] Calculate the actual values ​​of each participant under the initial control command: the initial control is due to the adjustment of equipment power, owner Actual carbon emissions Heavens, actual comfort level Part, actual operating costs Yuan / day; Property Actual carbon emissions Heavens, actual comfort level In terms of actual management costs Yuan / day; Energy companies Actual energy supply cost Yuan / day (no direct carbon emissions or comfort requirements) ).

[0040] Calculate the game utility value of each participant under the initial control: calculate step by step according to the utility function formula. owner : , ; Property : ; Energy companies : (When the denominator of the carbon emission item is 0, it is directly counted as 0; the comfort item is converted according to the target value).

[0041] Distributed iterative solution to Nash equilibrium: Each participant adjusts its own demands based on the current strategies of others, corrects the control parameters, and after 3 rounds of iteration, the rate of change of the strategy drops to 0.008 (below the convergence threshold), reaching Nash equilibrium and determining the final acceptable actual value for each participant: Owner Actual carbon emissions Heavens, actual comfort level Points, actual cost Yuan / day; Property Actual carbon emissions Heavens, actual comfort level Points, actual cost Yuan / day; Energy companies Actual energy supply cost Yuan / day, at which point the utility values ​​of each party are respectively All of these improvements are significant compared to the initial state, maximizing the interests of all parties.

[0042] Based on Nash equilibrium, the control instruction set is modified: the initial instructions are fine-tuned by combining the equilibrium results to generate an optimized control instruction set agreed upon by multiple stakeholders. abnormal nodes in the front end Energy carbon intensity from Adjustment to (The initial target should be relaxed appropriately, and the range of equipment adjustments should be reduced.) The carbon intensity of lobby node i was adjusted from 12.206946 comprehensive carbon intensity units to 10.8 comprehensive carbon intensity units (within the dynamic quota boundary [10.0, 11.0], to ensure indoor illuminance and ventilation efficiency); elevator hall node Fine-tuning the lighting power will optimize carbon intensity by 5%, while also meeting the needs of improving property energy efficiency; Energy suppliers This solution provides office buildings with a 10% increase in green electricity ratio, balancing the comfort of building owners with the need for carbon reduction. The optimized control instruction set retains the core optimization objectives of energy and carbon management while also considering the actual interests of all three participating parties, significantly improving the enforceability of the instructions and ensuring the effective implementation of energy and carbon control measures.

[0043] Step S3, the construction of the dynamic baseline, also includes: Identify renewable energy nodes and energy storage nodes within public building areas, and generate output confidence intervals for renewable energy nodes and dispatchable capacity intervals for energy storage nodes based on meteorological forecast data and historical output data. The output confidence interval and dispatchable capacity interval are used as dynamic constraints and embedded in the generation process of dynamic energy carbon benchmark values ​​to correct the rule logic of energy supply and demand balance in the association rule set. The process of generating the output confidence interval is as follows: ; in, The predicted power output of renewable energy nodes at time t is obtained through a weather-power output mapping model based on a long short-term memory network. The standard deviation of the prediction error is obtained through statistical analysis of historical prediction errors, and z is the coefficient corresponding to the confidence level. The schedulable capacity range is determined by both the state of charge and charge / discharge energy constraints, and its calculation process is as follows: ; in and The minimum and maximum dispatchable capacity of the energy storage node at time t (expressed as a percentage of state of charge). , These are the minimum and maximum allowable states of charge for the energy storage system. Let be the current state of charge at time t. , This represents the maximum discharge and charge power. For scheduling time intervals, This refers to the rated capacity of the energy storage system. The dynamic baseline construction incorporates the identification and quantitative analysis of renewable energy nodes and energy storage nodes within public building areas. Based on meteorological forecasts and historical data, it generates output confidence intervals for renewable energy nodes and determines the dispatchable capacity range for energy storage nodes by combining state of charge and charge / discharge constraints. This overcomes the technical limitations of traditional dynamic energy carbon baseline construction, which fails to consider regional renewable energy volatility and energy storage dispatch and regulation capabilities. By embedding these two types of intervals as dynamic constraints into the dynamic energy carbon baseline generation process and correcting the energy supply and demand balance logic of the associated rule set, the calculation of the dynamic energy carbon baseline not only considers node operation, environmental parameters, and the linkage of associated nodes, but also aligns with the actual energy supply capacity and dispatch potential of the region, significantly improving the adaptability of the baseline to the regional energy structure. The method achieves precise quantification of power output fluctuations by combining the LSTM weather-output mapping model with prediction error statistics. The energy storage dispatchable capacity range clarifies the actual dispatchable boundary through coupled calculation of state of charge, charging and discharging power, and time, making the consideration of energy supply and demand balance more scientific and practical. This method promotes the priority consumption and efficient utilization of renewable energy in public building areas, optimizes the regional energy consumption structure, reduces dependence on traditional fossil fuels, and further contributes to regional carbon emission reduction from the energy supply side. At the same time, it enables subsequent energy carbon situation prediction and multi-objective collaborative optimization to formulate strategies based on actual energy supply capacity, avoiding the problem of mismatch between control instructions and regional energy reality, and improving the rationality, sustainability, and energy utilization efficiency of the entire energy carbon management optimization system.

[0044] Continuing the aforementioned scenario of the public building area in the office building, in addition to the existing lobby node i and reception node... Elevator lobby node Based on this, the rooftop photovoltaic panels of the office building were identified as renewable energy nodes. The building's internal energy storage battery pack serves as an energy storage node. The calculation time is still set at t (9:00 AM) during the morning rush hour on weekdays. The dynamic range of the two nodes is used as a constraint to embed the dynamic energy-carbon baseline value of the lobby node i. The generation process involves revising the energy supply and demand logic of the association rule set. The specific parameter settings, calculations, and constraint embedding processes are as follows: Setting up renewable energy nodes Core parameters and output confidence interval calculation: Based on a meteorological-output mapping model using a Long Short-Term Memory (LSTM) network, combined with meteorological data such as light intensity and temperature at time t, the photovoltaic node is obtained. Predicted output at time t ; By comparing the predicted and actual power output of the photovoltaic panel over the past 30 days, the standard deviation of the prediction error was statistically calculated. ; Choosing a 95% confidence level, the corresponding coefficient z = 1.96; according to the formula... Substituting the data, we can calculate... photovoltaic nodes The confidence interval for output at time t is: .

[0045] Setting up energy storage nodes Core parameters and dispatchable capacity range calculation: Minimum allowable state of charge for energy storage system Maximum state of charge ; Energy storage node at time t The current state of charge (SOC(t)) is 60%. Maximum discharge power of energy storage battery Maximum charging power Scheduling time interval ; Rated capacity of energy storage system ; Calculate step by step according to the formula. Calculate the minimum schedulable capacity: , max(20%, 40%) = 40%, that is ; Calculate the maximum schedulable capacity: , ,Right now ; Finally, the energy storage node is obtained. The schedulable capacity range at time t is [40%, 76%].

[0046] Dynamic constraint embedding and association rule set modification: photovoltaic nodes Confidence interval of output and energy storage nodes The schedulable capacity range [40%, 76%] is used as a dynamic constraint, and the dynamic energy-carbon baseline value of lobby node i is embedded. Generation process; Corrected the rule logic in the original association rule set that "energy supply and demand balance only considers municipal power grid supply", and added "prioritize the absorption of photovoltaic nodes at time t". The output is 4.02-5.98kW, and the shortfall is covered by energy storage nodes. The rule is to "supplement power supply within the 40%-76% charge range, and then supply the remaining portion from the municipal power grid"; Simultaneously, based on the renewable energy consumption ratio, the adaptive weighting coefficients of the carbon efficiency feature vector of node i are dynamically adjusted, and the weight of energy intensity per unit area is adjusted. The weighting of carbon emission intensity per person-trip has been reduced from 0.4 to 0.35. The weighting of energy efficiency volatility was reduced from 0.4 to 0.35. The value was increased from 0.2 to 0.3 to better reflect the fluctuating characteristics of renewable energy output.

[0047] The revised dynamic energy-carbon baseline value is recalculated: Based on the adjusted weighting coefficients, the mean part of the historical energy-carbon efficiency characteristic of node i is recalculated. ; Spatial correlation correction factor Environmental dynamic correction factor Remain unchanged; according to the formula Substituting the data yields the corrected result. (Unit of total energy carbon intensity).

[0048] Effect verification: The revised dynamic energy carbon baseline value of node i, 9.312099, is closer to the actual energy supply of the office building at time t, which prioritizes the consumption of photovoltaic and energy storage dispatch supplements. It avoids the problem of the baseline value being too high due to the failure of the traditional baseline value to take into account the consumption of renewable energy. This makes the threshold for subsequent energy carbon anomaly judgment more reasonable and allows the energy carbon control instructions for node i to be formulated based on the actual clean energy consumption capacity. This effectively improves the operability of the control strategy and the regional energy utilization efficiency.

[0049] The specific process of updating the energy carbon digital twin based on execution feedback in step S6 includes: Calculate the response deviation between the execution feedback and the control instruction set. If the response deviation exceeds the preset deviation tolerance range, trigger the model self-correction process. The model self-correction process uses an incremental learning algorithm to fine-tune the internal parameters of the time series prediction model and the multi-objective collaborative optimization model online using response bias, and stores the response bias and its corresponding scene features in an anomaly case library for optimizing the subsequent construction of dynamic baselines. The parameter update process of the incremental learning algorithm is as follows: ; in, For the updated model parameter set, Here is the current model parameter set, and t is the number of exception cases that have been processed. The gradient of the loss function with respect to the model parameters is given. The loss function is constructed based on the weighted mean square error of response bias and scene features, and the weight coefficients are dynamically adjusted according to the similarity between scene features and historical cases in the abnormal case library. In the closed-loop control stage, the response deviation between the execution feedback and the control command set is accurately calculated, and a deviation tolerance range is set. This overcomes the technical limitations of traditional energy and carbon management, which only issues control commands without precise effect verification. When the deviation exceeds the limit, the model self-correction process is triggered, realizing the dynamic iterative optimization of the energy and carbon management model, rather than a static and fixed application. The use of incremental learning algorithms allows the internal parameters of the time-series prediction model and the multi-objective collaborative optimization model to be fine-tuned online without the need for full retraining of the model. This significantly reduces the computational and time costs of model updates and improves the real-time performance and efficiency of correction. The parameter update formula of incremental learning incorporates the number of processed abnormal cases as the denominator, allowing the parameter adjustment range to gradually become more rational with the accumulation of cases, avoiding drastic parameter fluctuations caused by a single deviation. The loss function is constructed based on the weighted mean square error of the response deviation and scene characteristics, with weight coefficients... The system dynamically adjusts parameters based on the similarity between the scenario and historical cases, ensuring that parameter updates better align with the characteristics of actual operating scenarios and accurately match energy and carbon management needs under different operating conditions. Simultaneously, response deviations and corresponding scenario characteristics are stored in an anomaly case library, supplementing the subsequent construction of dynamic baselines with real-world anomaly data. This continuously optimizes the generation logic of the dynamic baseline, making the baseline values ​​more adaptable to anomaly scenarios. The combination of model self-correction and the anomaly case library enables real-time synchronous updates of the energy and carbon digital twin and the physical entity's operating status, ensuring that the twin always possesses accurate mapping, analysis, and prediction capabilities. From a closed-loop optimization perspective, this achieves self-learning, self-optimization, and self-adaptation in energy and carbon management of public building areas, continuously improving the accuracy of subsequent energy and carbon situation prediction, anomaly tracing, and control command generation. This ensures the long-term stability, efficiency, and scenario adaptability of the entire energy and carbon management optimization system.

[0050] Continuing with the aforementioned scenario of the office building's public area, for the virtual energy and carbon node i in the lobby, an optimized control command has been generated based on multi-stakeholder negotiation: the energy and carbon intensity at time t is adjusted to 9.312099 (units of comprehensive energy and carbon intensity). The preset tolerance range for the response deviation of the comprehensive energy and carbon intensity of this node is ±0.2. The energy and carbon management model has currently processed 20 abnormal cases. After executing the control command, the physical entity execution feedback is obtained, and the response deviation calculation, model self-correction, and twin update are completed. The specific parameter settings, calculation, and execution process are as follows: Obtain execution feedback and calculate response deviation: After the control command is issued to the office building equipment execution mechanism, the energy carbon intensity of the actual execution feedback of node i at time t+1 is 9.652099 (comprehensive energy carbon intensity unit). The response deviation is calculated according to the formula = |actual execution value - control command value| = |9.652099 - 9.312099| = 0.35. This value exceeds the deviation tolerance range of ±0.2, triggering the model self-correction process and updating the energy carbon digital twin.

[0051] Setting the core parameters of the incremental learning algorithm: the current model parameter set of node i. (i.e., the adaptive weighting coefficients of the carbon efficiency eigenvector) The gradient of the loss function with respect to the model parameters is obtained by calculating the gradient of the loss function with respect to the model parameters. The gradient value reflects the degree of influence of each parameter on the response deviation. The scenario features are "weekday morning peak, photovoltaic output 4.02kW, energy storage state of charge 60%, and personnel density 150 people / hour". The scenario features are matched with historical cases in the abnormal case library for similarity, and the weight coefficient of the weighted mean square error is determined to be 0.95 (high similarity).

[0052] Model parameter updates for incremental learning: Formula for updating parameters based on incremental learning. Step-by-step, substitute the data to calculate the updated values ​​of each parameter: renew: ; renew: ; renew: ; Finally, the updated model parameter set is obtained. This parameter set will be directly applied to the subsequent calculations of the time series prediction model and the multi-objective collaborative optimization model.

[0053] Update the energy carbon digital twin and store it in the exception case library: update the model parameter set. The association rule set of node i in the energy carbon digital twin is written, and the energy supply and demand association logic between the photovoltaic and energy storage nodes and node i in the twin is simultaneously corrected to match the parameter system of the twin with the execution feedback of the physical entity; at the same time, the response deviation of 0.35, the corresponding scenario characteristics "weekday morning peak, photovoltaic output 4.02kW, energy storage state of charge 60%, personnel density 150 people / hour", and the parameters before and after the update are updated. and They are also stored in the exception case library to provide real exception data support for the subsequent construction of dynamic baselines in this scenario.

[0054] The updated model parameter set makes the calculation of the dynamic energy and carbon baseline value of node i more consistent with the abnormal operating conditions of "high population density during the morning peak + clean energy consumption". After parameter correction, the energy and carbon intensity of node i at the next moment is predicted to be 9.63, and the deviation from the actual operating value is reduced to 0.022, which is far below the deviation tolerance range. This proves that the self-correction of the incremental learning model effectively improves the accuracy of model prediction and regulation. At the same time, the synchronous update of the energy and carbon digital twin ensures its accurate mapping to the physical entity, providing more realistic digital support for each link of subsequent energy and carbon management optimization.

[0055] The method also includes step S7: generating and outputting an energy and carbon management optimization report based on the energy and carbon digital twin. The energy and carbon management optimization report includes: an energy and carbon leakage heat map generated based on the energy and carbon anomaly tracing path, an energy and carbon emission reduction potential analysis generated based on a multi-objective collaborative optimization model, and an energy and carbon efficiency improvement curve generated based on historical control command set and execution feedback. The carbon leakage heatmap is generated by visually rendering it on a building information model by associating the degree of carbon anomaly of each virtual carbon node with spatial coordinates. Its rendering parameters are obtained through the following process: ; in, Let i be the thermal rendering intensity of node i. Let i be the energy-carbon anomaly sensitivity coefficient. This is a normalization function that maps the degree of anomaly to the visual rendering range. The analysis of carbon emission reduction potential is generated by sensitivity analysis of the Pareto front of the multi-objective synergistic optimization model, identifying the combination of control measures that contributes the most to each objective function and their corresponding marginal emission reduction costs; the energy and carbon efficiency improvement curve is generated by fitting the cumulative change rate of unit energy consumption output or unit carbon emission output in each cycle after sorting the historical control instruction set by time series, where unit energy consumption output is the ratio of regional total output to total energy consumption, and unit carbon emission output is the ratio of regional total output to total carbon emissions; Based on a digital twin of energy and carbon data, this report generates and outputs an energy and carbon management optimization report that includes an energy and carbon leakage heat map, energy and carbon emission reduction potential analysis, and energy and carbon efficiency improvement curves. This overcomes the limitations of traditional energy and carbon management, which presents results only as scattered data and lacks intuitive visualization and systematic quantitative analysis. The energy and carbon leakage heat map correlates the degree of energy and carbon anomalies at each virtual energy and carbon node with spatial coordinates and visualizes it on a building information model. By combining the sigmoid normalization function with the node energy and carbon anomaly sensitivity coefficient, it transforms abstract energy and carbon anomaly values ​​into intuitive spatial thermal distributions, allowing managers to quickly locate high-risk nodes and areas, improving the efficiency and intuitiveness of anomaly detection. The energy and carbon emission reduction potential analysis, through sensitivity analysis of the Pareto front of a multi-objective collaborative optimization model, accurately identifies the combination of control measures that contributes most to carbon emissions, operating costs, and overall energy efficiency, along with their corresponding marginal emission reduction costs, providing a comprehensive view of energy efficiency. The report provides a quantitative basis for formulating phased and economical emission reduction strategies for public building areas. The energy and carbon efficiency improvement curve, by fitting the cumulative change rate of unit energy consumption output and unit carbon emission output within historical control cycles, intuitively demonstrates the long-term implementation effect of energy and carbon management control measures, enabling quantitative review and trend prediction of control effectiveness. The three types of reports form a complete analysis system for energy and carbon management from three dimensions: spatial anomaly location, future potential exploration, and historical effect review. At the same time, the parameters of heat map rendering, the marginal cost of emission reduction potential, and the cumulative change rate of efficiency curves are all calculated through standardized formulas, making the report content scientific, quantitative, and repeatable. This provides comprehensive and accurate decision support for short-term rectification, medium-term planning, and long-term optimization of energy and carbon management in public building areas, greatly improving the intuitiveness, scientificity, and feasibility of energy and carbon management decisions, and making the effects of energy and carbon management optimization quantifiable, traceable, and iterative.

[0056] Continuing with the aforementioned scenario of a public building area in an office building, let's take virtual energy and carbon node i (lobby). (front desk), Taking the elevator lobby as the core analysis object, and combining the full data of time t and historical control, the parameters for rendering the energy and carbon leakage heat map are calculated sequentially, sensitivity analysis of energy and carbon emission reduction potential is carried out, energy and carbon efficiency improvement curves are fitted, and a complete energy and carbon management optimization report is generated. The specific parameter setting, calculation and analysis process is as follows: Generation and rendering of carbon leakage thermal map intensity calculation: Setting basic parameters: real-time carbon intensity at node i at time t. Dynamic energy carbon benchmark ,node Real-time carbon intensity Dynamic energy carbon benchmark ,node Real-time carbon intensity Dynamic energy carbon benchmark ; The carbon anomaly sensitivity coefficient is set according to the importance of node functions. (Densely populated, highly sensitive) (Service core nodes, extremely high sensitivity) (Passage node, medium sensitivity); Calculated according to the rendering intensity formula, where the sigmoid function is: Calculate the rendering intensity of each node in sequence: Node i: , ; node : ; node : ; Visual rendering: The rendering intensity of each node is mapped to the spatial coordinates of the first floor of the office building in the Building Information Model (BIM). The rendering intensity ranges from 0 to 1, with the color being darker as the value approaches 1. Node i is rendered in light orange-red. Rendered with light orange-red, nodes A light yellow color scheme is used to generate a thermal map of energy and carbon leakage, which visually shows the spatial distribution characteristics of energy and carbon anomalies in the first-floor area, which gradually decrease from the lobby to the reception desk and then to the elevator hall.

[0057] Analysis of carbon emission reduction potential: Sensitivity analysis of the Pareto front of the multi-objective synergistic optimization model was conducted, and three types of core control measures combinations were selected: Combination 1 is the "front-end" "Equipment power optimization + lobby i-green electricity consumption increased by 20%", combination 2 is "elevator hall" Intelligent lighting control + energy storage Peak-valley scheduling", combination 3 is "photovoltaic" "Power output optimization + energy-saving retrofitting of equipment across the entire region"; Calculate the target contribution and marginal emission reduction cost for each combination: Combination 1 contributes 35% to minimizing total regional carbon emissions, 15% to minimizing operating costs, and 20% to maximizing overall energy efficiency, with a marginal emission reduction cost of 80 yuan / ; Combination 2 corresponds to contributions of 10%, 30%, and 15%, respectively, with a marginal emission reduction cost of 120 yuan. ; Combination 3 corresponds to contribution rates of 45%, 25%, and 50%, with a marginal emission reduction cost of 50 yuan / ; Analysis conclusion: Combination 3 is the optimal combination of control measures, with the highest contribution to each target and the lowest marginal emission reduction cost. In the short term, the region can prioritize the implementation of Combination 1 to quickly reduce carbon emissions at core nodes. In the medium term, Combination 3 can be used to optimize photovoltaic output and carry out energy-saving retrofits of equipment. In the long term, Combination 2's peak-valley scheduling strategy can be integrated to achieve a balance between emission reduction and cost.

[0058] Fitting the carbon efficiency improvement curve: Statistical Period and Indicators: The statistical period is set based on the energy and carbon control cycle of the office building over the past 6 months, divided into 6-month periods. The unit energy consumption output (total regional operating revenue / total energy consumption, unit: yuan / kWh) and unit carbon emission output (total regional operating revenue / total carbon emissions, unit: yuan / kWh) are calculated for each period. The data for each period are as follows: Cycle 1: Output per unit of energy consumption 2.0, output per unit of carbon emissions 15; Period 2: 2.2, 16.5; Period 3: 2.3, 17; Period 4: 2.5, 18.5; Period 5: 2.6, 19; Period 6: 2.8, 20.5; Fitting the cumulative rate of change curve: Plot the period on the horizontal axis and the cumulative rate of change of each period's index relative to period 1 on the vertical axis. The formula for the cumulative rate of change is: The cumulative change rates of unit energy consumption output for each period were calculated to be 10%, 15%, 25%, 30%, and 40%, and the cumulative change rates of unit carbon emission output were 10%, 13.33%, 23.33%, 26.67%, and 36.67%, respectively. Curve Fitting and Analysis: A univariate linear regression was used to fit the cumulative rate of change, resulting in the following curve fitting equation for the increase in output per unit of energy consumption: ( The equation for the curve fitting of the increase in output per unit of carbon emissions is: ( The fitted curve shows a significant linear upward trend, proving that the energy and carbon regulation measures in the past 6 months have achieved a steady improvement in energy and carbon efficiency. Moreover, the rate of increase in output per unit of energy consumption is slightly higher than that of output per unit of carbon emissions. In the future, the overall trend of energy and carbon efficiency improvement can be optimized by further increasing the proportion of green electricity consumption so that the rate of increase in output per unit of carbon emissions matches that of output per unit of energy consumption.

[0059] Report Integration: This report integrates the energy and carbon leakage heat map, emission reduction potential analysis, control combinations and marginal costs, energy and carbon efficiency improvement curves, and fitting equations into a comprehensive energy and carbon management optimization report for the office building's public areas. It clearly identifies the lobby and reception area as the key areas for short-term rectification. The energy and carbon anomaly control plan focuses on optimizing photovoltaic output and energy-saving equipment upgrades in the medium term, while the long-term optimization direction is to improve energy storage peak-valley scheduling and increase the proportion of green electricity consumption, providing comprehensive, quantitative, and intuitive support for subsequent energy and carbon management decisions for this office building.

[0060] A public building area energy and carbon management optimization system, the system comprising: The virtual node network construction module is used to construct a virtual energy and carbon node network in a public building area. The virtual energy and carbon node network consists of multiple virtual energy and carbon nodes. Each virtual energy and carbon node corresponds to a physical entity or a functional space in the public building area and is configured with a unique identifier, a set of energy and carbon attributes, and a set of association rules. The data mapping and twin generation module is connected to the virtual node network construction module. It is used to obtain real-time operation data and static feature data of each physical entity and functional space within the public building area. Based on the virtual energy and carbon node network, the real-time operation data and static feature data are mapped to the corresponding virtual energy and carbon nodes to generate an initialized energy and carbon digital twin. The dynamic baseline construction module, connected to the data mapping and twin generation module, is used to perform multi-source data fusion and dynamic baseline construction in the energy and carbon digital twin based on the association rule set of each virtual energy and carbon node, and generate dynamic energy and carbon baseline values ​​for each virtual energy and carbon node. The energy and carbon situation prediction and source tracing module is connected to the dynamic baseline construction module. Based on the dynamic energy and carbon baseline values, real-time operation data and static characteristic data of each virtual energy and carbon node, it performs multi-timescale prediction and spatial source tracing of energy and carbon situation by coupling the time series prediction model and the spatial transmission model, and generates energy and carbon situation prediction results and energy and carbon anomaly source tracing paths. The multi-objective collaborative optimization module is connected to the energy and carbon situation prediction and source tracing module. It is used to call the preset multi-objective collaborative optimization model based on the energy and carbon situation prediction results and the energy and carbon anomaly source tracing path to generate a set of control instructions for at least one virtual energy and carbon node in the virtual energy and carbon node network. The closed-loop control and twin update module, connected to the multi-objective collaborative optimization module, is used to send the control command set to the corresponding physical entity actuator, receive execution feedback, update the energy carbon digital twin according to the execution feedback, and form closed-loop control. The report generation and output module is connected to the closed-loop control and twin update module. Based on the energy and carbon digital twin, it generates and outputs an energy and carbon management optimization report. The energy and carbon management optimization report includes: an energy and carbon leakage heat map generated based on the energy and carbon anomaly tracing path, an energy and carbon emission reduction potential analysis generated based on the multi-objective collaborative optimization model, and an energy and carbon efficiency improvement curve generated based on the historical control command set and execution feedback.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for optimizing energy and carbon management in public building areas, characterized in that, Includes the following steps: Step S1: Construct a virtual energy and carbon node network for the public building area. The virtual energy and carbon node network consists of multiple virtual energy and carbon nodes. Each virtual energy and carbon node corresponds to a physical entity or a functional space within the public building area and is configured with a unique identifier, a set of energy and carbon attributes, and a set of association rules. Step S2: Obtain real-time operation data and static feature data of each physical entity and functional space within the public building area. Based on the virtual energy carbon node network, map the real-time operation data and static feature data to the corresponding virtual energy carbon nodes to generate an initialized energy carbon digital twin. Step S3: In the energy carbon digital twin, based on the association rule set of each virtual energy carbon node, perform multi-source data fusion and dynamic baseline construction to generate dynamic energy carbon baseline values ​​for each virtual energy carbon node. Step S4: Based on the dynamic energy and carbon baseline values, real-time operation data and static characteristic data of each virtual energy and carbon node, perform multi-timescale prediction and spatial source tracing of energy and carbon status by coupling the time series prediction model and the spatial transmission model, and generate energy and carbon status prediction results and energy and carbon anomaly source tracing paths. Step S5: Based on the energy and carbon situation prediction results and the energy and carbon anomaly tracing path, call the preset multi-objective collaborative optimization model to generate a set of control instructions for at least one virtual energy and carbon node in the virtual energy and carbon node network. Step S6: Send the control command set to the corresponding physical entity actuator, receive execution feedback, update the energy carbon digital twin according to the execution feedback, and form a closed-loop control.

2. The method for optimizing energy and carbon management in public building areas according to claim 1, characterized in that: Step S3 generates the dynamic energy-carbon baseline value for each virtual energy-carbon node, specifically including: For any virtual energy carbon node, obtain its historical concurrent operating data and the real-time operating data of its associated nodes that have spatial or functional connections with it. The energy and carbon efficiency feature vector is extracted from historical data of the same period through a sliding window mechanism. The energy and carbon efficiency feature vector includes energy consumption intensity per unit area, carbon emission intensity per person-time, and energy efficiency fluctuation rate. Then, by combining the real-time operating data of the associated nodes, the current external environmental parameters, and the personnel flow density data, a weighted regression algorithm is used to calculate the dynamic correction factor; A dynamic energy-carbon baseline value is generated based on the energy-carbon efficiency feature vector, the dynamic correction factor, and the static feature data of the node.

3. The method for optimizing energy and carbon management in public building areas according to claim 2, characterized in that: The specific process of performing multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4 includes: First, a time-series prediction model is constructed. The time-series prediction model adopts a combination architecture of temporal convolutional network and gated recurrent unit. The dynamic energy and carbon baseline value, real-time operation data and static feature data of each virtual energy and carbon node are used as input to generate multi-timescale energy and carbon situation prediction results for the ultra-short term, short term and medium term. Secondly, a spatial transmission model is constructed. The spatial transmission model is based on the topology and spatial coupling coefficient of the virtual energy carbon node network. When the deviation between the real-time energy carbon data of any virtual energy carbon node and the dynamic energy carbon benchmark value exceeds a preset threshold, the spatial source tracing mechanism is triggered. The spatial transmission model is used to analyze the transmission path and contribution of the deviation between related nodes and generate an energy carbon anomaly source tracing path. The spatial propagation model performs spatial source tracing calculations, including: Step 1: Calculate the instantaneous deviation of each virtual energy carbon node, define the deviation function, and construct a deviation vector from the deviation function values ​​of all nodes; Step 2: Construct a spatial propagation model of the deviation in the node network, assuming that the observed deviation vector is generated by the external disturbance vector through the coupling between nodes, and the propagation relationship meets the preset requirements; Step 3: Rewrite the transmission relationship in Step 2 to obtain the rewritten equation, and solve the rewritten equation to analyze the contribution of each potential source node; Step 4: Mark nodes whose contribution exceeds the preset threshold as anomaly sources, and trace back along the topological relationship of the adjacency matrix to form an energy and carbon anomaly tracing path.

4. The method for optimizing energy and carbon management in public building areas according to claim 3, characterized in that: The multi-objective collaborative optimization model preset in step S5 includes objective functions, constraints, and solution algorithms. The objective functions include at least minimizing total regional carbon emissions, minimizing total regional operating costs, and maximizing regional comprehensive energy efficiency. The constraints are constructed in the following manner: Based on the energy and carbon situation prediction results, dynamic energy and carbon quota boundaries are set for key nodes in the virtual energy and carbon node network; based on the energy and carbon anomaly tracing path, energy and carbon efficiency recovery targets and penalty coefficients are set for abnormal nodes. The multi-objective collaborative optimization model is solved using an improved particle swarm optimization algorithm, which introduces a topological distance weight based on the energy carbon anomaly tracing path into the velocity update formula.

5. The method for optimizing energy and carbon management in public building areas according to claim 4, characterized in that: The energy and carbon attribute set of the virtual energy and carbon node in step S1 also includes the real-time energy and carbon intensity, dynamic energy and carbon benchmark value, energy and carbon efficiency rating and life cycle carbon emission factor of the virtual energy and carbon node. The life-cycle carbon emission factor is obtained by calculating the carbon emissions of the energy, materials and water resources consumed by the physical entity or functional space corresponding to the virtual energy carbon node during the upstream production, transportation and downstream disposal stages. The life-cycle carbon emission factor is stored in the energy and carbon attribute set and used for multi-timescale prediction and spatial source tracing of the energy and carbon situation in step S4.

6. The method for optimizing energy and carbon management in public building areas according to claim 5, characterized in that: After generating the control instruction set in step S5, the following steps are also included: Based on game theory, a multi-party coordination mechanism is constructed, defining different owners, property management companies, and energy suppliers within the public building area as game participants, and constructing a game utility function based on each participant's energy and carbon quotas, energy comfort, and economic costs. Energy comfort level is obtained by weighted calculation based on a combination of indoor temperature, humidity, illuminance, and air quality index. By solving for the Nash equilibrium, the control instruction set is negotiated and corrected to generate an optimized control instruction set that is agreed upon by multiple parties. The solution process for Nash equilibrium employs a distributed iterative algorithm. In each iteration, each participant updates its own policy by maximizing its own utility function based on the current policies of other participants, until the rate of change of all participants' policies falls below a preset convergence threshold.

7. The method for optimizing energy and carbon management in public building areas according to claim 6, characterized in that: Step S3, the construction of the dynamic baseline, also includes: Identify renewable energy nodes and energy storage nodes within public building areas, and generate output confidence intervals for renewable energy nodes and dispatchable capacity intervals for energy storage nodes based on meteorological forecast data and historical output data. The output confidence interval and dispatchable capacity interval are used as dynamic constraints and embedded in the generation process of dynamic energy carbon benchmark values ​​to correct the rule logic of energy supply and demand balance in the association rule set. The schedulable capacity range is determined by both the state of charge and the charging / discharging energy constraints.

8. The method for optimizing energy and carbon management in public building areas according to claim 7, characterized in that: The specific process of updating the energy carbon digital twin based on execution feedback in step S6 includes: Calculate the response deviation between the execution feedback and the control instruction set. If the response deviation exceeds the preset deviation tolerance range, trigger the model self-correction process. The model self-correction process uses an incremental learning algorithm to fine-tune the internal parameters of the time series prediction model and the multi-objective collaborative optimization model online using response bias. The response bias and its corresponding scene features are stored in an anomaly case library for use in optimizing the subsequent construction of dynamic baselines.

9. A method for optimizing energy and carbon management in public building areas according to claim 8, characterized in that: The method also includes step S7: generating and outputting an energy and carbon management optimization report based on the energy and carbon digital twin. The energy and carbon management optimization report includes: an energy and carbon leakage heat map generated based on the energy and carbon anomaly tracing path, an energy and carbon emission reduction potential analysis generated based on a multi-objective collaborative optimization model, and an energy and carbon efficiency improvement curve generated based on historical control command set and execution feedback. The energy and carbon leakage heat map is generated by visually rendering it on a building information model by associating the degree of energy and carbon anomalies of each virtual energy and carbon node with spatial coordinates. The analysis of carbon emission reduction potential is generated by sensitivity analysis of the Pareto front of the multi-objective synergistic optimization model, identifying the combination of control measures that contributes the most to each objective function and their corresponding marginal emission reduction costs; the energy and carbon efficiency improvement curve is generated by fitting the cumulative change rate of unit energy consumption output or unit carbon emission output in each cycle after sorting the historical control command set by time series, where unit energy consumption output is the ratio of regional total output to total energy consumption, and unit carbon emission output is the ratio of regional total output to total carbon emissions.

10. A public building area energy and carbon management optimization system, wherein the system is applied in the method described in any one of claims 1-9, characterized in that: The system includes: The virtual node network construction module is used to construct a virtual energy and carbon node network in a public building area. The virtual energy and carbon node network consists of multiple virtual energy and carbon nodes. Each virtual energy and carbon node corresponds to a physical entity or a functional space in the public building area and is configured with a unique identifier, a set of energy and carbon attributes, and a set of association rules. The data mapping and twin generation module is connected to the virtual node network construction module. It is used to obtain real-time operation data and static feature data of each physical entity and functional space within the public building area. Based on the virtual energy and carbon node network, the real-time operation data and static feature data are mapped to the corresponding virtual energy and carbon nodes to generate an initialized energy and carbon digital twin. The dynamic baseline construction module, connected to the data mapping and twin generation module, is used to perform multi-source data fusion and dynamic baseline construction in the energy and carbon digital twin based on the association rule set of each virtual energy and carbon node, and generate dynamic energy and carbon baseline values ​​for each virtual energy and carbon node. The energy and carbon situation prediction and source tracing module is connected to the dynamic baseline construction module. Based on the dynamic energy and carbon baseline values, real-time operation data and static characteristic data of each virtual energy and carbon node, it performs multi-timescale prediction and spatial source tracing of energy and carbon situation by coupling the time series prediction model and the spatial transmission model, and generates energy and carbon situation prediction results and energy and carbon anomaly source tracing paths. The multi-objective collaborative optimization module is connected to the energy and carbon situation prediction and source tracing module. It is used to call the preset multi-objective collaborative optimization model based on the energy and carbon situation prediction results and the energy and carbon anomaly source tracing path to generate a set of control instructions for at least one virtual energy and carbon node in the virtual energy and carbon node network. The closed-loop control and twin update module, connected to the multi-objective collaborative optimization module, is used to send the control command set to the corresponding physical entity actuator, receive execution feedback, update the energy carbon digital twin according to the execution feedback, and form closed-loop control. The report generation and output module is connected to the closed-loop control and twin update module. Based on the energy and carbon digital twin, it generates and outputs an energy and carbon management optimization report. The energy and carbon management optimization report includes: an energy and carbon leakage heat map generated based on the energy and carbon anomaly tracing path, an energy and carbon emission reduction potential analysis generated based on the multi-objective collaborative optimization model, and an energy and carbon efficiency improvement curve generated based on the historical control command set and execution feedback.

Images