Method and device, system and storage medium for evaluating response potential of electric vehicle dispatching

Abstract

The application discloses an electric vehicle scheduling response potential evaluation method and device, system and storage medium, constructs a power grid-vehicle semantic matching mechanism, simultaneously introduces an adaptive membership parameter, establishes an individual response potential fuzzy inference model, and finally statistically evaluates the cluster potential, so that the scheduling matching precision and decision reliability are improved. The application solves the problems in the prior art, such as insufficient data feature fusion capability, low power grid demand and vehicle state matching precision, fuzzy inference parameter dependence on manual setting, and imperfect cluster potential evaluation capability quantization.

Description

Technical Field

[0001] This invention belongs to the field of electric vehicle dispatching technology, specifically relating to a method, device, system, and storage medium for evaluating the dispatching response potential of electric vehicles. Background Technology

[0002] Against the backdrop of accelerated construction driven by "dual-carbon" goals and the demand for new power systems, the new energy vehicle industry and technologies such as vehicle-to-grid (V2G) interaction are converging. Electric vehicle clusters have gradually become an important manifestation of large-scale, wide-area, and relatively flexible distributed resources with strong regulation capabilities in demand-side response, peak shaving and valley filling, and renewable energy grid connection. However, the inherent heterogeneity and dynamism of the operating status of individual electric vehicles lead to the diversification and complexity of data, posing challenges to grid dispatch data modeling and increasing the difficulty of grid dispatch. On the other hand, grid-side dispatch requires data to be expressed in structured and semi-structured forms, while vehicle-side data is often heterogeneous, with differences in syntactic and semantic spaces, making precise matching and coordination difficult. Existing cluster response potential assessment methods are mostly based on rule-based or simple statistical methods, such as the constant-weight feature synthesis method, the empirical rule-based cluster dispatch priority estimation method, and the manual fuzzy inference method based on fixed membership degrees. These methods can basically meet control requirements, but they have poor adaptability and low complexity in actual operation, making it difficult to achieve rapid and accurate assessment of the response potential of electric vehicle clusters.

[0003] Existing research primarily assesses the response potential of electric vehicle (EV) clusters from a local perspective, exhibiting significant shortcomings: In processing multi-feature data of EVs, a unified modeling approach is lacking, failing to comprehensively consider and effectively represent multi-dimensional information, thus affecting the accuracy of EV scheduling and matching; in matching EVs with the power grid, a unified semantic space is lacking, hindering effective coordination between scheduling demands and EV states, leading to unreasonable allocation of scheduling priorities; for fuzzy reasoning, changes in membership function parameters rely on subjective experience, resulting in high subjectivity, blindness, and poor adaptability, leading to inaccurate response evaluations; quantitative models for individual response potential fail to fully reflect data uncertainty and complex nonlinearity; and cluster evaluation fails to consider both consistency and reliability, lacking a scientifically unified objective index system for evaluating EV response potential, making practical application difficult. Summary of the Invention

[0004] To address the problems existing in current technologies, this invention provides a method, apparatus, system, and storage medium for assessing the dispatch response potential of electric vehicles (EVs). It solves issues such as insufficient data feature fusion capabilities, low accuracy in matching grid demand with vehicle status, reliance on manually set fuzzy inference parameters, and incomplete quantification of cluster potential assessment capabilities in existing EV cluster potential assessments. Through multi-dimensional feature coupling, semantic mapping of features from different domains, adaptive fuzzy inference, and aggregation of cluster response characteristics, it can accurately assess the response potential of each EV and reliably measure the EV cluster response capability, thereby improving the intelligence level of power network management and enhancing the optimization efficiency of resource utilization.

[0005] To achieve the above objectives, the present invention provides the following solution: A method for evaluating the dispatch response potential of electric vehicles includes: Step S1: Fusion of multi-dimensional operational characteristics of electric vehicles; Step S2: Construct a cross-semantic mapping model of power grid demand and vehicle status based on the fused features to obtain the matching degree between electric vehicles and the power grid; Step S3: Perform FCM clustering analysis based on the matching degree between electric vehicles and the power grid to obtain the centroids of the three clusters and the membership degrees of the fuzzy sets; Step S4: Based on the cluster centroid and the membership degree of the fuzzy set, the feature membership degree is mapped to the individual potential score to achieve nonlinear quantification of the response capability of EV individuals; Step S5: Based on the individual potential scores and the upper limit of the output power of each electric vehicle, obtain the total schedulable power of the cluster and the response confidence, and construct a comprehensive response capability evaluation index.

[0006] As a preferred option, in step S1, the battery status, operating constraints, and user intentions of the electric vehicle are first constructed into a standardized input matrix, and a query vector and a key vector are generated through the projection matrix; then the correlation weights between features are calculated, and the original information is used to complete the feature weighted fusion.

[0007] Preferably, in step S2, a demand-state matching mechanism is established using the CLIP semantic alignment model; a multilayer perceptron (MLP) network is trained using a manually labeled dataset, and the demand-state matching mechanism uses automatically generated semantic matching scores.

[0008] As a preferred option, in step S3, the FCM clustering algorithm is used to automatically classify the matching degree samples and iteratively calculate the cluster centroids. At the same time, fuzzy reasoning is used to quantify the scattered and random electric vehicle resources into overall flexible adjustment resources, and adaptively determine the fuzzy set and membership degree.

[0009] As a preferred option, in step S4, a potential mapping mechanism is introduced to transform the results of FCM clustering into fuzzy rules, thereby realizing intelligent interaction from cluster feature values ​​to electric vehicle response potential scores. This includes: fuzzy width calculation, cluster feature membership determination, and individual potential score mapping.

[0010] The present invention also provides an electric vehicle dispatch response potential assessment device, comprising: The first processing module is used to fuse the multi-dimensional operating characteristics of electric vehicles; The second processing module is used to construct a cross-semantic mapping model of power grid demand and vehicle status based on the fused features, so as to obtain the matching degree between electric vehicles and the power grid. The third processing module is used to perform FCM clustering analysis based on the matching degree between electric vehicles and the power grid, and obtain the centroids of the three clusters and the membership degree of the fuzzy set; The fourth processing module is used to map individual potential scores based on the feature membership degree obtained from the cluster centroid and fuzzy set membership degree, thereby realizing nonlinear quantification of the response capability of EV individuals. The fifth processing module is used to obtain the total schedulable power of the cluster and the response confidence based on the individual potential scores and the upper limit of the output power of each electric vehicle, and to construct a comprehensive response capability evaluation index.

[0011] The present invention also provides an electric vehicle dispatch response potential assessment system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an electric vehicle dispatch response potential assessment method when executed by the processor.

[0012] The present invention also provides a storage medium storing a computer program that executes an electric vehicle dispatch response potential assessment method when running.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) A self-attention mechanism is adopted to realize the coupling of electric vehicle data features and enhance privacy. A multi-dimensional data matrix covering battery status, operating constraints and user preferences is constructed. The correlation weights among the features are calculated by self-attention to complete the weighted feature coupling, thereby improving the data representation capability and privacy protection level.

[0014] (2) Construct a semantic matching mechanism between the power grid and vehicles to calculate the matching degree between power grid dispatching requirements and individual state vectors of electric vehicles. By training a semantic alignment network on manually labeled data, the power grid commands and vehicle states are accurately aligned in a unified semantic space, and the matching degree between the two is calculated.

[0015] (3) An adaptive learning method for membership parameters based on FCM clustering is proposed to achieve data-driven generation of Gaussian membership functions. This invention relies on the distribution of historical samples to carry out fuzzy clustering, which effectively avoids the subjective bias and complexity caused by manually setting parameters, thereby improving the generalization ability of the model in dynamic scenarios and enhancing the accuracy and adaptability of the fuzzy inference process.

[0016] (4) Construct a refined evaluation model for the individual response potential of EVs to achieve nonlinear mapping of response capabilities. By analyzing the clustering results and calculating the fuzzy width, a fuzzy inference model is constructed to carry out refined quantification of individual response potential, providing basic data support for subsequent cluster potential calculation.

[0017] (5) Based on the evaluation results of individual EVs, a comprehensive quantitative system for cluster response capability is constructed. By integrating the response potential of individual EVs and introducing the total power of the EV cluster and the response confidence, a quantitative method for group response capability is established, thereby improving the scientific rigor and reliability of power system dispatch decision-making. Attached Figure Description

[0018] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of the electric vehicle dispatch response potential assessment method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of multidimensional feature data fusion based on the self-attention mechanism. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1 like Figure 1As shown, this invention provides a method for evaluating the dispatch response potential of electric vehicles. Based on cross-modal semantic alignment and data-driven fuzzy reasoning, it achieves in-depth mining and precise profiling of the response potential of a single electric vehicle, and completes a comprehensive quantitative evaluation of the cluster response scale. Specifically, it includes: Step S1: To achieve the fusion of multi-dimensional operational features of electric vehicles and the elimination of intuitive and sensitive information, a unified modeling and association mining of multi-feature data is performed based on a self-attention mechanism. First, the original information of the electric vehicle, such as battery status, operational constraints, and user intentions, is constructed into a standardized input matrix. A query vector and key vector are generated through a projection matrix. Then, the association weights between features are calculated, and the original information is used to complete the weighted fusion of features, thereby achieving feature coupling while ensuring the integrity of the original data. In the modeling of the response characteristics of electric vehicle clusters in power systems, this invention can improve data representation capabilities while masking original data information, thus enhancing data privacy. This process includes four key steps: original data integration, feature vector mapping and correlation calculation, and weighted feature coupling. Figure 2 As shown. The specific modeling process is as follows: (1) Collect and organize individual electric vehicle operation data and construct an input matrix. This matrix is ​​used for subsequent self-attention mechanisms. It includes the EV's battery SOC state, battery temperature, maximum charge / discharge power, and the user's corresponding preferences. (1) in, For the first The state of charge of the battery of an EV; Battery temperature; and These are the upper limits for battery charging and discharging power, respectively. The probability of the user's corresponding intention; Representation matrix The Middle j These five features together determine the response potential of a single EV.

[0023] (2) Based on the matrix obtained in (1) The self-attention mechanism is used to perform related processing. This is achieved through the projection matrix. Each feature Mapped to , Vector pairs, representing the query matrix and the key matrix respectively: (2) in, The projection matrix; , The two work together to support subsequent relevance calculations: the query matrix is ​​used to initiate feature association queries, and the key matrix is ​​used to match related features.

[0024] (3) Based on (2) , Vectors are used to calculate the feature association weight matrix. First, the query matrix... AND key matrix Perform a dot product operation and adjust the results using a scaling factor to aggregate them into an attention score matrix. Subsequently, using The function normalizes the matrix, thus obtaining the association weight matrix of all five-dimensional features for each feature. To achieve multidimensional feature coupling: (3) (4) (5) (6) in, For the first The first in the EV data The feature is related to the first The greater the weight of the correlation of each feature, the greater the contribution of that modality to the current feature extraction. This is a scaling factor used to mitigate the vanishing gradient problem caused by high-dimensional features and ensure the stability of weight calculation. Since the feature sequence is five-dimensional, therefore... ; for Statistical matrix; These are the weights after normalization. For the activation function, on the matrix Normalization is performed to obtain the matrix. The sum of all its elements is 1, thus achieving a balance between weights.

[0025] (4) The correlation weight matrix obtained from (3) This achieves feature coupling. (Using...) Direct weighted raw data By summing the results, feature coupling was achieved while ensuring information integrity, resulting in a coupling matrix that integrates modal correlation information. This achieves the enhancement of core features.

[0026] (7) (8) in, The integrated feature matrix; For raw data Data after feature coupling.

[0027] This invention achieves data feature association modeling while avoiding direct exposure of raw sensitive data, thus improving privacy protection during data processing. At the same time, through the coupling between data, it breaks through the shallow semantic expression of a single feature, further enhancing data representation capabilities.

[0028] Step S2: To protect the privacy of power grid data and achieve accurate matching between power grid dispatching needs and the individual states of electric vehicles, this invention introduces a service provider mechanism to prevent direct contact between power grid data and user data. This mechanism utilizes the CLIP semantic alignment model to establish a demand-state matching mechanism. By training a multilayer perceptron (MLP) network using a manually labeled dataset, this mechanism can automatically generate semantic matching scores. Its core comprises four key stages: construction of the power grid demand vector, matrix normalization, MLP network training, and calculation of the semantic matching score for the current data. The specific modeling process is as follows: (1) Constructing the power grid demand vector This vector collects the core key parameters of 5-dimensional power grid dispatching requirements, providing comprehensive demand information support for subsequent semantic projection and matching calculations.

[0029] (9) in, for t The real-time grid demand response power, which is the total charging and discharging power that the grid currently needs to provide from the EV cluster, is the core indicator of dispatch demand and directly determines the scale of cluster response. for t The response time limit, which is the maximum time required by the power grid for the EV cluster to complete a response from receiving the instruction, determines the urgency of the response: the shorter the time limit, the higher the requirement for response speed. for t The types of real-time demand response mainly include three core scenarios: peak shaving, valley filling, and frequency regulation. Real-time electricity prices serve as an economic incentive factor influencing EV users' willingness to respond. for t The priority coefficient for real-time response is dynamically set by the power grid dispatch center based on the power grid operating status.

[0030] (2) For the power grid demand vector obtained in (1) Compared with the EV individual feature matrix obtained in step one First use The function restricts the value to the range [0,1]. (10) in, and The normalized EV individual feature matrix Matrix and Power Grid Demand Vector Then, based on their matching degree, they are manually labeled and given matching degree tags. Its value range is .when When the value is close to 1, it indicates that the individual EV state is highly matched with the grid demand, and the EV has strong response adaptability; when When the value is close to 0.5, it indicates that the matching degree between the two is average, and the EV can be used as a backup response resource; when A value close to 0 indicates extremely poor compatibility between the two, meaning the EV is unsuitable for the current scheduling task. Therefore, a training set needs to be built within the service provider. .

[0031] (3) The training set obtained from (2) D Establish a structurally symmetric but parameter-independent architecture: Network and The network, whose inputs are the feature matrices. With grid demand vector The outputs are the vehicle state projection vectors. Projection vector of grid demand .

[0032] (11) in, A dedicated projection network function for vehicle state characteristics; Dedicated projection network functions for power grid demand characteristics. , Maintaining the same dimensions as the original matrix ensures the feasibility and accuracy of subsequent matching degree calculations. Manually labeled values. The semantic correlation between EV and grid demand is defined abstractly, and therefore cosine similarity is used to simulate this relationship during model training, thus effectively measuring the directional consistency of high-dimensional vectors. (12) in, To output the matching score for the model, cosine similarity is used, which essentially measures the angle between two vectors; the smaller the angle, the higher the matching score. MSELoss is used as the unified loss function to minimize the deviation between the model's predicted similarity and the true label. (13) in, L For loss function, For the number of historical EVs, This represents the number of time periods selected. Finally, parameter iteration is performed, and the updated formula is: (14) (15) (16) (17) in, and They are respectively and Network parameters, k For update count, The learning rate is set manually. This represents updating the parameters of the MLP network.

[0033] (4) Integrating (1)-(3), the EV individual feature matrix can be directly input. With grid demand vector The matching degree is output by MLP network and cosine similarity. .

[0034] Step S3: To address the issue of the discrete and uncontrollable nature of electric vehicle (EV) cluster resources in power grid dispatch, based on the EV-grid matching degree obtained in Step S2, fuzzy inference is used to quantify the dispersed and random EV resources into overall flexible adjustment resources. Traditional evaluation methods are prone to dispatching bias and resource waste; therefore, the FCM clustering algorithm is used to automatically classify matching degree samples and iteratively calculate cluster centroids, thereby adaptively determining the fuzzy sets and membership degrees in the fuzzy inference system. This avoids the problems of strong subjectivity and insufficient precision in manual settings, providing standardized parameter support for subsequent fuzzy rule modeling and response reliability calculation. The specific modeling process is as follows: (1) Based on the semantic matching degree output in Step 2 Constructing a historical sample set This provides initial data support for FCM clustering, making the initialization parameters of the neural fuzzy system more closely match the actual operating scenario, thereby improving the accuracy of response potential evaluation. (18) in, This represents the historical sample set.

[0035] (2) For the historical sample set obtained in (1) The maximum, minimum, and mean of the samples are selected as the initial cluster centroids: (19) in, These are the initial centroids of clusters 1, 2, and 3, respectively.

[0036] (3) Based on the initial cluster centroid obtained in (2) Iterative calculation of the sample set Each matching degree sample The membership weights corresponding to each fuzzy cluster are: (20) in, Indicates the first n After the nth iteration, the th... l The centroid of the cluster, n=0 Time corresponds to the initial centroid ; a For the summation index, and l Same meaning; It is the Euclidean norm, used to measure the distance from a sample to the cluster centroid; Indicates the first n Samples in round iteration Belongs to the The membership degree of a fuzzy set satisfies and ; The fuzzy weighting index is usually taken as... , m The larger the cluster size, the more blurred the cluster boundaries become.

[0037] (4) Based on the membership degree obtained in (3), recalculate the centroid position of each fuzzy cluster: (twenty one) in, Represents the updated version For each cluster, the higher the membership degree of the sample, the greater its influence on the centroid's location. This represents the summation of all samples used within the set.

[0038] (5) Compare the deviation of the cluster centroids before and after iterations (3) and (4) to determine whether FCM clustering has converged: (twenty two) in, This represents the absolute deviation of the cluster centroid before and after the (n+1)th iteration; This is the preset convergence threshold. If If so, return to step (3) and continue iterating; if The clustering iteration terminates.

[0039] (6) Integrate the iterative convergence results of (3)-(5) and output the final clustering parameters of FCM: (twenty three) (twenty four) in, Represents the final cluster centroid, when l =1 ,l =2 and l When the value is 3, they represent the centroids of the best, medium, and worst matching clusters, respectively. Representative sample The corresponding final fuzzy set membership degree.

[0040] Step S4: To achieve accurate and nonlinear quantitative characterization of the individual response capabilities of electric vehicles, FCM clustering and fuzzy inference are employed, and an adaptive fuzzification and potential mapping for service items suitable for electric vehicles are proposed. A potential mapping mechanism is introduced to transform the results of FCM clustering into a series of fuzzy rules, enabling intelligent interaction from cluster feature values ​​to electric vehicle response potential scores. This includes fuzzy width calculation, cluster feature membership determination, and individual potential score mapping. For power dispatching and cluster response decisions, this mechanism can effectively address the uncertainty and fuzziness of the operational status data itself, and accurately describe the real-time response potential of a single electric vehicle, providing a basis for subsequent aggregated response power and reliability calculations. The specific modeling process is as follows: (1) Based on the final cluster centroid obtained in step S3 Membership degree with fuzzy set The fuzzy width of the Gaussian membership function is calculated. The cluster centers and membership distributions obtained through FCM clustering accurately match the actual data distribution, thus achieving data-driven adaptive fuzzification. Each cluster corresponds to a fuzzy subset, and the fuzzy width is determined using the cluster weighted standard deviation. (25) in, Represents all samples For the l The sum of the membership degrees of a fuzzy set; Representing the l The width of a fuzzy subset is used to characterize the coverage of the fuzzy set. The larger the width, the smoother the change in the membership function, and the wider the corresponding fuzzy coverage interval.

[0041] (2) Calculate the feature membership degree based on the fuzzy width obtained in (1): (26) in, Eigenvalues Belongs to the The membership degree of each fuzzy subset ranges from [0,1]. The closer the membership degree is to 1, the more the feature value conforms to the concept of the corresponding fuzzy subset; the closer the membership degree is to 0, the less the feature value conforms to the concept of the corresponding fuzzy subset. Through fuzzification, quantitative feature values ​​are transformed into qualitative fuzzy evaluations, realizing the combination of logic and quantitative data.

[0042] (3) Based on the feature membership degree obtained in (2) Mapping individual potential scores enables nonlinear quantification of an individual's EV response capability: (27) in, For the first A mapping function for fuzzy rules; It is a logical score, representing the first... i EVs in t The higher the score, the stronger the EV's responsiveness and the more response power it can handle, representing the percentage of its maximum output.

[0043] Step S5: To comprehensively quantify and evaluate the overall response capability and reliability of the electric vehicle cluster, the cluster response power and response confidence are calculated using the potential scores of individual electric vehicles. In the scenario of power system dispatching and resource optimization, this invention can provide accurate and reliable quantitative basis for upper-level decision-making. The specific modeling process is as follows: (1) Based on the individual potential score obtained in step S4 and the upper limit of output power for each EV The total power of the cluster response is obtained as follows: (28) in, for t The total power that the EV cluster can schedule at any given time.

[0044] (2) Individual potential scores obtained from Step 4 Build cluster response confidence to determine whether the cluster has the ability to safely execute scheduling instructions: (29) (30) in, for t Mean of individual potential score at any given moment; The overall confidence level of the EV cluster is denoted by [0,1]. A larger value indicates higher cluster response consistency, stronger stability, and lower scheduling execution risk.

[0045] (3) Integrate the total schedulable power of the cluster and the response confidence obtained from (2) and (3) to construct a comprehensive response capability evaluation index: (31) in, for t The comprehensive evaluation index of the EV cluster's response capability is determined by the total adjustable power of the EV cluster. and response confidence It consists of two parts. The table represents the adjustment capacity that the cluster can provide at the current moment. The two are used to measure the reliability and stability of the response results, and together they form a complete quantitative system for response potential, providing a reliable decision-making basis for power grid optimization and dispatch.

[0046] This invention aims to accurately model and reliably evaluate the response capabilities of EV clusters in complex, multi-feature data environments and dynamic, uncertain scheduling scenarios. Compared with traditional methods, the main innovative contributions of this invention include: First, by leveraging a self-attention mechanism, it achieves feature coupling of multi-dimensional information such as battery status, operational constraints, and user preferences, improving data representation capabilities while strengthening the protection of individual data privacy; Second, it introduces a semantic alignment network to map grid demand and vehicle status into a unified semantic space, generating semantic matching degrees and providing interpretable correlation metrics for subsequent inference; Third, it uses FCM clustering to automatically generate Gaussian membership function parameters, reducing subjectivity and uncertainty introduced by manual configuration and improving the adaptability of fuzzy models in dynamic scenarios; Fourth, it establishes a single-vehicle response potential mapping mechanism based on fuzzy inference, enabling nonlinear modeling and refined quantitative evaluation of electric vehicle response capabilities; Fifth, by aggregating the response potential of individual EVs, it proposes a comprehensive quantitative approach for cluster response capabilities, providing a reliable basis for grid scheduling.

[0047] Example 2 The present invention also provides an electric vehicle dispatch response potential assessment device, comprising: The first processing module is used to fuse the multi-dimensional operating characteristics of electric vehicles; The second processing module is used to construct a cross-semantic mapping model of power grid demand and vehicle status based on the fused features, so as to obtain the matching degree between electric vehicles and the power grid. The third processing module is used to perform FCM clustering analysis based on the matching degree between electric vehicles and the power grid, and obtain the centroids of the three clusters and the membership degree of the fuzzy set. The fourth processing module is used to map individual potential scores based on the feature membership degree obtained from the cluster centroid and fuzzy set membership degree, thereby realizing nonlinear quantification of the response capability of EV individuals. The fifth processing module is used to obtain the total schedulable power of the cluster and the response confidence based on the individual potential scores and the upper limit of the output power of each electric vehicle, and to construct a comprehensive response capability evaluation index.

[0048] Example 3 The present invention also provides an electric vehicle dispatch response potential assessment system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an electric vehicle dispatch response potential assessment method when executed by the processor.

[0049] Example 4 The present invention also provides a storage medium storing a computer program that executes an electric vehicle dispatch response potential assessment method when running.

[0050] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for evaluating the dispatch response potential of electric vehicles, characterized in that, include: Step S1: Fusion of multi-dimensional operational characteristics of electric vehicles; Step S2: Construct a cross-semantic mapping model of power grid demand and vehicle status based on the fused features to obtain the matching degree between electric vehicles and the power grid; Step S3: Perform FCM clustering analysis based on the matching degree between electric vehicles and the power grid to obtain the centroids of the three clusters and the membership degrees of the fuzzy sets; Step S4: Based on the cluster centroid and the membership degree of the fuzzy set, the feature membership degree is mapped to the individual potential score to achieve nonlinear quantification of the response capability of EV individuals; Step S5: Based on the individual potential scores and the upper limit of the output power of each electric vehicle, obtain the total schedulable power of the cluster and the response confidence, and construct a comprehensive response capability evaluation index.

2. The method for evaluating the dispatch response potential of electric vehicles as described in claim 1, characterized in that, In step S1, the battery status, operating constraints, and user intentions of the electric vehicle are first constructed into a standardized input matrix, and the query vector and key vector are generated through the projection matrix. Then, the correlation weights between features are calculated, and the original information is used to complete the feature weighted fusion.

3. The method for evaluating the dispatch response potential of electric vehicles as described in claim 2, characterized in that, In step S2, a demand-state matching mechanism is established using the CLIP semantic alignment model; a multilayer perceptron (MLP) network is trained using a manually labeled dataset, and the demand-state matching mechanism uses automatically generated semantic matching scores.

4. The method for evaluating the dispatch response potential of electric vehicles as described in claim 3, characterized in that, In step S3, the FCM clustering algorithm is used to automatically classify the matching degree samples and iteratively calculate the cluster centroids. At the same time, fuzzy reasoning is used to quantify the scattered and random electric vehicle resources into overall flexible adjustment resources, and adaptively determine the fuzzy set and membership degree.

5. The method for evaluating the dispatch response potential of electric vehicles as described in claim 4, characterized in that, In step S4, a potential mapping mechanism is introduced to transform the results of FCM clustering into fuzzy rules, realizing intelligent interaction from cluster feature values ​​to electric vehicle response potential scores. This includes: fuzzy width calculation, cluster feature membership determination, and individual potential score mapping.

6. An electric vehicle dispatch response potential assessment device, characterized in that, include: The first processing module is used to fuse the multi-dimensional operating characteristics of electric vehicles; The second processing module is used to construct a cross-semantic mapping model of power grid demand and vehicle status based on the fused features, so as to obtain the matching degree between electric vehicles and the power grid. The third processing module is used to perform FCM clustering analysis based on the matching degree between electric vehicles and the power grid, and obtain the centroids of the three clusters and the membership degree of the fuzzy set. The fourth processing module is used to map individual potential scores based on the feature membership degree obtained from the cluster centroid and fuzzy set membership degree, thereby realizing nonlinear quantification of the response capability of EV individuals. The fifth processing module is used to obtain the total schedulable power of the cluster and the response confidence based on the individual potential scores and the upper limit of the output power of each electric vehicle, and to construct a comprehensive response capability evaluation index.

7. An electric vehicle dispatch response potential assessment system, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executed by the processor, the computer program executing the electric vehicle dispatch response potential assessment method as described in any one of claims 1-5 when run by the processor.

8. A storage medium, characterized in that, The storage medium stores a computer program that, when executed, performs the electric vehicle dispatch response potential assessment method as described in any one of claims 1-5.

Images