A Method for Power Allocation and Voltage Risk Suppression in Charging Stations Based on Master-Slave Game Theory

Abstract

This invention relates to a power allocation and voltage risk suppression method for charging stations based on master-slave game theory, belonging to the field of charging control technology. It includes: S1, collecting multi-source operational data, constructing a state dataset, and calculating the station-level schedulable power budget considering capacity, thermal stability, and voltage safety boundaries; S2, quantifying the session urgency index and constructing a sub-Bruker set based on Wasserstein distance; S3, constructing a master-slave game model, with the upper-level coordinator generating policy variables and the lower-level charging piles using the MM-ADMM algorithm and the water-filling method to solve for power allocation; S4, performing robust optimization enhancement based on the sub-Bruker set and conditional risk value; S5, performing grid security constraint verification and multi-objective collaborative correction; S6, issuing instructions and forming a closed-loop adaptive scheduling based on execution feedback. This invention uses voltage safety as the core game objective, co-optimizing it with quality of service, improving charging service and distribution network voltage safety under uncertain environments, and achieving soft capacity expansion of charging station power.

Description

Technical Field

[0001] This invention belongs to the field of charging control technology and relates to a method for power allocation and voltage risk suppression in charging stations based on master-slave game theory. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the number of electric vehicles on the road continues to grow rapidly, and users are increasingly reliant on charging infrastructure. Especially in urban core areas, highway entrances and exits, and large transportation hubs, high-power DC fast charging stations have gradually become an important way to recharge electric vehicles. However, limited by factors such as power distribution capacity, site conditions, and construction time, the construction and expansion speed of charging stations cannot keep pace with the growth in the number of electric vehicles, leading to an increasingly prominent supply-demand imbalance.

[0003] Under current technological conditions, improving the service capacity of charging stations mainly relies on hardware methods such as increasing the number of charging piles, increasing transformer capacity, or expanding and upgrading power distribution lines. However, these hardware expansion solutions generally suffer from high investment costs, long construction periods, and difficulties in construction coordination. Especially in existing urban areas or areas with dense power loads, further expansion is often limited by spatial conditions and grid access capacity, making rapid implementation difficult. Therefore, simply relying on hardware expansion is no longer sufficient to meet the actual operational needs in the context of large-scale electric vehicle access.

[0004] To alleviate the conflict between charging demand and power supply capacity, some existing technologies propose coordinating the management of multiple charging piles within a station through charging control and power scheduling. For example, Chinese patent application CN118644026A discloses a "vehicle-station bilateral benefit game-theoretic charging scheduling method considering distribution network security." This method constructs a master-slave game model between charging stations and electric vehicle users, where the charging station, as the leader, adjusts electricity prices, and electric vehicle users, as followers, choose charging stations to achieve a bilateral benefit equilibrium. However, this scheme mainly focuses on the game theory at the "vehicle-station" selection guidance level. Its voltage risk handling is achieved by imposing punitive costs on overload operation and voltage deviation, which is a post-event penalty mechanism and is difficult to preventively control voltage risks in uncertain environments. Furthermore, this scheme does not address the continuous power allocation problem among multiple charging piles within a station, nor does it model the uncertainty of user behavior.

[0005] On the other hand, as fast charging power levels continue to increase, the instantaneous load fluctuations at charging stations become significantly stronger, easily leading to operational risks such as voltage exceeding limits in the distribution network and line overload. Some existing research, such as Chinese patent application CN121036014A, employs demand response signal-based scheduling optimization, allocating charging power through centralized optimization, fixed-rule scheduling, or single-layer control strategies. However, these methods often require global information, have high computational complexity, and are prone to scheduling failures or amplified operational risks when facing uncertain demand or extreme charging scenarios, making stable application in practical engineering difficult.

[0006] Furthermore, existing technologies often focus on optimizing single performance indicators, such as minimizing charging completion time or maximizing user satisfaction, with less emphasis on comprehensively balancing charging station operation from the perspective of synergistic improvement of grid security and service quality. Under complex operating conditions such as high load and short dwell time, there remains a lack of effective technical solutions for achieving "soft expansion" of charging stations through intelligent control without adding new hardware, thereby improving the user charging experience while ensuring the safe operation of the distribution network.

[0007] Therefore, there is an urgent need for a method that can perform hierarchical coordination and risk perception control of multiple charging piles in a charging station, taking into account the uncertainty of charging demand and the constraints of power grid operation, so as to achieve a soft improvement in the service capacity of the charging station and improve its safety and stability in complex operating scenarios. Summary of the Invention

[0008] In view of this, in order to address the bottleneck of charging station service capacity without adding or with minimal addition of hardware facilities, this invention proposes a charging pile power allocation and voltage risk suppression method based on master-slave game theory. The dynamic power scheduling and risk assessment mechanism enables the power grid and charging stations to jointly participate in scheduling decisions, thereby improving grid stability and charging station operating efficiency while ensuring users' charging needs are met.

[0009] To achieve the above objectives, this invention provides a method for power allocation and voltage risk suppression in charging stations based on master-slave game theory, comprising the following steps: S1. In response to the scheduling cycle trigger, collect multi-source operation data of station-network-pile-vehicle, construct a status dataset and calculate the station-level schedulable power budget, which simultaneously considers capacity constraints, thermal stability constraints and voltage safety boundary constraints.

[0010] The voltage safety boundary constraints include: Obtain the lower limit of the allowable voltage at critical nodes and the current minimum phase voltage Calculate voltage margin ; Based on voltage sensitivity coefficient Calculate the available power for voltage margin: ;in, To prevent small constants from being divided by zero; voltage risk. ,but ; Calculate available power under capacity constraints ;in, To the maximum available active power capacity, This refers to the active power of the station's basic load. Calculate the available power for thermal stability constraints ;in, This is the equivalent power factor conversion factor. The maximum allowable load for the station transformer. ; The minimum of the available power with voltage margin, available power with capacity constraint, and available power with thermal stability constraint is used to obtain the upper limit of the station-level charging budget. ; Introducing safety margin factor The station-level schedulable power budget is obtained. .

[0011] S2. Based on the state dataset, construct a demand assessment and robust representation model, quantify the urgency index of the session, and construct a sub-Bruker set describing uncertainty based on Wasserstein distance.

[0012] Constructing Uncertain Random Vectors The empirical sample set, in which, For the deviation of the remaining stay time, This represents the deviation from the maximum available power.

[0013] Wherein, the sub-Bruker set is defined as based on the empirical distribution The distribution uncertainty set centered at the Wasserstein distance: ;in, For Wasserstein distance, Let be the robust radius.

[0014] The quantitative indicators of session urgency include: For any charging session within the scheduling set , Calculate the remaining energy demand With remaining stay time ;in, Energy is required to achieve the goal. The energy that has already been charged, For the estimated departure time, The scheduling period; Calculate the demand intensity index ;in, It indicates the intensity of energy replenishment required per unit time; the larger the value, the more urgent the need. Map the demand intensity index to demand weights and then smooth them out:

[0015]

[0016]

[0017] in,( , ) represents the upper and lower limits of the weight. For smoothing coefficients; The final weight vector is obtained as follows: .

[0018] S3. Construct a master-slave game scheduling model: The station-level / coordinator layer generates Leader policy variables and drives the Follower response of the charging pile layer to solve the problem, outputting the station-level power command vector that satisfies the station-level power and voltage safety boundaries.

[0019] The Leader objective function is constructed as follows: ; in, For service quality weighting coefficient, This is a statistic representing the overall service quality within the station. This is the voltage safety penalty factor. It is a voltage penalty function; This is the power fluctuation suppression coefficient. Penalty for power variation; The robustness coefficient is... Placeholder for robust risk measurement; The objective function of the Follower is constructed as follows: ;in, This is a function used to characterize the effect of power allocation on the quality of charging services, and the function increases with the increase of power allocation; For time Assigned to the The charging power of a single charging session; This is the power fluctuation suppression coefficient, used to constrain the power variation amplitude between adjacent scheduling cycles. For scheduling time The station power allocation vector.

[0020] The Follower uses MM-ADMM to decompose the coupled constraints and employs the water-filling method to solve each subproblem quickly, enabling it to converge and output within a finite number of iterations. The expression used in the water injection method is: ;in, The water level induced by total power constraint, This is the interval projection operator.

[0021] Preferably, in step S3, the strategy variable Includes at least one of the following: Service quality weighting coefficient, tending to improve overall charging completion rate; Voltage safety penalty factor, which imposes a penalty on the risk of voltage drop / exceeding the limit; Power fluctuation suppression coefficient, which suppresses command jumps and enhances smoothness; Station-level safety margin / risk preference coefficient, with a margin reserved for station-level budget; : Robustness coefficient, used for defensive scheduling in the face of uncertainty / distribution shift.

[0022] Preferably, the Leader generates a station-level charging power upper limit. ,in, From S1, From S2, function It is used to translate risk appetite and safety boundaries into an executable budget.

[0023] S4. Construct an in-station scheduling enhancement mechanism based on sub-Bruker bar optimization, and suppress the impact of uncertainty on service quality and voltage safety through Wasserstein distance and conditional risk measurement; The in-station scheduling enhancement mechanism based on distributed bar optimization includes: The Leader objective function is constructed as follows:

[0024] Introducing the Conditional Value at Risk (CVaR):

[0025]

[0026]

[0027] in, , For voltage safety threshold, The objective function for optimizing the original nominal Leader layer is not considered for uncertainties and extreme scenarios.

[0028] S5. Based on grid operation safety constraints, the scheduling results are verified and multi-objective collaborative correction is performed to form a final power control scheme that takes into account both service quality and voltage safety.

[0029] The multi-objective collaborative correction includes: The power distribution results within the station output by S4 will be used. Mapping to the station-to-network interface model to evaluate the actual minimum node voltage. ;in, The voltage per unit value of the monitored node. This is a set of key nodes related to the charging station access point; when When approaching a safety boundary, a multi-target collaborative correction mechanism is triggered, the correction mechanism including at least one of the following: Fine-tune the Leader layer strategy variables; Implement safe compression of the station's power budget: ; Power is selectively reduced for some low-urgency sessions, and power resources are concentrated on high-urgency sessions.

[0030] S6. Issue power control commands and form an in-station closed-loop adaptive scheduling mechanism based on execution feedback.

[0031] The closed-loop adaptive scheduling mechanism includes: Collect actual execution power Calculate the target power Deviation from actual execution power and station-level power execution deviation ; Update the cumulative charging volume for each charging session. ; Based on the updated remaining energy demand With remaining available stay time Update the state dataset and trigger the next scheduling cycle.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) Achieve multi-objective collaborative optimization of voltage safety and service quality This invention places voltage security as a core game objective alongside quality of service, through voltage margin. and voltage sensitivity coefficient The voltage safety boundary is directly embedded into the station-level power budget calculation, and service quality and voltage safety are explicitly weighed in the upper-level objective function of the master-slave game. This mechanism enables the system to pursue an improved user charging experience without sacrificing voltage safety, achieving "soft expansion" and fair scheduling of charging power in uncertain environments.

[0033] (2) Uncertainty suppression mechanism based on Wasserstein blobs optimization To address uncertainties such as vehicle arrival time, departure time, and energy demand, this invention introduces Wasserstein decomposed bar sets and conditional risk value, requiring the Leader strategy to consider worst-case scenarios under uncertainty during its formulation. This mechanism ensures that even under the most unfavorable disturbances, the minimum node voltage remains above the minimum threshold, effectively suppressing voltage exceedance risks and improving the long-term stability and reliability of the system.

[0034] (3) Hierarchical real-time solution architecture based on MM-ADMM and water injection method This invention employs a hierarchical collaborative architecture, with an upper-level coordinator performing robust risk decisions and lower-level charging piles using the MM-ADMM algorithm and water-filling method to achieve rapid power allocation. By using the ADMM iterative format and water-filling equations, the complex optimization problem involving multiple coupled charging piles is decomposed into sub-problems that can be solved quickly, meeting the real-time requirement of second-level scheduling cycles and demonstrating significant engineering application value.

[0035] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0036] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a general framework diagram of the present invention; Figure 2 This is a flowchart illustrating the execution process of the present invention. Detailed Implementation

[0037] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0038] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0039] like Figure 1 As shown in Example 1, a method for power allocation and voltage risk suppression in charging stations based on master-slave game theory is provided. The method is executed by a coordinator in the charging station as the master control unit and adopts a three-layer collaborative architecture, including a charging station layer, a coordinator layer and a charging pile layer.

[0040] Charging station layer: Responsible for station-level sensing and power boundary calculation, including station-level sensing / collection, base load-transformer margin calculation, adjustable power budget generation, and voltage safety alerts.

[0041] Coordinator layer: As the core of decision-making, it is responsible for demand aggregation, leader strategy generation, follower solution, three-phase coordination, robust correction, and instruction shaping.

[0042] Charging station layer: Responsible for execution and status feedback, including executing power commands, measuring and reporting actual power, SOC and status information, and session updates.

[0043] The three layers interact through a closed-loop information flow, forming a collaborative control mechanism of perception, game theory decision-making, power coordination, robust correction, and execution feedback. Based on this, this embodiment executes specific processes according to S1 to S6 within each scheduling cycle.

[0044] S1: In response to the scheduling cycle trigger, collect multi-source operation data of station-network-pile-vehicle, construct a status dataset and calculate the station-level schedulable power budget, which simultaneously considers capacity constraints, thermal stability constraints and voltage safety boundary constraints.

[0045] In this embodiment, the system enters a scheduling round in response to a periodic trigger request from the local coordinator of the charging station. The scheduling period can be configured from 1s to 60s, preferably 5s. At each scheduling time t, the operating status of the station and the upstream distribution side is collected and organized to form a station status dataset for subsequent master-slave game decision-making and robust optimization.

[0046] (I) Multi-source data acquisition and state dataset construction The system acquires the following two types of information at scheduling time t: A. Station / Grid-side operation data: including but not limited to: maximum station-level access capacity, non-charging basic load within the station, active power, reactive power, voltage / current of station service transformers / incoming line switches and metering points, station-level three-phase voltage and imbalance index, and voltage lower limit margin of upstream nodes, etc.

[0047] B. Pile / Session-side data: including but not limited to: the current set of online charging sessions, the maximum rated power of each device in each session, the actual output power, the remaining charging demand, the remaining dwell time, the maximum power supported by the vehicle / gun, phase information (A / B / C phases), session status (normal / limited power / fault / about to leave the station), etc.

[0048] In this embodiment, the above information is organized into a site operation status dataset:

[0049] in: Maximum active power capacity that can be connected (kW); The active power (kW) of the station's basic load (lighting, air conditioning, supporting facilities, etc.); , These are the station-level three-phase voltage / current measurement vectors; The current load level of station service transformers or incoming line equipment can be expressed as apparent power or equivalent current carrying capacity. For a moment The set of charging sessions within the scheduling set. ; For the first Each session access phase is different; Allowed power lower / upper limit; For the session The measured output power; These represent the remaining energy required for the session and the remaining dwell time, respectively.

[0050] (ii) Multi-constraint calculation of station-level dispatchable power budget To ensure that the station's dispatching operates within the safety boundaries, this embodiment considers capacity constraints, thermal stability constraints, and voltage safety boundary constraints simultaneously.

[0051] (1) Available power obtained from capacity / base load: .

[0052] (2) Thermally stable usable power obtained from transformer or incoming line thermal limit constraints: If the maximum allowable load of the station service transformer is The equivalent occupancy excluding charging load is Therefore, the thermal limit budget for charging is: ;in, It is the equivalent power factor conversion factor, used to convert apparent margin into distributable active power margin.

[0053] (3) Voltage margin available power obtained from voltage safety boundary constraints: In order to explicitly reflect voltage safety, this embodiment introduces station-level voltage boundary estimation.

[0054] Let the allowable lower voltage limit for the upstream or critical node be... (e.g., 0.95 pu), the current minimum phase voltage is The voltage margin is: .

[0055] When the value is greater than 0, theoretically, more additional charging power is allowed. When the power output approaches zero, the amount of new power should be limited.

[0056] This embodiment uses voltage sensitivity / equivalent voltage drop to approximately construct the upper limit of usable power under voltage constraints: ;in, The equivalent sensitivity coefficient of the station load to the voltage at critical nodes; To prevent small constants from being divided by zero.

[0057] When voltage risk ,but No further load increases are allowed during this cycle, and this will trigger load reduction / redistribution in subsequent steps.

[0058] (4) Comprehensive station-level dispatchable power budget: The minimum of the above three types of available power is taken to obtain the upper limit of the station-level charging budget that satisfies multiple hard constraints. .

[0059] (III) Introducing safety margins and implementation lag Considering uncertainties such as measurement noise, short-term fluctuations in base load, control delay, and communication packet loss, this invention introduces a safety margin factor. This forms the final station-level schedulable power budget used to be distributed to the coordinator layer: ;in, The larger it is, the more conservative and stable it is; Smaller is more aggressive, but it is also more likely to hit the lower voltage limit.

[0060] (iv) Prepare empirical distribution sample packs for robust optimization of Wasserstein distance To support robust optimization of the subsequent coordinator layer based on Wasserstein distance, this embodiment synchronously constructs and updates the uncertainty sample set in S1, such as the basic load disturbance sample. Arrival / Departure and Demand Error Samples Voltage sensitivity error samples Combine them into an empirical sample set at time t: .

[0061] The above sample set will be used to construct the empirical distribution in subsequent steps. and its Distance neighborhood.

[0062] (V) S1 Output After completing S1, the system output should include at least: Internal operating status dataset ; Station-level dispatchable power budget As a constraint or strategy input from the upper level / leader; Voltage risk key quantity Voltage sensitivity ; Experience sample set required for robust optimization .

[0063] S2: Based on the state dataset, construct a demand assessment and robust representation model, quantify the urgency index of the session, and construct a robust set describing uncertainty based on Wasserstein distance.

[0064] This step, under the premise that the station-level schedulable power budget has been given by S1, simultaneously characterizes the conflict relationship between two types of objectives for concurrent charging sessions within the station: user-side charging demand (urgency / fairness) and grid-side operational safety (lowest voltage point, etc.). It then transforms these into parameterized inputs that can be directly called by the subsequent coordinator layer, thereby supporting the real-time solution of master-slave game and robust power allocation.

[0065] (a) Calculation of Residual Demand and Time Constraints For any charging session within the scheduling set In this embodiment, the remaining energy demand and remaining dwell time of S1 are calculated based on the station operation dataset:

[0066]

[0067] in, Energy is required to achieve the target; This refers to the energy currently charged or the energy corresponding to the equivalent SOC. This is the estimated departure time; The scheduling period is used to avoid a denominator of 0 and to ensure consistency of time dimensions.

[0068] Based on this, a demand intensity (urgency) index is constructed:

[0069] in, It indicates the intensity of energy replenishment required per unit time; the larger the value, the more urgent the need.

[0070] (II) Construction of Service Quality Demand Function To ensure the system is not only fast but also fair, this embodiment constructs a session-oriented architecture. The Quality of Service (QoS) metric. QoS is defined in the form of energy delivery satisfaction and time pressure penalties, turning service quality into an optimizable objective:

[0071] in, To assign to a session The charging power; Charging efficiency / conversion factor; Optimize window length for scrolling; Time pressure penalty coefficient; It is a very small positive number, avoiding division by zero and enhancing numerical stability.

[0072] Furthermore, this embodiment provides statistics on the overall service quality within the station:

[0073] in, This represents the number of active sessions within the current scheduling period. A smaller value usually indicates better service fairness.

[0074] (III) Demand Weight Generation This embodiment will emphasize urgency. Mapped to demand weights for power allocation The preferred approach is a combination of normalization, saturation limiting, and smoothing to prevent extreme sessions from monopolizing power. (1) Normalization and saturation:

[0075]

[0076] in, These are the upper and lower limits of the weight.

[0077] (2) Weight smoothing:

[0078] in, This is the smoothing coefficient.

[0079] The final weight vector is obtained as follows:

[0080] This weight vector will serve as one of the core inputs for subsequent MM-ADMM and water injection power allocation.

[0081] (iv) Construction of power grid security proxy quantity This invention constructs a power grid security proxy in S2 for use in upper-level decision-making at the coordinator layer:

[0082] in, For the set of distribution network nodes, This represents the per-unit value of the node voltage.

[0083] To facilitate real-time calculations, a sensitivity approximation is used to construct a voltage lower bound proxy:

[0084] in, This is the minimum reference voltage when there is no charging load. Let be the equivalent sensitivity coefficient of session i to the weakest node voltage.

[0085] This form enables improvement By adjusting the power vector The penalties are directly reflected in the penalties.

[0086] (v) Robust requirement representation based on W distance Considering the randomness and bias in vehicle arrival / departure, energy demand, and maximum available power, this invention introduces a robust representation based on W distance to construct a distributionally RobustSet of demand random variables.

[0087] Let a random vector represent the perturbation related to session demand, and construct an uncertain random vector: .

[0088] The system constructs an empirical distribution based on historical samples. (Formed from samples within a sliding window), and a distribution uncertainty set with a radius of W is defined:

[0089] in, The Wasserstein distance; The robust radius is obtained from the confidence level, sample size, or empirical parameter tuning.

[0090] When evaluating upper-level strategies, the coordinator layer uses worst-case expectation / risk metrics: ;in, The loss function is a combination of voltage over-limit penalty, QoS loss, and power fluctuation penalty:

[0091] in, For voltage safety threshold, These are the weighting coefficients.

[0092] (vi) Fault tolerance rules for abnormal / missing data When the following situations occur: missing departure time, abnormal energy demand, interruption of pile terminal communication, or In response to fluctuations and sudden changes, the system adopts a conservative rollback strategy. like If missing, then ; like If it is abnormal, then let ; If a session is interrupted, its weight is set to [value missing]. And limit its power to no more than a conservative upper limit. .

[0093] The above fault tolerance ensures that the S2 output parameter set is always available, preventing subsequent solvers from encountering NaN / Inf and causing system interruption.

[0094] (vii) S2 output S2 ultimately forms the parameter set required for robust optimization: .

[0095] S3: Construct a master-slave game scheduling model: The station-level / coordinator layer generates Leader policy variables and drives the charging pile layer Follower to respond and solve, outputting an in-station power command vector that satisfies the station-level power and voltage safety boundaries.

[0096] After completing S2, the system enters the master-slave game (Stackelberg) scheduling phase.

[0097] The coordinator layer acts as the leader to generate in-station scheduling strategy variables, and the charging pile layer acts as the follower to solve for feasible power allocation under the given strategy. Thus, under the coexistence of "station-level power constraints, user demand differences, and distribution side voltage safety constraints", an executable set of in-station power commands is formed.

[0098] (I) Leader-Follower Role Definition and Information Interaction Leader (Coordinator Layer): Generates on-site scheduling strategy variables in each scheduling cycle t and publishes them to the charging pile layer as a response basis; Follower (charging pile layer): Receives Leader strategy variables and station-level schedulable power budget, solves the target power of each session under pile-level constraints, forms an executable power vector, and sends back key feedback quantities.

[0099] The system has obtained a set of session urgency / service demand parameters in S2, such as demand weights. ), surplus energy demand Remaining stay time Equipment power boundary and the set of distribution / uncertainty description parameters required for robust characterization. In S3, the above parameters serve as the basic inputs for solving the master-slave game.

[0100] (II) Construction of Leader Strategy Variables The system uses parameterized policy vectors as the Leader decision variables: ; Wherein, K is the strategy dimension, and in the preferred embodiment, K=3 to 8, and its meaning may include, but is not limited to: Service quality weighting coefficient, tending to improve overall charging completion rate; Voltage safety penalty factor, which imposes a penalty on the risk of voltage drop / exceeding the limit; Power fluctuation suppression coefficient, which suppresses command jumps and enhances smoothness; Station-level safety margin / risk preference coefficient, with a margin reserved for station-level budget; : Robustness coefficient, used for defensive scheduling in the face of uncertainty / distribution shift.

[0101] In a preferred embodiment, the Leader does not directly allocate power to each charging session individually, but instead generates three types of strategy information: "constraint boundaries, priority rules, and risk preferences," including at least the station-level charging power limit.

[0102] in, From S1, From S2, function It is used to translate risk appetite and safety boundaries into an executable budget.

[0103] (III) Leader Objective Function: Optimizing the trade-off between service quality and voltage security The system defines two types of core performance indicators and uses them as the target trade-offs for the Leader.

[0104] (1) Quality of Service (QoS) Define a quality of service metric for each session i: ;in, This indicates the actual energy delivered within the assessment window. Energy is required to achieve the target; For truncation operators.

[0105] Further, provide overall metrics for the site:

[0106] in, Reflecting the overall service completion rate This reflects fluctuations in service consistency / fairness.

[0107] (2) Voltage safety index ( ) The system uses the lowest voltage on the distribution side as a safety profile: , ;in, The set of monitored nodes This represents the per-unit value of the node voltage.

[0108] Based on this, the optimization objective of the Leader is written as:

[0109] in, It is a voltage penalty function; Penalty for power variation; This is a placeholder term for robust risk measurement.

[0110] (iv) Follower Response Problem: Solving for feasible power allocation under a given Leader policy Follower receives and Then, the power vector is calculated for all active sessions within the site:

[0111] It must satisfy the following basic constraints:

[0112] The objective function of the Follower is preferably driven by demand urgency and constrained by smoothness.

[0113] in, Available from S2 Alternatively, a piecewise concave function can be used to ensure that the most urgent tasks are given priority and that the allocation property of diminishing marginal returns is maintained.

[0114] To balance real-time performance and scalability, the Follower employs MM-ADMM to decompose coupled constraints and uses the water-filling method to solve each subproblem quickly, enabling it to converge and output within a finite number of iterations. The KKT (Kalush-Kun-Tucker condition) form corresponding to the water-filled structure is expressed as follows: ;in, The water level induced by the total power constraint (Lagrange multiplier). This is the interval projection operator.

[0115] This structure ensures that when the station-level budget is tightened, allocation automatically tilts towards high-weight / high-urgency sessions; when the station-level budget is relaxed, more sessions receive higher power while still being constrained by device limits.

[0116] (V) S3 Output use , , Describe the scheduling effect without relying on external human evaluation.

[0117] To ensure the system has an interpretable closed-loop operation, after the Follower outputs the power vector, the system synchronously calculates and records it: This will be used as feedback for the next cycle's Leader strategy update.

[0118] Therefore, the above indicators are not post-experimental statistics, but rather performance metrics and closed-loop feedback variables built into the scheduling system, used to support: when When the level is too low, the strategy automatically enhances the service preference to improve it. Or relax part of the budget; when When the voltage is close to the threshold, the strategy automatically increases the voltage penalty to improve performance. Tighten Thus present An upward trend in operations; when When the value is too large (service fluctuation / unfairness), consistency suppression or smoothing constraints are introduced to make the service quality distribution more stable.

[0119] S4: Construct an in-station scheduling enhancement mechanism based on distributed bar optimization, and suppress the impact of uncertainty on service quality and voltage safety through Wasserstein distance and conditional risk measurement.

[0120] In S3, the system has already solved the in-station power scheduling strategy and response power allocation based on a master-slave game structure. However, in actual operation, the load environment faced by AC charging stations has significant uncertainties, including but not limited to vehicle arrival time deviations, users leaving the station early, energy demand estimation errors, base load fluctuations, and grid operation disturbances. If scheduling is based solely on nominal parameters or a single predicted value, the scheduling results are prone to performance degradation under disturbance conditions, leading to service quality deterioration or voltage safety risks.

[0121] Therefore, this embodiment enhances the master-slave game scheduling framework by introducing a distributed bar optimization.

[0122] (I) Sources of Uncertainty and Random Variable Modeling The following key factors are considered as uncertain random variables: deviation of actual vehicle dwell time, deviation of actual deliverable charging energy, short-term fluctuation of base load within the station, and measurement errors in the scheduling execution and feedback process.

[0123] The system abstracts the aforementioned uncertainties into a unified random disturbance vector. ,in, This represents an uncertain parameter space. Correspondingly, the system performance indicators induced by the Leader-Follower scheduling results in S3 can be regarded as random variables:

[0124] (ii) Characterization of Uncertain Bruker Bars Based on Wasserstein Distance A bibru bar modeling method based on Wasserstein distance is adopted. The system constructs an empirical distribution based on historical operating data or short-term observation samples. Introduce a set of uncertain distributions with the Wasserstein distance as the radius:

[0125] in, Represents the Wasserstein distance; The robust radius is used to characterize how conservative the system is with respect to distribution shifts.

[0126] (III) Robust target construction: Defensive optimization of performance under the worst-case distribution In S3, the original objective function of the Leader layer is constructed based on nominal performance metrics. The Leader objective is enhanced as follows:

[0127] This objective means that when selecting a scheduling strategy, the Leader not only focuses on performance in the nominal scenario but also explicitly considers service quality and voltage safety performance under the most unfavorable conditions within the distribution offset range. Through this mechanism, when the system faces a sudden surge in demand or abnormal load fluctuations, the scheduling strategy will automatically exhibit a more conservative power allocation behavior, thereby avoiding a precipitous drop in service quality or voltage exceeding limits.

[0128] (iv) Extreme risk mitigation mechanism based on conditional value at risk (CVaR) In a preferred embodiment, the system further introduces Conditional Value-at-Risk (CVaR) as an auxiliary robustness metric for voltage safety and quality of service lower bounds.

[0129] Taking the minimum node voltage as an example, its risk function is defined as follows: ;in, This is the voltage safety threshold. The system constructs the corresponding CVaR metric: ;in, Indicates the level of confidence in the risk.

[0130] By introducing a CVaR term into the Leader objective function: ,in, The original nominal optimization objective function of the Leader layer is not considered for uncertainties and extreme scenarios; the system can explicitly suppress the significant impact of a small number of extreme scenarios on voltage safety and service quality, making the scheduling results more robust.

[0131] (v) Explanation of the structural impact of robust scheduling on system performance indicators By embedding Wasserstein distance and CVaR risk metric into a master-slave game scheduling framework, this invention achieves the following effects at the mechanism level: When uncertainty increases, the Leader strategy automatically tightens the station-level power budget or adjusts the demand weight mapping, enabling the system to maintain a high level of power even in the most unfavorable scenarios. ; Because extreme risks are penalized in advance, fluctuations in service quality across different sessions are effectively suppressed, thereby reducing... ; During periods of manageable uncertainty or stable load, the system can still release power resources through strategic adjustments, enabling... It remains at a high level.

[0132] The aforementioned performance improvement is not due to offline parameter tuning, but rather a natural result of the distributed decision-making mechanism during the online scheduling process.

[0133] (vi) S4 output Through S4, the system obtains: the robustly corrected Leader strategy parameters, the corresponding Follower robust power allocation results, and better service quality statistics and voltage safety indicators under uncertain disturbance conditions.

[0134] The robust scheduling results will serve as direct input to the final power command generation and execution control (S5 / S6), thereby ensuring the stability, security, and engineering feasibility of AC charging station power scheduling in complex operating environments.

[0135] S5: Based on grid operation safety constraints, the scheduling results are verified and multi-objective collaborative correction is performed to form a final power control scheme that takes into account both service quality and voltage safety.

[0136] After completing the master-slave game scheduling in S3 and the sub-bar optimization correction in S4, the system has obtained a set of in-station power allocation results with good robustness under uncertain disturbance conditions. In actual engineering practice, the final execution effect of in-station power scheduling still needs to meet the operational safety constraints of the distribution network side.

[0137] (I) Mapping and evaluation of dispatch results to power grid status The system will output the station power allocation results from S4. Mapping to the station-network interface model, its impact on the distribution network operation status can be evaluated in any of the following ways: fast power flow calculation based on the distribution network power flow model; linear approximation calculation based on the sensitivity matrix or equivalent impedance model; and state estimation method based on station-level measurement data and historical operation characteristics.

[0138] In the preferred embodiment, the system focuses on extracting the following key power grid security indicators, namely, the actual minimum node voltage: ;in, The voltage per unit value of the monitored node. This refers to the set of key nodes related to the charging station access point.

[0139] (II) Voltage safety judgment and operating boundary detection The system will compare the minimum node voltage obtained from the evaluation with a preset safety threshold: ;in, This is the lower limit for safe operation.

[0140] When the above conditions are met, the current scheduling result is considered acceptable in terms of voltage safety, and the system can directly enter the power command issuance and execution stage; When detected When the system approaches the safety boundary, it triggers a multi-objective collaborative correction mechanism to make a secondary adjustment to the scheduling results with a safety orientation.

[0141] (III) Construction of a multi-objective collaborative correction mechanism After the correction condition is triggered, the system does not simply roll back or reduce power as a whole. Instead, based on the aforementioned scheduling structure, it performs directional and coordinated correction of the power allocation results. Its basic principles include: prioritizing the protection of voltage safety boundaries to prevent voltage overshoot or undershoot risks; maintaining the quality of service structure as much as possible to avoid proportional and brutal compression of all sessions; and maintaining the continuity and interpretability of scheduling results to avoid drastic changes in power commands.

[0142] Therefore, the system adopts one or a combination of the following correction strategies: ① Fine-tune the Leader-level strategy variables, such as increasing the voltage penalty coefficient and tightening the station-level power limit; ② Implement safe compression of the station's power budget: ; ③Targeted power reduction for some low-urgency sessions, and power resources are concentrated on high-urgency sessions.

[0143] (iv) Service Quality and Voltage Safety: Coordination and Stability Assurance During the correction process, the system continuously monitors the changing trends of service quality and voltage safety indicators, including: .

[0144] By coordinating with the strategies in S3 and S4, the system achieves the following collaborative behavior: when voltage safety risks increase, the system trades off significant performance gains with minimal loss of service quality. Enhancement; when the power grid has sufficient operational margin, the system gradually releases power resources, enabling... Restore or enhance; Since the correction process is guided by both demand weighting and robustness risk constraints. The system remains within a controllable range, preventing drastic fluctuations in service quality. This mechanism establishes a continuous and adjustable trade-off between service quality and voltage safety, rather than a discrete and uncontrollable jump.

[0145] (V) S5 Output Through S5, the system ultimately forms: (1) The final power allocation scheme that satisfies the station-level power constraints, user demand constraints and distribution network voltage security constraints; (2) Corresponding service quality statistics and voltage safety indicators; (3) A set of safe power control instructions that can be directly sent to the execution layer of the charging pile.

[0146] S6: Issue power control commands and form an intra-station closed-loop adaptive scheduling mechanism based on execution feedback.

[0147] After completing station-level power budget calculation, demand assessment and robust characterization, master-slave game decision-making, robust optimization correction, and three-phase power coordination in S1-S5, the system obtains a final set of charging power commands that satisfy multiple operational constraints. The final power control command is denoted as:

[0148] in, This represents the final execution power distributed to the i-th charging pile or charging session within the current scheduling cycle.

[0149] (a) Issuance of power control commands and execution by charging piles The system transmits the final power control command through the charging station's local control network or station control system. The command is sent to the corresponding charging station execution unit. Each charging station, based on the received power command and its own control strategy and hardware characteristics, adjusts the charging output power in real time to bring the actual output power closer to the target power.

[0150] In a preferred embodiment, the power control command can be executed in either of the following ways: continuous power setting mode, i.e., directly according to... Controlled as a target power value; discrete gear mapping method, i.e. Map to the nearest power level supported by the charging station.

[0151] (II) Collection of Execution Status and Operational Feedback During the power adjustment process of the charging piles, the system acquires operational feedback information in real time through the station status acquisition module. This feedback information includes, but is not limited to, the actual output power of each charging pile. The cumulative charging power or equivalent state of charge update value for each charging session, the charging pile operation status flag (normal, power limited, abnormal or fault), the total charging power at the station level and the three-phase current, and the power distribution.

[0152] Based on the feedback information mentioned above, the system updates the station's operational status dataset in real time for power budget assessment and demand parameter calculation in the next scheduling cycle.

[0153] (III) Preparation for Deviation Calculation and Closed-Loop Correction The system calculates the deviation between the target power and the actual executed power: ; Furthermore, station-level power execution deviations were statistically analyzed: .

[0154] (iv) Vehicle status and session set update Based on power execution and feedback acquisition, the system synchronously updates the cumulative charging power of each charging session. And update the remaining energy requirement accordingly. With remaining available stay time .

[0155] The system will switch or remove the corresponding charging session when any of the following conditions are met: the vehicle has reached the target energy requirement and completed charging; the vehicle leaves the station early; or the charging station malfunctions or the session is interrupted.

[0156] The power resources released by the removed session will be reallocated in the station-level power budget in subsequent scheduling cycles.

[0157] (V) Closed-loop scheduling triggering and adaptive iteration After completing power execution, feedback acquisition, and status update, the system enters a waiting state until the triggering conditions for the next scheduling cycle are met. The scheduling cycle can be triggered in one of the following ways: fixed time interval triggering; event triggering mechanism, including new vehicle access, vehicle departure, station-level power budget changes, or three-phase imbalance exceeding limits, etc.

[0158] Within each scheduling cycle, the system re-executes S1-S5 based on the latest operating status data, forming a closed-loop scheduling mechanism of decision-making-execution-feedback-correction.

[0159] Through S6, this embodiment constructs a real-time closed-loop power scheduling system covering station-level power constraints, differences in demand urgency, robust uncertainty, and three-phase load coordination. This enables the proposed master-slave game and robust optimization strategies to be stably executed in complex and multi-disturbance AC charging station operating environments, and continuously improves station-level operational safety and voltage stability, overall service quality under multi-vehicle concurrency conditions, and the robustness and interpretability of charging power allocation results; thus forming a complete, closed-loop, and adaptive AC charging station power scheduling method.

[0160] like Figure 2 As shown, the method described in this embodiment uses a coordinator within the charging station as the main control unit, and executes the following process sequence in each scheduling cycle: Start or Trigger: In response to a preset scheduling cycle or event triggering conditions, such as new vehicle access, vehicle departure, or station-level power boundary changes, a power scheduling process is initiated.

[0161] Initialization: Read system parameters and constraint information, including station-level power capacity, voltage safety threshold, robust parameters, master-slave game-related control parameters, etc.

[0162] Data Acquisition: Real-time operational data is obtained from the charging station and charging pile layers, including station-level adjustable power boundaries, base load, voltage risk indicators, as well as status information, remaining demand, and dwell time for each charging session.

[0163] Session filtering: Remove sessions that have completed charging, left the station early, or are in an abnormal state, and retain the controllable set of sessions.

[0164] Demand aggregation: The demand information of each valid charging session is aggregated and evaluated to form a set of demand parameters.

[0165] Leader decision-making: Under the constraints of station-level power, voltage safety, and robust risk, determine the station-level power regulation strategy and related control parameters.

[0166] Follower Response: Under predetermined policy parameters, each charging session responds to its own charging power through a distributed power allocation mechanism, forming a preliminary session-level power allocation result.

[0167] Three-phase coordination: Based on the phase information corresponding to each charging session, the power distribution results are aggregated and coordinated in three phases to correct the power imbalance between phases.

[0168] Robust Correction: Based on load uncertainty and execution deviation, robust assessment mechanisms such as Wasserstein distance or conditional risk value (CVaR) are introduced to correct the power allocation results for risk.

[0169] Issuing instructions: The finalized power control instructions are sent to the execution units of each charging pile.

[0170] Feedback collection: Real-time collection of execution feedback information from each charging station, including actual output power, power changes, and operating status.

[0171] Constraints satisfied: Determine whether the current scheduling result meets the station-level power constraints, voltage safety constraints, and quality of service requirements; if the constraints are not met, return to the Leader decision stage to continue iterating; if the constraints are met or the preset convergence conditions are met, proceed to the next step.

[0172] Output results: Output the power allocation results and corresponding operating indicators within the current scheduling cycle, complete a complete scheduling process, and wait for the triggering of the next scheduling cycle.

[0173] The above process corresponds to S1 to S6, where S1 corresponds to data collection and session filtering, S2 corresponds to requirement aggregation, S3 corresponds to leader decision-making and follower response, S4 corresponds to robust correction, S5 corresponds to three-phase coordination and constraint verification, and S6 corresponds to issuing instructions and feedback collection.

[0174] Example 2, building upon Example 1, introduces an adaptive parameter adjustment mechanism. Based on historical operational data, the system employs reinforcement learning or Bayesian optimization methods to optimize the coefficients in the Leader policy vector online, enabling the scheduling strategy to adapt to changes in the charging station's operational mode.

[0175] For large-scale charging station clusters, this embodiment, based on Embodiment 1, sets up a two-layer architecture: an upper-level regional coordinator and a lower-level intra-station coordinator. The regional coordinator is responsible for inter-station power allocation, while the intra-station coordinator executes S1-S6 of this application, forming a hierarchical collaborative scheduling system.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for power allocation and voltage risk mitigation in charging stations based on a master-slave game, characterized in that, Includes the following steps: S1. In response to the scheduling cycle trigger, collect multi-source operation data of station-network-pile-vehicle, construct a status dataset and calculate the station-level schedulable power budget, which simultaneously considers capacity constraints, thermal stability constraints and voltage safety boundary constraints. The voltage safety boundary constraint comprises: acquiring a key node allowable voltage lower limit and a current minimum phase voltage , calculating a voltage margin ; calculating the voltage margin available power based on the voltage sensitivity coefficient ; wherein, is a small constant to prevent division by zero; and the voltage risk , then ;​ S2. Based on the state dataset, construct a demand assessment and robust representation model, quantify the urgency index of the session, and construct a sub-Bruker set describing uncertainty based on Wasserstein distance; Wherein, the sub-Bruker set is defined as based on the empirical distribution The distribution uncertainty set centered at the Wasserstein distance: ;in, For Wasserstein distance, The robust radius; S3. Construct a master-slave game scheduling model: The station-level / coordinator layer generates Leader policy variables and drives the charging pile layer Follower to solve the response, outputting the station power command vector that satisfies the station-level power and voltage safety boundaries; S4. Construct an in-station scheduling enhancement mechanism based on sub-Bruker bar optimization, and suppress the impact of uncertainty on service quality and voltage safety through Wasserstein distance and conditional risk measurement; S5. Based on the power grid operation safety constraints, the scheduling results are verified and multi-objective collaborative correction is performed to form a final power control scheme that takes into account both service quality and voltage safety. S6. Issue power control commands and form an in-station closed-loop adaptive scheduling mechanism based on execution feedback.

2. The method according to claim 1, characterized in that, In S1, the calculation method for the station-level schedulable power budget also includes: Calculate available power under capacity constraints ;in, To maximize the available active power capacity, This refers to the active power of the station's basic load. Calculate the available power for thermal stability constraints ;in, This is the equivalent power factor conversion factor. The maximum allowable load for the station transformer. ; The minimum of the available power with voltage margin, available power with capacity constraint, and available power with thermal stability constraint is used to obtain the upper limit of the station-level charging budget. ; Introducing safety margin factor The station-level schedulable power budget is obtained. .

3. The method according to claim 2, characterized in that, In S2, the quantitative session urgency index includes: For any charging session within the scheduling set Calculate the remaining energy demand With remaining stay time ;in, Energy is required to achieve the goal. The energy that has already been charged, For the scheduling period, This is the estimated departure time; Calculate the demand intensity index ;in, It indicates the intensity of energy replenishment required per unit time; the larger the value, the more urgent the need. Map the demand intensity index to demand weights and then smooth them out: in, These are the upper and lower limits of the weight. For smoothing coefficients; The final weight vector is obtained as follows: .

4. The method according to claim 3, characterized in that, In S2, the construction of the sub-Bruker set based on Wasserstein distance further includes: constructing an uncertain random vector. The empirical sample set, in which, For the deviation of the remaining stay time, This represents the deviation from the maximum available power.

5. The method according to claim 4, characterized in that, In S3, The Leader objective function is constructed as follows: ； in, For service quality weighting coefficient, This is a statistic representing the overall service quality within the station. This is the voltage safety penalty factor. It is a voltage penalty function; This is the power fluctuation suppression coefficient. Penalty for power variation; The robustness coefficient is... Placeholder for robust risk measurement; The objective function of the Follower is constructed as follows: ;in, This is a function used to characterize the effect of power allocation on the quality of charging services, and the function increases with the increase of power allocation; For scheduling time Assigned to the Charging power per charging session; This is the power fluctuation suppression coefficient, used to constrain the power change amplitude between adjacent scheduling cycles; For scheduling time The station power allocation vector; The Follower uses MM-ADMM to decompose the coupled constraints and employs the water-filling method to solve each subproblem quickly, enabling it to converge and output within a finite number of iterations. The expression for the water injection method is: ;in, The water level induced by total power constraint, This is the interval projection operator.

6. The method according to claim 5, characterized in that, In S4, the in-station scheduling enhancement mechanism based on distributed bar optimization includes: The Leader objective function is constructed as follows: Introducing the Conditional Value at Risk (CVaR): in, , For voltage safety threshold, The objective function for optimizing the original nominal Leader layer is not considered for uncertainties and extreme scenarios.

7. The method according to claim 6, characterized in that, In S5, the multi-objective cooperative correction includes: The power distribution results within the station output by S4 will be used. Mapping to the station-network interface model to evaluate the actual minimum node voltage. ;in, The voltage per unit value of the monitored node. This is a set of key nodes related to the charging station access point; when When approaching a safety boundary, a multi-target collaborative correction mechanism is triggered, the correction mechanism including at least one of the following: Fine-tune the Leader layer strategy variables; Implement safe compression of the station's power budget: ; Power is selectively reduced for some low-urgency sessions, and power resources are concentrated on high-urgency sessions.

8. The method according to claim 7, characterized in that, In S6, the closed-loop adaptive scheduling mechanism includes: Collect actual execution power Calculate the target power Deviation from actual execution power and station-level power execution deviation ; Update the cumulative charging volume for each charging session. ; Based on the updated remaining energy demand With remaining available stay time Update the state dataset and trigger the next scheduling cycle.

9. The method according to claim 1, characterized in that, In step S3, the policy variable Includes at least one of the following: Service quality weighting coefficient, tending to improve overall charging completion rate; Voltage safety penalty factor, which imposes a penalty on the risk of voltage drop / exceeding the limit; Power fluctuation suppression coefficient, which suppresses command jumps and enhances smoothness; Station-level safety margin / risk preference coefficient, with a margin reserved for station-level budget; : Robustness coefficient, used for defensive scheduling in the face of uncertainty and / or distribution shift.

10. The method according to claim 1, characterized in that, In step S3, the Leader generates the upper limit of station-level charging power. ,in, From S1, From S2, function It is used to translate risk appetite and safety boundaries into an executable budget.

Images