A cloud-edge collaboration-based charging facility resource allocation method and system

Abstract

The application relates to the technical field of cloud-edge collaborative computing, in particular to a charging facility resource deployment method and system based on cloud-edge collaboration; the method comprises the following steps: time alignment and fusion of multi-source operation data are carried out on the edge side of a charging station; service tension reflecting real-time pressure of the station is constructed through a load evolution rate, power constraints and queuing density; a user charging request is converted into a packet elastic scheduling unit, so that the demand has the characteristics of migration and extension; service tension and elastic units uploaded by each edge node are converged on the cloud side; a cross-site tension difference collaborative potential field is established; global optimization is carried out with the overall tension balance as the target; structured instructions containing priority, migration suggestion and time correction are generated; charging resources are abstracted into time service slots on the edge side; queues are dynamically reconstructed and local migration is executed according to the instructions; meanwhile, execution deviation is collected and fed back to the cloud side to realize closed-loop optimization. The application realizes efficient self-adaptive deployment of charging facility resources through cloud-edge collaboration.

Description

Technical Field

[0001] This invention relates to the field of cloud-edge collaborative computing technology, specifically to a method and system for allocating charging facility resources based on cloud-edge collaboration. Background Technology

[0002] As the penetration rate of electric vehicles continues to increase and the scale of charging infrastructure networks continues to expand, charging demand exhibits significant imbalances in both time and space. During peak hours within urban areas, charging resources are concentrated, while the utilization rate of charging stations is relatively low in other times or areas. This dynamic difference places higher demands on the unified coordination capabilities of charging resources. Electric vehicle users consider multiple factors during charging, such as waiting time, driving distance, and station service status, resulting in dynamic adjustment and migration characteristics in their decision-making. Internal operating parameters of the charging system, such as charging power, power capacity, and queue length, are interconnected, causing the overall system state to exhibit complex nonlinear characteristics. The cloud-edge collaborative architecture, by organically combining global computing in the cloud with local response on the edge, provides a technical foundation for collaborative resource management across multiple stations.

[0003] Chinese invention patent application CN118418822A discloses a cloud-edge collaborative intelligent charging station orderly charging optimization scheduling method. The method includes: on the cloud side, determining the orderly charging scheduling curve of each charging station based on the power information in the area and the operation data of the charging piles corresponding to multiple charging stations in the area; the orderly charging scheduling curve includes the charging and discharging power of the charging station changing over time; on the edge side, determining the charging and discharging scheduling instructions for the charging piles in the charging station based on the charging and discharging data of the current moment in the charging station and the orderly charging scheduling curve fed back from the cloud side, the charging and discharging scheduling instructions are used to control the charging piles to charge and discharge the charging object so that the charging and discharging power of the charging station at each moment meets the orderly charging scheduling curve.

[0004] Under this architecture, effectively expressing the real-time pressure of stations, transforming user demand into flexible variables that can participate in scheduling, and achieving cross-station resource balancing have become the key to improving the operational efficiency of the charging service network. Therefore, researching a cloud-edge collaborative method for allocating charging facility resources, forming a complete technical link from pressure perception and demand elasticity to global balanced scheduling, has significant engineering application value. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a method and system for allocating charging facility resources based on cloud-edge collaboration.

[0006] The technical solution of this invention: a method for allocating charging facility resources based on cloud-edge collaboration, comprising the following specific implementation steps: S1. Construct a multi-source status acquisition and unified time alignment mechanism at the edge of the charging station, structure and integrate charging pile operation data, power reserve and queue information, introduce load evolution rate and power constraint coupling analysis, characterize pressure change characteristics through nonlinear disturbance amplification mechanism, and integrate queue density to form service tension output; S2. Transform charging requests into elastic scheduling units. By constructing a user charging intention trigger domain, a time-space coupled elastic interval, and a behavior drift probability model, the dynamic structural transformation of user needs is achieved. S3. On the cloud side, the elastic scheduling unit and service tension are integrated and processed. By constructing a tension transmission relationship across sites, the collaborative potential field formed by the tension difference is used to guide the demand to be adaptively distributed among different charging stations. Global optimization is carried out with tension balance as the goal, and structured scheduling instructions containing priority, migration suggestions and time correction are generated. S4. Execute cloud scheduling results at the edge, abstract charging resources into service slots with time elasticity, realize dynamic reconstruction and continuous adjustment of charging queues, flexibly adjust user allocation based on slot matching and local migration mechanism, perceive load and queue changes in real time and quickly repair local disturbances, and collect execution deviations to feed back to the cloud to achieve closed-loop optimization.

[0007] Preferably, step S1 further includes: At the edge, the power, occupancy status, charging time, queue length and power reserve of the charging pile are collected in real time. Data from different sampling frequencies are aligned and reconstructed through a unified time window to eliminate the time deviation of heterogeneous data and form a standardized state vector. Based on the unified state data, an overall load function for the site is constructed, and a sliding time window is used to calculate the load evolution rate to characterize the growth or decline trend of charging load in a short period of time, providing trend input for stress modeling. Based on the load evolution rate, a dual-factor coupling structure of power resource approach ratio and change amplitude is introduced to construct a service disturbance enhancement factor, which is used to reflect the nonlinear pressure amplification effect of charging stations when they are close to the capacity limit and under conditions of severe load fluctuation. By adjusting the weight of different factors through the sensitivity coefficient, the early enhanced identification of critical congestion state can be achieved. By fusing the disturbance enhancement factor with the queue density, a final service tension index is constructed, which simultaneously reflects the equipment load pressure and the user waiting aggregation effect. The standardized state vector and dynamic tension value are then uploaded to the cloud to provide a unified input for cross-site resource scheduling and realize the structured compressed expression of edge-side pressure information.

[0008] Preferably, the standardized state vector includes the actual output charging power of the charging pile, the load occupancy status, the duration of the charging session, the local queue length corresponding to the charging pile (i.e., the number of vehicles waiting for the charging pile to serve), and the instantaneous remaining power capacity at the charging station level. The load evolution rate is calculated by dividing the overall load change of the site within the sliding time window by the length of the time window, and is used to reflect the concentration and rate of change of charging demand. The service disturbance enhancement factor is jointly adjusted by the power constraint sensitivity coefficient and the load change sensitivity coefficient. The power constraint sensitivity coefficient is set according to the power grid security level or historical stability analysis and is used to adjust the intensity of the impact on system pressure when the power is close to the limit. The load change sensitivity coefficient is obtained through historical load fluctuation statistics or learning and is used to adjust the sensitivity of the system to dynamic fluctuations. The final service tension index introduces a queue sensitivity enhancement coefficient to characterize the amplification effect of user queuing and aggregation on system pressure. It is obtained by coupling the disturbance enhancement factor with the current total number of queued vehicles.

[0009] Preferably, step S2 further includes: Based on the service tension generated at the edge, a user charging intention trigger domain is constructed. The trigger strength of user requests entering the system is adjusted through the tension function, so that whether a user enters the scheduling system changes dynamically with the system load, thereby realizing pressure diversion and load pre-control at the demand entry stage. Based on the trigger domain, a flexible interval combining time and space is constructed. The time dimension dynamically expands with system tension to absorb queuing pressure, while the spatial dimension contracts through trigger intensity to limit long-distance migration, forming a dynamic coupling structure of time extension and spatial convergence, enabling user needs to adapt to changes in system load. By combining system load evolution rate and user waiting time, a behavior drift probability model is introduced to dynamically characterize user behavior such as rerouting, abandoning, or continuing to wait during the waiting process. This transforms user demand from static execution to a probabilistic evolution process, thereby improving the scheduling model's adaptability to real-world behavioral fluctuations. By structurally integrating time elasticity intervals, spatial candidate sets, and user behavior stability weights, a unified elastic scheduling unit expression is generated, enabling each user request to be portable and schedulable, and providing standardized input for cross-site resource matching and stress balancing in the cloud.

[0010] Preferably, the user charging intention trigger domain is controlled by a tension sensitivity coefficient to determine the steepness of the impact of service tension changes on the trigger intensity, and is set according to the standard tension threshold based on the historical stable operating range of the charging station. The time elasticity interval is defined by the earliest start time, the latest acceptable time, and the delay compensation function driven by service tension. The time expansion function outputs the time compensation amount for the system's allowable delay under the current service tension. The spatial elasticity range is the set of candidate charging stations acceptable to the user. It is determined by comparing the distance from the user's current location to the candidate station with the user's maximum acceptable driving distance. When the system tension increases, the trigger strength decreases, causing the acceptable distance range to shrink. That is, the user is guided to prioritize choosing a closer or more vacant station. Under low load conditions, a greater range of spatial selection freedom is allowed. The behavior drift probability model uses the congestion sensitivity coefficient and patience decay coefficient to characterize the weight of the impact of system load changes on user behavior and the weight of the impact of waiting time on user behavior, respectively. After the input is compressed to the 0 to 1 range by a nonlinear mapping function, the output is the probability of the user changing the original charging plan. User stability weights are obtained by transforming behavior drift probabilities, with values ​​ranging from 0 to 1, and are used to characterize the reliability of user needs during the scheduling process.

[0011] Preferably, step S3 further includes: Each user elastic scheduling unit is expanded, and its time interval and spatial candidate set are mapped to specific sites and time slices to form a three-dimensional candidate relationship set of user-site-time. In combination with the service tension threshold, sites under high pressure are filtered, thereby compressing the computing scale and eliminating infeasible allocation paths. Based on the service tension of each site, a tension difference relationship is constructed to form a cross-site collaborative potential field. By introducing user stability weights and tension suppression and enhancement mechanisms, the attraction strength of users migrating to different sites is calculated, forming a scheduling trend of natural diversion from high tension to low tension areas. A global optimization model is constructed under the guidance of the tension potential field. The overall tension equilibrium of the system is taken as the objective function. The incremental impact of user allocation on the tension of the site is introduced to constrain the solution. The coordinated allocation between multiple users and multiple sites is achieved by combining capacity, power and time-space constraints. The optimization results are transformed into structured scheduling instructions, including user-assigned priority sequences to determine the service order within the target site, cross-site migration suggestions to provide alternative sites and corresponding time windows, and time fine-tuning to make fine-grained adjustments to the reservation time within the user's elastic range. These instructions are then sent to edge nodes, enabling the edge side to dynamically execute and locally optimize based on actual changes during real-time operation.

[0012] Preferably, the tension difference relationship is obtained by calculating the service tension difference between any two different stations at the same time, so as to form a resource flow trend from high tension station to low tension station; The attraction intensity is controlled by the site tension suppression coefficient, which controls the rate at which the attraction of high-tension sites to users decays. The attraction of low-tension sites to demand is amplified by the tension difference diffusion coefficient. It is also calculated in conjunction with the user stability weight, so that the system can automatically guide users to migrate to low-tension sites with negative tension differences. The global optimization model takes minimizing the unevenness of the tension distribution at each site as the objective function and solves the user allocation for each site jointly. The incremental impact of each user being assigned to a site on the tension of that site is calculated by the user demand parameters and the site load model. In structured scheduling instructions, user allocation priority sequence is used to determine the service order within the target site, cross-site migration suggestion provides alternative sites and corresponding time windows for users who originally planned to enter high-tension sites, and time fine-tuning is used to make fine-grained adjustments to the reservation time within the user's elastic range.

[0013] Preferably, step S4 further includes: The charging piles in the station are logically reconstructed, and their future service capabilities are divided into several service slots according to the time dimension. The schedulability of each slot is evaluated based on the current occupancy status and user behavior stability, thereby transforming rigid physical resources into time-elastic scheduling units. Based on the user priority and time adjustment instructions issued by the cloud, local queued users are sorted in layers and matched and allocated according to the schedulability of each slot. This allows highly stable users to obtain continuous and stable service intervals first, while users with greater flexibility are arranged in slots with greater adjustment space, thereby ensuring service experience while reserving room for future adjustments. During operation, load and queue changes are continuously monitored. When a disturbance is detected that exceeds the threshold, the user order, cross-slot transfer, or time fine-tuning is adjusted locally through the slot migration mechanism to repair the scheduling structure at the lowest cost and avoid overall congestion caused by local anomalies. The system continuously collects data on actual execution results, including waiting time deviations, changes in user behavior, and slot utilization. It then generates a unified deviation index and feeds it back to the cloud to correct the service tension model and scheduling parameters. This allows the system to continuously iterate and gradually optimize the scheduling strategy, achieving adaptive closed-loop control for cloud-edge collaboration.

[0014] Preferably, the schedulability of service slots is determined by the ratio of the remaining adjustable time of the slot to the total duration of the slot, combined with the user occupancy stability coefficient. The remaining adjustable time of the slot refers to the length of time that the current slot is not occupied and can be adjusted or redistributed. The user occupancy stability coefficient ranges from 0 to 1. Slot matching and allocation is achieved by maximizing the matching degree between users and slots. The matching degree is affected by the user's time adjustment amount and the time offset sensitivity coefficient. The time offset sensitivity coefficient controls the intensity of the impact of time adjustment on the matching degree. The larger the value, the more the system tends to reduce time offset. The smaller the value, the more flexible the time adjustment is allowed. The disturbance intensity is obtained by multiplying the load change deviation and the queue change by the load disturbance weight and the queue disturbance weight, respectively, and then summing them. The load change deviation refers to the difference between the current load and the expected load, and the queue change refers to the change in the number of people in the queue relative to the predicted value. The overall execution deviation is composed of the deviation between the actual waiting time and the planned waiting time multiplied by the time deviation weight and the user behavior deviation multiplied by the behavior deviation weight. It is used to measure the difference between the actual system operation results and the scheduling expectations, and serves as an important basis for cloud model correction to adjust service tension calculation parameters, user behavior models, and scheduling weight settings.

[0015] The technical solution of this invention: A cloud-edge collaborative charging facility resource allocation system, used to execute the aforementioned cloud-edge collaborative charging facility resource allocation method, comprising: The edge-side service tension generation module is deployed at the edge nodes of each charging station. It is used to collect and process multi-source data on the charging pile operation status, queuing structure and power carrying capacity in a time-consistent manner. Based on the load evolution trend and power constraint relationship, it constructs a service tension expression to form a structured output that can reflect the real-time pressure status of the station. The charging intention elastic modeling module is used to receive user charging requests, reconstruct user needs under service tension constraints, transform the original rigid needs into elastic scheduling units that include time elastic intervals, spatial optional sets and behavioral stability weights, and combine user behavior drift characteristics to realize the dynamic expression of needs. The cloud-based tension balancing scheduling module is deployed on the central cloud platform. It is used to aggregate service tension information uploaded by various edge nodes and user elastic scheduling units. By building a cross-site tension difference collaborative relationship, it realizes the transmission and balanced allocation of resource pressure among multiple sites, and outputs structured scheduling instructions including user priority sequence, cross-site migration suggestions and time adjustment information. The edge dynamic execution and feedback module is deployed at the edge nodes of each charging station. It is used to map cloud scheduling instructions to the local service slot system. By abstracting charging resources into time slices and reconstructing the queue structure, it realizes dynamic matching between users and slots. During operation, it performs local migration and real-time repair based on load changes and queue disturbances. At the same time, it continuously collects execution deviation and user behavior change information and feeds it back to the cloud, forming a closed-loop adjustment mechanism for cloud-edge collaboration.

[0016] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: This invention designs a cloud-edge collaborative method and system for allocating charging facility resources. By constructing a service tension expression at the edge, it integrates multi-dimensional information such as equipment load, power capacity, and queuing status within charging stations into a unified pressure index. This allows the cloud to perceive the real-time and accurate operational status of each station, avoiding the data lag and information distortion problems of traditional centralized scheduling, and providing a reliable foundation for cross-station resource balancing. It transforms user charging requests into dynamic scheduling units with time elasticity, spatial mobility, and behavioral drift probability, breaking the strict constraints of traditional rigid demand on scheduling. This enables the system to proactively guide users to postpone charging, reroute to idle stations, or adjust service order when load fluctuates, thereby achieving pressure distribution at the user end and effectively alleviating station congestion during peak hours. The cloud, based on cross-station tension differentials... By establishing a collaborative potential field and performing global equilibrium optimization, demand can be automatically shifted from high-tension areas to low-tension areas, achieving adaptive load distribution among multiple sites and significantly improving the overall resource utilization efficiency of charging facilities while avoiding service degradation caused by local overload. At the edge, a service slot abstraction and dynamic queue reconstruction mechanism is adopted to refine cloud scheduling instructions into flexibly adjustable execution schemes. During operation, local disturbances are quickly repaired and slots are migrated, greatly enhancing the system's robustness and adaptability in complex dynamic environments. Through continuous collection of execution deviations and closed-loop correction of the cloud model, the system can continuously approximate real-world operating patterns, achieving self-evolution and self-optimization of scheduling strategies. This ensures long-term, efficient, stable, and fair resource allocation capabilities, comprehensively improving the operational efficiency and user experience of the charging service network. Attached Figure Description

[0017] Figure 1 This is a flowchart of a cloud-edge collaborative charging facility resource allocation method proposed in this invention. Figure 2 This is a system architecture diagram of a cloud-edge collaborative charging facility resource allocation system proposed in this invention. Detailed Implementation

[0018] Example 1, as Figure 1 As shown, the present invention proposes a method for allocating charging facility resources based on cloud-edge collaboration, which includes the following specific implementation steps: S1. By constructing a multi-source status acquisition and unified time alignment mechanism at the edge of the charging station, the charging pile operation data, power reserve and queue information are structurally integrated. On this basis, load evolution rate and power constraint coupling analysis are introduced. Furthermore, the pressure change characteristics of the system when it approaches the resource limit are characterized by a nonlinear disturbance amplification mechanism. Finally, the queue density is integrated to form a service tension output, realizing the transformation from raw operation data to a unified pressure expression that can be used for cloud scheduling. S2. Using the service tension and load evolution information generated in step S1 as input, the traditional fixed charging request is transformed into an elastic scheduling unit that can be dynamically adjusted according to the system's operating status. By constructing a user charging intention trigger domain, a time-space coupled elastic interval, and a behavior drift probability model, the user demand is transformed from a static and deterministic input to a dynamic structure that is transferable, extensible, and reorderable. This provides a unified expression with flexible constraints for cloud-based cross-site resource balancing, enabling the scheduling system to maintain stability and adaptability under high fluctuating loads. S3. On the cloud side, the elastic scheduling unit formed in step S2 and the service tension in step S1 are integrated. By constructing a cross-site tension transmission relationship, the originally discrete user demand is mapped into scheduling resources that can flow between multiple sites. On this basis, the cooperative potential field formed by the tension difference guides the demand to be adaptively distributed among different charging stations, and the overall system load balance is achieved through a global optimization mechanism aimed at tension balancing. Finally, a structured scheduling instruction containing priority, migration suggestions and time correction is generated, providing a basis for dynamic execution and real-time adjustment on the edge side. S4. Execute cloud scheduling results at the edge. By abstracting charging resources into time-elastic service slots, dynamic reconstruction and continuous adjustment of the charging queue are achieved. During execution, user allocation is flexibly adjusted based on slot matching and local migration mechanisms. Load and queue changes are perceived in real time during operation, and local disturbances are quickly repaired. At the same time, by continuously collecting execution deviations and feeding them back to the cloud, closed-loop optimization of the scheduling model is achieved, so that the system gradually approaches the real operating state in multiple rounds of operation.

[0019] In an optional embodiment, step S1 is completed at the edge of each charging station, that is, deployed within the edge computing nodes of each charging station. It transforms the discrete operating state into a service tension expression that can be used for scheduling decisions, providing a comprehensive characterization of equipment constraints, power boundaries, queue backlog, and dynamic trends. The specific implementation process is as follows: S11. Real-time collection of multi-source data such as charging pile power, occupancy status, charging duration, queue length, and remaining power is performed at the edge. Data from different sampling frequencies are aligned and reconstructed through a unified time window to form a standardized state vector at the same time granularity, thereby eliminating the time deviation of heterogeneous data. Specifically: On the edge side, each charging station is equipped with a local status acquisition unit to monitor the status of all charging piles within the station at the millisecond level. The acquired information includes not only electrical parameters such as charging power, current, and voltage, but also vehicle access / departure events, charging session duration, current queue length, and instantaneous load capacity of the station's power distribution unit. Because different devices have different sampling periods—for example, power data is sampled at high frequencies while queuing status is updated by events—direct use would cause timing misalignment. Therefore, a unified time alignment mechanism is introduced to map all data to a fixed time window. The data within each time slice is reconstructed to ensure that the data in each time slice has consistent semantics. Based on this, the state of each charging pile at time t is organized into a unified state vector: ; In the formula, Let represent the unified state vector of the i-th charging pile at time t; This represents the actual output charging power of the i-th charging pile at time t, reflecting the current energy transmission intensity of the device; This represents the load occupancy status of the i-th charging pile, used to characterize whether the charging pile is in service or its occupancy ratio. It is expressed as: 0 for idle state, 1 for fully occupied state, and [0,1] for continuous occupancy ratio (depending on equipment capacity). This indicates the duration of the current charging session for the i-th charging station; This represents the length of the local queue corresponding to the i-th charging pile, which is the number of vehicles waiting for the service of that charging pile. This indicates the instantaneous remaining power capacity at the charging station level, used to reflect the current available power supply redundancy on the grid side; S12. Based on the unified state data, a site-wide load function is constructed, and a sliding time window is used to calculate the load evolution rate to characterize the growth or decline trend of charging load in a short period of time. This transforms static load information into dynamic evolution characteristics, reflecting the concentration and rate of change of charging demand, and providing trend input for stress modeling. Specifically: After achieving state unification, a sliding time window mechanism is introduced to continuously track changes in the overall load of the site in order to depict its dynamic evolution trend. Define the overall site load function: ; Further introduce load evolution rate: ; In the formula, The total load of the charging station at time t is the weighted sum of the power of all charging piles currently in use; N represents the total number of charging piles in a single charging station. This represents the load evolution rate of the charging station, used to characterize the rate of load change. This represents the length of the time sliding window, i.e., the time interval between adjacent state samples; S13. Based on the load evolution rate, a dual-factor coupling structure of power resource approximation ratio and change amplitude is introduced to construct a service disturbance enhancement factor. This factor reflects the nonlinear pressure amplification effect of charging stations approaching their capacity limits and experiencing severe load fluctuations. By adjusting the weights of different factors through sensitivity coefficients, early enhanced identification of critical congestion states is achieved. Specifically: After obtaining the load change trend, it is necessary to further characterize the nonlinear stress amplification effect when the system approaches the resource limit. For this purpose, a service disturbance enhancement factor is introduced to reflect the sensitivity of the system when it transitions from normal operation to congestion. The definition is as follows: ; In the formula, This represents the service disturbance enhancement factor, used to describe the degree of nonlinear amplification of system stress under critical conditions; This represents the power constraint sensitivity coefficient, which is used to adjust the intensity of the impact on system pressure when the power is close to its limit. It is set according to the power grid security level or historical stability analysis. It represents the instantaneous remaining power carrying capacity of a site and is used to measure power supply redundancy; This represents the load change sensitivity coefficient, used to adjust the system's sensitivity to dynamic fluctuations. It is obtained through historical load fluctuation statistics or learning. S14. The disturbance enhancement factor and queue density are fused and calculated to construct the final service tension index, which simultaneously reflects the equipment load pressure and the user waiting aggregation effect. A standardized state vector and dynamic tension value are then uploaded to the cloud to provide a unified input for cross-site resource scheduling, achieving a structured compressed expression of edge-side pressure information. Specifically: Based on the disturbance enhancement factor, queueing structure information is further integrated to construct the final service tension, which is used to uniformly characterize the overall pressure state of charging stations. Define service tension by coupling device-side pressure with user-side aggregation effects: ; In the formula, Indicates the final service tension value; This represents the queue sensitivity enhancement coefficient, used to characterize the amplification effect of user queuing and aggregation on system pressure; This represents the total number of vehicles currently queuing at station t at time t.

[0020] In an optional embodiment, step S2 is the service tension output in step 1. Load evolution rate State vector Based on this, user charging requests are transformed into a flexible behavioral range structure that can be dynamically adjusted according to system pressure, making user behavior a flexible variable that can participate in scheduling games. The specific implementation process is as follows: S21. Based on the service tension generated at the edge, a user charging intention trigger domain is constructed. The trigger strength of user requests entering the system is adjusted through a tension function, so that whether a user enters the scheduling system no longer depends on fixed rules, but changes dynamically with the system load. This achieves pressure diversion and load pre-control at the demand entry stage. Specifically: When a user initiates a charging request, the edge node first reads the service tension generated in step S1. And use it as the core variable for regulating the scheduling space that users can enter; And according to The initial serviceable range of users is dynamically shrunk or expanded to form a charging intention trigger domain; this trigger domain is essentially an reachability space that changes with system pressure, and is a service boundary defined in reverse by the system's acceptability. Define the trigger strength function for the user at the current moment: ; when When T(t) is in a relaxed state, users are more likely to be connected immediately; when T(t) rises above the threshold, the system automatically reduces the trigger strength to suppress new entry demands. In the formula, This indicates the user trigger strength, representing the feasibility of a user being connected to the scheduling system under the current system state. This represents the tension sensitivity coefficient, which controls the steepness of the impact of changes in service tension on trigger intensity. It is obtained through historical congestion data or empirical calibration. This indicates the standard tension threshold, which is set based on the historical stable operating range of the charging station. S22. Based on the trigger domain, a flexible interval combining time and space is constructed. The time dimension expands with system tension to absorb queuing pressure, while the spatial dimension contracts through trigger intensity to limit long-distance migration. This forms a dynamically coupled structure with temporal extension and spatial convergence, enabling user needs to adapt to changes in system load. Specifically: After the trigger domain is determined, the user's charging needs are further structured and broken down into two coupled flexible intervals: time dimension and space dimension. Elasticity Range 1 (Time Elasticity Range): Dynamically expanded based on system pressure. ; Elasticity Range Two (Spatial Elasticity Range): By triggering strength Apply reverse constraints: ; When the system tension increases A decrease in load leads to a contraction in the acceptable distance range, meaning users are guided to prioritize closer or less busy sites; conversely, under low load conditions, the system allows for a greater range of spatial choices. In the formula, This indicates the time flexibility range, which is the range of charging time that the user can accept. This indicates the earliest start time, the earliest time a user can begin charging. This indicates the latest acceptable time, the latest time the user expects to start charging. This represents the time spread function, which is the amount of time compensation for the system's allowed delay under the current service load, and is a functional relationship driven by service load. This represents the spatially flexible set, i.e., the set of candidate charging stations acceptable to the user. This represents the j-th candidate station, which is the charging station that the user can choose. This represents the distance from the user to the station, i.e., the distance from the user's current location to the candidate station; u represents the user identifier, i.e., the user corresponding to the current charging request; This indicates the maximum acceptable distance, which is the farthest driving distance that the user can accept. S23. Combining system load evolution rate and user waiting time, a behavior drift probability model is introduced to dynamically characterize user behaviors such as rerouting, abandoning, or continuing to wait during the waiting process. This transforms user demand from static execution to a probabilistic evolution process, thereby improving the scheduling model's adaptability to real-world behavioral fluctuations. Specifically: Based on the elastic range, we should further consider an important point that cannot be ignored in actual operation: user behavior will not strictly follow the initial plan, but will deviate with the waiting time and changes in system state. A behavior drift mechanism is introduced to characterize the actual changes in user decisions during the waiting process, and the behavior drift probability function is defined as follows: ; It should be noted that the above mechanism reflects two real behavioral patterns: when the system load increases rapidly, users will feel the pressure of queuing and are more inclined to give up or switch to another platform; when the waiting time increases, even if the system load remains unchanged, users' patience will gradually decrease. Furthermore, behavioral drift is broken down into three actual paths: waiting (behavioral stability); rerouting to other stations (spatial migration); and canceling charging (demand withdrawal). In the formula, This represents the probability of behavioral drift, which is the probability that a user will change their original charging plan in the current state. The value ranges from 0 to 1. This indicates the user's waiting time, which is the cumulative waiting time from when the user initiated the request to the current moment. This represents the congestion sensitivity coefficient, which is the weight of the impact of system load changes on user behavior, and is fitted from historical data. This represents the patience decay coefficient, which is the weight of the impact of waiting time on user behavior, based on user behavior statistics. This represents a nonlinear mapping function that compresses the input to the range of 0 to 1. S24. The time elasticity interval, spatial candidate set, and user behavior stability weights are structurally integrated to generate a unified elastic scheduling unit expression, ensuring that each user request is portable and schedulable. This provides standardized input for cross-site resource matching and stress balancing in the cloud, specifically: After completing the time-space elasticity interval and behavior drift modeling, user requirements are finally structured and compressed to form elastic scheduling units that can be directly computed in the cloud; user stability weights are defined: ; The final output user scheduling expression is as follows: ; In the formula, This represents the user stability weight, which is the reliability of user demand during the scheduling process, and its value ranges from 0 to 1. This represents the user scheduling unit, which is the structured expression of user requirements ultimately used for cloud scheduling.

[0021] In an optional embodiment, step S3 is executed in the cloud scheduling center, and its input comes from two types of core information formed in the first two steps: one is the service load, load evolution rate, and operating status structure uploaded by each charging station edge node; the other is the reconstructed user scheduling unit on the user side. Furthermore, a cross-site tension transmission and release mechanism is constructed, and resource allocation is completed under the constraints of this mechanism to maintain overall balance during the operation of the entire system. The specific implementation process is as follows: S31. Expand each user elastic scheduling unit, map its time interval and spatial candidate set to specific sites and time slices, forming a three-dimensional candidate relationship set of user-site-time. Combine this with a service stress threshold to filter sites under high pressure, thereby compressing the computational scale and eliminating infeasible allocation paths. Specifically: The cloud first analyzes each user scheduling unit, dividing its time elasticity range. With spatial candidate set Mapped onto a unified scheduling timeline and site set, a three-dimensional candidate space of user-site-time slice is formed, and a layer-by-layer filtering process is introduced during the unfolding process; Filtering is performed based on the time dimension, retaining only time slices that meet the user's time constraints; and in the spatial dimension, candidate sites that deviate significantly from the reasonable path are eliminated based on the actual travel distance between the user's current location and the site; and then, combined with the site service tension in step S1, sites that are continuously under high pressure and have no trend of release in the short term are filtered a second time. Its formal expression relationship is as follows: ; In the formula, Indicates whether user u can be assigned to a site at time t. The feasibility assessment result takes a value of 0 or 1, where 1 means all constraints are met. This indicates an indicator function that takes the value 1 if the condition is true and 0 otherwise. Indicates site The service tension value at time t; This indicates the tension screening threshold, which is set based on system experience and historical operating data. After the above step-by-step filtering, the originally huge combination space is compressed into a sparse but effective candidate set; S32. Based on the service tension of each site, a tension difference relationship is constructed to form a cross-site collaborative potential field. By introducing user stability weights and tension suppression and enhancement mechanisms, the attraction strength of user migration to different sites is calculated, forming a scheduling trend of natural diversion from high-tension to low-tension areas, specifically: After obtaining the candidate space, the tension relationship between each site is analyzed, and a collaborative potential field describing the trend of resource flow is constructed, which is a virtual guidance mechanism: when some sites are in a high-tension state, the system will naturally guide some demand to transfer to low-tension areas, thereby achieving pressure distribution on a global scale. Calculate the tension difference between any two stations: ; Based on this, an attraction strength is constructed for each candidate assignment relationship: ; In the formula, This represents the set of possible sites for user u, excluding the current target site j itself. This represents the tension difference between station j and station k at time t; This represents the attraction strength of user u being assigned to station j at time t; This represents the site tension suppression coefficient, used to control the rate at which the attractiveness of high-tension sites to users diminishes. It is calibrated based on system operation experience or historical scheduling data. It represents the tension difference diffusion coefficient, which is used to amplify the ability of low-tension sites to attract demand, and is adjusted according to the regional site distribution density and load differences; S33. Under the guidance of the tension potential field, a global optimization model is constructed, with the overall tension equilibrium of the system as the objective function. Constraints are applied by introducing the incremental impact of user allocation on site tension, and capacity, power, and time-space constraints are combined to achieve coordinated allocation between multiple users and multiple sites. Specifically: Guided by the potential field, a global optimization model is further constructed to achieve overall tension equilibrium. In this embodiment, the equilibrium of tension distribution at each station is directly taken as the optimization objective, and the objective function can be expressed as: ; In the formula, S represents the set of all charging stations (the complete set of stations within the scope of cloud-based scheduling and management). This represents the set of users who are assigned or candidate to be assigned to site j at time t; Indicates whether user u was assigned to station j at time t; This represents the incremental impact of user u being assigned to site j on the site's load capacity, calculated from user demand parameters and the site load model. S34. The optimization results are transformed into structured scheduling instructions, including user priority sequences, cross-site migration suggestions, and time adjustment amounts, and issued to edge nodes in the form of instructions. This enables the edge side to dynamically execute and locally optimize according to actual changes during real-time operation, thereby achieving the feasibility and continuous adaptability of the scheduling results. Specifically: After completing the global allocation, the cloud does not directly generate a fixed queuing result, but instead outputs a set of scheduling instructions with adjustable space and distributes them to each edge node for execution; this instruction system mainly includes three categories: User-assigned priority sequences are used to determine the order of services within the target site; Cross-site migration recommendations: For users who originally planned to migrate to the high-tension site, alternative sites and corresponding time windows are provided. Time fine-tuning is used to make fine-grained adjustments to the appointment time within the user's flexible range.

[0022] In an optional embodiment, step S4 is executed at the edge nodes of each charging station. Its input comes from the structured scheduling instructions output from the cloud in step S3, including user allocation priority sequences, cross-station migration suggestions, and time fine-tuning amounts. It emphasizes reserving sufficient adjustment space on the edge side so that the scheduling results can be continuously corrected according to subtle changes during actual operation. Through four stages—service slot abstraction, local queue reconstruction, operational disturbance response, and execution deviation feedback—the edge nodes not only assume the execution role but also have a certain degree of adaptive adjustment capability. The specific implementation process is as follows: S41. Logically refactor the charging piles within the station, dividing their future service capabilities into multiple service slots based on a time dimension. Assess the schedulability of each slot based on its current occupancy status and user behavior stability, thereby transforming the originally rigid physical resources into time-flexible scheduling units. Specifically: After receiving the cloud scheduling instruction, the edge node performs a logical reconstruction of the charging resources in the station, dividing the service capabilities that each charging pile can provide in the future into several continuous segments according to the time dimension, and abstracting these segments into service slots. It should be noted that a service slot emphasizes the capability unit that can carry a charging task within a certain time interval. When the actual operation changes, the time slices can be recombined without having to rearrange the entire queue. Furthermore, a schedulability metric is introduced for each slot to assess whether a slot is suitable for dynamic adjustment, such as whether a new user can be inserted or time compression can be performed. ; In the formula, The slot schedulable coefficient represents the adjustability of the i-th charging pile at time t in the k-th time slot. It is used to measure whether the slot is suitable for dynamic rearrangement or insertion of new charging tasks. This indicates the remaining adjustable time for a slot, which is the length of time that the slot is not currently occupied and can be adjusted or redistributed. This indicates the total duration of the slot, which is the time span corresponding to the initial division of the slot. The stability coefficient of user occupancy represents the stability of the behavior of the user currently occupying the i-th charging pile. Its value ranges from [0,1]. The closer it is to 1, the more stable the user's behavior is (not easy to end or change the plan in advance). The closer it is to 0, the higher the uncertainty. When a charging slot is in its initial stage and user behavior is relatively stable, its available space for adjustment is large; while when charging is nearing completion or user behavior is stable, the slot tends to remain unchanged. S42. Based on the user priority and time adjustment instructions issued by the cloud, the local queued users are sorted in layers and matched according to the schedulability of each slot. This ensures that highly stable users receive continuous and stable service intervals first, while users with greater flexibility are assigned to slots with greater adjustment space. This ensures service experience while reserving room for future adjustments. Specifically: After completing the slot abstraction, the edge nodes perform a non-destructive reconstruction of the local queuing structure based on the user priority and time adjustment information sent from the cloud. In practice, users are stratified according to their stability weights, and within each stratum, users are mapped to appropriate slot ranges based on their time elasticity intervals. Specifically, user needs are mapped to the corresponding slot set and sorted according to their time elasticity intervals and stability weights. Users with higher priority and stronger stability are given priority in being assigned continuous and stable slots, while users with greater elasticity are given priority in being assigned to slot ranges with higher adjustability. To ensure reasonable allocation, a slot matching function is introduced: ; In the formula, This represents the matching degree between the user and the charging slot, that is, the degree of matching between user u and the kth charging slot of the i-th charging pile; This indicates the amount of time adjustment for the user, that is, the extent of the adjustment to the user's original charging time. The time offset sensitivity coefficient controls the strength of the impact of time adjustment on the matching degree: the larger the value, the more the system tends to reduce time offset; the smaller the value, the more flexible the time adjustment is allowed. By maximizing the matching function, users and slots are allocated locally to the optimal value, thereby achieving dynamic reconstruction of the queue structure. S43. During operation, continuously monitor load and queue changes. When a disturbance exceeds a threshold, adjust user order, perform cross-slot transfer, or fine-tune time within a local area through a slot migration mechanism to repair the scheduling structure at minimal cost and avoid overall congestion caused by local anomalies. Specifically: In actual operation, charging behavior is often highly uncertain. For example, users may end charging early, cancel their reservations temporarily, or suddenly have new emergency charging needs. If these changes are not handled properly, they can easily disrupt the existing scheduling structure and cause local congestion. To address the aforementioned issues, edge nodes introduce disturbance sensing and slot migration mechanisms to locally correct abnormal changes during operation. This is achieved by continuously monitoring site load and queue changes to calculate disturbance intensity. ; In the formula, It indicates the intensity of system disturbance, that is, the degree to which the system's operating state deviates from the normal level at the current moment; This represents the load change deviation, i.e., the difference between the current load and the expected load, which comes from the load model in step S1. This represents the change in the queue size, i.e., the change in the current number of people in the queue relative to the predicted value, and is used to reflect fluctuations in user-side demand. This indicates the load disturbance weight, used to measure the degree of impact of load changes on the overall disturbance; This represents the queue disturbance weight, used to measure the contribution of queue changes to system disturbances; When the disturbance intensity exceeds a set threshold, a slot migration operation is performed within a local area. The migration methods include, but are not limited to: Adjust the order of users within the same charging station, and appropriately postpone the users with more flexible time. Lateral migration between different charging stations will redirect some demand to charging stations with lower loads. Slightly compress or extend the service time for some users to rebalance the overall queuing structure; S44. Continuously collect actual execution results, obtain information such as waiting time deviation, changes in user behavior, and slot utilization, and generate a unified deviation index to feed back to the cloud. This index is used to correct the service tension model and scheduling parameters, enabling the system to gradually optimize the scheduling strategy through continuous iteration, and achieve adaptive closed-loop control of cloud-edge collaboration. Specifically: After completing the above dynamic execution, the edge nodes continuously monitor the execution effect and feed back key deviation information to the cloud for the next round of scheduling optimization; Feedback should include, but is not limited to: the difference between actual waiting time and planned time; deviations in user behavior, such as leaving early or not arriving as scheduled; and the difference between slot utilization and expected utilization. To provide a unified description of execution deviations, a comprehensive deviation index is introduced: ; In the formula, This represents the overall execution deviation, used to measure the difference between the actual operating results of the system and the scheduling expectations; This indicates the actual waiting time, which is the real waiting time from when the user enters the system to when charging begins; This indicates the planned waiting time, which is the waiting time predicted during cloud scheduling. This indicates the weight of the time deviation, used to measure the degree of influence of the waiting time error on the overall deviation; This represents the behavior offset weight, used to measure the degree of impact of changes in user behavior on system stability; This deviation metric will serve as an important basis for cloud model correction, used to adjust service tension calculation parameters, user behavior models, and scheduling weight settings, so that the system gradually approaches the real scenario after multiple rounds of operation.

[0023] Example 2, as Figure 2 As shown, the present invention proposes a cloud-edge collaborative charging facility resource allocation system, which is used to execute a cloud-edge collaborative charging facility resource allocation method proposed in Embodiment 1. It includes: an edge-side service tension generation module, a charging intention elastic modeling module, a cloud-side tension balancing scheduling module, and an edge dynamic execution and feedback module.

[0024] The edge-side service tension generation module is deployed at the edge nodes of each charging station. It is used to collect and process multi-source data on the charging pile operation status, queuing structure and power carrying capacity in a time-consistent manner. Based on the load evolution trend and power constraint relationship, it constructs a service tension expression, thereby forming a structured output that can reflect the real-time pressure status of the station. The charging intention elastic modeling module is used to receive user charging requests, reconstruct user needs under service tension constraints, transform the original rigid needs into scheduling units that include time elastic intervals, spatial optional sets and behavioral stability weights, and combine user behavior drift characteristics to realize the dynamic expression of needs, making user-side input adjustable. The cloud-based tension balancing scheduling module is deployed on the central cloud platform. It is used to aggregate service tension information uploaded by various edge nodes and user elastic scheduling units. By building a cross-site tension difference collaborative relationship, it realizes the transmission and balanced allocation of resource pressure among multiple sites, and outputs structured scheduling instructions including user priority sequence, cross-site migration suggestions and time adjustment information, thereby completing the global resource optimization configuration. The edge dynamic execution and feedback module is deployed at the edge nodes of each charging station. It maps cloud scheduling instructions to the local service slot system. By abstracting charging resources into time slices and reconstructing the queue structure, it realizes dynamic matching between users and slots. During operation, it performs local migration and real-time repair based on load changes and queue disturbances. At the same time, it continuously collects execution deviation and user behavior change information and feeds it back to the cloud, thus forming a closed-loop adjustment mechanism of cloud-edge collaboration, enabling the system to maintain stable and efficient operation in complex dynamic environments.

[0025] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for allocating charging facility resources based on cloud-edge collaboration, characterized in that, The specific implementation steps include the following: S1. Construct a multi-source status acquisition and unified time alignment mechanism at the edge of the charging station, structure and integrate charging pile operation data, power reserve and queue information, introduce load evolution rate and power constraint coupling analysis, characterize pressure change characteristics through nonlinear disturbance amplification mechanism, and integrate queue density to form service tension output; S2. Transform charging requests into elastic scheduling units. By constructing a user charging intention trigger domain, a time-space coupled elastic interval, and a behavior drift probability model, the dynamic structural transformation of user needs is achieved. S3. On the cloud side, the elastic scheduling unit and service tension are integrated and processed. By constructing a tension transmission relationship across sites, the collaborative potential field formed by the tension difference is used to guide the demand to be adaptively distributed among different charging stations. Global optimization is carried out with tension balance as the goal, and structured scheduling instructions containing priority, migration suggestions and time correction are generated. S4. Execute cloud scheduling results at the edge, abstract charging resources into service slots with time elasticity, realize dynamic reconstruction and continuous adjustment of charging queues, flexibly adjust user allocation based on slot matching and local migration mechanism, perceive load and queue changes in real time and quickly repair local disturbances, and collect execution deviations to feed back to the cloud to achieve closed-loop optimization.

2. The method for allocating charging facility resources based on cloud-edge collaboration according to claim 1, characterized in that, Step S1 further includes: At the edge, the power, occupancy status, charging time, queue length and power reserve of the charging pile are collected in real time. Data from different sampling frequencies are aligned and reconstructed through a unified time window to eliminate the time deviation of heterogeneous data and form a standardized state vector. The overall load function of the site is constructed based on the standardized state vector, and the load evolution rate is calculated using a sliding time window to characterize the growth or decline trend of the charging load in a short period of time, providing trend input for stress modeling. Based on the load evolution rate, a dual-factor coupling structure of power resource approach ratio and change amplitude is introduced to construct a service disturbance enhancement factor, which is used to reflect the nonlinear pressure amplification effect of charging stations when they are close to the capacity limit and under severe load fluctuations. By adjusting the weight of different factors through the sensitivity coefficient, the early enhancement identification of critical congestion state is completed. By fusing service disturbance enhancement factors with queue density, a final service tension index is constructed, which simultaneously reflects equipment load pressure and user waiting aggregation effect. The standardized state vector and dynamic tension value are then uploaded to the cloud to provide a unified input for cross-site resource scheduling and realize the structured compressed expression of edge-side pressure information.

3. The method for allocating charging facility resources based on cloud-edge collaboration according to claim 2, characterized in that, The standardized state vector includes the actual output charging power of the charging pile, the load occupancy status, the duration of the charging session, the local queue length corresponding to the charging pile (i.e., the number of vehicles waiting for the charging pile to serve), and the instantaneous remaining power capacity at the charging station level. The load evolution rate is calculated by dividing the overall load change of the site within the sliding time window by the length of the time window, and is used to reflect the concentration and rate of change of charging demand. The service disturbance enhancement factor is jointly adjusted by the power constraint sensitivity coefficient and the load change sensitivity coefficient. The power constraint sensitivity coefficient is set according to the power grid security level or historical stability analysis and is used to adjust the intensity of the impact on system pressure when the power is close to the limit. The load change sensitivity coefficient is obtained through historical load fluctuation statistics or learning and is used to adjust the sensitivity of the system to dynamic fluctuations. The final service tension index introduces a queue sensitivity enhancement coefficient to characterize the amplification effect of user queuing and aggregation on system pressure. It is obtained by coupling the disturbance enhancement factor with the current total number of queued vehicles.

4. The method for allocating charging facility resources based on cloud-edge collaboration according to claim 2, characterized in that, Step S2 further includes: Based on the service tension generated at the edge, a user charging intention trigger domain is constructed. The trigger strength of user requests entering the system is adjusted through the tension function, so that whether a user enters the scheduling system changes dynamically with the system load, thereby realizing pressure diversion and load pre-control at the demand entry stage. Based on the user charging intention trigger domain, a flexible interval combining time and space is constructed. The time dimension dynamically expands with system tension to absorb queuing pressure, while the spatial dimension contracts through trigger intensity to limit long-distance migration, forming a dynamic coupling structure of time extension and spatial convergence, enabling user demand to adapt to changes in system load. By combining system load evolution rate and user waiting time, a behavior drift probability model is introduced to dynamically characterize user behavior such as rerouting, abandoning, or continuing to wait during the waiting process. This transforms user demand from static execution to a probabilistic evolution process, thereby improving the scheduling model's adaptability to real-world behavioral fluctuations. By structurally integrating time elasticity intervals, spatial candidate sets, and user behavior stability weights, a unified elastic scheduling unit expression is generated, enabling each user request to be portable and schedulable, and providing standardized input for cross-site resource matching and stress balancing in the cloud.

5. A method for allocating charging facility resources based on cloud-edge collaboration according to claim 4, characterized in that, The user charging intention trigger domain controls the steepness of the impact of service tension changes on the trigger intensity through a tension sensitivity coefficient, and is set according to the standard tension threshold based on the historical stable operating range of the charging station. The time elasticity interval is defined by the earliest start time, the latest acceptable time, and the delay compensation function driven by service tension. The time expansion function outputs the time compensation amount for the system's allowable delay under the current service tension. The spatial elasticity range is the set of candidate charging stations acceptable to the user. It is determined by comparing the distance from the user's current location to the candidate station with the user's maximum acceptable driving distance. When the system tension increases, the trigger strength decreases, causing the acceptable distance range to shrink. That is, the user is guided to prioritize choosing a closer or more vacant station. Under low load conditions, a greater range of spatial selection freedom is allowed. The behavior drift probability model uses the congestion sensitivity coefficient and patience decay coefficient to characterize the weight of the impact of system load changes on user behavior and the weight of the impact of waiting time on user behavior, respectively. After the input is compressed to the 0 to 1 range by a nonlinear mapping function, the output is the probability of the user changing the original charging plan. User stability weights are obtained by transforming behavior drift probabilities, with values ​​ranging from 0 to 1, and are used to characterize the reliability of user needs during the scheduling process.

6. A method for allocating charging facility resources based on cloud-edge collaboration according to claim 5, characterized in that, Step S3 further includes: Each user elastic scheduling unit is expanded, and its time interval and spatial candidate set are mapped to specific sites and time slices to form a three-dimensional candidate relationship set of user-site-time. In combination with the service tension threshold, sites under high pressure are filtered, thereby compressing the computing scale and eliminating infeasible allocation paths. Based on the service tension of each site, a tension difference relationship is constructed to form a cross-site collaborative potential field. By introducing user stability weights and tension suppression and enhancement mechanisms, the attraction strength of users migrating to different sites is calculated, forming a scheduling trend of natural diversion from high tension to low tension areas. A global optimization model is constructed under the guidance of the tension potential field. The overall tension equilibrium of the system is taken as the objective function. The incremental impact of user allocation on the tension of the site is introduced to constrain the solution. The coordinated allocation between multiple users and multiple sites is achieved by combining capacity, power and time-space constraints. The optimization results are transformed into structured scheduling instructions, including user-assigned priority sequences to determine the service order within the target site, cross-site migration suggestions to provide alternative sites and corresponding time windows, and time fine-tuning to make fine-grained adjustments to the reservation time within the user's elastic range. These instructions are then sent to edge nodes, enabling the edge side to dynamically execute and locally optimize based on actual changes during real-time operation.

7. A method for allocating charging facility resources based on cloud-edge collaboration according to claim 6, characterized in that, The tension difference relationship is obtained by calculating the service tension difference between any two different stations at the same time, which is used to form the resource flow trend from high tension stations to low tension stations; The attraction intensity is controlled by the site tension suppression coefficient, which controls the rate at which the attraction of high-tension sites to users decays. The attraction of low-tension sites to demand is amplified by the tension difference diffusion coefficient. It is also calculated in conjunction with the user stability weight, so that the system can automatically guide users to migrate to low-tension sites with negative tension differences. The global optimization model takes minimizing the unevenness of the tension distribution at each site as the objective function and solves the user allocation for each site jointly. The incremental impact of each user being assigned to a site on the tension of that site is calculated by the user demand parameters and the site load model. In structured scheduling instructions, user allocation priority sequence is used to determine the service order within the target site, cross-site migration suggestion provides alternative sites and corresponding time windows for users who originally planned to enter high-tension sites, and time fine-tuning is used to make fine-grained adjustments to the reservation time within the user's elastic range.

8. A method for allocating charging facility resources based on cloud-edge collaboration according to claim 7, characterized in that, Step S4 further includes: The charging piles in the station are logically reconstructed, and their future service capabilities are divided into several service slots according to the time dimension. The schedulability of each slot is evaluated based on the current occupancy status and user behavior stability, thereby transforming rigid physical resources into time-elastic scheduling units. Based on the user priority and time adjustment instructions issued by the cloud, local queued users are sorted in layers and matched and allocated according to the schedulability of each slot. This allows highly stable users to obtain continuous and stable service intervals first, while users with greater flexibility are arranged in slots with greater adjustment space, thereby ensuring service experience while reserving room for future adjustments. During operation, load and queue changes are continuously monitored. When a disturbance is detected that exceeds the threshold, the user order, cross-slot transfer, or time fine-tuning is adjusted locally through the slot migration mechanism to repair the scheduling structure at the lowest cost and avoid overall congestion caused by local anomalies. The system continuously collects data on actual execution results, including waiting time deviations, changes in user behavior, and slot utilization. It then generates a unified deviation index and feeds it back to the cloud to correct the service tension model and scheduling parameters. This allows the system to continuously iterate and gradually optimize the scheduling strategy, achieving adaptive closed-loop control for cloud-edge collaboration.

9. A method for allocating charging facility resources based on cloud-edge collaboration according to claim 8, characterized in that, The schedulability of a service slot is determined by the ratio of the remaining adjustable time of the slot to the total duration of the slot, combined with the user occupancy stability coefficient. The remaining adjustable time of the slot refers to the length of time that the current slot is not occupied and can be adjusted or reallocated. The user occupancy stability coefficient ranges from 0 to 1. Slot matching and allocation is achieved by maximizing the matching degree between users and slots. The matching degree is affected by the user's time adjustment amount and the time offset sensitivity coefficient. The time offset sensitivity coefficient controls the intensity of the impact of time adjustment on the matching degree. The larger the value, the more the system tends to reduce time offset. The smaller the value, the more flexible the time adjustment is allowed. The disturbance intensity is obtained by multiplying the load change deviation and the queue change by the load disturbance weight and the queue disturbance weight, respectively, and then summing them. The load change deviation refers to the difference between the current load and the expected load, and the queue change refers to the change in the number of people in the queue relative to the predicted value. The overall execution deviation is composed of the deviation between the actual waiting time and the planned waiting time multiplied by the time deviation weight and the user behavior deviation multiplied by the behavior deviation weight. It is used to measure the difference between the actual system operation results and the scheduling expectations, and serves as an important basis for cloud model correction to adjust service tension calculation parameters, user behavior models, and scheduling weight settings.

10. A cloud-edge collaborative charging facility resource allocation system, used to execute the cloud-edge collaborative charging facility resource allocation method according to any one of claims 1 to 9, characterized in that, include: The edge-side service tension generation module is deployed at the edge nodes of each charging station. It is used to collect and process multi-source data on the charging pile operation status, queuing structure and power carrying capacity in a time-consistent manner. Based on the load evolution trend and power constraint relationship, it constructs a service tension expression to form a structured output that can reflect the real-time pressure status of the station. The charging intention elastic modeling module is used to receive user charging requests, reconstruct user needs under service tension constraints, transform the original rigid needs into elastic scheduling units that include time elastic intervals, spatial optional sets and behavioral stability weights, and combine user behavior drift characteristics to realize the dynamic expression of needs. The cloud-based tension balancing scheduling module is deployed on the central cloud platform. It is used to aggregate service tension information uploaded by various edge nodes and user elastic scheduling units. By building a cross-site tension difference collaborative relationship, it realizes the transmission and balanced allocation of resource pressure among multiple sites, and outputs structured scheduling instructions including user priority sequence, cross-site migration suggestions and time adjustment information. The edge dynamic execution and feedback module is deployed at the edge nodes of each charging station. It is used to map cloud scheduling instructions to the local service slot system. By abstracting charging resources into time slices and reconstructing the queue structure, it realizes dynamic matching between users and slots. During operation, it performs local migration and real-time repair based on load changes and queue disturbances. At the same time, it continuously collects execution deviation and user behavior change information and feeds it back to the cloud, forming a closed-loop adjustment mechanism for cloud-edge collaboration.

Images