New energy logistics network intelligent management system and method based on digital twinning

Abstract

The invention discloses a new energy logistics network intelligent management system and method based on digital twinning, and relates to the technical field of logistics network management.The system senses abnormal events in a logistics network in real time, maps the events to a digital twinning body through semantic analysis, and adjusts the topological structure and operation parameters of the digital twinning network; on the basis of the digital twin network, a seepage model of abnormal event influence propagation is constructed, the infection probability and seepage centrality of network nodes are analyzed, and an abnormal influence thermodynamic diagram is generated; constructing a scene tree according to the abnormal influence thermodynamic diagram and the current scheduling plan, performing parallel simulation on each branch of the scene tree in a digital twin environment, and selecting an optimal scheduling strategy through target evaluation; and tracking an execution effect in real time, constructing an experience case library, and matching similar historical scenes through analogy reasoning to realize self-adaptive optimization of an exception coping strategy. The response speed is improved; and the robustness of the system is improved.

Description

Technical Field

[0001] This invention relates to the field of logistics network management technology, specifically to an intelligent management system and method for new energy logistics networks based on digital twins. Background Technology

[0002] In current new energy logistics network management systems, the application of digital twin technology is mainly focused on monitoring and simulating routine operations. Existing solutions typically rely on periodically collected vehicle trajectories, charging pile status, and order data to construct static or quasi-static twin models for route planning and scheduling optimization. However, these methods have significant drawbacks when facing unexpected events. First, existing systems exhibit severe lag in responding to abnormal events, often requiring manual confirmation or waiting for the next data collection cycle to detect faults, accidents, or order cancellations, leading to a disconnect between the twin model and the real-time state of the physical world. When vehicle malfunctions or charging pile disconnections occur, the system still makes scheduling decisions based on outdated network topology, potentially leading other vehicles to already failed charging nodes or congested road sections, not only failing to alleviate the problem but also exacerbating the spread of the anomaly's impact. Second, traditional methods lack the ability to quantitatively analyze the propagation mechanisms of abnormal events, often treating anomalies as isolated local events and ignoring the cascading effects between traffic congestion, energy shortages, and order delays. A failure in one node can spread rapidly through road network connections, leading to a wider range of network performance degradation, and existing technologies struggle to identify such propagation paths and the extent of their impact in advance.

[0003] Therefore, this invention discloses a new energy logistics network intelligent management system and method based on digital twins to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent management system and method for new energy logistics networks based on digital twins, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a new energy logistics network intelligent management method based on digital twins, the corresponding method comprising the following steps:

[0006] S1: Real-time perception of abnormal events in the logistics network, mapping abnormal events to digital twins through semantic parsing, and adjusting the topology and operating parameters of the digital twin network; S2: Based on digital twin networks, construct a seepage model for the propagation of the impact of abnormal events, analyze the infection probability and seepage centrality of network nodes, and generate a heat map of the impact of abnormal events. S3: Based on the heatmap of the impact of anomalies and the current scheduling plan, construct a scenario tree, perform parallel simulation on each branch of the scenario tree in the digital twin environment, and select the optimal scheduling strategy through target evaluation; S4: Track execution results in real time, build an experience case library, and match similar historical scenarios through analogical reasoning to achieve adaptive optimization of anomaly response strategies.

[0007] S1 includes the following: S101: Real-time collection of multi-source data in the logistics network, including abnormal event data and regular operation data; the abnormal event data includes at least new energy vehicle fault data, charging pile disconnection alarm flag, order cancellation flag, and traffic accident data; the regular operation data includes at least new energy vehicle location, remaining battery power, charging pile power, and order origin and destination coordinates. By using predefined event semantic parsing templates, abnormal event data is mapped to the attributes of the corresponding entities in the digital twin. This includes assigning fault type attributes to the virtual new energy vehicle model based on fault data of new energy vehicles and recording the start time of the anomaly; mapping traffic accident data to the corresponding road segments in the road network model and generating temporary obstacle attributes; setting the status field of charging piles with a disconnection alarm flag of 1 to unavailable; and setting the status field of orders with an order cancellation flag of 1 to cancelled, thus completing the structured representation of abnormal events and their association with twin entities. S102: Adjust the topology and operating parameters of the digital twin network based on the extracted abnormal event types, locations, and impact parameters; For charging pile disconnection alarms, the corresponding charging pile node is temporarily removed from the charging service network, and the charging network adjacency relationship is updated; for traffic accident data, the affected road segment is located based on the accident coordinates, and the traffic capacity of the corresponding road segment is downgraded: the road network edge weights are updated; for new energy vehicle fault data, the fault status is marked on the new energy vehicle model according to the fault type and the radius of influence, and a temporary node is added to indicate that the fault occupies the lane; for order cancellations, the order pool status is updated, and the transportation demand of the corresponding order is removed from the scheduling task list. The reconstructed digital twin network is represented as a graph structure, which includes a set of entity nodes, a set of edges that change over time, and a node attribute matrix. The node attribute matrix includes regular operating data and abnormal event feature vectors. The abnormal event feature vectors include at least the event type code, intensity level, influence radius, and start time fields. This application collects abnormal event data (fault codes, alarms, cancellation markers, accident coordinates) and regular operational data, and uses semantic parsing templates to map anomalies to twin entity attributes. This achieves unified representation of heterogeneous data and precise binding of event semantics, providing structured anomaly features for subsequent analysis. The topology of the twin network is adjusted based on abnormal events (e.g., removing faulty charging pile nodes) and operational parameters (e.g., reducing the traffic capacity of accident sections), and the dynamically changing network is output in graph form. This reflects the changes in the network's physical structure caused by anomalies in real time, ensuring that subsequent propagation analysis is based on the current real network state. For charging pile disconnections, nodes are temporarily removed and adjacency relationships are updated; for traffic accidents, the edge weights of the corresponding road sections are downgraded; for vehicle malfunctions, the malfunction status is marked and temporary nodes are added to indicate lane occupancy. These operations enable the twin network to accurately simulate changes in the physical world caused by anomalies, such as connection interruptions, reduced traffic capacity, and new obstacles, providing a realistic graph structure input for the propagation model. For order cancellations, the order pool status is updated and the corresponding transportation demand is removed, ensuring that the scheduling task list is always consistent with the actual situation, avoiding invalid order dispatches and resource waste. Different adjustment methods are adopted for different anomalies (such as charging pile failures and traffic accidents), which reflects the differentiated modeling of the physical characteristics of the anomalies and makes the subsequent analysis more realistic.

[0008] S2 includes the following: S201: Define the initial infection probability of each node on the digital twin network; to describe the propagation dynamics of anomalous events in the network, this invention adopts an extended SIR model based on percolation theory, and the node state is divided into three categories: susceptible state (S): the node has not yet been affected by the anomalous event, but may be propagated by neighboring infected nodes; infected state (I): the node is currently affected by the anomalous event and may propagate the anomalous event to neighboring susceptible nodes; recovered state (R): the node's anomalous impact has been eliminated, it has returned to normal function, and no longer participates in propagation; Define the probability β of an infected node i propagating an anomaly to its neighboring susceptible node j. （i，j）= β0 / w （i，j） Where β0 is the basic propagation coefficient, w （i，j） Define the current road network edge weight connecting edge (i, j); define the recovery rate γ of adjacent susceptible node j. j The inverse of the time required for abnormal recovery; the infection probability I of adjacent susceptible nodes. j (t) and recovery probability R j The evolution of (t) is described by the following seepage model: ; Where N(j) represents the set of neighboring nodes of node j, and the initial condition is I. j (0)=p j (0), R j (0) = 0; S202: Solve the seepage model using the fourth-order Runge-Kutta method or the Euler method to obtain the infection probability of each adjacent susceptible node j at time t; extract nodes whose infection probability exceeds a preset infection probability threshold and delineate the abnormal influence area; analyze the seepage centrality K of node j. j (t), ; Nodes with seepage centrality exceeding the seepage centrality threshold are marked as critical nodes, and a list of critical nodes and the boundary information of the affected area are output. S203: Divide the monitored area into a uniform geographical grid, where the heat value of each grid is equal to the average infection probability of all nodes within the grid, and generate an anomaly impact heat map.

[0009] This application constructs an extended SIR seepage model by defining initial infection probability, transmission probability, and recovery rate. This model quantitatively describes how anomalous events gradually spread along the network topology, revealing the propagation dynamics of anomalies such as congestion and failures, surpassing traditional static influence circle analysis. Solving the model yields the infection probability of each node, delineates the anomalous impact area, and calculates seepage centrality to identify key propagation nodes, helping managers identify "super-spreaders" of anomalous spread and providing a basis for precise intervention. Mapping the infection probability to a geographic grid generates a heatmap, visually displaying the spatiotemporal distribution of anomalous impacts, facilitating rapid location of high-risk areas, and supporting intuitive decision visualization.

[0010] S3 includes the following: S301: Using the heat map of abnormal impacts and the current scheduling plan as input, construct a scenario tree; the current scheduling plan includes new energy vehicle routes, charging task allocation, and order new energy vehicle matching relationships; the root node of the scenario tree represents the network state at the current moment, and multiple branches are generated from the root node based on a predefined intervention rule base, each branch represents a scheduling intervention measure, and each branch corresponds to a set of intervention operations; The intervention rule base includes at least the following: when the grid heat value is greater than the grid heat value threshold, the corresponding geographic grid is marked as a high-impact area, and a detour branch is generated. The intervention operation set includes applying Dijkstra's algorithm or A* algorithm that considers dynamic weights for the affected new energy vehicles. The algorithm replans the path; when a charging pile is unavailable due to an anomaly or is located in a high-impact area, a charging load transfer branch is generated. The intervention operation set includes obtaining a list of affected new energy vehicles, using bipartite graph matching or linear programming to solve the adjustment scheme of the charging allocation matrix based on the charging pile status and the remaining power of the new energy vehicles, and updating the charging pile allocation target for each new energy vehicle; when the order origin and destination are located in a high-impact area or cannot be fulfilled due to a new energy vehicle malfunction, an order reassignment branch is generated. The intervention operation set includes releasing unfinished orders from the original matching relationship, updating the order pool, calling the new energy vehicle order matching algorithm, including but not limited to the assignment model that considers time windows, and generating a new order new energy vehicle matching relationship; S302: In a digital twin environment, perform parallel simulation on each branch of the scenario tree to deduce the evolution of the network state within a future time window from the current moment. The specific simulation mechanism is as follows: Within each simulation step, the topology and node attributes of the digital twin network are updated according to the set of intervention operations of the branch, and the node infection probability is updated by calling the permeation model based on the updated network state. At the same time, logistics operation indicators are calculated according to the new new energy vehicle routes, charging plans, and order matching. The logistics operation indicators include at least the location of new energy vehicles, remaining power, and order completion status. The above process is repeated until the simulation ends. After the simulation, the target evaluation index vector for each branch is recorded. The target evaluation index vector includes total delay time, total energy consumption, and anomaly recovery time. The total delay time is the sum of the differences between the actual completion time and the planned completion time of all orders. The total energy consumption represents the total electricity consumed by all new energy vehicles during the simulation. The anomaly recovery time represents the time required from the current moment until the thermal value of all grids is lower than the preset recovery threshold. S303: Construct a set of reference points, which consist of historical best target values, business expected targets, or ideal points. Each reference point contains a corresponding optimal evaluation index vector. For each branch, analyze the fitness with all reference points. Take the minimum value as the optimal fitness of the corresponding branch. The branch with the lowest fitness is selected as the optimal branch, and its corresponding set of intervention operations is the optimal scheduling strategy. The optimal strategy is then transformed into a specific set of scheduling instructions and issued to the actual logistics execution system.

[0011] This application uses three key indicators—total delay, total energy consumption, and anomaly recovery time—to quantitatively evaluate branches, taking into account efficiency, cost, and stability. Through fitness calculation based on reference points, it can balance multiple conflicting objectives and select the overall optimal strategy. The reference points consist of historical best practices or business expectations, ensuring that strategy selection respects the actual feasible domain while guiding the system towards a better goal, avoiding the limitations of simply relying on fixed weights.

[0012] S4 includes the following: S401: During execution, continuously collect real-time operational data of the logistics network, including but not limited to: real-time location, speed, remaining battery power, and fault status of new energy vehicles; working status and output power of charging piles; actual order completion status; and newly occurring abnormal events, including but not limited to secondary faults, traffic accidents, and road congestion; simultaneously, record the performance indicators of the actual execution effect: actual total delay time and actual total energy consumption; compare with the predicted values ​​of the corresponding branches, and calculate the execution deviation vector; S402: Combine the current abnormal event feature vector, the selected intervention strategy, the execution deviation vector, and the final recovery flag into a case and store it in the experience case library; The final recovery flag is a binary variable. If the anomaly is resolved within a preset time window and no new chain of anomalies is triggered, the final recovery flag is 1; otherwise, it is 0. S403: When a new abnormal event occurs, extract the corresponding abnormal event feature vector and calculate the Euclidean distance between the new abnormal event feature vector and the abnormal event feature vector of each historical case in the knowledge base; if the minimum distance is less than the preset threshold and the final recovery flag of the corresponding case is 1, then directly retrieve the intervention strategy of the corresponding case as a candidate solution; otherwise, construct a scenario tree to generate a new strategy.

[0013] This application continuously collects actual execution data and compares it with the predicted data to generate an execution deviation vector, thereby quantifying the model's prediction error and providing a basis for subsequent optimization. When a new anomaly occurs, the most similar historical case is retrieved by calculating the Euclidean distance of the feature vectors. If a match is found, the strategy is directly invoked, avoiding repeated deductions and significantly improving response speed. At the same time, warnings from failed cases prevent repeating the same mistakes.

[0014] Another aspect of this application provides a digital twin-based intelligent management system for new energy logistics networks. The system is applied to the aforementioned digital twin-based intelligent management method for new energy logistics networks. The system includes an anomaly adjustment module, an anomaly impact propagation analysis module, a scheduling strategy analysis module, and a strategy optimization module. The anomaly adjustment module senses abnormal events in the logistics network in real time, maps the events to the digital twin through semantic parsing, and adjusts the topology and operating parameters of the digital twin network. The abnormal impact propagation analysis module is based on a digital twin network, constructs a seepage model for the propagation of abnormal event impacts, analyzes the infection probability and seepage centrality of network nodes, and generates an abnormal impact heatmap. The scheduling strategy analysis module constructs a scenario tree based on the heatmap of anomaly impacts and the current scheduling plan. In the digital twin environment, it performs parallel simulations on each branch of the scenario tree and selects the optimal scheduling strategy through target evaluation. The strategy optimization module tracks the execution effect in real time, builds an experience case library, and matches similar historical scenarios through analogical reasoning to achieve adaptive optimization of the anomaly response strategy.

[0015] The anomaly adjustment module includes a data acquisition and mapping unit and a digital twin network update unit; The data acquisition and mapping unit is used to collect multi-source data in the logistics network in real time, and to map abnormal events to the attributes of the corresponding entities in the digital twin through a predefined event semantic parsing template. The digital twin network update unit adjusts the topology and operating parameters of the digital twin network based on the extracted abnormal event types, locations, and impact parameters. The reconstructed digital twin network is represented as a graph structure, which includes a set of entity nodes, a set of edges that change over time, and a node attribute matrix.

[0016] The abnormal impact propagation analysis module includes a parameter definition unit, an infection analysis unit, and a heat map generation unit; The parameter definition unit is used to define the initial infection probability of each node on the digital twin network; define the probability of an infected node spreading anomalies to neighboring susceptible nodes; define the recovery rate of neighboring susceptible nodes as the reciprocal of the time required for anomaly recovery; and describe the evolution of the infection probability and recovery probability of neighboring susceptible nodes by a percolation model. The infection analysis unit is used to solve the seepage model to obtain the infection probability of each adjacent susceptible node at time t; extract nodes whose infection probability exceeds a preset infection probability threshold and delineate the abnormal influence area; analyze the seepage centrality of the nodes; mark nodes whose seepage centrality exceeds the seepage centrality threshold as key nodes, and output a list of key nodes and the boundary information of the influence area. The heatmap generation unit is used to divide the monitoring area into uniform geographical grids, where the heat value of each grid is equal to the average infection probability of all nodes within the grid, thereby generating an anomaly impact heatmap.

[0017] The scheduling strategy analysis module includes a scenario tree construction unit, a simulation evolution unit, and a scheduling strategy filtering unit; The scenario tree construction unit is used to construct a scenario tree with the heat map of abnormal impacts and the current scheduling plan as inputs. The current scheduling plan includes the new energy vehicle route, charging task allocation, and order new energy vehicle matching relationship. The root node of the scenario tree represents the network state at the current moment. Starting from the root node, multiple branches are generated based on a predefined intervention rule base. Each branch represents a scheduling intervention measure, and each branch corresponds to a set of intervention operations. The simulation evolution unit is used to perform parallel simulation of each branch of the scenario tree in a digital twin environment, and to deduce the evolution of the network state within a future time window from the current moment; after the deduction is completed, the target evaluation index vector of each branch is recorded. The scheduling strategy filtering unit is used to construct a set of reference points. The reference points consist of historical best target values, business expected targets, or ideal points. Each reference point contains a corresponding preferred evaluation index vector. For each branch, the fitness with all reference points is analyzed. The minimum value is taken as the optimal fitness of the corresponding branch. The branch with the lowest fitness is selected as the optimal branch, and its corresponding set of intervention operations is the optimal scheduling strategy. The optimal strategy is then transformed into a specific set of scheduling instructions and issued to the actual logistics execution system.

[0018] The strategy optimization module includes an execution deviation analysis unit and a strategy optimization unit; The execution deviation analysis unit is used to continuously collect real-time operating data of the logistics network during the execution process, including but not limited to: real-time location, speed, remaining battery power, and fault status of new energy vehicles; working status and output power of charging piles; actual order completion status; and newly occurring abnormal events, including but not limited to secondary faults, traffic accidents, and road congestion; at the same time, it records the performance indicators of the actual execution effect: actual total delay time and actual total energy consumption; and compares them with the inferred prediction values ​​of the corresponding branches to calculate the execution deviation vector. The strategy optimization unit is used to combine the current abnormal event feature vector, the selected intervention strategy, the execution deviation vector, and the final recovery flag into a case and store it in the experience case library. When a new abnormal event occurs, the corresponding abnormal event feature vector is extracted, and the Euclidean distance between the new abnormal event feature vector and the abnormal event feature vector of each historical case in the knowledge base is calculated. If the minimum distance is less than a preset threshold and the final recovery flag of the corresponding case is 1, the intervention strategy of the corresponding case is directly retrieved as a candidate solution. Otherwise, a scenario tree is constructed to generate a new strategy.

[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This application collects abnormal event data and regular operational data, and uses semantic parsing templates to map anomalies to twin entity attributes, achieving unified representation of heterogeneous data and precise binding of event semantics, providing structured anomaly features for subsequent analysis. Adjusting the topology and operational parameters of the twin network based on abnormal events and outputting the dynamically changing network in graph structure form can reflect the changes in the network's physical structure caused by anomalies in real time, allowing subsequent propagation analysis to be based on the current real network state. This application constructs an extended SIR seepage model by defining initial infection probability, propagation probability, and recovery rate, which can quantitatively describe how abnormal events gradually spread along the network topology, revealing the propagation dynamics of anomalies such as congestion and failures, surpassing traditional static influence circle analysis. Solving the model yields the infection probability of each node, delineates the anomaly impact area, and calculates seepage centrality to identify key propagation nodes, helping managers identify "super-spreaders" of anomaly propagation and providing a basis for precise intervention. Mapping the infection probability to a geographic grid generates a heat map, intuitively displaying the spatiotemporal distribution of anomaly impacts, facilitating rapid location of high-risk areas, and supporting intuitive decision visualization. This application uses three key indicators—total delay, total energy consumption, and anomaly recovery time—to quantitatively evaluate branches, balancing efficiency, cost, and stability. Through fitness calculation based on reference points, it can balance multiple conflicting objectives and select the optimal comprehensive strategy. Reference points consist of historical best practices or business expectations, ensuring that strategy selection respects the feasible domain while guiding the system towards a better goal, avoiding the limitations of relying solely on fixed weights and providing system robustness. This application continuously collects actual execution data and compares it with predicted scenarios to generate an execution deviation vector, quantifying model prediction errors and providing a basis for subsequent optimization. When a new anomaly occurs, the most similar historical case is retrieved by calculating the Euclidean distance of the feature vectors. If a match is found, the strategy is directly invoked, avoiding repeated deductions and significantly improving response speed; simultaneously, warnings from failed cases prevent repetition. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the intelligent management method for new energy logistics networks based on digital twins according to the present invention. Figure 2 This is a schematic diagram of the structure of the intelligent management system for new energy logistics networks based on digital twins according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 This invention provides a technical solution: an intelligent management method for new energy logistics networks based on digital twins, the method comprising the following steps: S1: Real-time perception of abnormal events in the logistics network, mapping abnormal events to digital twins through semantic parsing, and adjusting the topology and operating parameters of the digital twin network; S1 includes the following: S101: Real-time collection of multi-source data in the logistics network, including abnormal event data and routine operation data; abnormal event data includes at least new energy vehicle fault data, charging pile disconnection alarm flag, order cancellation flag, and traffic accident data; routine operation data includes at least new energy vehicle location, remaining battery power, charging pile power, and order origin and destination coordinates. By using predefined event semantic parsing templates, abnormal event data is mapped to the attributes of the corresponding entities in the digital twin. This includes assigning fault type attributes to the virtual new energy vehicle model based on fault data of new energy vehicles and recording the start time of the anomaly; mapping traffic accident data to the corresponding road segments in the road network model and generating temporary obstacle attributes; setting the status field of charging piles with a disconnection alarm flag of 1 to unavailable; and setting the status field of orders with an order cancellation flag of 1 to cancelled, thus completing the structured representation of abnormal events and their association with twin entities. S102: Adjust the topology and operating parameters of the digital twin network based on the extracted abnormal event types, locations, and impact parameters; For charging pile disconnection alarms, the corresponding charging pile node is temporarily removed from the charging service network, and the charging network adjacency relationship is updated; for traffic accident data, the affected road section is located according to the accident coordinates, and the traffic capacity of the corresponding road section is downgraded: the road network edge weights are updated. Example 1: In this example, the road network edge weight w （u，v） =w1 （u，v） ×(1 η acc ), where w1 （u，v） For the original passage capacity, η acc This is the accident reduction factor; For fault data of new energy vehicles, the fault status is marked on the new energy vehicle model according to the fault type and the radius of influence, and a temporary node is added to indicate that the fault is blocking the road; for order cancellation, the order pool status is updated, and the transportation requirements of the corresponding order are removed from the scheduling task list. The reconstructed digital twin network is represented as a graph structure, which includes a set of entity nodes, a set of edges that change over time, and a node attribute matrix. The node attribute matrix includes regular operating data and abnormal event feature vectors. The abnormal event feature vectors include at least the event type code, intensity level, influence radius, and start time fields. S2: Based on digital twin networks, construct a seepage model for the propagation of the impact of abnormal events, analyze the infection probability and seepage centrality of network nodes, and generate a heat map of the impact of abnormal events. S2 includes the following: S201: Define the initial infection probability of each node on the digital twin network; Example 2: In this example, if node v is the node where the primary abnormal event occurred, then the initial infection probability p v (0) is marked as 1, otherwise... Where α is the maximum propagation intensity, λ is the spatial attenuation coefficient, and d （v，u） Let v be the distance between node v and the nearest primary anomalous node u. To describe the propagation dynamics of anomalous events in a network, this invention adopts an extended SIR model based on percolation theory. Node states are divided into three categories: susceptible state (S): the node has not yet been affected by the anomalous event, but may be propagated by neighboring infected nodes; infected state (I): the node is currently affected by the anomalous event and may propagate the anomalous event to neighboring susceptible nodes; and recovered state (R): the node's anomalous impact has been eliminated, it has returned to normal function, and no longer participates in propagation. Define the probability β of an infected node i propagating an anomaly to its neighboring susceptible node j. （i，j）= β0 / w （i，j） Where β0 is the basic propagation coefficient, w （i，j） Define the current road network edge weight connecting edge (i, j); define the recovery rate γ of adjacent susceptible node j. j The inverse of the time required for abnormal recovery; the infection probability I of adjacent susceptible nodes. j (t) and recovery probability R j The evolution of (t) is described by the following seepage model: ; Where N(j) represents the set of neighboring nodes of node j, and the initial condition is I. j (0)=p j (0), R j (0) = 0; S202: By solving the seepage model using the fourth-order Runge-Kutta method or the Euler method, the infection probability of each adjacent susceptible node j at time t is obtained; nodes with infection probabilities exceeding a preset infection probability threshold are extracted, and abnormal influence areas are delineated; the seepage centrality K of node j is analyzed. j (t), ; Nodes with seepage centrality exceeding the seepage centrality threshold are marked as critical nodes, and a list of critical nodes and the boundary information of the affected area are output. S203: Divide the monitored area into a uniform geographical grid, where the heat value of each grid is equal to the average infection probability of all nodes within the grid, and generate an anomaly impact heat map.

[0023] S3: Based on the heatmap of the impact of anomalies and the current scheduling plan, construct a scenario tree, perform parallel simulation on each branch of the scenario tree in the digital twin environment, and select the optimal scheduling strategy through target evaluation; S3 includes the following: S301: Using the heatmap of abnormal impacts and the current scheduling plan as input, construct a scenario tree; the current scheduling plan includes the new energy vehicle route, charging task allocation, and order new energy vehicle matching relationship; the root node of the scenario tree represents the network state at the current moment, and multiple branches are generated from the root node based on the predefined intervention rule base, each branch represents a scheduling intervention measure, and each branch corresponds to a set of intervention operations; The intervention rule base should include at least the following: when the grid heat value is greater than the grid heat value threshold, the corresponding geographic grid should be marked as a high-impact area, and a detour branch should be generated. The intervention operation set should include applying Dijkstra's algorithm or A* algorithm that considers dynamic weights for the affected new energy vehicles. The algorithm replans the path; when a charging pile is unavailable due to an anomaly or is located in a high-impact area, a charging load transfer branch is generated. The intervention operation set includes obtaining a list of affected new energy vehicles, using bipartite graph matching or linear programming to solve the adjustment scheme of the charging allocation matrix based on the charging pile status and the remaining power of the new energy vehicles, and updating the charging pile allocation target for each new energy vehicle; when the order origin and destination are located in a high-impact area or cannot be fulfilled due to a new energy vehicle malfunction, an order reassignment branch is generated. The intervention operation set includes releasing unfinished orders from the original matching relationship, updating the order pool, calling the new energy vehicle order matching algorithm, including but not limited to the assignment model that considers time windows, and generating a new order new energy vehicle matching relationship; S302: In a digital twin environment, perform parallel simulation on each branch of the scenario tree to deduce the evolution of the network state within a future time window from the current moment. The specific simulation mechanism is as follows: Within each simulation step, the topology and node attributes of the digital twin network are updated according to the set of intervention operations of the branch, and the node infection probability is updated by calling the permeation model based on the updated network state. At the same time, logistics operation indicators are calculated according to the new new energy vehicle routes, charging plans, and order matching. The logistics operation indicators include at least the location of new energy vehicles, remaining power, and order completion status. The above process is repeated until the simulation ends. After the simulation, the target evaluation index vector for each branch is recorded. The target evaluation index vector includes total delay time, total energy consumption, and anomaly recovery time. The total delay time is the sum of the differences between the actual completion time and the planned completion time of all orders. The total energy consumption represents the total amount of electricity consumed by all new energy vehicles during the simulation. The anomaly recovery time represents the time required from the current moment until the thermal value of all grids is lower than the preset recovery threshold. S303: Construct a set of reference points, which consist of historical best target values, business expected targets, or ideal points. Each reference point contains a corresponding optimal evaluation index vector. For each branch, analyze the fitness with all reference points. Take the minimum value as the optimal fitness of the corresponding branch. Example 3: In this example: the optimal fitness F of the b-th branch b The specific calculation method is as follows: ; Where a represents the reference point number, A represents the set of reference points; D represents the total delay time, E represents the total energy consumption, and T represents the total energy consumption. rec Indicates the abnormal recovery time; Q （m，b） This represents the value of branch b on target m; r （m，b） This represents the value of reference point a on target m; The branch with the lowest fitness is selected as the optimal branch, and its corresponding set of intervention operations is the optimal scheduling strategy. The optimal strategy is then transformed into a specific set of scheduling instructions and issued to the actual logistics execution system.

[0024] S4: Track execution results in real time, build an experience case library, and match similar historical scenarios through analogical reasoning to achieve adaptive optimization of anomaly response strategies.

[0025] S4 includes the following: S401: During execution, continuously collect real-time operational data of the logistics network, including but not limited to: real-time location, speed, remaining battery power, and fault status of new energy vehicles; working status and output power of charging piles; actual order completion status; and newly occurring abnormal events, including but not limited to secondary faults, traffic accidents, and road congestion; at the same time, record the performance indicators of the actual execution effect: actual total delay time and actual total energy consumption; compare with the inferred prediction value of the corresponding branch, and calculate the execution deviation vector; S402: Combine the current abnormal event feature vector, the selected intervention strategy, the execution deviation vector, and the final recovery flag into a case and store it in the experience case library; The final recovery flag is a binary variable. If the anomaly is resolved within the preset time window and no new chain of anomalies is triggered, the final recovery flag is 1; otherwise, it is 0. S403: When a new abnormal event occurs, extract the corresponding abnormal event feature vector and calculate the Euclidean distance between the new abnormal event feature vector and the abnormal event feature vector of each historical case in the knowledge base; if the minimum distance is less than the preset threshold and the final recovery flag of the corresponding case is 1, then directly retrieve the intervention strategy of the corresponding case as a candidate solution; otherwise, construct a scenario tree to generate a new strategy.

[0026] Please see Figure 2 The present invention provides a technical solution: a new energy logistics network intelligent management system based on digital twins, which includes an anomaly adjustment module, an anomaly impact propagation analysis module, a scheduling strategy analysis module, and a strategy optimization module; The anomaly adjustment module senses abnormal events in the logistics network in real time, maps the events to the digital twin through semantic parsing, and adjusts the topology and operating parameters of the digital twin network. The abnormal impact propagation analysis module is based on digital twin networks. It constructs a seepage model for the propagation of abnormal event impacts, analyzes the infection probability and seepage centrality of network nodes, and generates an abnormal impact heat map. The scheduling strategy analysis module constructs a scenario tree based on the heatmap of anomaly impacts and the current scheduling plan. In the digital twin environment, it performs parallel simulations on each branch of the scenario tree and selects the optimal scheduling strategy through target evaluation. The strategy optimization module tracks the execution effect in real time, builds an experience case library, and matches similar historical scenarios through analogy reasoning to achieve adaptive optimization of anomaly response strategies.

[0027] The anomaly adjustment module includes a data acquisition and mapping unit and a digital twin network update unit; The data acquisition and mapping unit is used to collect multi-source data in the logistics network in real time. Through a predefined event semantic parsing template, abnormal events are mapped to the attributes of the corresponding entities in the digital twin. The digital twin network update unit adjusts the topology and operating parameters of the digital twin network based on the extracted abnormal event types, locations, and impact parameters. The reconstructed digital twin network is represented as a graph structure, which includes a set of entity nodes, a set of edges that change over time, and a node attribute matrix.

[0028] The abnormal impact propagation analysis module includes a parameter definition unit, an infection analysis unit, and a heatmap generation unit; The parameter definition unit is used to define the initial infection probability of each node on the digital twin network; define the probability of an infected node spreading anomalies to neighboring susceptible nodes; define the recovery rate of neighboring susceptible nodes as the reciprocal of the time required for anomaly recovery; the evolution of the infection probability and recovery probability of neighboring susceptible nodes is described by the percolation model. The infection analysis unit is used to solve the seepage model and obtain the infection probability of each adjacent susceptible node at time t; extract nodes whose infection probability exceeds the preset infection probability threshold and delineate the abnormal influence area; analyze the seepage centrality of the nodes; mark the nodes whose seepage centrality exceeds the seepage centrality threshold as key nodes, and output the list of key nodes and the boundary information of the influence area. The heatmap generation unit is used to divide the monitored area into uniform geographical grids. The heat value of each grid is equal to the average infection probability of all nodes within the grid, generating an anomaly impact heatmap.

[0029] The scheduling strategy analysis module includes a scenario tree construction unit, a simulation evolution unit, and a scheduling strategy selection unit; The scenario tree construction unit is used to construct a scenario tree with the heat map of abnormal impacts and the current scheduling plan as input. The current scheduling plan includes the new energy vehicle route, charging task allocation, and order matching relationship of new energy vehicles. The root node of the scenario tree represents the network state at the current moment. Starting from the root node, multiple branches are generated based on the predefined intervention rule base. Each branch represents a scheduling intervention measure, and each branch corresponds to a set of intervention operations. The simulation evolution unit is used to perform parallel simulations of each branch of the scenario tree in a digital twin environment, and to deduce the evolution of the network state within a future time window from the current moment; after the deduction is completed, the target evaluation index vector of each branch is recorded. The scheduling strategy filtering unit is used to construct a set of reference points. The reference points consist of historical best target values, business expected targets, or ideal points. Each reference point contains a corresponding optimal evaluation index vector. For each branch, the fitness with all reference points is analyzed. The minimum value is taken as the optimal fitness of the corresponding branch. The branch with the lowest fitness is selected as the optimal branch, and its corresponding set of intervention operations is the optimal scheduling strategy. The optimal strategy is then transformed into a specific set of scheduling instructions and issued to the actual logistics execution system.

[0030] The strategy optimization module includes an execution deviation analysis unit and a strategy optimization unit; The execution deviation analysis unit is used to continuously collect real-time operational data of the logistics network during execution, including but not limited to: real-time location, speed, remaining battery power, and fault status of new energy vehicles; working status and output power of charging piles; actual order completion status; and newly occurring abnormal events, including but not limited to secondary faults, traffic accidents, and road congestion. At the same time, it records the performance indicators of the actual execution effect: actual total delay time and actual total energy consumption; compares them with the extrapolated prediction values ​​of the corresponding branches, and calculates the execution deviation vector. The strategy optimization unit combines the current abnormal event feature vector, the selected intervention strategy, the execution deviation vector, and the final recovery flag into a case and stores it in the experience case library. When a new abnormal event occurs, the corresponding abnormal event feature vector is extracted, and the Euclidean distance between the new abnormal event feature vector and the abnormal event feature vector of each historical case in the knowledge base is calculated. If the minimum distance is less than a preset threshold and the final recovery flag of the corresponding case is 1, the intervention strategy of the corresponding case is directly retrieved as a candidate solution; otherwise, a scenario tree is constructed to generate a new strategy.

[0031] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0032] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A new energy logistics network intelligent management method based on digital twins, characterized in that, The corresponding method includes the following steps: S1: Real-time perception of abnormal events in the logistics network, mapping abnormal events to digital twins through semantic parsing, and adjusting the topology and operating parameters of the digital twin network; S2: Based on digital twin networks, construct a seepage model for the propagation of the impact of abnormal events, analyze the infection probability and seepage centrality of network nodes, and generate a heat map of the impact of abnormal events. S3: Based on the heatmap of the impact of anomalies and the current scheduling plan, construct a scenario tree, perform parallel simulation on each branch of the scenario tree in the digital twin environment, and select the optimal scheduling strategy through target evaluation; S4: Track execution results in real time, build an experience case library, and match similar historical scenarios through analogical reasoning to achieve adaptive optimization of anomaly response strategies.

2. The intelligent management method for new energy logistics networks based on digital twins according to claim 1, characterized in that: S1 includes the following: S101: Real-time collection of multi-source data in the logistics network, including abnormal event data and regular operation data; the abnormal event data includes at least new energy vehicle fault data, charging pile disconnection alarm flag, order cancellation flag, and traffic accident data; the regular operation data includes at least new energy vehicle location, remaining battery power, charging pile power, and order origin and destination coordinates. By using predefined event semantic parsing templates, abnormal event data is mapped to the attributes of the corresponding entities in the digital twin; S102: Adjust the topology and operating parameters of the digital twin network according to the extracted abnormal event types, locations, and impact parameters; the reconstructed digital twin network is represented as a graph structure, including a set of entity nodes, a set of edges that change over time, and a node attribute matrix. The node attribute matrix includes regular operating data and abnormal event feature vectors. The abnormal event feature vectors at least include the event type code, intensity level, location of the anomaly, radius of influence, and start time fields.

3. The intelligent management method for new energy logistics networks based on digital twins according to claim 2, characterized in that: The method of mapping abnormal events to the corresponding entities in the digital twin includes: assigning fault type attributes to the virtual new energy vehicle model for new energy vehicle fault data and recording the start time of the abnormality; mapping traffic accident data to the corresponding road segments in the road network model and generating temporary obstacle attributes; setting the status field of charging piles with a disconnection alarm flag of 1 to unavailable; and setting the status field of orders with an order cancellation flag of 1 to cancelled.

4. The intelligent management method for new energy logistics networks based on digital twins according to claim 3, characterized in that: The adjustments to the topology and operating parameters of the digital twin network include: for charging pile disconnection alarms, temporarily removing the corresponding charging pile node from the charging service network and updating the charging network adjacency relationship; for traffic accident data, locating the affected road segment based on the accident coordinates, downgrading the traffic capacity of the corresponding road segment, and updating the road network edge weights; for new energy vehicle fault data, marking the fault status on the new energy vehicle model according to the fault type and impact radius, and adding temporary nodes to indicate fault-occupied lanes; for order cancellations, updating the order pool status and removing the corresponding order's transportation demand from the scheduling task list.

5. The intelligent management method for new energy logistics networks based on digital twins according to claim 2, characterized in that: S2 includes the following: S201: Define the initial infection probability of each node on the digital twin network; Define the probability of an infected node propagating an anomaly to neighboring susceptible nodes; define the recovery rate of neighboring susceptible nodes as the reciprocal of the time required for anomaly recovery; the evolution of the infection probability and recovery probability of neighboring susceptible nodes is described by the seepage model. S202: Solve the seepage model to obtain the infection probability of each adjacent susceptible node at time t; extract nodes whose infection probability exceeds the preset infection probability threshold and delineate the abnormal influence area; analyze the seepage centrality of the nodes; Nodes with seepage centrality exceeding the seepage centrality threshold are marked as critical nodes, and a list of critical nodes and the boundary information of the affected area are output. S203: Divide the monitored area into a uniform geographical grid, where the heat value of each grid is equal to the average infection probability of all nodes within the grid, and generate an anomaly impact heat map.

6. The intelligent management method for new energy logistics networks based on digital twins according to claim 5, characterized in that: S3 includes the following: S301: Using the heat map of abnormal impacts and the current scheduling plan as input, construct a scenario tree; the current scheduling plan includes new energy vehicle routes, charging task allocation, and order new energy vehicle matching relationships; the root node of the scenario tree represents the network state at the current moment, and multiple branches are generated from the root node based on a predefined intervention rule base, each branch represents a scheduling intervention measure, and each branch corresponds to a set of intervention operations; S302: In a digital twin environment, perform parallel simulation on each branch of the scenario tree to deduce the evolution of the network state within a future time window from the current moment. After the simulation is completed, the target evaluation index vector for each branch is recorded. The target evaluation index vector includes total delay time, total energy consumption, and anomaly recovery time. The total delay time is the sum of the differences between the actual completion time and the planned completion time of all orders; the total energy consumption represents the total electricity consumed by all new energy vehicles during the simulation; the anomaly recovery time represents the time required from the current moment until the thermal values ​​of all grids are lower than the preset recovery threshold. S303: Construct a set of reference points, which consist of historical best target values, business expected targets, or ideal points. Each reference point contains a corresponding optimal evaluation index vector. For each branch, analyze the fitness with all reference points. Take the minimum value as the optimal fitness of the corresponding branch. The branch with the lowest fitness is selected as the optimal branch, and its corresponding set of intervention operations is the optimal scheduling strategy. The optimal strategy is then transformed into a specific set of scheduling instructions and issued to the actual logistics execution system.

7. The intelligent management method for new energy logistics networks based on digital twins according to claim 6, characterized in that: The intervention rule base includes at least the following: when the grid heat value is greater than the grid heat value threshold, the corresponding geographical grid is marked as a high-impact area, and a detour branch is generated. The intervention operation set includes replanning routes for affected new energy vehicles; when charging piles are unavailable due to abnormalities or are located in high-impact areas, a charging load transfer branch is generated. The intervention operation set includes obtaining a list of affected new energy vehicles, using bipartite graph matching or linear programming to solve the adjustment scheme of the charging allocation matrix based on the charging pile status and the remaining power of the new energy vehicles, and updating the charging pile allocation target for each new energy vehicle; when the order origin and destination are located in high-impact areas or cannot be fulfilled due to new energy vehicle malfunctions, an order reassignment branch is generated. The intervention operation set includes releasing unfinished orders from the original matching relationship, updating the order pool, calling the new energy vehicle order matching algorithm, including but not limited to an assignment model that considers time windows, and generating a new order new energy vehicle matching relationship.

8. The intelligent management method for new energy logistics networks based on digital twins according to claim 7, characterized in that: The simulation mechanism is as follows: within each simulation step, the topology and node attributes of the digital twin network are updated according to the set of intervention operations of the branch, and the node infection probability is updated by calling the permeation model based on the updated network state. At the same time, logistics operation indicators are calculated according to the new new energy vehicle routes, charging plans, and order matching. The logistics operation indicators include at least the location of new energy vehicles, remaining power, and order completion status. Repeat the above process until the simulation ends.

9. The intelligent management method for new energy logistics networks based on digital twins according to claim 7, characterized in that: S4 includes the following: S401: During execution, continuously collect real-time operational data of the logistics network, including but not limited to: real-time location, speed, remaining battery power, and fault status of new energy vehicles; and the working status and output power of charging piles. The actual completion status of orders; and newly occurring abnormal events, including but not limited to secondary failures, traffic accidents, and traffic congestion; Record the actual performance metrics: actual total delay time and actual total energy consumption; compare with the predicted values ​​of the corresponding branches, and calculate the execution deviation vector; S402: Combine the current abnormal event feature vector, the selected intervention strategy, the execution deviation vector, and the final recovery flag into a case and store it in the experience case library; The final recovery flag is a binary variable. If the anomaly is resolved within a preset time window and no new chain of anomalies is triggered, the final recovery flag is 1; otherwise, it is 0. S403: When a new abnormal event occurs, extract the corresponding abnormal event feature vector and calculate the Euclidean distance between the new abnormal event feature vector and the abnormal event feature vector of each historical case in the knowledge base; if the minimum distance is less than the preset threshold and the final recovery flag of the corresponding case is 1, then directly retrieve the intervention strategy of the corresponding case as a candidate solution. Otherwise, construct a scenario tree to generate a new strategy.

10. A new energy logistics network intelligent management system based on digital twins, wherein the system is applied to the new energy logistics network intelligent management method based on digital twins as described in any one of claims 1-9, characterized in that, The system includes an anomaly adjustment module, an anomaly impact propagation analysis module, a scheduling strategy analysis module, and a strategy optimization module; The anomaly adjustment module senses abnormal events in the logistics network in real time, maps the events to the digital twin through semantic parsing, and adjusts the topology and operating parameters of the digital twin network. The abnormal impact propagation analysis module is based on a digital twin network, constructs a seepage model for the propagation of abnormal event impacts, analyzes the infection probability and seepage centrality of network nodes, and generates an abnormal impact heatmap. The scheduling strategy analysis module constructs a scenario tree based on the heatmap of anomaly impacts and the current scheduling plan. In the digital twin environment, it performs parallel simulations on each branch of the scenario tree and selects the optimal scheduling strategy through target evaluation. The strategy optimization module tracks the execution effect in real time, builds an experience case library, and matches similar historical scenarios through analogical reasoning to achieve adaptive optimization of the anomaly response strategy.

Images