Highway service area carbon emission prediction method based on space-time diagram embedding

Abstract

The invention relates to the technical field of energy exchange transportation and low-carbon operation and maintenance, in particular to an expressway service area carbon emission prediction method based on space-time diagram embedding, and the method comprises the steps: obtaining a portal frame and vehicle passing counting streams of entrances and exits of all service areas, constructing a dynamic topology and space-time prediction frame based on corridor-level tensors, and carrying out the prediction of the vehicle passing counting streams. Generating a traffic flow embedding vector and a comprehensive parking intensity curve; the method comprises the following steps: mapping a service area into an intelligent agent by combining a comprehensive parking intensity curve, dividing a scheduling cluster, carrying out three-section carbon pulse penalty and distributed strategy gradient training, freezing an intelligent agent strategy, generating an emission matrix and a thermodynamic diagram based on a power plan and an emission factor library, and comparing predicted and actually measured emission splitting errors to obtain a scheduling result. By generating the corridor-level emission matrix and the thermodynamic diagram, the predicted emission and the actually measured data are compared grid by grid, the traffic flow embedding distortion and the equipment strategy drift error are split by adopting the double-pointer method, and the long-term stability of the carbon emission prediction precision and the scheduling effect is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy exchange transportation and low-carbon operation and maintenance, and in particular to a highway service area carbon emission prediction method based on space-time graph embedding. BACKGROUND

[0002] As a key node of traffic energy consumption and carbon emission, the energy consumption and emission control of highway service area is of great significance to the realization of double carbon target in the field of transportation. At present, the energy consumption and emission control of highway service area in the industry mainly develops along two technical paths: one is based on single node traffic volume prediction, through seasonal difference ARIMA, support vector regression, LSTM and other time series models, combined with rainfall, temperature and humidity, holiday marking and other external variables, to extrapolate the short-term passenger flow of the service area, and then assist in energy consumption prediction; the second is to implement central dispatching type time-sharing power limit on the load side, relying on traditional building energy management system to divide high-consumption equipment into independent sub-circuits, and through PLC bus to execute load control based on manually set time-sharing table or rule-based start-stop logic.

[0003] However, the existing technical solutions still have significant defects, which cannot meet the needs of accurate prediction and dynamic control of service area carbon emission under corridor scale: on the one hand, the single node prediction model ignores the dynamic conduction characteristics of vehicle flow between the gantry and the service area node, the differences in data caliber and sampling period lead to inconsistent measurement of the same vehicle flow by upstream and downstream, which cannot depict the cross-node advancing law of vehicle flow peak, and the response lags when facing holiday long peak or sudden change of vehicle flow caused by bad weather; on the other hand, the load side control scheme mainly uses static energy consumption coefficient to estimate the power demand of equipment, which cannot reflect the differentiated energy response caused by different service areas and different vehicle passenger flow, and the centralized power limit framework regards the adjustable load of service area as a discrete variable, ignoring the impact of nonlinear change of vehicle flow on energy consumption, which needs to reserve a large energy consumption safety margin to avoid risks, resulting in low energy utilization efficiency. At present, a highway service area carbon emission prediction method based on space-time graph embedding is needed. SUMMARY

[0004] In order to solve the problem that the existing technology cannot meet the needs of accurate prediction and dynamic control of service area carbon emission under corridor scale, the present application provides a highway service area carbon emission prediction method based on space-time graph embedding.

[0005] The highway service area carbon emission prediction method based on space-time graph embedding provided by the present application adopts the following technical scheme: A highway service area carbon emission prediction method based on space-time graph embedding, comprising: S1, obtaining gantry and vehicle counting flow at each service area entrance and exit, generating vehicle equivalent sequence and personnel equivalent sequence through vehicle classification processing, and aggregating into a corridor level tensor; S2, constructing a dynamic topology and a space-time prediction framework based on the corridor-level tensor, including establishing a dynamic directed graph containing gantries and service areas, and generating a vehicle prediction equivalent using a space-time graph neural network embedded with a gating module; S3, calculating a dynamic time delay to split an upstream conduction amount and a local fluctuation amount according to the vehicle prediction equivalent, generating a traffic flow embedding vector and a comprehensive parking intensity curve; S4, mapping the service area to an agent and dividing a dispatch cluster in combination with the comprehensive parking intensity curve, calculating a cluster-level power baseline and broadcasting it to the agent, and the agent executing device discrete control and reporting operation data; S5, constructing a cold start rule table according to historical operation data, and controlling adjacent area channels in combination with traffic impact probability, training and freezing the agent strategy through a three-section carbon pulse penalty and a distributed policy gradient, obtaining a power plan and a carbon emission curve; S6, generating an emission matrix and a heat map based on the power plan and an emission factor library, comparing the prediction and measured emission split errors, and completing the final output.

[0006] Further, step S1 specifically includes the following sub-steps: S11, reading the vehicle count flow, sequentially performing vehicle length differentiation, axle load label supplementation and license plate attribute mapping, classifying vehicles through a three-dimensional cross-research table of length, axle load and license plate, and merging and accumulating each class to generate a vehicle type count sequence 、 、 ; S12, calling a conversion coefficient set mapping the vehicle type count sequence and the corresponding coefficient product to obtain a vehicle equivalent sequence and a personnel equivalent sequence; S13, aggregating the vehicle equivalent sequence and the personnel equivalent sequence according to the corridor number, the lane direction and the inner procedure number, accumulating through a single-day accumulation window to obtain a corridor-level daily aggregated vector, and writing into a unified tensor structure; wherein, 、 and represent the conversion coefficients of small cars, large buses and trucks, 、 and represent the category labels of small cars, large buses and trucks.

[0007] Further, the dynamic directed graph containing gantry and service area is established, including writing the gantry and service area in the node list in increasing order of kilometer post and appending a virtual boundary node, recording node attributes, creating a directed edge between the nodes of adjacent mileage coordinates in the node list V according to the actual travel direction of the vehicle, forming a directed edge set E, constructing an adjacency list A combining the node list V and the directed edge set E to obtain an initial directed graph structure, extracting the inter-node mileage difference, window average speed and traffic event level from each directed edge of the adjacency list, and calculating the propagation impedance of each directed edge by using a weight normalization superposition method, and the propagation impedance expression is: , wherein, is the edge propagation impedance at time , are three factor weights respectively, is the kilometer difference from node to node , is the maximum kilometer difference of adjacent nodes in the same direction, is the free flow reference speed, is the average speed of edge at time , is a very small positive number to prevent zero denominator, is the event level of edge at time , is an event amplification coefficient to control the lifting amplitude of disasters on impedance.

[0008] Further, the time-space graph neural network embedded with a gating module is used to generate vehicle prediction equivalent, including extracting the daily aggregated vehicle equivalent vector of each node based on a unified tensor structure, mapping it into an input feature matrix of the time-space graph neural network in the order of the node list, constructing a time-space graph neural network containing a graph convolution layer and a time convolution layer, embedding a gating module between the graph convolution layer and the time convolution layer, inputting the input feature matrix X into the time-space graph neural network, capturing spatial correlation through the graph convolution layer combined with real-time propagation impedance, capturing time sequence dependence through the time convolution layer, executing hour-by-hour reasoning after modulation by the gating module, and outputting a vehicle equivalent prediction sequence, and the vehicle equivalent prediction sequence expression is: , wherein, is the time-space graph neural network reasoning function of the gating module, is the corresponding vehicle type feature sub-matrix, A is the adjacency list, is the propagation impedance matrix at time t, is the gating module state.

[0009] Furthermore, the step of calculating the dynamic delay separation of upstream transmission and local fluctuation based on vehicle prediction equivalents includes aggregating service area nodes within the corridor according to the vehicle equivalent prediction sequence in ascending order of kilometer markers, and establishing a local prediction sequence for each target service area node. Overlapping sequence with upstream node translation The system collects the associated physical parameters between the target service area node and each upstream node, calculates the dynamic delay for each vehicle model, and then shifts the prediction sequence of the upstream nodes clockwise based on the dynamic delay, comparing it with the local prediction sequence of the target service area. The upstream conduction amount is obtained by taking the minimum value hourly. The dynamic delay calculation formula is as follows: , in, upstream node To the target node The road distance, Vehicle category The reference velocity in free flow. Vehicle category At any moment The average speed of the corridor For the edge At any moment The severity level of the accident, The overall weather severity level for the corridor. The weighting coefficients are based on four factors: distance, speed, accidents, and weather. It is a very small positive number, used to avoid the denominator being zero. Vehicle category From node Drive to the node Dynamic latency.

[0010] Furthermore, the generation of the traffic flow embedding vector and the comprehensive dwell intensity curve includes calculating the upstream transmission ratio and the local fluctuation ratio based on the upstream transmission amount, generating the cross-regional traffic flow embedding vector based on the upstream transmission ratio and the local fluctuation ratio, and calling the historical offline dwell probability surface. By combining the time delay correction coefficients corresponding to the vehicle flow embedding vector, future time intervals are generated. The vehicle arrived on the timeline. Search for car models At any moment Dwell probability factor ,Will Multiplying the signal strength component in the traffic flow embedding vector yields the single-vehicle model parking intensity sequence. For three types of vehicles By performing time alignment and intensity overlay, the comprehensive dwelling intensity curve for the next decision window of the target service area is obtained.

[0011] Furthermore, the step of mapping service areas to intelligent agents and dividing them into scheduling clusters by combining the comprehensive parking intensity curve includes constructing a service area index table based on the comprehensive parking intensity curve of each service area and the static attribute data of all service areas in the corridor, converting the service area corresponding to each row in the index table into multiple independent control intelligent agents according to the mapping relationship, clustering the intelligent agents according to the driving direction of the target service area and the adjacency of the kilometer markers, and grouping several intelligent agents with the same direction and consecutively adjacent kilometer markers into the same scheduling cluster.

[0012] Furthermore, the calculation of the cluster-level power baseline and its broadcast to the agent includes calling upon the service area average load factor and corridor total power limit from the historical operation database to calculate the cluster-level power baseline for each scheduled cluster. Construct a power baseline broadcast channel, and The scheduling cluster affiliation is used to broadcast to the corresponding intelligent agent and generate a list of discrete actions for the device. The formula for calculating the cluster-level power baseline is as follows: , Where m is the scheduling cluster The number of intelligent agents within, For intelligent agents The rated transformer capacity of the corresponding service area, The historical average load factor, M represents the upper limit of the total power of the corridor, and M represents the total number of scheduling clusters within the corridor.

[0013] Furthermore, step S5 specifically includes the following sub-steps: S51. Combine historical operational data and establish an index based on the three-dimensional keys of season, time period and service area type to generate a start-stop rule table. Map the start-stop rule table to a combination of power levels of multiple high-power devices and load the corresponding cold start strategy set for each intelligent agent. S52. Calculate the probability of traffic flow impact in the next four hours based on the vehicle equivalent prediction sequence and the comprehensive parking intensity curve, set the traffic flow impact probability threshold, and determine the opening and closing status of the agent's neighboring cell capacity transfer channel based on the probability threshold. S53. Based on the measured carbon emissions reported by the agent in the previous hour at the start of the current decision cycle, calculate the rolling median value of carbon emissions for the entire corridor. Using the median value as a benchmark, calculate the emission exceedance, spatial synchronization indication, and rate of increase exceeding the threshold. Construct a three-segment carbon pulse penalty function based on the emission exceedance, spatial synchronization indication, and rate of increase exceeding the threshold. S54, training the agent by using a distributed policy gradient algorithm, constructing a reward function based on carbon pulse punishment, performing normalization processing on the reward value of each agent after each iteration, and feeding the normalized reward back to the value network of the corresponding agent to update the policy parameters.

[0014] Further, step S6 specifically comprises the following sub-steps: S61, mapping and calculating the power values of the five types of equipment in the power gear plan, i.e., lighting, heating, catering, charging piles and refueling islands, by hour granularity and the emission factor of the corresponding energy path to obtain the hourly carbon emission value of each service area; S62, writing the hourly carbon emission value of each service area into a corridor-level emission matrix, the rows of the corridor-level emission matrix correspond to time stamps, the columns correspond to service area node indexes, and a color gradient is used to render the emission matrix to generate a real-time carbon emission heat map; S63, collecting hourly measured carbon emission data of each service area to form a measured emission matrix, comparing the emission matrix with the measured emission matrix grid by grid, splitting the error by using a double-pointer method to obtain traffic flow embedding distortion error and equipment strategy drift error; S64, accumulating the traffic flow embedding distortion error and the equipment strategy drift error according to the service area nodes respectively to obtain the total distortion error and the total drift error of each node, and taking the emission matrix, the carbon emission heat map and the error splitting result as the final output.

[0015] In summary, the present application has the following beneficial technical effects: The present application performs three-dimensional cross classification on the vehicle length distinction, axle load label supplement and license plate attribute mapping of the vehicle counting flow of the portal of the gantry and the service area, classifies the heterogeneous vehicle data of different nodes into three standard categories of small passenger cars, large passenger cars and trucks, and then aggregates the vehicle equivalent and personnel equivalent sequences according to the corridor number, lane direction and lane number to write into a unified tensor structure. This process unifies the sampling granularity, statistical caliber and spatial correlation logic of the whole corridor data, avoids the dimensional confusion and time-space mismatch defects in the traditional scheme, and provides a homogeneous and comparable data basis for subsequent dynamic topology construction and collaborative scheduling.

[0016] 2, the present application constructs a dynamic directed graph containing gantry and service area based on corridor-level tensor, dynamically calculates the propagation impedance of directed edges by real-time collecting the mileage difference between nodes, driving speed and traffic event level, and describes the attenuation and lag law of the node advancement of the traffic flow wave crest; at the same time, a gating module is embedded in the space-time graph neural network, which automatically opens the far-end information channel during holiday peak or bad weather, strengthens the cross-area traffic flow information aggregation, and avoids the problem of response lag of traditional static models to sudden traffic flow. This design enables the vehicle equivalent prediction to match the spatial transmission characteristics and time-dependent characteristics of the traffic flow simultaneously, significantly improving the adaptation accuracy of the prediction result to the actual traffic flow changes.

[0017] 3、The application calculates the dynamic time delay according to the vehicle prediction equivalent, translates the upstream node prediction sequence according to the actual travel rhythm, compares it with the service area prediction sequence in the same period, splits the upstream conduction amount and the local fluctuation amount, and generates a comprehensive parking intensity curve by combining historical parking probability data, thereby clearly distinguishing the differentiated needs of transit traffic and local traffic for energy, compared with the traditional scheme of estimating energy consumption by a static coefficient, the design can accurately depict the energy use characteristics of traffic from different sources, so that the subsequent device power regulation is more in line with the actual passenger flow demand.

[0018] 4、In step S4, each service area is mapped as an independent agent, and the dispatch cluster is divided according to the driving direction and the kilometer post adjacent relationship, the cluster-level power baseline is calculated and distributed through the broadcast channel, and the whole corridor power is dynamically allocated within the safe capacity; step S5 further introduces a three-section carbon pulse punishment mechanism, and combines a distributed policy gradient algorithm to train the agent, so that each service area dynamically balances between controlling carbon emissions and ensuring service quality, avoiding the problem of global mismatch of traditional centralized scheduling, and the cooperative mechanism can effectively smooth the corridor-level power peak, reduce the carbon emission pulse caused by the centralized start and stop of equipment, and improve the energy utilization efficiency.

[0019] 5、The application generates a corridor-level emission matrix and a heat map, compares the predicted emissions with the measured data, uses a double-pointer method to split the vehicle flow embedding distortion and equipment strategy drift error, and according to the error result, reversely corrects the model parameters of the previous step, which can continuously adapt to the changes of vehicle flow and fluctuations of equipment working conditions, and ensures the long-term stability of carbon emission prediction accuracy and scheduling effect. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is the overall flow schematic diagram of a highway service area carbon emission prediction method based on space-time graph embedding according to an embodiment of the application.

[0021] Figure 2 is a scheme comparison diagram of the embodiment of the application and the traditional technology.

[0022] Figure 3 is a flowchart of step S3 in the embodiment of the application.

[0023] Figure 4 is a flowchart of step S4 in the embodiment of the application.

[0024] Figure 5 is a flowchart of step S5 in the embodiment of the application.

[0025] Figure 6 is a flowchart of step S6 in the embodiment of the application. DETAILED DESCRIPTION

[0026] The present invention will be further described in detail below with reference to the accompanying drawings.

[0027] Example 1 Reference Figure 1 This embodiment of a method for predicting carbon emissions from highway service areas based on spatiotemporal graph embedding includes: S1. Obtain the vehicle count streams at the gantry and entrances / exits of each service area, process the vehicles to generate vehicle equivalent sequences and personnel equivalent sequences, and aggregate them into corridor-level tensors; S2. Based on corridor-level tensor, construct a dynamic topology and spatiotemporal prediction framework, including establishing a dynamic directed graph containing gantry and service area, and using a spatiotemporal graph neural network with embedded gating module to generate vehicle prediction equivalents. S3. Calculate the dynamic delay based on the vehicle prediction equivalent, separate the upstream transmission amount from the local fluctuation amount, and generate the vehicle flow embedding vector and the comprehensive parking intensity curve. S4. Combine the comprehensive parking intensity curve to map the service area into intelligent agents and divide it into scheduling clusters. Calculate the cluster-level power baseline and broadcast it to the intelligent agents. The intelligent agents perform discrete control of equipment and report operating data. S5. Construct a cold start rule table based on historical operation data, and control the neighboring lanes by combining the probability of traffic flow impact. Train and freeze the agent policy through three-stage carbon pulse penalty and distributed policy gradient to obtain the power plan and carbon emission curve. S6. Generate emission matrix and heat map based on power plan and emission factor library, compare the emission splitting error between prediction and actual emission, and complete the final output.

[0028] Specifically, a method for predicting carbon emissions from highway service areas based on spatiotemporal graph embedding includes the following steps: S1. Obtain the vehicle count streams at the gantry and entrances / exits of each service area, process the vehicles to generate vehicle equivalent sequences and personnel equivalent sequences, and aggregate them into corridor-level tensors; like Figure 1 , Figure 2 As shown, S11 is the total vehicle counting stream where the gantry and the entrances / exits of each service area are on the same corridor. Read in chronological order, and perform three mapping operations on each vehicle record sequentially: First, based on the vehicle length threshold This embodiment differentiates between long and short vehicles, defining passenger cars as ≤6m, buses as 6-12m, and trucks as >12m. By comparing the length data of each vehicle passage record, the vehicle category range can be preliminarily determined; Second, according to the axle load range, the load label is supplemented to the division result. For the preliminarily divided vehicles, the load label is supplemented according to the axle load range. The axle load of a small passenger car is less than or equal to 10 t, the axle load of a large passenger car is 10-20 t, and the axle load of a truck is greater than 20 t. If the length of a vehicle is 5 m (preliminarily classified into the small passenger car range) but the axle load is 12 t (exceeding the axle load threshold of the small passenger car), the vehicle is corrected to be classified into the large passenger car candidate range, so as to realize the cross verification of the length and the axle load. Third, the attribute of the vehicle with the label is supplemented according to the license plate classification code. The attribute of the vehicle with the supplemented axle load label is mapped according to the license plate attribute, and the final vehicle attribute is confirmed. After the three mappings are completed, the mapping result is positioned by using a three-dimensional cross-research table of length, axle load and license plate, and each record is uniquely classified into one of three categories of small passenger cars, large passenger cars and trucks. Then, the counting of the same type of vehicles in the same node and the same time interval is combined and accumulated, to generate a vehicle type counting sequence that maintains the original sampling interval, including a small passenger car counting sequence , a large passenger car counting sequence , and a truck counting sequence , wherein t is a time interval identifier, the sequence is locked when the sampling granularity is generated , and does not change with the node difference.

[0029] S12, based on the operation data of the highway corridor in the past 12 months (including the energy consumption experiment data of the vehicle type and the statistical data of the typical vehicle passenger capacity), a set of conversion coefficients is preset , wherein , and represent the conversion coefficients of small passenger cars, large passenger cars and trucks respectively. Specifically, taking small passenger cars as the reference, the vehicle equivalent contribution degree is set to 1.0, i.e. , the energy consumption demand of large passenger cars is 3.5 times that of small passenger cars due to the large passenger capacity and long stay time, and the embodiment sets , the truck has high load, high maintenance and high refueling frequency, and the embodiment sets . Similarly, based on the historical data statistics, the personnel equivalent conversion coefficients of the three types are determined.

[0030] The three counting sequences obtained in sub-step S11 are respectively called the fixed vehicle and personnel conversion coefficient set ; the , , at the same time are matched with the corresponding coefficients and the product mapping is performed, to directly obtain the vehicle equivalent sequence and the personnel equivalent sequence . The calculation formula of the vehicle equivalent sequence is: , in, , and These are the category markers for passenger cars, buses, and trucks, respectively. The physical dimensions of the vehicle are used for subsequent calculations of the electrical and energy modules; Based on typical vehicle passenger-energy consumption test values, a dual-dimensional scale is established to ensure that both vehicle and passenger perspectives are available at the same time. S13. Convert the vehicle equivalent sequence generated by each node in sub-step S12. Human equivalent sequence The sequence is aggregated and sorted using a three-element ordered key-value pair: corridor number, lane direction, and inner program number; a daily cumulative window is then applied to the sorted sequence. Perform the accumulation to obtain the corridor-level daily aggregation vector. and ;in, For the cumulative number of passenger cars on a single day, For the cumulative number of large passenger buses on a single day, This refers to the cumulative vehicle equivalent of trucks on a single day. This refers to the cumulative passenger volume of passenger vehicles on a single day. This refers to the cumulative passenger equivalent of a single day's large passenger bus. The daily cumulative number of personnel equivalent for trucks will be... and The corresponding kilometer marker index and direction identifier are also written into the unified tensor structure. Tensors are directly used as the node inputs of the spatiotemporal graph neural network and as the common data benchmark for subsequent collaborative energy consumption decisions. Through aggregation, upstream and downstream service areas reference vehicle and personnel equivalent information of the same source, same dimension, and same direction at any subsequent time point.

[0031] S2. Based on corridor-level tensor, construct a dynamic topology and spatiotemporal prediction framework, including establishing a dynamic directed graph containing gantry and service area, and using a spatiotemporal graph neural network with embedded gating module to generate vehicle prediction equivalents. First, read the corridor-level tensor. The coordinates and node types of all gantry and service area kilometer markers recorded in the system are written into node list V in ascending order of kilometer marker values. Simultaneously, one virtual boundary node is added to the beginning and end of the list to close the actual vehicle travel path. For each node in node list V, three immutable attributes are recorded simultaneously: type identifier. (Value taken from gantry service area or virtual), traffic flow direction (Value taken as up or down), mileage coordinates (kilometers), then, according to the actual travel direction of the vehicle (the uplink direction is increased by the kilometer post, and the downlink direction is decreased by the kilometer post), a directed edge is created between the nodes of the adjacent kilometer coordinates in the node list V, for example, a directed edge is created between the K100 gantry (node i) and the K105 gantry (node j) in the uplink direction , a directed edge is created between the K105 gantry (node j) and the A service area (node k) , thereby forming a directed edge set E; in combination with the node list V and the directed edge set E, an adjacency list A is constructed in the form of an adjacency list, each row in the adjacency list A corresponds to a node, records the adjacent nodes, the directed edge identifier and the basic attributes of the edge of the node, and then an initial directed graph structure is obtained .

[0032] Subsequently, for each directed edge in the adjacency list A , three physical parameters are collected in real time: first, the kilometer difference between node i and node j (unit, km), second, the average driving speed in the last 10-minute sliding window (unit, km·h ), and third, the traffic incident level of the directed edge at time t , taking values 1-5, corresponding to normal to severe blockage, 1 for normal traffic, 2 for slight congestion, 3 for moderate congestion, 4 for severe congestion, and 5 for severe blockage, reported in real time by the traffic incident monitoring system), and finally, the real-time propagation impedance of each directed edge at time t is calculated using the weight normalization superposition method , the propagation impedance expression is: , where, is the propagation impedance of edge at time , and are the three factor weights, respectively, is the kilometer difference between node and node , is the maximum kilometer difference between adjacent nodes in the same direction, is the free flow reference speed, is the average speed of edge at time , is a very small positive number to prevent a zero denominator, is the incident level of edge at time , is the incident amplification coefficient, which controls the lifting amplitude of the disaster on the impedance, and by calculating the propagation impedance of each directed edge in real time, the initial directed graph structure​ Transformed into a dynamically changing directed graph over time. ,in, Let be the propagation impedance matrix at time t, according to a two-level threshold. and right Perform dynamic segmentation if Then Marking a graph as weakly connected causes graph convolutions to retain only first-order information and block further downstream propagation; if Immediately restore the strong connection, allowing information to be transmitted without attenuation.

[0033] Through the aforementioned continuously evolving impedance weights, the framework maintains lightweight calculations for nearest neighbors under normal conditions, while automatically weakening long-range diffusion paths during peak holiday periods or severe weather, ensuring that downstream service areas obtain traffic flow prediction benchmarks that better match the physical rhythm of travel in advance.

[0034] The corridor-level tensor generated in step S1 Extract the daily aggregated vehicle equivalent vector for each node (including gantries and service areas). The daily aggregated vehicle equivalent data (including passenger cars, buses, and trucks) is mapped to the input feature matrix X of the spatiotemporal graph neural network according to the order of the node list V. The dimension of the input feature matrix X is the number of nodes × time step × feature dimension. The "number of nodes" is the total number of nodes in the node list V. In this embodiment, the corridor node list V contains 143 nodes, so the number of nodes = 143. The time step is the number of time intervals for historical data sampling (taking 10-minute data from the past 7 days, 144 time intervals per day, so the time step = 7 × 144 = 1008). The feature dimension is 3, corresponding to the vehicle equivalent data of the three types of vehicles: passenger cars, buses, and trucks.

[0035] Next, a spatiotemporal graph neural network is constructed, comprising graph convolutional layers and temporal convolutional layers: the graph convolutional layers use the ChebNet graph convolution operator, with the input being the input feature matrix X and the adjacency list A, combined with the propagation impedance matrix at time t. The system captures spatial correlation features between nodes and adjusts the information transmission weights between nodes by dynamically changing the propagation impedance. For example, directed edges with high propagation impedance correspond to reduced information transmission strength between nodes, simulating the delayed transmission of traffic flow information under congestion conditions. The temporal convolutional layer uses a 1D convolution operator with a kernel size of 3, and the input is the output features of the graph convolutional layer, used to capture the temporal dependencies of vehicle equivalent data, such as the traffic flow time sequence patterns during morning and evening peak hours. A holiday-weather gating module is embedded between the graph convolutional layer and the temporal convolutional layer. The gating module The control logic is driven by real-time holiday information and weather data: when the current period is detected to be a statutory holiday or the weather level is ≥2 (1 for sunny, 2 for cloudy / light rain, 3 for heavy rain / strong wind, 4 for heavy rain / snowstorm), the gating module... Output control signal "1" to open remote jump connection channels across 3 or more nodes, allowing traffic flow information from upstream nodes to be directly transmitted to remote downstream nodes, enhancing the rapid aggregation of cross-regional traffic flow information in abnormal scenarios; when it is a normal time period and the weather level is <2, the gating module... Output control signal "0" to close the remote jump connection channel, retaining only the information transmission between adjacent nodes to reduce unnecessary computational redundancy. Then, input feature matrix X is input to the above-mentioned embedded gating module. The spatiotemporal graph neural network, through graph convolutional layers and propagation impedance matrix The synergistic effect accurately captures the spatial correlation features of traffic flow between different nodes; the temporal convolutional layer captures the dependence of vehicle equivalent in the time dimension; and the holiday-weather gating module... The information transmission paths are dynamically modulated under different scenarios to ensure the network's responsiveness to abnormal traffic flow. After multiple rounds of inter-layer iterative inference (10 iterations, with the Adam optimizer updating network parameters in each iteration and a learning rate of 0.001), the spatiotemporal graph neural network outputs a node-level vehicle equivalent prediction sequence for the next 24 hours, categorized by vehicle type. Finally, the network output is formatted to generate a passenger car vehicle equivalent prediction sequence. Bus vehicle equivalent prediction sequence Truck vehicle equivalent prediction sequence Where t is a 10-minute time interval identifier for the next 24 hours (t=1 corresponds to the first 10 minutes of the prediction start time, and t=144 corresponds to the last 10 minutes of the prediction end time), the above prediction sequence is generated according to the formula: , in, They represent passenger cars, buses, and trucks, respectively. This is the spatiotemporal graph neural network inference function for the gating module. Let A be the feature submatrix corresponding to the vehicle model, and let A be the adjacency list. Let be the propagation impedance matrix at time t. As the state of the gating module, this prediction sequence provides accurate prior traffic data for the subsequent steps S3, including traffic source segmentation and parking intensity calculation.

[0036] S3. Calculate the dynamic delay based on the vehicle prediction equivalent, separate the upstream transmission amount from the local fluctuation amount, and generate the vehicle flow embedding vector and the comprehensive parking intensity curve. like Figure 3As shown, S31 is the 24-hour prediction tensor output from step S2, which has been sorted by vehicle type by node. Aggregation is performed in ascending order of kilometer markers; for each service area node in the aggregation... Two buffer queues are established simultaneously to record the predicted sequence for this region, denoted as . ; Another summary is the sequence obtained by translating and superimposing all upstream nodes, denoted as... .

[0037] To maintain a consistent inference pace, the scrolling window length is set uniformly for both queues. And lock step size Subsequently, vehicle type indexes were established for the three categories of vehicles: passenger cars, buses, and trucks. , , To ensure that subsequent disassembly actions are performed under the same vehicle model specifications, and finally, to explicitly mark the time intervals in the predicted sequence that fall within the holiday or official accident duration ranges. This provides peak position clues for the decomposition operator and reserves high-risk warning positions for subsequent energy-consuming agents.

[0038] S32. Decomposition of execution latency-source for upstream-target node pairs: 1. Combine all upstream nodes within the aggregation. Target service area By vehicle type Extract the 24-hour prediction sequence one by one Then, the driving distance was used. Real-time average speed in the corridor Free flow velocity Accident level Weather rating Five-item generation dynamic delay The distance and speed terms provide the base values ​​for the time scale, while the accident and weather terms are multiplicatively increased under abnormal scenarios.

[0039] II. Based on each upstream sequence After performing a clockwise shift, sum the results for the same vehicle model, and then combine this sum with the predicted sequence for this region. By taking the minimum value hourly, the conduction quantity explained by the upstream is obtained. The upstream transmission ratio can be obtained by summing the values ​​over 24 hours. Local fluctuation ratio ,in .

[0040] The two-stage action described above, through a progressive path of physical time delay correction and minimum value truncation, explicitly separates explainable external shocks from unexplainable local fluctuations, providing clear leading signals for subsequent agents: , , in, upstream node To the target node Road distance, in km; Vehicle category Reference velocity in free flow, in km· ; Vehicle category At any moment The average speed of the corridor; For the edge At any moment Accident severity level (1–5); The overall weather severity level for the corridor is (0–3). The four weighting coefficients for distance, speed, accident, and weather are preset and meet industry safety standards; It is a very small positive number, used to avoid the denominator being zero; Vehicle category From node Drive to the node Dynamic latency, in hours (h); , Vehicle category At the node or The The estimated equivalent for the next hour; To predict the time step, it is fixed at 1 hour; Vehicle category At the node The The conduction equivalent that can be explained upstream within an hour; Vehicle category Upstream transmission ratio; Vehicle category Local fluctuation ratio, always satisfying .

[0041] S33, Output the code from sub-step S32 , and holiday markers Accident markers The four types of dominant factors are concatenated in the order of transmission intensity, fluctuation intensity, and special state to generate the cross-regional traffic flow embedding vector. .

[0042] Replicating sequentially across three vehicle model dimensions And add a vehicle model index to form This vector set provides interpretable components while maintaining directionality, enabling subsequent energy agents to instantly identify the proportion of external peaks and the degree of local fluctuations during reading.

[0043] The meanings and origins of the symbols in S33 are as follows: This represents the conduction ratio sequence that can be explained by the upstream. This represents the local fluctuation ratio sequence; For hourly holiday markers, For accident status labeling, these four types of factors are all directly obtained from the time delay-source decomposition of S32. Specifically, for service areas... With car model S32 based on dynamic delay Shift each upstream prediction sequence and compare it with the prediction sequence for this region. The upstream conduction quantity is obtained by taking the minimum value hourly. and cause local volatility ,then , and ; Provided by the holiday calendar, Based on the traffic event grading flow at the hourly granularity (normalized values ​​can be used), the cross-regional traffic flow embedding vector is defined as follows: And add a vehicle index form ,gather This is the result described in S33.

[0044] S34, will Corresponding time delay correction coefficient Combine them, and call the linear time-shift mapping for each vehicle type; first, based on Generate future The arrival timeline of a moment, and then the probability surface of offline dwell in history. Searching for corresponding probability factors Finally Multiply by the conduction strength component to output the dwell strength. sequence The three types of vehicles By performing time alignment and intensity overlay, the service area is obtained. Overall dwell curve of the next decision window and complete embedding pairs .

[0045] S4, mapping service areas as agents and dividing scheduling clusters based on the comprehensive parking intensity curve, broadcasting cluster-level power baseline to the agents, and the agents performing device discrete control and reporting operation data; As shown in Figure 4 , an index table is constructed based on the comprehensive parking intensity curve and the static attributes of the service area, and the agents are mapped: from the output results of step S3, the comprehensive parking intensity curves of multiple pairs of service areas are extracted, and the static attribute data of all service areas in the corridor are called, including service area type identifier , rated transformer capacity (unit kVA), main building function , the comprehensive parking intensity curve is integrated with the above static attribute data in order of service area number, driving direction and kilometer post coordinates, and a service area index table is constructed, each row in the index table corresponds to a service area, and contains number, direction, kilometer post, parking intensity curve , type , rated transformer capacity , main building function , then, according to the one-to-one mapping relationship, the service area corresponding to each row in the index table is converted into an independent control agent to form an agent set , wherein each agent , , corresponds to a service area, has the core functions of receiving power instructions, executing device control, and reporting operation data, and the number and direction attributes of the agent are consistent with the information of the corresponding service area in the index table.

[0046] Next, based on the driving direction and the adjacent kilometer post, the scheduling cluster is divided: for the generated agent set A, first, the uplink and downlink agents are completely separated according to the driving direction, to ensure that the agents in the same scheduling cluster correspond to service areas with the same driving direction of the vehicle flow, and to avoid the difference in the time-space transmission characteristics of cross-direction vehicle flow interfering with the power coordination scheduling logic; then, for the agents in the same direction, the kilometer post coordinates of the associated service area are sorted in ascending order, so that the order of the agents is consistent with the actual driving path of the vehicle in the corridor, laying a spatial foundation for subsequent power coordination based on vehicle flow transmission; then, according to the clustering principle of the same direction and continuous kilometer post without interval, the sorted agents are divided into several scheduling clusters , each scheduling cluster contains several continuous and adjacent agents, and the kilometer post interval of all agents in the cluster has no breakpoint, and each scheduling cluster records the number of agents m in the cluster and the rated transformer capacity of each agent.

[0047] Subsequently, cluster-level power baselines were calculated and broadcast channels were constructed for targeted distribution. Service area average load factors were extracted based on historical operational databases. This factor is defined as the average ratio of the actual operating power of the service area to the rated transformer capacity during the same historical period. It reflects the load utilization level of the service area under normal operating conditions, with a value range of 0-1. It is calculated from the load data of the same period in the past 6-12 months, providing historical operating basis for the power baseline and avoiding the baseline setting from deviating from the actual operating scenario. At the same time, it retrieves the upper limit of the total power of the corridor. (Determined by the power supply contract signed between the regional substation and the highway operator, a global constraint upper limit is set for the total power consumption of all service areas in the entire corridor to prevent the risk of power outages caused by exceeding the power supply capacity); Based on the above parameters, a calculation method with dual constraints of intra-cluster equipment capacity constraint + corridor total power constraint is adopted, and the cluster-level power baseline calculation formula is as follows: , Where m is the scheduling cluster The number of intelligent agents within, For intelligent agents The rated transformer capacity of the corresponding service area The historical average load factor, M represents the upper limit of the total power of the corridor, and M represents the total number of scheduling clusters in the corridor. The total power is allocated according to the proportion of the number of intelligent agents in the cluster to ensure that the power demand of each cluster matches the total capacity and avoid some clusters having excessive power while others have insufficient power.

[0048] Create a broadcast channel at the top of the corridor This allows the scheduling layer to forward power adjustment commands to all agents at once; whenever traffic flow embeds... When the wave crest is advancing upstream and the downstream clusters have not yet been impacted, the scheduling layer first tightens control on the upstream clusters where the wave crest has not yet arrived. Then, the reserved capacity is released to the downstream clusters that are about to be impacted, and finally the increase or decrease is calculated. Broadcast to This enables cross-node power rolling, allowing all service areas to share a consistent total power constraint at the same time, thus suppressing serial overflow from the source.

[0049] Finally, the agents perform discrete control of the devices and report operational data: each agent receives the cluster-level power baseline of the corresponding scheduling cluster. With additional instructions, the system maps the five types of high-consumption equipment it controls—lighting, HVAC, catering, charging stations, and refueling islands—to a set of discrete power levels preset in this document. Within each intelligent agent, five types of high-consumption equipment—lighting, HVAC, catering, charging stations, and refueling islands—are mapped to a fixed power level set. Each gear level and its duration Combinations are discrete actions , and write to the action list . For Generate device layer channels so that any action in the list can trigger the closing or opening of the scene shunt switch in the corresponding period, avoiding strategies that only stay at the digital layer.

[0050] For each agent Assemble the observation vector , and record the sampling time stamp for all information segments. The vector is concatenated in the following order: local room temperature and humidity, local electricity price, service area type segment, vehicle flow embedding generated in step S3 , the latest broadcast power baseline of the scheduling layer, so that At the same time, present the cross-area vehicle flow impact, station operating cost and real-time power constraint, provide multi-dimensional decision basis for the strategy network, so that the agent can balance carbon peak suppression and customer comfort in actual execution.

[0051] Add a feedback channel on the broadcast channel At the end of each decision cycle, the agent Voluntarily reports two indicators: actual power usage And the proportion of unmet demand, after the scheduling layer collects all the reports, uses a linear penalty function To modify the cluster-level baseline for the next cycle , and reissue the new baseline through .

[0052] Among them, the penalty function is data-driven calibrated, used to make monotonic and stable closed-loop correction to the cluster-level baseline in the corridor total power constraint for the next cycle.

[0053] Specifically, the slope is obtained by least squares fitting of the samples of the previous decision cycles , where is the proportion of unmet demand, and capacity clamping and smoothing are used to avoid oscillation to form the update formula , where , is the upper limit of the capacity of the cluster and the corridor, is the smoothing coefficient. The linear form is derived from the empirical monotonic marginal relationship and convexity requirement, which is convenient for online rapid iteration; Through grid search and cross-validation of the historical validation set, the unmet rate is reduced and the baseline does not overshoot, and then ​Only for baseline correction, independent of the three-stage carbon pulse penalty in S5 training; the latter is weighted by the rolling median reference to the over-amplitude, spatial synchronization and rate threshold segments, used for policy learning, not directly involved in baseline real-time correction.

[0054] On full-corridor power cap Internal dynamic allocation margin, so that each service area can embed flexible capacity according to real-time traffic, and suppress serial overflow and carbon emission pulses caused by information island from the source.

[0055] S5, according to historical operation data to build cold start rule table, and combine with traffic impact probability control adjacent area channel, through three-stage carbon pulse penalty and distributed strategy gradient training and freeze agent strategy, get power plan and carbon emission curve; As Figure 5 shown, first, combined with historical operation data to build cold start rule table and load cold start strategy set: call the historical operation database defined in this file, extract three types of core historical data, historical passenger flow log, branch power curve (including hourly power data of five types of high-consumption equipment: lighting, heating, catering, charging pile, refueling island, reflecting the correlation between equipment power and passenger flow), emission account (including hourly carbon emission measured value, used to correlate the corresponding rules of power consumption and carbon emission); index the above data according to three dimensions of season, time period and service area type, where "season" is divided into four categories according to spring, summer, autumn and winter, time period is divided into five categories according to 24 hours a day, morning, morning peak, flat peak, evening peak and night, service area type is consistent with step S4, historical data is classified by three-dimensional index, and start-stop rule table is generated; for each index record in the start-stop rule table , according to the historical branch power curve, it is mapped to the power gear combination of five types of high-consumption equipment, that is , to ensure that each rule corresponds to the optimal equipment gear scheme under the specific season, specific time period and specific service area type; then load the exclusive cold start strategy set for each agent generated in step S4 according to service area type, that is, the comprehensive agent loads the rule subset corresponding to the comprehensive service area, and the refueling side agent loads the rule subset corresponding to the refueling side service area, so that each agent has a basic control strategy that meets its functional attributes in the initial stage of training, avoiding equipment operation disorder or service quality decline caused by random strategy in the initial stage of training.

[0056] Based on the traffic impact probability, dynamically control the agent adjacent area capacity channel: call the future 24-hour vehicle equivalent prediction sequence output by step S2 and the comprehensive stop intensity curve generated by step S3, take the next four hours as the prediction window, and calculate the traffic impact probability The calculation logic of the traffic flow impact probability is to first determine the 90th percentile of the historical comprehensive parking intensity of the service area corresponding to the agent , as the traffic flow impact determination reference, and then count the predicted parking intensity every hour within the next four hours The calculation formula of the traffic flow impact probability is: , wherein, is an indicator function, which takes 1 when the condition in the parentheses is met, and 0 otherwise, is the 90th percentile of the historical comprehensive parking intensity of the service area corresponding to the agent, and the traffic flow impact probability threshold is set When , it is determined that the service area corresponding to the agent will face traffic flow impact in the next four hours, and the capacity transfer channel of the adjacent area is automatically opened , allowing the agent to report the standby request to adjacent agents within the same dispatch cluster, and allowing adjacent agents to report the remaining capacity to it, realizing the capacity complementation within the cluster, and when , it is determined that the traffic flow is at a regular level, and the capacity transfer channel of the adjacent area is closed.

[0057] Subsequently, a three-section carbon pulse penalty function is constructed and the penalty calculation logic is determined: the measured carbon emissions reported by each agent in the entire corridor in the last hour is used as the basic data, and a rolling median algorithm is used to calculate the dynamic benchmark For each agent in the current hour h, three core derived indicators are calculated: emission amplitude ; spatial synchronization indicator , which is 1 when the same direction of three consecutive service areas simultaneously satisfies , otherwise 0; rising rate threshold , wherein, is the slope of the carbon emissions in this area within one hour, the rising rate of the carbon emissions of the agent in the current hour is first calculated , and then compared with the maximum allowed rising rate , the formula is: , A three-section carbon pulse penalty function is constructed based on the above three indicators : , wherein, , is the emission amplitude segmented threshold, is the base penalty coefficient, which satisfies , is the spatial synchronization amplification coefficient, For the rate over-threshold amplification coefficient (used to amplify the punishment intensity of the steep rise of carbon emissions), the segmented punishment is used to achieve precise control of mild over-amplitude mild suppression and severe over-amplitude severe suppression, while considering spatial coordination and rate control, and avoiding carbon emission pulse omission caused by single-dimensional punishment.

[0058] As shown in Figure 1 , error backfill refers to the automatic triggering of model correction and weight update when the deviation between the prediction result and the actual monitoring value exceeds the set threshold during continuous operation (when the average error or relative error exceeds the set range (generally 5%-20% of the maximum power or 5%-15% of the average error)). The "penalty coefficient enhancement" refers to the system applying higher penalty weights to periods with larger errors during carbon emission prediction or dispatch optimization to promote faster model convergence.

[0059] The penalty coefficient is divided into three intervals: The first interval is the safe interval, corresponding to normal deviation, and the basic weight is used; The second interval is the warning interval, when the error exceeds the first set threshold (usually 2%-5% of the maximum power), the system gradually increases the penalty coefficient; The third interval is the over-standard interval, when the error exceeds the second threshold (usually 5%-10% of the maximum power), the penalty coefficient will quickly rise to the upper limit range (generally 50-200 times of the basic value) to force the algorithm to converge and reduce subsequent deviations.

[0060] Finally, the distributed policy gradient algorithm is used to train the agent and freeze the strategy: the distributed policy gradient algorithm (PG) defined in this document is used to synchronize training of the agent set, and the training process is constructed to maximize the long-term comprehensive income as the goal , the reward function is composed of three parts, which are the actual power reported in step S4 , the proportion of unsatisfied demand , and the three-stage carbon pulse punishment , the specific expression is: , , and are different weight coefficients, after each iteration, the corridor dispatch layer collects the reward values of all agents , uses the min-max normalization method to map the reward values to the [0,1] interval, and performs normalization processing on the reward values of each agent to avoid training deviation caused by the magnitude difference of the reward values of different agents, and the reward item comes from the formal modeling of the corridor level operation target. The power constraint compliance, service quality guarantee, and carbon pulse suppression are mapped into three observable and dimensionless quantities, resulting in ,in, The effect of power bias (the deviation between actual power and cluster-level baseline, normalized by min–max). The proportion of unmet demand. A three-segment carbon pulse penalty based on a rolling median benchmark. Obtained through offline calibration, representing the weights of different coefficients: in the past Seeking answers in the historical log. ,in, Let q be the expected reward value of the agent, constrained by the upper limit of the total power of the corridor and the service quality threshold. Bayesian optimization is used and cross-validation is performed to select stable weights; all quantities are uniformly calculated. Interval normalization is used to eliminate dimensional differences; the required data are derived from the actual power reported by the agent. On-site service completion rate And the amplitude exceeding the limit, spatial synchronization indication and slope exceeding the threshold constructed from the emission measurement sequence, combined as .

[0061] The normalized reward is fed back into the value network of the corresponding agent. The agent's policy parameters are updated using a policy gradient descent algorithm, enabling the agent to gradually learn the optimal control policy that balances power constraints, service quality, and carbon emission suppression; a policy convergence threshold is set. The change in policy parameters of each agent during three consecutive iterations. Upon successful convergence of the policy training, the policy network parameters of the agent are frozen to prevent parameter drift during subsequent operation. The frozen policy network then performs rolling inference on traffic demand for the next 24 hours, outputting power level plans for five types of high-consumption equipment. Simultaneously, it calls the emission factor library defined in this document (containing carbon emission factors for three energy pathways: fuel, electricity, and self-generated power) to generate the power level plans. The formula for mapping hourly particle size to carbon emissions is as follows: , in, Let h be the planned power of equipment of type i in hour h. Hourly carbon emission curves are generated based on the carbon emission factors corresponding to the energy pathways.

[0062] The policy network is obtained by training and converging a multi-agent PG network, with each agent observing... (Considering factors such as dwell intensity, cluster-level power baseline, weather conditions, and electricity prices) Input to both ends of the network: Strategy With value The network uses a single temporal convolutional layer for extraction. The time characteristics are represented by two fully connected layers, outputting discrete power level distributions; parameters are initialized using Xavier, the optimizer is Adam, and the policy follows the time characteristics. Updated, value is trained using TD error. Training is performed in distributed synchronous rounds. At the end of each round, the reward for each agent is normalized to min–max and backpropagated; when three consecutive rounds meet the condition... Convergence is determined and parameters are frozen. The frozen strategy network is used to generate power tier plans for the next 24 hours on a rolling basis, and mapped with the emission factor library to obtain hourly carbon emission curves.

[0063] S6. Generate emission matrix and heat map based on power plan and emission factor library, compare the emission splitting error between prediction and actual emission, and complete the final output.

[0064] like Figure 6 As shown, S61, the power level plan for each service area output in step S5 is... With emission factor library A step-by-step mapping was performed. During the mapping process, factor groups were selected for each of the three energy pathways: oil-fired boilers, purchased electricity, and self-generated electricity. The carbon emission values ​​for five types of equipment were calculated at the hourly level. Aggregate all nodes in the same direction of travel according to the kilometer marker order and write them into the corridor emission matrix. The rows of the matrix represent timestamps. The column represents the node index. .right Color gradient rendering is used to generate a real-time emission heat map, which is then projected onto a large scheduling screen.

[0065] S62, the predicted emission matrix formed in sub-step S61 With real-time meter reading matrix Grid-by-grid comparison is performed, and the source and load dual-pointer method is used to split the error. The first pointer is anchored to the source of traffic flow, which is the stopping intensity of the same node in step S3. On-site passenger flow count Difference attribution to vehicle flow embedding distortion The second pointer anchors the equipment load by comparing the planned power at the same node. Power collected on site Difference attribution to device policy drift The two types of errors are accumulated separately as follows: and When the cumulative value exceeds the threshold or At that time, for the node Mark the source error or load error to clarify the direction of subsequent backtracking.

[0066] S63, difference analysis is performed on the marked nodes, and the upstream energy-saving down-regulation rate is calculated using the proportion method and the downstream energy preparation rate in advance . Write the future 24-hour cooperative start-stop schedule in the order of increasing kilometer pile and At the same time, the feedback to step S2, the proportion down-regulation of false-weather gating threshold , enhance the sensitivity of the model to abnormal traffic flow; the feedback to step S5, linearly increase the carbon pulse penalty amplification coefficient and , strengthen the strategy to suppress device drift. Finally, the new cluster-level power baseline and the start-stop schedule are issued to all agents through the broadcast channel , realize the corridor-level closed-loop calibration and start the next round of cooperative operation.

[0067] Embodiment 2 provides a specific application of a highway service area carbon emission prediction method based on spatiotemporal graph embedding: In the data preparation phase, access 107 sets of original vehicle records of highway gantry, 28 sets of entrance weighing stations and 62 sets of ETC gantry, a total of 112 million records. All records are first mapped to vehicle categories according to length, axle load and license plate, and vehicle equivalent and personnel equivalent double sequences are generated with fixed conversion coefficients; then, through the mileage difference, window average speed and traffic event level, the edge weight function is constructed, and the dynamic directed graph covering 34 pairs of service areas and 296 directional edges is generated, and the impedance matrix is refreshed every five minutes and written into the time series graph database. Based on this graph structure, the spatiotemporal graph neural network completes 14-day sliding window training on an A100 GPU cluster, and the holiday-weather gating module focuses on the return peak from the sixth to the eighth day of the lunar New Year, and through the gating weight amplification, the information channel across the 120-kilometer section is amplified.

[0068] ​In the joint debugging stage, the system is deployed with cluster-level control on the EMS master station side, 34 pairs of service areas are divided into 6 load clusters, and the power baseline broadcast channel is integrated in each cluster; the local PLC of the service area loads a multi-agent strategy execution module, and is connected in real time with the lighting, heating, catering, fast charging and refueling five fixed gear executions through Modbus-TCP. From 17:00 on February 17, 2024 to 02:00 on February 19 (main peak return), the system outputs a 24x12 window vehicle equivalent prediction matrix, and then obtains the profile results of the upstream conduction amount and the local fluctuation amount through the time delay and source double-level operators; the multi-agent strategy takes this as the input and calculates the power increase and decrease amount to each cluster. The power distribution SCADA record shows that the 10 pairs of service areas in the upstream Shantou-Fenjiang section reduced the heating and lighting power by an average of 17.4% during the period of 18:00-22:00, the 12 pairs of service areas in the downstream Jingzhu section simultaneously obtained 3.7MW of transferable capacity, the peak load was reduced from 41.3MW in the same period to 34.6MW, and the peak duration was compressed from 110 minutes to 48 minutes.

[0069] Table 1 is a specific data table;

[0070] To verify the effect, the core indicators of the technical method of the application and the three existing common schemes in the same period are compared, and the results are as follows: The data table shows that the application technology has better performance in prediction accuracy, peak load control and carbon emission reduction, while maintaining minute-level response. Centralized planning has lower prediction error, but the solving time is long and it is difficult to support five-minute rolling prediction; the single-node model and manual scheduling lack awareness of cross-node peak conduction during the peak period, resulting in limited peak shaving and emission reduction effect.

[0071] As shown in Figure 2 , the existing schemes mostly use single-agent regression prediction models, such as time series regression (such as ARIMA) or ensemble learning algorithms (such as XGBoost). Such models can only rely on historical power or carbon emission single-variable data for short-term prediction and cannot effectively capture the spatial correlation between service areas and the conduction relationship between upstream and downstream. The output of such models is usually a static prediction curve, which needs to rely on manual correction of periodic adjustment parameters. The prediction accuracy of such models decreases significantly during holiday peaks or sudden traffic incidents, and they cannot adaptively optimize according to external environmental changes.

[0072] The algorithm system proposed in the application introduces a dynamic spatio-temporal graph neural network framework in structure, embeds the spatial relationship nodes of gantry and service area, and combines multi-source dynamic data such as traffic flow, energy consumption and emission to form node-edge-time three-dimensional information coupling.

[0073] Based on the real-time closed loop of load and emission, the technical method of the application still reduces the overall demand electricity fee of the corridor by 11.6% under the condition that the total equivalent of the vehicle flow rises by 8.2%. The emission heat map and power migration log automatically generated by the operation center are recognized by the third-party auditing organization as a complete traceable accounting chain, which also provides a standardized interface for subsequent replication to the port collection and distribution channel or intercity railway complex. As can be seen, the embodiment verifies the practicability and promotion potential of the application technology in large-scale, multi-node and high-traffic energy systems.

[0074] The above are preferred embodiments of the application, and do not limit the protection scope of the application, therefore: any equivalent changes made in structure, shape, principle, etc. according to the application should be covered within the protection scope of the application.

Claims

1. A method for predicting carbon emissions of a highway service area based on spatio-temporal graph embedding, characterized in that, Comprise: S1, obtain gantry and each service area entrance and exit vehicle counting flow, generate vehicle equivalent sequence and personnel equivalent sequence through vehicle classification processing, and aggregate into corridor level tensor; S2, based on the corridor level tensor, construct a dynamic topology and a space-time prediction framework, including establishing a dynamic directed graph containing the gantry and the service area, and generating vehicle predicted equivalent using a space-time graph neural network with an embedded gating module; S3, according to the vehicle predicted equivalent, calculate the dynamic time delay to split the upstream conduction quantity and the local fluctuation quantity, generate the traffic flow embedding vector and the comprehensive parking intensity curve; S4, combine the comprehensive parking intensity curve to map the service area to an agent and divide the dispatch cluster, calculate the cluster level power baseline and broadcast it to the agent, and the agent executes device discrete control and reports operation data; S5, according to the historical operation data, construct a cold start rule table, and control the adjacent area channel combined with the traffic impact probability, train and freeze the agent strategy through three-stage carbon pulse punishment and distributed policy gradient, obtain the power plan and carbon emission curve; S6, based on the power plan and the emission factor library, generate the emission matrix and the heat map, compare the prediction and the measured emission split error, and complete the final output.

2. The highway service area carbon emission prediction method based on space-time graph embedding according to claim 1, characterized in that Step S1, specifically comprising the following substeps: S11, read the vehicle counting flow, sequentially execute vehicle length distinction, axle load label supplement and license plate attribute mapping, classify vehicles through three-dimensional cross-retrieval table of length, axle load and license plate, and generate sub-vehicle type counting sequence by merging and accumulating each class 、 、 ; S12, calling a set of conversion coefficients The vehicle equivalent sequence and the personnel equivalent sequence are obtained by mapping the vehicle type count sequence and the corresponding coefficient product. S13, aggregate the vehicle equivalent sequence and the personnel equivalent sequence according to the corridor number, the lane direction and the inner procedure number, accumulate the corridor level daily aggregated vector through the single day accumulation window, and write it into a unified tensor structure; wherein, , and represent the conversion factors for small passenger cars, large passenger cars, and trucks, respectively, , and represent the class markers for small passenger cars, large passenger cars, and trucks, respectively.

3. The method of claim 1, wherein, The dynamic directed graph containing the gantry and the service area includes writing the gantry and the service area into the node list in increasing order of kilometer post and appending virtual boundary nodes, recording node attributes, creating directed edges between nodes with adjacent mile coordinates in the node list V according to the actual travel direction of vehicles, forming a directed edge set E, constructing an adjacency list A combining the node list V and the directed edge set E, obtaining an initial directed graph structure, extracting the inter-node mile difference, the window average speed and the traffic event level from each directed edge of the adjacency list, and calculating the propagation impedance of each directed edge by weight normalization superposition. The propagation impedance expression is: ， in, For the edge At any moment Propagation impedance, These are the weights of the three factors. For nodes To node The difference in kilometers, The maximum kilometer difference between adjacent nodes in the same direction. For free flow reference velocity, For the edge At any moment average speed, To ensure the denominator is a very small positive number, and to prevent the denominator from being zero, For the edge At any moment Event level, This is the event amplification factor, which controls the magnitude of the increase in impedance caused by the disaster.

4. The method of claim 1, wherein, The space-time graph neural network with an embedded gating module generates vehicle predicted equivalent, including extracting the daily aggregated vehicle equivalent vector of each node based on the unified tensor structure, mapping the input feature matrix of the space-time graph neural network according to the order of the node list, constructing a space-time graph neural network containing a graph convolution layer and a time convolution layer, embedding a gating module between the graph convolution layer and the time convolution layer, inputting the input feature matrix X into the space-time graph neural network, capturing spatial correlation through the graph convolution layer combined with real-time propagation impedance, capturing time sequence dependence through the time convolution layer, executing time-by-time reasoning after modulation by the gating module, and outputting the vehicle equivalent prediction sequence. The vehicle equivalent prediction sequence expression is: ， wherein, is a spatio-temporal graph neural network inference function of a gating module, is a corresponding vehicle model feature sub-matrix, A is an adjacency list, is a propagation impedance matrix at time t, is a gating module state.

5. The method of claim 1, wherein, The dynamic time delay is calculated according to the vehicle prediction equivalent, the upstream conduction quantity is split from the local fluctuation quantity, which comprises aggregating the service area nodes in the corridor according to the vehicle equivalent prediction sequence in the order of kilometer post increment, and establishing the local prediction sequence for each target service area node The upstream node translation superposition sequence The associated physical parameters of the target service area node and each upstream node are collected, the dynamic time delay is calculated for each vehicle type one by one, and the prediction sequence of the upstream node is translated clockwise after the dynamic time delay, and then superposed with the local prediction sequence of the target service area The minimum value is taken every hour to obtain the upstream conduction quantity every hour, and the dynamic time delay calculation formula is: ， wherein, is the road distance from the upstream node to the target node , is the vehicle category reference speed in free flow state, is the vehicle category corridor average speed at time , is the accident severity level at time for the edge , is the corridor overall weather severity level, are the distance, speed, accident, and weather weight coefficients, is a very small positive number to avoid zero denominator, is the dynamic delay for the vehicle category from the node to the node .

6. The method of claim 1, wherein, The process of generating the traffic flow embedding vector and the comprehensive dwell intensity curve includes calculating the upstream transmission ratio and the local fluctuation ratio based on the upstream transmission volume, generating the cross-regional traffic flow embedding vector based on the upstream transmission ratio and the local fluctuation ratio, and calling the historical offline dwell probability surface. By combining the time delay correction coefficients corresponding to the vehicle flow embedding vector, future time intervals are generated. The vehicle arrived on the timeline. Search for car models At any moment Dwell probability factor ,Will Multiplying the signal strength component in the traffic flow embedding vector yields the single-vehicle model parking intensity sequence. For three types of vehicles By performing time alignment and intensity overlay, the comprehensive dwelling intensity curve for the next decision window of the target service area is obtained.

7. The method of claim 1, wherein, The combined comprehensive parking intensity curve maps the service area as an agent and divides the scheduling cluster, including constructing a service area index table based on the comprehensive parking intensity curve of each service area and the static attribute data of all service areas in the corridor, converting the corresponding service area of each row in the index table into multiple independent control agents according to the mapping relationship, clustering and dividing the agents according to the driving direction of the target service area and the adjacent nature of the kilometer post, and grouping the agents with the same direction and kilometer post into the same scheduling cluster.

8. The method of claim 1, wherein, The cluster-level power baseline is broadcast to the intelligent agent, including calling the average load factor of the service area and the upper limit of the total power of the corridor in the historical operation database, and calculating the cluster-level power baseline of each dispatch cluster , a power baseline broadcast channel is constructed, and is broadcast to the corresponding intelligent agent according to the dispatch cluster affiliation relationship and a device discrete action list is generated, and the cluster-level power baseline calculation formula is: ， wherein m is the number of agents in the cluster , is the rated transformation capacity of the corresponding service area of the agent , is the historical average load factor is the corridor total power cap, and M is the total number of clusters in the corridor.

9. The method of claim 1, wherein, Step S5 specifically includes the following sub-steps: S51, combine historical operation data and establish an index according to three dimensions of season, time period and service area type to generate a start-stop rule table, map the start-stop rule table to a plurality of high-consumption device power gear combinations, and load the corresponding cold start strategy set for each agent; S52, calculate the future four-hour traffic impact probability according to the vehicle equivalent prediction sequence and the comprehensive parking intensity curve, set a traffic impact probability threshold, and determine the switch state of the neighbor area capacity transfer channel of the agent according to the probability threshold; S53, take the carbon emission measured value reported by the agent in the last hour at the start time of the current decision cycle as the basis data, calculate the rolling median value of the carbon emission of the entire corridor, and take the median value as the reference to calculate the emission amplitude, spatial synchronization indication and rising rate threshold, and construct a three-section carbon pulse penalty function according to the emission amplitude, spatial synchronization indication and rising rate threshold; S54, train the agent using a distributed policy gradient algorithm, construct a reward function based on the carbon pulse penalty, perform normalization processing on the reward value of each agent after each iteration, and feed the normalized reward back to the value network of the corresponding agent to update the policy parameters.

10. The method of claim 1, wherein, Step S6 specifically includes the following sub-steps: S61, map and calculate the power values of the five types of devices in the power gear plan, i.e. lighting, heating, catering, charging piles and fueling islands, by hour granularity and corresponding emission factors of energy paths to obtain the hourly carbon emission value of each service area; S62, write the hourly carbon emission value of each service area into a corridor-level emission matrix, which corresponds to the time stamp in the row and the service area node index in the column, and generate a real-time carbon emission heat map by rendering the emission matrix using a color gradient; S63, collect hourly measured carbon emission data of each service area to form a measured emission matrix, compare the emission matrix with the measured emission matrix by grid, split the error using the double-pointer method, and obtain the traffic embedding distortion error and device strategy drift error; S64, accumulate the traffic embedding distortion error and the device strategy drift error according to the service area nodes respectively to obtain the total distortion error and the total drift error of each node, and take the emission matrix, the carbon emission heat map and the error splitting result as the final output.

Images