Dual-objective optimization method for electric vehicle charging scheduling based on incentive game

Abstract

The application discloses a double-target optimization method for electric vehicle charging scheduling based on incentive game, relates to the technical field of electric vehicle charging scheduling, and comprises the following steps: S100, a user behavior dynamic response identification model is constructed, historical charging behavior data and incentive trigger time nodes are acquired, response changes of users under different incentive conditions are extracted, and a response feature vector is generated.The application realizes dynamic allocation and differentiated scheduling of incentive resources by constructing a user response feature vector and response level division; a response feedback matrix is constructed by introducing behavior feedback and load data, strategy effect evaluation and adaptive optimization are supported, a closed-loop control mechanism of identification-guiding-feedback-correction is formed, and the incentive response efficiency, power grid scheduling adaptability and system intelligent level are improved, so that the power grid operation safety and user charging utility are effectively ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric vehicle charging scheduling, specifically to a double-objective optimization method for electric vehicle charging scheduling based on incentive game. BACKGROUND

[0002] The double-objective optimization of electric vehicle charging scheduling based on incentive game refers to designing reasonable strategies using the incentive mechanism in game theory to guide users to autonomously choose appropriate charging time or power in the context of multiple electric vehicle users participating in charging together, so as to ensure the charging willingness of users while reducing the peak load pressure of the power grid. This method models the electric vehicle charging scheduling as a game process with multiple participants, and sets double optimization objectives: one is to maximize the charging satisfaction or benefit of electric vehicle users, and the other is to minimize the load fluctuation or operating cost of the power grid side. By designing incentive strategies (such as dynamic pricing, subsidy mechanism, etc.) and establishing game models (such as non-cooperative game, Stackelberg game, etc.), the collaborative optimization between the regulation of user behavior and the improvement of system efficiency is achieved, so that the electric vehicle charging process can meet the individual interests while also considering the safe and efficient operation of the overall power system.

[0003] The existing technology has the following shortcomings:

[0004] In the charging scheduling process for a large number of electric vehicle users, the existing regulation mechanism based on incentive game generally relies on a unified price signal or incentive strategy to guide users to concentrate charging in a specific time period to achieve load peak shaving. However, in actual operation, multiple users form synchronous perception of the same incentive information and make "game optimal" responses based on the principle of maximizing individual benefits, which often leads to a large amount of charging demand instantaneously flowing into the incentive period, forming a new concentrated load peak. This group behavior lacks time-based guidance and load peak-shaving regulation mechanisms at the scheduling level, which easily induces secondary load peak concentration, leading to deviation of power grid load prediction and imbalance of scheduling resource allocation.

[0005] Further, under the condition of secondary load peak, the substations and distribution branches in the regional power grid may be overloaded due to exceeding the design capacity, triggering abnormal operating conditions such as voltage drop and current surge, and in severe cases, causing protection tripping of the distribution system, or even equipment failure such as insulation breakdown or thermal loss burning of local transformers. In addition, the power regulation mechanism at the system level is difficult to complete the supply-demand adaptive regulation in a short time, which easily causes real-time frequency fluctuation, voltage disturbance, and response misalignment of energy storage systems, thereby disrupting the dynamic balance of the entire charging scheduling system, seriously affecting the operation safety of the power system and the charging reliability of electric vehicle users.

[0006] The above information disclosed in the BACKGROUND section merely to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0007] The purpose of the present application is to provide a double-objective optimization method for electric vehicle charging scheduling based on incentive game, which realizes dynamic allocation and differentiated scheduling of incentive resources by constructing user response feature vectors and response level division; introduces behavior feedback and load data to construct a response feedback matrix, supports policy effect evaluation and adaptive optimization, forms a closed-loop control mechanism of identification-guidance-feedback-correction, improves the efficiency of incentive response, the adaptability of power grid scheduling and the intelligent level of the system, effectively guarantees the safety of power grid operation and the charging utility of users, to solve the problems in the above background technology.

[0008] In order to achieve the above purpose, the present application provides the following technical scheme: a double-objective optimization method for electric vehicle charging scheduling based on incentive game, comprising the following steps:

[0009] S100, a user behavior dynamic response identification model is constructed, historical charging behavior data and incentive trigger time nodes are obtained, response changes of users under different incentive conditions are extracted, and a response feature vector is generated;

[0010] S200, the users are grouped based on the response feature vector, a plurality of response level groups with different incentive sensitivity gradients are formed, and corresponding incentive control priorities are set;

[0011] S300, based on the response level groups and the incentive control priorities, the response features of each response level group and the current power grid operating state are combined to generate a distributed incentive scheduling table;

[0012] S400, when a user initiates a charging connection request, differential incentive information is pushed according to the parameters in the response level group and the distributed incentive scheduling table, and the user is guided to generate an access path sequence;

[0013] S500, in the process of the user performing the charging task according to the access path sequence, the charging behavior data of the user and the power grid load state data in the corresponding period are collected, and a feedback matrix between the user path behavior and the power grid response state is constructed;

[0014] S600, based on the feedback matrix, the deviation degree of the user behavior and the recommended path, and the difference between the power grid state and the scheduling expectation are identified, the incentive amplitude, the access period and the control strategy parameters are dynamically corrected, the scheduling table is updated, and the user behavior identification model is simultaneously optimized, to realize closed-loop adaptive optimization control.

[0015] Preferably, step S100 comprises:

[0016] Obtaining historical charging behavior data of electric vehicle users and incentive trigger time nodes, and constructing a sequence of behavior-incentive interaction data pairs;

[0017] Extracting index data of behavior changes of users before and after incentives, including behavior fluctuation index, relative difference ratio, and incentive action window delay;

[0018] Based on the extracted index data, a response feature vector is constructed, normalized encoding is adopted, and a covariance weight adjustment mechanism is introduced to enhance the feature discrimination;

[0019] In subsequent charging operations, the values of each dimension of the response feature vector are continuously updated to realize dynamic correction and adaptive learning of the user behavior model.

[0020] Preferably, the covariance weight adjustment mechanism comprises:

[0021] A covariance matrix between each dimension of a plurality of user response feature vectors is calculated to determine the degree of linear correlation between each feature dimension;

[0022] Feature dimensions with covariance values below a set threshold are assigned high weights to enhance their contribution to the differentiation of user behavior patterns;

[0023] Feature dimensions with covariance values above a set threshold are assigned low weights to suppress their interference in user behavior feature recognition;

[0024] The feature vector after weight adjustment is reconstructed to improve the clustering clarity and grouping effectiveness of the user response features in the feature space.

[0025] Preferably, step S200 comprises:

[0026] Performing unsupervised clustering analysis on the weighted user response feature vector, and forming a plurality of behavior response categories according to the feature space distribution;

[0027] Calculating the average response feature vector of each behavior response category to determine its incentive sensitivity index;

[0028] According to the incentive sensitivity index, a plurality of response level groups are divided;

[0029] Setting incentive control priorities for each response level group for subsequent incentive resource allocation and strategy pushing timing configuration.

[0030] Preferably, step S300 comprises:

[0031] Based on the historical response features of the response level groups and the current power grid operating state, a dispatch scenario recognition data set is constructed;

[0032] According to the scheduling scene recognition dataset, the incentive amplitude parameters of each response level group are determined by combining the user historical response function and the target scheduling demand;

[0033] The charging access time period parameters of each response level group are set by analyzing the target area power grid load fluctuation curve;

[0034] The corresponding regulation intervention strategy parameters are configured to form a distributed incentive scheduling table containing incentive amplitude, access time period and intervention strategy.

[0035] Preferably, step S400 comprises:

[0036] When the electric vehicle user initiates a charging connection request, the corresponding distributed incentive scheduling table parameters are extracted by identifying the response level group to which the user belongs;

[0037] According to the distributed incentive scheduling table parameters and the current power grid load state, a differentiated incentive strategy with the minimum incentive cost and the maximum behavior adjustment amplitude is generated;

[0038] The incentive information containing the incentive amplitude, the access time window, the access location and the effective time limit is pushed to the user terminal;

[0039] After the user selects, the user registers the access path sequence and records the current path as the behavior execution record for subsequent behavior feedback and strategy iteration analysis.

[0040] Preferably, step S500 comprises:

[0041] The charging behavior data of the electric vehicle user in the process of executing the access path sequence are obtained, and the user identifier and the path number are bound;

[0042] The operating state data of the corresponding power grid node and its adjacent area in the user charging period are collected, and time alignment and abnormal elimination processing are performed;

[0043] A response feedback matrix is constructed with the user path behavior as the row vector and the power grid operating state as the column vector;

[0044] The response feedback matrix is input into the data feedback engine for incentive strategy evaluation and regulation parameter update analysis.

[0045] Preferably, step S600 comprises:

[0046] The response feedback matrix is periodically analyzed to identify the deviation degree between the user behavior response and the recommended access path, and the difference information between the power grid load state and the scheduling expectation;

[0047] According to the identified deviation degree and difference information, the incentive strategy correction condition is determined;

[0048] When the correction condition is met, the incentive amplitude parameter, the access period parameter and the regulation intervention strategy parameter are dynamically adjusted, and the distributed incentive schedule table is updated.

[0049] The new user behavior data is fused with the original training data, the user behavior recognition model is optimized, and the response level grouping result is updated.

[0050] In the above technical solution, the technical effects and advantages provided by the present application are as follows:

[0051] The present application constructs a user response feature vector and divides the response level, so that the system can dynamically allocate incentive resources according to user sensitivity, and form a differentiated and distributed scheduling strategy in combination with the power grid operation state. At the same time, the response feedback matrix is constructed by introducing real-time behavior feedback and power grid load data, so that the strategy execution effect is visual and quantifiable, and on this basis, the adaptive iteration optimization of incentive parameters and behavior models is realized. Finally, a closed-loop control mechanism with "recognition-guidance-feedback-correction" as the core is formed, which takes into account the individual charging utility of electric vehicle users and the overall operation safety of the power system, effectively improves the response efficiency of incentive investment, the dynamic adaptability of scheduling strategy and the load balancing of power grid operation, and significantly enhances the intelligent level and robustness of large-scale electric vehicle charging scheduling system. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only represent some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0053] Figure 1 The method flowchart of the double-objective optimization method for electric vehicle charging scheduling based on incentive game of the present application. DETAILED DESCRIPTION

[0054] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the gist of each example to those skilled in the art.

[0055] The present application provides a double-objective optimization method for electric vehicle charging scheduling based on incentive game as shown in Figure 1 The double-objective optimization method for electric vehicle charging scheduling based on incentive game includes the following steps:

[0056] S100, a user behavior dynamic response identification model is constructed, historical charging behavior sequences and incentive trigger time nodes are obtained, response changes of each electric vehicle user under different incentive conditions are extracted, and corresponding response feature vectors are generated;

[0057] To effectively identify the charging behavior response of electric vehicle users under different incentive conditions, a user behavior dynamic response identification method is constructed. Through deep analysis of multi-source data and behavior modeling, the sensitivity features of individual users to external incentive information are extracted, which provides the basis for behavior grouping and parameter setting for subsequent differentiated charging regulation. The method includes the following steps:

[0058] The historical charging behavior data of electric vehicle users and the associated incentive trigger information are obtained. In actual operation, by statistically analyzing the charging behavior records of users within a certain time period, key data including but not limited to charging initiation time, duration, actual access power, charging location, charging completion rate, historical price information, etc. are collected. At the same time, the time nodes of all incentive trigger events and the strategy content are labeled to form the time matching relationship between charging behavior and incentive events. In order to ensure the continuity and correlation of the data, the behavior data is normalized in time sequence, and the data bias introduced by abnormal behavior (such as power failure, fault interruption) is eliminated, and the behavior-incentive interaction data pair sequence under the unified time reference is constructed.

[0059] On the basis of obtaining the alignment of historical charging behavior sequences and incentive information, the response of each user's behavior change is extracted. By constructing a time period comparison window, the behavior feature differences of users in adjacent time windows before and after the incentive are counted respectively, and the response behavior features such as whether to initiate charging, whether to adjust power, whether to change location, etc. are extracted. To enhance the description of the behavior change trend, further introduce behavior fluctuation indicators, relative difference ratios, incentive action window delays, and other advanced behavior statistics. Through these multi-dimensional indicators, the response direction, response strength, and response lag of users under different types of incentive conditions can be described, so as to mine the individual behavior dynamic change rules.

[0060] In the process of identifying the dynamic response of electric vehicle user charging behavior, in order to more accurately describe the change trend of user behavior patterns before and after the implementation of incentive strategies, three types of representative statistics are introduced: behavior fluctuation indicators, relative difference ratios, and incentive action window delays, which are used to construct multi-dimensional quantitative expressions of behavior changes.

[0061] Behavior volatility is the numerical variation range of user charging behavior within a certain time window, usually reflected as the standard deviation, coefficient of variation or sliding difference of charging initiation time, duration or power level. Its role is to measure whether the user's charging behavior is stable before and after the incentive intervention, and then evaluate the predictability and response jitter characteristics of their behavior patterns. For example, if a user's charging period fluctuates significantly over multiple incentive periods, their volatility index is larger, indicating that the incentive has a strong but unstable effect on their behavior.

[0062] Relative difference ratio is the proportional difference between the average value of a key behavior variable (such as access period or power value) in the adjacent time period before and after the incentive event and the baseline state, usually expressed as "(average value after incentive - average value before incentive) / average value before incentive". Its role is to measure the strength, direction and significance of user response, which is the core indicator of whether the user has significantly adjusted their behavior to respond to the incentive. The larger the index, the more likely the user will adjust their strategy based on their own profit maximization principle.

[0063] Incentive action window delay is the time interval from the time the incentive is triggered to the time the user's behavior actually changes, usually measured in minutes or hours. This indicator reflects the time lag characteristics of user behavior response, especially for users who do not respond to incentives immediately. Its role is to help distinguish between "agile users" and "lagging users", and further set the advance amount and duration of incentive information push.

[0064] In actual implementation, the acquisition of the above three indicators is based on the time alignment of user historical behavior data and incentive strategy records, through the construction of a sliding time window for statistical analysis: first, the data is divided into equal length time windows before and after the incentive event; second, the mean, variance and time difference of each user's behavior variable within the window are calculated; finally, the statistical results are normalized and encoded to construct a response feature vector. These indicators complement each other, together depicting the amplitude, direction and time characteristics of user behavior response, providing a detailed behavior data basis for subsequent user clustering and behavior modeling.

[0065] After extracting the behavior response indicators, a feature expression structure is constructed to generate a response feature vector for each electric vehicle user. Specifically, the behavior response indicators extracted above are taken as feature dimensions, and a high-dimensional vector representation structure is constructed using a unified normalization coding method. In the response feature vector, each dimension represents the typical behavior response of the user under a certain incentive condition (such as peak-valley electricity price, instant subsidy, integral return, etc.), and the vector as a whole reflects the incentive sensitivity distribution of the user. To improve the separability of the feature vector in the subsequent grouping process, a covariance weight adjustment mechanism is introduced to assign different weights to different response dimensions, so that in the feature space, users with similar behavior patterns are more easily clustered, thereby enhancing the robustness and stability of the identification method.

[0066] The "covariance weight adjustment mechanism" refers to dynamically assigning a weight to each dimension according to the covariance structure between different dimensions (i.e., behavior response features under different incentive conditions) after constructing the user response feature vector, to reflect the importance of the dimension in distinguishing user behavior patterns. In the present application, the role of this mechanism is to enhance the separability and clustering effectiveness of user features in high-dimensional space, so that users with similar behavior patterns are more concentrated in subsequent gradient response grouping, and users with significant behavior differences are more easily distinguished, thereby improving the robustness and stability of the identification results.

[0067] Specifically, in a large user sample response feature data set, the covariance matrix between each dimension is calculated to evaluate the degree of linear dependence between different feature dimensions. For dimensions with small covariance (indicating higher independence and significant differences in user behavior), a higher weight is given to emphasize their important contribution to behavior pattern division; while for dimensions with high covariance, strong redundancy, or high concentration in multiple user distributions, a lower weight is given to reduce their interference with clustering boundaries. Weight adjustment can be achieved through principal component analysis (PCA) or feature covariance normalization transformation, which remaps the feature vector to a feature space with compressed weights. This process not only avoids the influence of invalid or weak discriminant features on the grouping results, but also helps to build a more structured and response trend obvious user classification system, thereby providing a more stable grouping basis for subsequent differentiated configuration of distributed incentive scheduling table.

[0068] To ensure that the response feature vector can dynamically adapt to the changing trend of user behavior, a response vector updating mechanism is established. By continuously collecting new round of user charging behavior data in subsequent charging scheduling operation, and comparing and analyzing the time series with the original behavior characteristics, the values of each dimension of the response vector are dynamically adjusted, so as to realize the continuous learning and evolution of the user behavior model. This updating mechanism not only can identify the long-term drift of user behavior patterns, but also can quickly respond to the triggering of short-term abnormal behavior, so as to maintain the timeliness and accuracy of the identification results. Finally, based on the constructed response feature vector, a personalized behavior response label can be labeled for each user, providing a data basis and behavior basis for the subsequent response gradient grouping and differential incentive regulation strategy formulation.

[0069] The core role of the step of "constructing a dynamic response identification model of user behavior, obtaining a historical charging behavior sequence and an incentive trigger time node, extracting the response changes of each electric vehicle user under different incentive conditions, and generating a corresponding response feature vector" is to provide a precise, dynamic, and personalized behavior portrait basis for the subsequent differential regulation of electric vehicle user behavior. In the large-scale charging scheduling scenario, the responses of electric vehicle users to incentive strategies are significantly different, which is manifested as some users being sensitive and quickly adjusting their behavior, while some users being slow or even not responding. If individual differences are ignored and a unified incentive signal is directly applied, it is easy to lead to user strategy convergence and cause risks such as concentrated access and secondary load peak. Therefore, this step analyzes the fusion of historical charging behavior data and incentive trigger time series data, quantifies the response characteristics of each user under different incentive conditions, constructs a high-dimensional response feature vector from the response direction, response strength, and response lag, and realizes the comprehensive characterization of user behavior sensitivity, stability, and consistency. This feature vector not only serves as an input basis for subsequent user grouping and incentive priority setting, but also participates in dynamic scheduling strategy optimization and behavior evolution modeling, has long-term adaptability and renewability, and is the behavior-driven core of the entire dual-objective optimization control framework. Through this step, the traditional "uniform treatment according to users" scheduling thinking can be broken through, and a new paradigm of intelligent scheduling of "precise regulation according to behavior differences" can be achieved.

[0070] S200, grouping electric vehicle users based on the response feature vector to form multiple response level groups with different incentive sensitivity gradients, and setting corresponding incentive regulation priorities for each response level group;

[0071] To realize the precise classification of electric vehicle user behaviors and the priority configuration of response strategies, a grouping method based on response feature vectors is constructed to divide a large number of electric vehicle users into multiple response level groups with different gradient sensitivities to incentives, and set corresponding incentive regulation priorities for each group to support the differentiated strategy configuration of the subsequent distributed incentive scheduling table. This method has good individual recognition ability and group scheduling adaptability, which includes the following steps:

[0072] The generated user response feature vector set is collected as the input data source of this step. Each response feature vector is composed of multiple dimensions, corresponding to the response behavior characteristics of users under different incentive conditions, such as response intensity, response direction, response volatility, response time delay, etc. In order to improve the discrimination of grouping, all response feature vectors are normalized to analyze the spatial distribution under a unified numerical scale. At the same time, according to the actual distribution characteristics of each dimension, a covariance weight adjustment mechanism is introduced to weight the response feature dimensions, improving the influence of discriminative features in the clustering process, and ensuring that users with similar behavior patterns have higher aggregation in space.

[0073] Based on the weighted response feature vector set, multi-level behavior response clustering analysis is performed. Unsupervised learning methods such as Gaussian mixture model, fuzzy C-means algorithm or spectral clustering method are used to automatically divide the user samples in the feature space. This clustering process does not pre-set the number of users, but adaptively determines the optimal number of clusters based on the natural clustering trend in the response feature space, forming several behavior response mode categories. In this process, to prevent boundary samples from being misclassified, a response stability index is designed as an auxiliary evaluation basis to perform secondary correction classification on edge users, improving the overall clustering quality.

[0074] The "multi-level behavior response clustering analysis" refers to the response feature vector of the electric vehicle user. Without relying on artificial label setting, the non-supervised learning method is used to classify the differences of the user in the behavior response dimension, so as to cluster the users with similar behavior patterns into the same category, and form multiple response levels. The core role of this analysis is to automatically identify different types of user groups from the response data of the user charging behavior to the incentive strategy, to provide a fine grouping basis for subsequent implementation of differentiated incentive regulation strategy, and to avoid the problem of strategy inadaptation or resource waste caused by traditional coarse-grained division. This process usually constructs a high-dimensional feature space, and uses non-supervised clustering algorithms such as Gaussian mixture model, fuzzy C-means algorithm or spectral clustering in this space for analysis. Gaussian mixture model is suitable for handling continuous probability distribution of behavior characteristics, fuzzy C-means allows users to belong to multiple behavior categories at the same time, and is suitable for scenarios with fuzzy behavior boundaries, while spectral clustering identifies non-convex group distribution through graph partitioning idea, and adapts to complex behavior similarity relationship. In the implementation process, the weighted response feature vector is first input into the clustering algorithm, and the optimal clustering number is adaptively determined through internal evaluation indicators such as silhouette coefficient and Davies-Bouldin index, and then the behavior category label to which each user belongs is output. The multiple behavior response categories formed finally, that is, the multi-level response gradient grouping, provides operational and dynamically updateable data support for subsequent assignment of incentive regulation priority and generation of distributed scheduling strategy.

[0075] After completing the clustering division, the overall sensitivity index of the external incentive signal is calculated according to the average response feature vector of each cluster group. The sensitivity index comprehensively considers the average response amplitude, response direction consistency and response lag time of the users in the group, and the higher the value, the more likely the users of this type are to adjust their behavior under the incentive intervention. According to the size of the sensitivity index, all cluster groups are sorted and divided into multiple response gradient level groups, defined as "high sensitivity group", "medium sensitivity group", "low sensitivity group", etc., to clearly define the response ability level of different user groups.

[0076] Based on the strength of the incentive sensitivity of each response level group, the corresponding incentive regulation priority is set. Specifically, the response level group with high priority will be allocated more incentive resources, earlier incentive push time or stronger price elasticity to fully stimulate its regulation potential; the group with low priority will appropriately weaken the incentive intensity, or use the lag strategy to avoid incentive waste and strategy congestion. This regulation priority not only serves as the basis for setting the subsequent scheduling strategy, but also can be dynamically adjusted. When the behavior characteristics of a response group change significantly during operation, its priority can be automatically updated to realize the linkage evolution of grouping and regulation strategy.

[0077] The core role of this step is to break the inherent mode of "homogeneous treatment" of users in traditional charging scheduling, and to realize the hierarchical classification and differential precise delivery of user regulation strategies. In practical applications, there are significant individual differences in the behavior response of electric vehicle users when faced with different forms of incentives such as price, subsidy, and integral return. Some users are highly sensitive and immediately adjust their charging behavior upon receiving incentives, while others are slow to react and show weak or no response to incentives. If these response differences are not identified and grouped, and a unified incentive strategy is directly applied, it will not only fail to achieve effective regulation of the power grid load, but also may cause resource waste, secondary peak, and behavior congestion, among other negative effects. Therefore, by analyzing the response feature vector, constructing a high-dimensional behavior representation space, and using clustering algorithms to automatically group users in this space, users can be divided into high-sensitivity, medium-sensitivity, and low-sensitivity behavior level groups, thereby mastering the actual response capabilities of different user groups to regulation signals. Based on this division result, different level groups are set with incentive regulation priorities, which can achieve the optimal allocation of incentive resources: preferentially inputting incentives to groups with high response potential to achieve maximum benefit guidance; for groups with weak response, delay the push and reduce the frequency to avoid ineffective stimulation. This step not only improves the overall scheduling efficiency and response flexibility, but also lays the foundation for a behavior-driven distributed regulation strategy, embodies the transition from "average incentive" to "precise incentive", and is a key link in building an intelligent, multi-level, and self-adaptive charging control system.

[0078] S300, on the basis of obtaining the response level groups and the incentive regulation priorities, combining the behavior response characteristics of each response level group and the current power grid operation state, generating a distributed incentive scheduling table for regulation, the incentive scheduling table including an incentive amplitude parameter, a charging access time period parameter, and a regulation intervention strategy parameter;

[0079] To achieve fine and differentiated guidance for different types of electric vehicle users, based on the established response level groups and their incentive regulation priorities, combining the behavior response characteristics of each level group and the real-time power grid operation state, a distributed incentive scheduling table for charging regulation is constructed. The scheduling table can be dynamically generated and is scalable, containing core scheduling information such as incentive amplitude parameters, charging access time period parameters, and regulation intervention strategy parameters. This scheduling table not only realizes the customized configuration of incentive delivery content and path, but also supports the linkage adjustment with time, power grid state, and user behavior feedback, including the following steps:

[0080] Based on the obtained user response level grouping results, the response behavior distribution characteristics of each response level group in different historical incentive scenarios are counted, and indexes including response rate, behavior adjustment amplitude, lag time, and unexpected behavior frequency are extracted. In combination with current power grid operation state parameters (such as regional load prediction value, time period power margin, electricity price floating interval, and voltage safety margin), a dispatch scenario recognition data set is constructed. In order to ensure the coordination between the strategy design and the power grid demand, the user group behavior characteristics and the power supply capacity matching relationship need to be fused in the data set, so as to identify which level users have the maximum regulation potential in the current or future time period and can be incentivized without triggering load risk.

[0081] According to the above identification result, the incentive amplitude parameter of each response level group is determined. The parameter is used to quantify the incentive intensity that should be given to the level user group under the current power grid operation state, including but not limited to price discount ratio, integral return rate, and subsidy amount. The parameter design process adopts a response elasticity matching mechanism, that is, according to the historical response function of users to different incentive amplitudes, in combination with the current load peak shaving target or load valley filling demand, the most cost-effective incentive amplitude is reversely deduced, so as to ensure the maximum user behavior guidance efficiency within the incentive cost control range. For high-sensitive group users, a smaller incentive amplitude can trigger expected behavior, while for low-sensitive group users, the incentive amplitude can be moderately increased and the delivery time can be delayed to share the response burden and improve the regulation coverage.

[0082] The "response elasticity matching mechanism" refers to a strategy design method for dynamically determining the optimal incentive amplitude according to the relationship between the historical behavior data of users and the behavior change under the incentive condition. In the present application, the core role of the mechanism is to measure the "behavior change degree brought by unit incentive intensity" (i.e. response elasticity) according to the historical response function of users to different incentive amplitudes, and match it with the current dispatch target of the power grid side (such as peak shaving or valley filling), to reversely deduce the most cost-effective incentive intensity value that can trigger user expected behavior under the current condition without causing waste of incentive resources. The implementation process of the mechanism includes three key links: first, based on historical data, a behavior response function model of each response level group user under different incentive amplitudes is constructed, which can be represented as a nonlinear function of charging behavior variables (such as access probability and power adjustment ratio) with respect to incentive amplitude; second, in combination with the load regulation demand of the current dispatch time period, the behavior change amount to be guided is calculated, and the target response value is set accordingly; finally, by inversely calculating the response function, the minimum incentive amplitude required to make the target response value true is solved, so as to determine the incentive intensity adapted to the response group in the current time period. This method makes the incentive strategy take into account the user behavior difference and the power grid target demand, effectively improves the cost-benefit ratio of incentive guidance, and is an important strategy for realizing precise regulation and optimal allocation of resources.

[0083] The "historical response function" refers to a mathematical function expression reflecting the relationship between incentive intensity and user behavior response, which is established based on the charging behavior data of electric vehicle users in different time periods and different incentive conditions in the past. Its core function is to quantify the sensitivity and response mode of user behavior to incentive changes, and to predict whether users will adjust their charging behavior and to what extent under a given incentive amplitude. Specifically, the independent variable of the historical response function is usually the incentive parameter (such as price discount, subsidy amount), and the dependent variable is a certain behavior variable (such as initiation charging probability, access time delay, power selection change, etc.). By collecting historical behavior data of users under multiple incentive scenarios, the response function model is trained using curve fitting, nonlinear regression or machine learning methods (such as gradient boosting tree, support vector regression, etc.). This function not only reveals the behavioral flexibility of users to incentive changes, but also serves as a basic tool for subsequent calculation of "incentive amplitude", "control strategy design" and "hierarchical response classification". Through the historical response function, more accurate prediction of future user behavior can be made, which is the key foundation for achieving fine-grained behavior guidance and efficient allocation of dispatch resources.

[0084] To further fine-tune user behavior control, configure corresponding charging access time period parameters for each response level group. The specific approach is as follows: First, analyze the time period load fluctuation curve and adjustable capacity distribution of the target regional power grid, combine with the user behavior prediction curve, and identify multiple target access time windows with sufficient low load or standby capacity. Then match these time windows with different response level groups, preferentially guide high-sensitivity groups with high stability and strong behavior consistency to critical peak shaving periods, and guide response-lagging users to buffer time bands. Such access time period parameters are set in interval form, combined with incentive information pushing and request guidance mechanisms, to provide input basis for subsequent path recommendation.

[0085] After setting the incentive amplitude parameters and access time period parameters, set the matching control intervention strategy parameters for each response level group. This parameter is used to define the real-time intervention means and restriction conditions that should be taken during the execution of charging dispatch, including load constraint boundary, response failure re-adjustment strategy, user behavior deviation warning mechanism, etc. The core of the control intervention strategy parameter is to ensure the safety and stability of dispatch execution. When some users do not behave according to the recommended path or the power grid load prediction deviates, the strategy can be triggered immediately to adjust the response, such as limiting subsequent access, temporarily adjusting the incentive scheme or redistributing the access time window, so as to maintain the dynamic balance of the overall dispatch structure.

[0086] The core role of this step is to build a set of fine-grained charging guidance and load control execution strategy system that can be based on user behavior differences and power grid dynamic state, in order to realize the unification of individual response control and load balancing goals in the charging scheduling process. This step is a key link that connects the previous user response grouping and priority identification behavior cognition basis, and directly outputs the operational incentive strategy table as the scheduling execution basis. By introducing the incentive amplitude parameter, the minimum effective incentive strength can be dynamically set according to the user behavior flexibility, thereby realizing the balance of resource conservation and guidance efficiency; by setting the charging access time period parameter, users with different response capabilities can be reasonably distributed to time windows with different load bearing capacities, avoiding strategy clustering, response congestion and secondary peak risk; and the control intervention strategy parameter provides security for the execution process, and once the user behavior deviates from the expected or the power grid load appears abnormal, the parameter can trigger the corresponding restriction, adjustment or emergency strategy compensation mechanism. The generation of distributed incentive scheduling table not only makes the incentive control have visual, configurable and dynamically updated operation structure, but also realizes the transformation from "global unified scheduling" to "group fine guidance", which is the core technical support for building an adaptive, flexible and high-response intelligent charging scheduling system.

[0087] S400, when the electric vehicle user initiates a charging connection request, according to the response level group to which it belongs and the control parameters in the distributed incentive scheduling table, push differential incentive information to guide the user to generate an actual charging access path sequence;

[0088] In order to realize the differential real-time guidance of electric vehicle users and avoid the problem of secondary load peak caused by the concentrated access of users to the power grid, a differential incentive pushing and path guidance method combining user response level group and distributed incentive scheduling table is proposed. This method triggers the behavior recognition and strategy matching mechanism at the key node of the user initiating the charging connection request, and finally guides the user to select the access path that best meets the power grid control target. The process includes the following steps:

[0089] When the electric vehicle user initiates a charging connection request, the corresponding response feature vector is automatically identified, and the response level group to which it belongs is retrieved. The level group is obtained by multidimensional clustering according to the user's historical incentive response behavior, indicating the user's response ability level when facing incentive information, such as high sensitivity, medium sensitivity, or low sensitivity group. After identification, the user's corresponding distributed incentive scheduling table content under the current scheduling period is extracted, which contains the incentive amplitude parameters, access time period parameters, and control strategy constraint parameters matching the response group, constituting the basis for individualized configuration of user control strategies.

[0090] According to the extracted scheduling parameters, the optimal incentive strategy is calculated under the current grid load state and user demand conditions. In the incentive strategy generation process, not only the incentive control priority of the user's level group is considered, but also the load prediction value, voltage margin, and power adjustable margin of the grid node in the current time period are combined to evaluate whether the target user can be guided to a low load period or a region with high available capacity. On this basis, the incentive cost-effectiveness of the user under different time windows and access node combinations is calculated, and according to the incentive response function in the user's historical behavior, the strategy combination that meets the minimum incentive cost and can achieve the maximum behavior adjustment amplitude is selected as the differentiated incentive push content.

[0091] After the generation of the incentive strategy, the differentiated incentive information is pushed to the user terminal in a parameterized structure. The incentive information not only contains the charging price discount, integral return, or reward subsidy value, but also explicitly specifies the recommended access time window range, suggested access location range, and corresponding effective time limit. To enhance the user's execution probability, the push information also combines the response success case prompts and dynamic risk prompt mechanisms of similar behavior users, such as informing the user that "accessing in the recommended time period will obtain XX points and reduce the queuing risk." At the same time, the strategy execution conditions are bound in the background, and once the user selects and confirms the recommended path and accesses, the behavior record and feedback monitoring process is automatically entered.

[0092] After the user receives the differentiated incentive information and autonomously selects the charging period and access path, the user's confirmed access path information is registered as the "path sequence" in the current scheduling period. The path sequence includes user identification, access time point, access location, and electric energy request amount, which serves as the input basis for subsequent grid-side load adjustment, capacity allocation, and dynamic prediction modeling. At the same time, the path sequence is also included in the user behavior database as a behavior execution record, providing actual execution feedback data for the next response feature identification and incentive strategy iteration. Through the above mechanism, a closed-loop behavior guidance chain from user grouping identification to differentiated incentive push to path decision support to actual path landing is formed, realizing the bidirectional collaborative optimization between scheduling targets and user autonomous selection.

[0093] The core role of this step is to achieve precise closed-loop conversion from user classification and identification to individual control execution. The user behavior characteristics identified in the early stage and the strategy optimization results are applied to the actual scheduling operation, thereby completing the coordinated interaction between electric vehicle users and the power grid. This process breaks the traditional "uniform rules, passive access" electric vehicle charging mode and adopts an active control mechanism of "response identification - hierarchical push - behavior guidance". It can accurately push incentive content according to the user's behavior response ability, and then guide users to charge at a reasonable time, in a reasonable place, and at a reasonable power level. This step not only makes the allocation of incentive resources more accurate and targeted, avoiding over-incentivization or resource waste, but also effectively disperses charging load and reduces local congestion and sudden increases in power grid pressure caused by the concentrated access of homogeneous users. At the system level, this step realizes the pre-control and capacity management support of the power grid load through the dynamic generation of user access path sequences, making the load distribution more uniform and the scheduling more controllable. At the user level, this differentiated guidance strategy improves individual participation and response rate while preserving user decision-making freedom, enhancing the acceptance and behavior stability of incentive execution. Therefore, this step is the key bridge to achieve the goal of "differentiating treatment and adapting to the situation" in intelligent charging scheduling, plays a core role between connection scheduling strategy and actual behavior, and is the execution carrier of strategy application landing and behavior transformation in the entire double-target optimization system.

[0094] S500, in the process of electric vehicle users performing charging tasks according to the access path sequence, collecting their charging behavior data and power grid load state data in the corresponding time period, and constructing a response feedback matrix between user path behavior and power grid response state;

[0095] To realize the dynamic correlation analysis between the behavior response effect of electric vehicle users and the actual operation state of the power grid, a response feedback evaluation mechanism is established to construct a response feedback matrix with user access path as the main line and power grid load fluctuation as the control. This matrix is not only used to evaluate the effect of differentiated incentive strategy guidance, but also provides real-time data support for subsequent incentive parameter correction and identification model optimization. This process includes the following steps:

[0096] In the process of charging task performed by electric vehicle users according to the recommended access path, the charging behavior data of the users is continuously recorded. The collected data includes but is not limited to charging initiation time, access location, charging duration, average power level, maximum power value, power demand completion rate, task interruption frequency and other key indicators. At the same time, in order to ensure the continuity and accuracy of data collection, a high-frequency sampling mechanism is used to periodically scan the behavior process, ensuring that the behavior change trajectory of the entire charging process is captured completely. In addition, in order to take into account the differences between users, all data records are bound to user identification and path number to form a clear behavior attribution relationship.

[0097] The power grid load state data in the time period corresponding to the user's charging task is synchronously collected. The data includes node load value, load fluctuation rate, voltage offset, current change, power factor, power grid frequency and other operating state information in the dispatching period. In particular, the real-time operating state of the user access location node and its surrounding area should be collected, and the time window is divided, which is strictly aligned with the user behavior data in time sequence, so as to realize "one-to-one" state correspondence. In order to improve the consistency and comparability of the data, the power grid state data is smoothed and the abnormal data is removed to eliminate the errors caused by external disturbance factors.

[0098] Based on the above two data sources, a two-dimensional response feedback matrix between user path behavior and power grid response state is constructed. The row vector of the matrix represents the behavior sequence of the user under different access paths, and the column vector represents various indicators of the power grid state in the time period of the path. Each matrix element reflects the actual feedback of the power grid operating state caused by a certain behavior selection, for example, the "high power - short period access" behavior may correspond to the "current surge - voltage drop" and other index responses. In order to enhance the expression ability of the matrix, normalization processing, behavior label mapping and load disturbance weight correction mechanism are introduced, so that the matrix not only has a visual behavior-load mapping relationship, but also can be used as an input basis for calculating comprehensive indicators such as guided deviation degree, strategy effectiveness and power grid stability contribution rate.

[0099] The constructed response feedback matrix is incorporated into the data feedback engine for subsequent incentive strategy evaluation and adaptive correction of the scheduling mechanism. The response feedback matrix can be used as a superimposed record of multiple charging behaviors. Through cross-cycle comparison and analysis of user behavior stability and consistency, and the long-term impact trend of the incentive scheme on power grid operation quality, the stability and consistency of user behavior and the long-term impact trend of the incentive scheme on power grid operation quality are identified. At the same time, by clustering and tracking the high deviation units in the matrix, phenomena such as "behavior deviating from the recommended path" and "abnormal load response" are identified, and feedback is given to the incentive parameter control mechanism to trigger iterative optimization of the incentive amplitude, access period and intervention strategy. Finally, a closed-loop evaluation mechanism is constructed with real behavior as the core and system response as the basis, realizing quantitative evaluation of incentive guidance effect, risk warning of power grid stability and continuous evolution of behavior identification model, and significantly improving the intelligence, adaptability and robustness level of the entire charging scheduling system.

[0100] The "data feedback engine" refers to a logic processing mechanism in the electric vehicle charging scheduling system, which is specially used for receiving, integrating and analyzing user charging behavior data and power grid operation state data. It is equivalent to the "feedback center" in the entire scheduling system, responsible for converting the actual operation results in the execution stage into the basis for strategy evaluation and model updating. In the present application, the role of the data feedback engine is reflected in three aspects: first, the behavior of the user after executing the recommended access path and the load response of the power grid are dynamically captured and aligned to form a response feedback matrix; second, the matrix is analyzed and the trend is judged to identify whether the incentive guidance achieves the expected goal, such as whether the user deviates from the path, whether the power grid has load abnormalities due to behavior concentration, etc.; third, these analysis results are fed back to the front end of the strategy control chain to trigger parameter correction of the incentive amplitude, access period or behavior identification model. Through this mechanism, a closed-loop control logic from "push - execution - perception - correction" is realized, ensuring that the incentive scheduling strategy is not only based on theoretical model optimization, but also dynamically adapts to actual operation results, enhances self-learning ability and environmental adaptability, thereby continuously improving the accuracy, stability and intelligence level of charging scheduling.

[0101] The core role of this step is to establish a data loop and causal mapping relationship between user behavior and actual operation effect of the power grid, to realize quantitative evaluation and adaptive optimization of scheduling strategy effect. In the previous steps, through response identification and incentive guidance, customized charging access paths are pushed to different users, but whether these strategies really work, whether they can stably achieve peak load shifting, and whether the power grid is under pressure can only be accurately judged through synchronous observation and linkage analysis of behavior and load. Therefore, this step synchronously collects behavior data (such as access time, power level, and duration) of users when they execute the access path and key state data (such as load change, voltage, current, frequency, etc.) of the power grid in the same time period, and builds a high-resolution response feedback matrix. Each cell of the matrix represents the specific response of a certain type of user behavior to the power grid state under a specific scheduling background. This mechanism not only identifies which behavior combinations are beneficial to the power grid and which behaviors are risky, but also reveals the degree of deviation of strategy guidance, i.e., whether users strictly follow the recommended path, whether their behavior deviates from the expected goal, and whether it triggers abnormal power grid operation. By continuously building and iteratively updating the feedback matrix, the effectiveness of incentive control can be dynamically judged, and problems such as strategy failure, user resistance, and group congestion can be discovered in a timely manner, providing detailed and reliable data basis for the next step of incentive parameter correction and behavior model update. Therefore, this step is the key link to realize "verifiable" behavior guidance, "traceable" power grid impact, and "evolvable" scheduling strategy, and is the core support for building a globally optimal, distributed intelligent, and closed-loop adaptive electric vehicle charging scheduling system.

[0102] S600, based on the analysis results of the response feedback matrix, identifying the deviation degree between user behavior response and the recommended access path, and the difference information between the power grid load state and the scheduling expectation in the corresponding period, dynamically correcting the incentive amplitude parameter, the access time period parameter, and the control intervention strategy parameter, updating the distributed incentive scheduling table, and synchronously optimizing the user behavior identification model, to realize closed-loop adaptive optimization control between electric vehicle user behavior control and power grid system operation;

[0103] To realize dynamic matching between electric vehicle user behavior control strategy and power grid operation state, a closed-loop adaptive optimization control based on response feedback matrix is constructed. Based on long-term accumulated user behavior data and power grid operation state data, behavior deviation in incentive strategy execution and load abnormality are identified, and strategy parameters and behavior models are corrected accordingly to realize dynamic optimization of control strategy. This process includes the following steps:

[0104] The constructed response feedback matrix is periodically analyzed. The response feedback matrix is composed of user behavior data after executing the recommended access path and power grid load state data in the time period. Each row of the matrix represents a charging path behavior sample, and each column represents a power grid operation state index corresponding thereto. In the analysis process, two aspects of data deviation are identified: one is the deviation degree between the actual behavior of the user and the recommended path, and the other is the difference between the actual load state of the power grid and the expected load change. Among them, the behavior deviation degree can be calculated by using quantitative indicators such as path similarity, execution rate, and time period deviation value; and the power grid load deviation can be expressed by using the residual mean and peak deviation between the actual load curve and the expected curve.

[0105] According to the above deviation identification result, the incentive strategy correction condition determination rule is established. When the user deviation degree exceeds the set threshold, or when the strategy fails for a long time on some access paths, even if the incentive is pushed, the user cannot be guided to access the target time period; or when the actual load of the power grid deviates from the expected trend for many times, secondary load concentration and equipment overload risk signals appear, the strategy correction mechanism is automatically triggered. At this time, the effectiveness and cost-effectiveness of the current incentive strategy need to be re-evaluated by comprehensively considering the elasticity model of user behavior response and the load response characteristics. In order to avoid excessive adjustment caused by abnormal individual cases, the behavior confidence interval and load disturbance filtering mechanism are introduced in the determination process to improve the stability and robustness of the strategy correction.

[0106] After the correction mechanism is triggered, the corresponding incentive amplitude parameter, access time period parameter and regulation intervention strategy parameter are dynamically adjusted according to the deviation degree and load difference level. Specifically, for the response level group with serious deviation behavior, the incentive amplitude can be increased to enhance the willingness to participate, or the access time period can be adjusted to a window with higher behavior acceptance. At the same time, for the area where load abnormalities frequently occur, the regulation intervention strategy can be strengthened, for example, limiting the access of low-response users in sensitive time periods, or releasing incentives to high-response users in advance to share the adjustment pressure. All the adjusted parameters are written into the distributed incentive scheduling table in real time to replace the original strategy item, ensuring that the regulation strategy always reflects the current running environment and actual behavior.

[0107] While updating the incentive scheduling table, the user behavior identification model is also optimized. The specific operation is to fuse the newly added user actual behavior data with the original training data, update the user response feature vector and behavior label weight, and retrain the identification model, so that the model can adapt to the change trend of user behavior over time. The optimized identification model will be used to update the behavior level grouping when the user initiates the next charging request, promoting the continuous co-evolution of the scheduling strategy and the actual behavior of the user.

[0108] The role of this step is to build a closed-loop optimization mechanism that can dynamically update the scheduling strategy and behavior model according to the actual behavior feedback of users and the operation state of the power grid, so as to realize the coordinated stability of behavior and electricity on both sides, with the ability of self-correction, continuous learning and real-time control. In actual operation, although the user has been given a personalized incentive strategy and access path in advance, and guided through response grouping and incentive pushing, the actual charging behavior may deviate from the recommended path due to factors such as personal preference, temporary travel demand, price sensitivity fluctuations, etc. At the same time, due to external environment, power load prediction error or user concentrated response, the operation state of the power grid may also deviate from the original scheduling expectation. Therefore, this step analyzes the response feedback matrix formed between the user path behavior and the power grid state data, identifies which paths and which user groups have sustained deviation behavior, and which scheduling time periods frequently appear load abnormalities, so as to judge the problems or ineffective links in the existing strategy. On the basis of identifying the problems, the corresponding incentive amplitude (such as appropriately increasing the subsidy intensity of the low response group), the access time period configuration (such as guiding some users to the time window with more stable load), and the adjustment of the control intervention strategy (such as shortening the response lag window, increasing the strategy response fault tolerance range, etc.) can be modified. More importantly, this step also simultaneously updates the behavior recognition model, incorporates new behavior data into feature training, so that the model adapts to the long-term evolution trend of user behavior. Through this closed-loop process, it changes from "prediction-driven" to "feedback-driven", realizes the dynamic coordination between behavior control, power grid safety and strategy resource utilization efficiency, and is the most intelligent, evolutionary and control depth key link in the whole incentive game scheduling method.

[0109] By the above-mentioned double-objective optimization method of electric vehicle charging scheduling based on incentive game, the individual behavior differences of users can be deeply identified and guided, thereby effectively solving the user strategy convergence and secondary load peak problems caused by the existing unified incentive strategy. The scheme constructs a user response feature vector and performs response level division, so that the system can dynamically allocate incentive resources according to user sensitivity, and form a differentiated and distributed scheduling strategy combined with the grid operation state. At the same time, the response feedback matrix is constructed by introducing real-time behavior feedback and grid load data, so that the strategy execution effect is visual and quantifiable, and on this basis, the adaptive iteration optimization of incentive parameters and behavior model is realized. Finally, a closed-loop control mechanism with “identification-guidance-feedback-correction” as the core is formed, which takes into account the individual charging utility of electric vehicle users and the overall operation safety of the power system, effectively improves the response efficiency of incentive investment, the dynamic adaptability of scheduling strategy and the load balancing of grid operation, and significantly enhances the intelligent level and robustness of large-scale electric vehicle charging scheduling system.

[0110] The above merely describes certain exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be construed as limiting the scope of the claims of the present application.

[0111] It should be noted that in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without more limitations, an element preceded by "comprises... a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0112] It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0113] Those skilled in the art can clearly understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0114] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0115] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment according to actual needs.

[0116] In addition, the functional units in each embodiment of the present application can be integrated into one processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit.

[0117] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0118] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without deviating from the spirit and scope of the present application. Therefore, the above figures and description are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. A double-objective optimization method for electric vehicle charging scheduling based on incentive game, characterized in that, The method comprises the following steps: S100, constructing a user behavior dynamic response identification model, obtaining historical charging behavior data and incentive trigger time nodes, extracting user response changes under different incentive conditions, and generating a response feature vector; S200, grouping users based on the response feature vector to form multiple response level groups with different incentive sensitivity gradients, and setting corresponding incentive control priorities; S300, based on the response level groups and the incentive control priorities, combining the response features of each response level group and the current power grid operating state to generate a distributed incentive scheduling table; S400, when a user initiates a charging connection request, pushing differentiated incentive information according to the response level group and the parameters in the distributed incentive scheduling table to guide the user to generate an access path sequence; S500, in the process of the user performing the charging task according to the access path sequence, collecting the charging behavior data and the power grid load state data in the corresponding period to construct a feedback matrix between the user path behavior and the power grid response state; S600, based on the feedback matrix, identifying the deviation degree of the user behavior and the recommended path, and the difference between the power grid state and the scheduling expectation, dynamically correcting the incentive amplitude, the access period and the control strategy parameters, updating the scheduling table, and synchronously optimizing the user behavior identification model. 2.The double-objective optimization method for electric vehicle charging scheduling based on incentive game according to claim 1, wherein, Step S100 comprises: Obtaining historical charging behavior data and incentive trigger time nodes of electric vehicle users, and constructing a behavior-incentive interaction data pair sequence; Extracting index data of behavior changes before and after the user is incentivized, including behavior fluctuation index, relative difference ratio and incentive action window time delay; Based on the extracted index data, a response feature vector is constructed, normalized coding is adopted, and a covariance weight adjustment mechanism is introduced to enhance the feature discrimination; In the subsequent charging operation, the values of each dimension of the response feature vector are continuously updated to realize dynamic correction and adaptive learning of the user behavior model. 3.The double-objective optimization method for electric vehicle charging scheduling based on incentive game according to claim 2, wherein, The covariance weight adjustment mechanism comprises: Calculate the covariance matrix between each dimension of the response feature vector of multiple users to determine the linear correlation degree between each feature dimension; Assign high weight to the feature dimension with a covariance value below a set threshold to enhance the contribution of the dimension in user behavior pattern differentiation; Assign low weight to the feature dimension with a covariance value above a set threshold to suppress its interference in user behavior feature identification; Reconstruct the feature vector after weight adjustment to improve the clustering clarity and grouping effectiveness of the user response feature in the feature space. 4.The double-objective optimization method of electric vehicle charging scheduling based on incentive game according to claim 1, wherein, Step S200 comprises: Performing unsupervised clustering analysis on the weighted user response feature vector to form multiple behavior response categories according to the feature space distribution; Calculate the average response feature vector of each behavior response category to determine its incentive sensitivity index; Divide into multiple response level groups according to the incentive sensitivity index; Set the incentive control priority for each response level group for subsequent incentive resource allocation and strategy pushing timing configuration.

5. The double-objective optimization method for incentive game-based electric vehicle charging scheduling according to claim 1, wherein, Step S300 comprises: Based on the historical response features of the response level groups and the current power grid operating state, a scheduling scenario identification data set is constructed; According to the scheduling scene recognition dataset, the user historical response function and the target scheduling demand are combined to determine the incentive amplitude parameters of each response level group; The target area power grid load fluctuation curve is analyzed, and the charging access time period parameters of each response level group are set; The corresponding control intervention strategy parameters are configured to form a distributed incentive scheduling table containing incentive amplitude, access time period and intervention strategy.

6. The double-objective optimization method for incentive game-based electric vehicle charging scheduling according to claim 1, wherein, Step S400 includes: When the electric vehicle user initiates a charging connection request, identify the corresponding response level group and extract the corresponding distributed incentive scheduling table parameters; According to the distributed incentive scheduling table parameters and the current power grid load state, a differentiated incentive strategy with minimum incentive cost and maximum behavior adjustment amplitude is generated; The incentive information containing the incentive amplitude, access time window, access location and effective time limit is pushed to the user terminal; After the user selects, register the access path sequence, and record the current path as the behavior execution record for subsequent behavior feedback and strategy iteration analysis.

7. The double-objective optimization method for incentive game-based electric vehicle charging scheduling according to claim 1, wherein, Step S500 includes: Obtain the charging behavior data of electric vehicle users during the execution of the access path sequence, and bind the user identifier and path number; Collect the operating state data of the corresponding power grid node and its adjacent area within the user charging period, and perform time alignment and abnormal elimination processing; Build a response feedback matrix with user path behavior as row vector and power grid operating state as column vector; Input the response feedback matrix into the data feedback engine for incentive strategy evaluation and control parameter update analysis. 8.The double-objective optimization method of electric vehicle charging scheduling based on incentive game according to claim 1, wherein, Step S600 includes: Periodically analyze the response feedback matrix to identify the deviation between user behavior response and recommended access path, as well as the difference between power grid load state and scheduling expectation; According to the identified deviation and difference information, determine the incentive strategy correction condition; When the correction condition is met, dynamically adjust the incentive amplitude parameters, access time period parameters and control intervention strategy parameters, and update the distributed incentive scheduling table; Fuse the new user behavior data with the original training data, optimize the user behavior recognition model, and update the response level grouping result.

Images