User demand response potential evaluation method and device based on GMM clustering and information entropy

Abstract

The invention provides a user demand response potential evaluation method and device based on GMM clustering and information entropy, and relates to the technical field of smart power grids. The method comprises the following steps: constructing a charging behavior original feature set of an electric vehicle user based on a time flexible index and a power adjustable index; clustering the charging behavior original feature set based on an SOM model and a GMM clustering algorithm to obtain multiple user portraits; wherein each user portrait corresponds to one type of electric vehicle users; and for each user portrait, calculating the closeness between the charging behavior characteristics of the electric vehicle user corresponding to the user portrait and the ideal scheme through a TOPSIS method, and obtaining the demand response potential of the electric vehicle user corresponding to the user portrait. According to the invention, quantitative evaluation of the demand response potential of the electric vehicle user can be realized, and a basis is provided for demand response scheduling.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method and apparatus for assessing user demand response potential based on GMM clustering and information entropy. Background Technology

[0002] Electric vehicles have not only promoted the deep integration of energy decarbonization and transportation electrification, but also provided key support for my country to achieve its sustainable development goals.

[0003] Currently, despite the rapid popularization of electric vehicles and significant progress in the construction of charging infrastructure, many technical challenges remain in the accuracy of electric vehicle charging load forecasting and load regulation. Existing research largely ignores individual differences in user charging behavior, particularly regarding the demand response potential of different user groups, lacking refined analysis based on user behavioral characteristics. Charging load is closely related to users' travel patterns, charging habits, mileage, vehicle characteristics, and environmental factors. Therefore, accurate profiling and classification of charging behavior for different types of users are needed to provide strong data support for grid dispatching, demand response strategy optimization, and the construction of electric vehicle charging infrastructure. Summary of the Invention

[0004] This invention provides a method and apparatus for evaluating user demand response potential based on GMM clustering and information entropy, in order to solve the problem of evaluating the user demand response potential of electric vehicles.

[0005] In a first aspect, embodiments of the present invention provide a method for evaluating user demand response potential based on GMM clustering and information entropy, including: Based on time-flexible and power-adjustable indicators, we construct the original feature set of electric vehicle users' charging behavior. Based on the SOM model and GMM clustering algorithm, the original feature set of charging behavior is clustered to obtain multiple user profiles; each user profile corresponds to a type of electric vehicle user. For each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle user corresponding to that user profile and the ideal solution, thereby obtaining the demand response potential of the electric vehicle user corresponding to that user profile.

[0006] Secondly, embodiments of the present invention provide a user demand response potential assessment device based on GMM clustering and information entropy, comprising: The feature construction module is used to construct the original feature set of electric vehicle users' charging behavior based on time-flexible and power-adjustable indicators. The user profiling module is used to cluster the original feature set of charging behavior based on the SOM model and GMM clustering algorithm to obtain multiple user profiles; each user profile corresponds to a type of electric vehicle user. The potential assessment module is used to calculate the degree of similarity between the charging behavior characteristics of electric vehicle users corresponding to each user profile and the ideal solution using the TOPSIS method, thereby obtaining the demand response potential of electric vehicle users corresponding to that user profile.

[0007] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0008] The present invention provides a method and apparatus for profiling electric vehicle user behavior based on GMM clustering and information entropy. Through the SOM-GMM clustering method, it performs multi-level and multi-dimensional analysis of electric vehicle user charging behavior, overcoming the limitations of traditional clustering algorithms in adapting to complex behavioral patterns and improving clustering accuracy and flexibility. By profiling electric vehicle user charging behavior and analyzing demand response potential as an important dimension, it reveals significant differences among different user groups in terms of active periods, load adjustment capabilities, and peak-valley distribution, providing data support for subsequent personalized demand response strategy design. Finally, the TOPSIS method is used to calculate the degree of closeness between various user groups and the ideal solution, thereby achieving a quantitative assessment of demand response potential and providing a basis for demand response scheduling. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the implementation of the user demand response potential assessment method based on GMM clustering and information entropy provided in this embodiment of the invention. Figure 2 This is a charging behavior profile radar map provided in an embodiment of the present invention; Figure 3 This is a four-quadrant diagram of demand response potential profile provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the user demand response potential assessment device based on GMM clustering and information entropy provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0010] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0011] See Figure 1The flowchart illustrating the implementation of the user demand response potential assessment method based on GMM clustering and information entropy provided in this embodiment of the invention is described in detail below: Step 101: Construct the original feature set of electric vehicle users' charging behavior based on time-flexible and power-adjustable indicators.

[0012] In this embodiment, from the perspective of demand response, this embodiment proposes an analysis index for the demand response characteristics of electric vehicles from two dimensions: time flexibility and power adjustability, targeting the charging load characteristics.

[0013] (1) Time flexibility indicators These metrics reflect the adjustability and portability of users' "when to charge," that is, their flexible participation in terms of time. These metrics mainly involve characteristics such as charging time distribution, duration, frequency, time period selection, and responsive window.

[0014] ① Average charging time

[0015] Average charging time It refers to the average time a user takes to perform a charging operation, representing the time consumed by a user during a single charging process.

[0016] (1) In the formula, For the first i The duration of each charge, N This represents the total number of charging cycles.

[0017] ② Weekly charging frequency

[0018] Weekly charging frequency This refers to the number of times a user charges their phone per week. It is an important indicator for measuring user charging behavior and can reflect the periodicity of a user's charging needs.

[0019] (2) In the formula, For the first d Number of times to charge per day.

[0020] ③ Charging location preference

[0021] Charging site preference This refers to the frequency or preference of users charging at different charging locations (such as parking lots, shopping malls, and offices). This indicator reflects users' habits in choosing different charging locations by statistically analyzing the distribution of charging locations in users' historical charging data.

[0022] (3) In the formula, For users who use charging sites during the statistical period s Number of charging cycles This represents the total number of times a user charges their phone during the statistical period. S Number the charging locations, for example, parking lot number 1, shopping mall number 2, etc.

[0023] ④ Charging start-up time distribution

[0024] Charging start-up time distribution This refers to the distribution of time when users initiate charging operations throughout the 24 hours of a day.

[0025] (4) ⑤ Standard deviation of charging time

[0026] Standard deviation of charging time This refers to the degree of fluctuation in the charging time for each user, reflecting the stability of the user's charging time. A larger standard deviation indicates greater variation in charging time, while a smaller standard deviation indicates relatively stable charging time.

[0027] (5) In the formula, For the first i First charge duration This is the average charging time. This represents the total number of charging cycles.

[0028] ⑥ Charging active window length

[0029] Charging active window length It refers to the length of the time period during which users' charging behavior occurs in a concentrated manner.

[0030] (6) In the formula, The end time of the active charging window. The start time of the active charging window.

[0031] ⑦ Concentration of charging start time

[0032] Concentration of charging start time This refers to the concentration of time users initiate charging throughout the day. Higher concentration indicates that users' charging time is more concentrated, for example, most users charge at night; while lower concentration indicates that users' charging time is more evenly distributed.

[0033] (7) In the formula, The standard deviation of the charging start time distribution. For at any time t The frequency at which charging is initiated.

[0034] ⑧ Peak hour electricity consumption ratio

[0035] Off-peak electricity consumption ratio This refers to the proportion of electricity a user charges during periods of low electricity demand (such as off-peak hours at night) out of the user's total charging volume.

[0036] (8) In the formula, The set of time points during the valley period. This represents the total charge amount.

[0037] ⑨ Response Period

[0038] Response period It refers to the overlapping time period during which electric vehicle users initiate demand response incentives on the grid side, and where they themselves have the conditions for grid connection. In other words, it is the theoretical time range during which electric vehicle users can participate in demand response regulation.

[0039] (9) In the formula, This refers to the time period for demand response control initiated by the power grid side. For users l The grid connection period (i.e., the actual time period during which electric vehicles connect to charging facilities).

[0040] (2) Power adjustability index These metrics reflect a user's adjustability and responsiveness in "how much power to charge," i.e., flexibility or agility at the power level. They reflect the range, speed, stability, and peak-valley distribution of charging load variations.

[0041] ① Average charging power

[0042] Average charging power It refers to the average power consumed by the user during the charging process, reflecting the power consumption intensity of the user's charging equipment.

[0043] (10) In the formula, For users l exist t Charging power at any time T This represents the total charging time interval.

[0044] ② Total charging amount

[0045] Total charging capacity It refers to the total amount of electricity consumed by a user within a certain period of time (such as a day, a week, or a month), and is a basic indicator for evaluating the intensity of a user's charging behavior.

[0046] (11) In the formula, This represents the sampling time interval.

[0047] ③ Peak hour electricity consumption ratio

[0048] Peak electricity consumption ratio This refers to the peak hours in a user's daily load curve. The proportion of peak-hour electricity consumption to total daily electricity consumption. A higher peak-hour proportion indicates a more concentrated period of electricity consumption, resulting in better peak-shaving and valley-filling regulation. The expression is as follows: (12) In the formula, For users l exist t The charging load at any time, The sampling time interval, It is a set of sampling time points.

[0049] ④ Load fluctuation rate

[0050] Load volatility This is the ratio of the standard deviation to the mean of the user load curve, reflecting the degree of dispersion of charging load over time. A higher volatility indicates greater fluctuations in the user charging load curve and greater user dispatch potential. The expression is as follows: (13) In the formula, and users respectively l Standard deviation and mean of charging load, The number of user load sampling points.

[0051] ⑤ Adjustment range

[0052] Adjustment range This refers to the maximum range within which a user's charging load can fluctuate without affecting their basic usage needs during load regulation. This indicator reflects the user's maximum adjustable load capacity and is one of the key parameters for evaluating their participation in peak shaving and demand response. A larger adjustment range indicates a stronger flexible adjustment capability and a higher potential for participating in system operation optimization. The adjustment range can be defined as the ratio of the difference between the user's maximum and minimum charging power within the adjustment time window to the maximum power, expressed as follows: (14) In the formula, For users l Maximum charging power within the adjustable time period For users l Minimum charging power within the adjustable time period.

[0053] ⑥ Adjustment rate

[0054] Adjustment rate This refers to the rate of change of user charging load per unit time, used to measure the agility of user load response adjustment commands. In actual demand response scheduling, users with a faster adjustment rate can respond to control commands more quickly, improving the real-time performance and safety of system operation. It is suitable for high-frequency control scenarios such as emergency response and ancillary services. The adjustment rate is usually characterized by the derivative of the load change with respect to the time interval, as shown in the following expression: (15) In the formula, This represents the charging power after the sampling time interval.

[0055] Ultimately, the original feature set of electric vehicle user charging behavior .

[0056] Step 102: Based on the SOM model and GMM clustering algorithm, the original feature set of charging behavior is clustered to obtain multiple user profiles; each user profile corresponds to a type of electric vehicle user.

[0057] In this embodiment, user profiling is a data analysis tool that uses user data to define labels to characterize users as a whole and develop precise marketing strategies for target users. The charging and discharging load of electric vehicles is closely related to factors such as users' travel patterns, charging habits, mileage, and vehicle characteristics. Through user profiling technology, the various factors affecting charging and discharging load can be abstracted into labels, allowing for precise characterization and classification of electric vehicle user behavior characteristics, providing a foundation for precise incentive research. Currently, scholars' profiling of electric vehicle user behavior characteristics mainly supports adjustment potential prediction and charging load optimization, and their research methods can be broadly divided into two categories: statistical analysis methods based on historical data and data-driven load forecasting methods.

[0058] In terms of data statistical analysis, Li Xiaohui et al. (2024) mainly analyzed key charging activity chains and constructed a user charging behavior profile based on the sequence of parking and charging activities between adjacent trips using clustering methods. This profile was used to analyze the charging patterns of different users. Bian Haihong et al. (2025) considered users' social characteristics based on travel characteristics, conducted hierarchical quantum clustering analysis, constructed a user profile, and combined it with real-time traffic flow to predict charging load. Yang Aixing et al. (2024) first used the FCM method to identify charging behavior characteristics, and then classified the number of features of electric vehicle users according to the feature aggregation method, constructed a user profile suitable for parking lots, and distinguished the response potential priority of different users accordingly. Liu Yanli and Wang Junyi et al. (2023) mined household electricity consumption behavior and electric vehicle travel patterns based on household electricity consumption data and electric vehicle charging pile data, proposed a multi-source data-driven electric vehicle user identification method, and constructed a resident electric vehicle user profile to achieve targeted management and control of electric vehicles. Wang Yangyang et al. (2024) combined density space-based clustering algorithms and improved self-organizing mapping deep clustering algorithms to effectively integrate the temporal distribution of electric vehicle power and the spatial distribution characteristics of charging piles. They constructed more detailed charging profiles such as "morning type", "noon type" and "evening type" from the time dimension, laying a data foundation for predicting the adjustment potential of different users.

[0059] In terms of data-driven load forecasting, Ahmadian Amirhossein et al. (2023) mainly constructed user profiles by jointly predicting the energy consumption and charging time of electric vehicles through a trainable artificial deep neural network, which supports the prediction of charging load adjustment potential.

[0060] In summary, domestic and international scholars primarily employ data analysis to mine basic user behavioral characteristics. Through data-driven methods, they automatically identify potential patterns and complex relationships within user behavior, constructing user profiles from dimensions such as travel characteristics and charging time. This supports the prediction of adjustment potential and charging load, and the relevant methods and models provide a theoretical foundation for this project. However, to promote the deep participation of electric vehicles in grid interaction, several issues require further in-depth research: First, in addition to charging behavior, given the increasing maturity of vehicle-grid interaction technology, it is crucial to consider users' discharge intentions and conduct in-depth research on the charging and discharging behavior characteristics of electric vehicle users to provide a basis for fully exploring their adjustment potential. Second, besides spatiotemporal characteristics such as charging time and driving trajectory, it is necessary to combine individual user characteristics and external environmental factors to deeply analyze the main factors influencing charging and discharging behavior, providing a basis for user group classification. Third, in addition to classifying users on a time scale, it is also necessary to consider their charging and discharging frequency and flexibility, conducting refined classification to construct more comprehensive and accurate user profiles.

[0061] Based on this, this embodiment proposes a comprehensive clustering model based on SOM–GMM (Self-Organizing Map – Gaussian Mixture Model) to achieve nonlinear identification and probabilistic classification of charging behavior patterns. First, the SOM network maps high-dimensional behavioral features to a two-dimensional topological space through a competitive learning mechanism, achieving nonlinear dimensionality reduction and topology preservation, effectively alleviating the curse of dimensionality and clustering instability problems faced by traditional clustering algorithms in high-dimensional feature spaces. Subsequently, the GMM model establishes a Gaussian mixture distribution in the mapped feature space, achieving soft clustering and smooth boundary classification of charging behavior types, thereby obtaining statistically significant user category classification results.

[0062] Step 103: For each user profile, calculate the degree of closeness between the charging behavior characteristics of the electric vehicle user corresponding to that user profile and the ideal solution using the TOPSIS method, and obtain the demand response potential of the electric vehicle user corresponding to that user profile.

[0063] This invention employs the SOM-GMM clustering method to conduct multi-level and multi-dimensional analysis of electric vehicle users' charging behavior, overcoming the limitations of traditional clustering algorithms in adapting to complex behavioral patterns and improving clustering accuracy and flexibility. By profiling electric vehicle users' charging behavior and analyzing demand response potential as an important dimension, it reveals significant differences among different user groups in terms of active periods, load regulation capabilities, and peak-valley distribution, providing data support for subsequent personalized demand response strategy design. Finally, the TOPSIS method is used to calculate the degree of closeness between various user groups and the ideal solution, thereby achieving a quantitative assessment of demand response potential and providing a basis for demand response scheduling.

[0064] In one possible implementation, based on the SOM model and GMM clustering algorithm, the original feature set of charging behavior is clustered to obtain multiple user profiles, including: Based on the SOM model, the data in the original feature set of charging behavior is mapped to a low-dimensional space to obtain the mapped charging behavior feature set. The target number of clusters is determined using the sum of squared errors as the clustering evaluation index. The GMM clustering algorithm is used to cluster charging behavior feature sets after mapping the target cluster number pairs, resulting in multiple clusters; each cluster corresponds to a user profile.

[0065] In this embodiment, Self-Organizing Map (SOM) is an unsupervised learning algorithm that can map high-dimensional data to a low-dimensional space while preserving the topological relationships between the data. The SOM model groups similar data points together by mapping data in the input space to a two-dimensional grid.

[0066] (1) Initialize the network Neuron weight vector of SOM Dimensions and input data The values ​​are taken from the original feature values ​​in Section 3.1. During initialization, the weight vector is usually randomly set to values ​​similar to the input samples. Assume there are... M If there are neurons, then the weight vector of each neuron... for n -dimensional vector , representing the neuron's number. During initialization, the weight vector... Typically, small random values ​​are selected from the input data.

[0067] (2) Calculate the distance between the sample and the neuron. For each input sample Calculate its relationship with the weight vector of each neuron in the network. Euclidean distance: (16) in, Indicates input sample With the The distance between neurons.

[0068] (3) Select the best matching unit (BMU) Select distance samples The nearest neuron is selected as the Best Matching Unit (BMU): (17) in, It is distance The weight vector of the nearest neuron.

[0069] (4) Update weights To enable the network to adapt to the input data, the weight vectors of the BMU and its neighbors are updated. The update formula is: (18) (19) in, It's the learning rate. For neighborhood functions, The width of the neighborhood.

[0070] (5) Iterative training Through multiple iterations of training, the weights of neurons are continuously adjusted until the weight vector converges. In each iteration, the weights of neurons are adjusted based on the current input sample, forming a self-organizing map of the data.

[0071] The GMM clustering algorithm must first determine the number of clusters. k Based on the principles of the GMM algorithm, the sum of squared errors (SSE) can be used as an evaluation metric for determining the number of clusters. SSE The formula is as follows: (20) In the formula, k Indicates the first k Clusters are the number of clusters. Indicates the first k The number of data points in the class It's a data point. It is the first k The average value of the cluster.

[0072] pass SSE Determine the number of clusters k Then, assume that each class of data follows a Gaussian distribution, where the probability density function of the Gaussian distribution is: (twenty one) In the formula, and The first k The mean and standard deviation of the class.

[0073] Then, the probability density function and weight parameters of each cluster are estimated using the Expectation-Maximization (EM) algorithm. Finally, the probability of each sample belonging to each Gaussian distribution is calculated, and each sample is assigned to the cluster corresponding to the Gaussian distribution with the highest probability. The specific classification formula is as follows: (twenty two) In the formula, Indicates sample Belongs to the k The probability of a class It is the first k Class weight.

[0074] Finally, by comparing the probabilities of each sample in various Gaussian distributions, the samples are classified into the cluster with the highest probability values.

[0075] In one possible implementation, before calculating the approximation of the charging behavior characteristics of the electric vehicle user corresponding to each user profile to the ideal solution using the TOPSIS method, and thus obtaining the demand response potential of the electric vehicle user corresponding to that user profile, the following steps are also included: By combining time-flexible and power-adjustable indicators from the original feature set of charging behavior, multiple candidate behavior feature sets are obtained; The feature set adaptability evaluation coefficient of each candidate behavior feature set is calculated based on the preset feature set adaptability function, and the optimal feature set is selected based on the feature set adaptability evaluation coefficient. Accordingly, for each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle user corresponding to that user profile and the ideal solution, thereby obtaining the demand response potential of the electric vehicle user corresponding to that user profile, including: For each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle users corresponding to that user profile in the optimal feature set and the ideal solution, thereby obtaining the demand response potential of the electric vehicle users corresponding to that user profile.

[0076] In this embodiment, feature selection is a crucial step in electric vehicle user behavior profiling. Obtaining a reasonably sized feature set that accurately characterizes key user traits is the ideal goal of electric vehicle user behavior profiling. The characteristics inherent in electric vehicle user load data can be characterized by various feature indicators; however, different features may describe user characteristics in a repetitive way, meaning there may be correlations between different features. Therefore, when selecting features, the effectiveness and redundancy between features should be considered. Furthermore, to enhance the understandability of user features, quantitative analysis should be performed on the user features.

[0077] To enhance the interpretability and discriminative power of clustering, this embodiment introduces a feature optimization and redundancy measurement mechanism. Key feature sets are selected through information gain and correlation analysis to ensure that feature dimensions maintain high discriminative power while preserving information content. Finally, based on the clustering results and feature quantification indicators, a charging behavior profile model for various user types is constructed, achieving a quantitative description from data to cognition, and providing data support for subsequent demand response strategy design and time-of-use pricing mechanisms.

[0078] The method for selecting and quantifying the electricity consumption characteristics of electric vehicle users adopted in this embodiment is as follows: First, a multi-dimensional original feature set of electricity users is constructed. On this basis, the degree of information gain of the addition of each feature to the overall information of the electricity user is measured. Then, the Spearman correlation coefficient is introduced to measure the similarity between different features. Finally, taking into account the information gain after the feature is introduced and the similarity between features, the feature set adaptability evaluation coefficient is obtained, and the feature set with the highest evaluation coefficient is selected as the optimal electricity consumption feature set of the electricity user.

[0079] In one possible implementation, the feature set fitness function is:

[0080] Where L is the feature set fitness evaluation coefficient of candidate feature set S, G(S) is the information gain index of all features in candidate feature set S, and R(S) is the redundancy index of all features in candidate feature set S.

[0081] In this embodiment, the information gain of a feature can effectively measure the effectiveness of different features in characterizing the behavior of electric vehicle users. To better understand this metric, the concept of information entropy must be introduced. Information entropy measures the degree of uncertainty of a system; the larger the value, the more chaotic the system and the greater the amount of information. The amount of information measures whether an event has occurred; the greater the amount of information, the lower the probability of occurrence and the greater the uncertainty.

[0082] For random variables X Assuming all its possible values ​​are The corresponding probability of occurrence is Then the random variable X The information entropy is defined as shown in equation (23).

[0083] (twenty three) In the analysis of electric vehicle user charging characteristics, random variables X All possible values ​​are all clusters formed by the clustering, and the corresponding probability is the proportion of the number of samples contained in each cluster to the total number of samples.

[0084] Introducing feature sets Z It will change the random variable XInformation entropy. Assuming features Z All possible values The corresponding probability of occurrence is This refers to the proportion of samples falling within each feature value range out of the total sample size. For a given feature value... The probability that a sample belongs to a different cluster is given by the internal structure. Then the characteristics Z After adding random variables X The formula for calculating information entropy is shown in equation (24).

[0085] (twenty four) For a certain feature Z Its information gain can be measured Z The introduction of random variables X The degree of effectiveness in reducing uncertainty. Information gain is numerically defined as the feature... Z Before and after introducing random variables X The interpolation of information entropy is calculated using the formula shown in equation (25).

[0086] (25) However, the traditional information gain expression mentioned above has the problem of favoring high-cardinality features (i.e., features with a large number of value types), which may lead to high-value features obtaining higher information gain due to their finer-grained segmentation, thus misleading feature selection. Therefore, this embodiment introduces normalized information gain (Gain Ratio) based on the feature's own information entropy. The information gain is normalized and corrected; the larger the value, the stronger the ability to distinguish variables.

[0087] (26) To measure redundancy among features and thus avoid potential nonlinear relationships, a distance correlation coefficient is used. This method effectively captures nonlinear relationships between features. Given a set of feature vectors... and ,in , , n Features and The number of samples. Feature vector sample set. and The distance matrices are respectively A and B Each element within it is a normalized and centered Euclidean distance matrix between each pair of samples.

[0088] (27) (28) (29) (30) (31) In the formula, and For the first i and j Euclidean distance between samples and The elements are the distance matrix after bi-centering. Let the distance covariance be... and Let be the distance variance. The distance correlation coefficient ranges from [0,1] and is true only when... and When independent, it takes the value of 0, thus enabling the detection of statistical dependencies of arbitrary forms, both linear and nonlinear. It is an important tool for capturing nonlinear correlations when improving feature redundancy measures.

[0089] Therefore, for a given feature set U Define redundancy index , can be represented as: (32) In the formula, The total number of features contained in the feature set. For the first feature set i and j One characteristic.

[0090] For a certain feature set Based on the effectiveness of the comprehensive feature information gain and the redundancy measured by the correlation coefficient between features, a feature set fitness evaluation coefficient is constructed, which is defined as follows: (33) In the formula, For feature set S Information gain index of all features in For feature set S Redundancy indicators for all features.

[0091] Searching using traversal method S The feature subset with the largest feature evaluation function value is the optimal feature subset sought in this embodiment, which is the optimal behavioral feature for the category classification problem of user behavior profiling.

[0092] In one possible implementation, after using the GMM clustering algorithm to cluster the charging behavior feature set based on the target cluster number pairs to obtain multiple clusters, the following is also included: The scores of electric vehicle users in each cluster across all feature dimensions are calculated based on a preset formula, and the optimal charging feature set is semantically represented based on each score; the preset formula is as follows:

[0093] in, For the first i Class User j Individual tag scores, For all belonging to the first i Class User j The average value of each label, For the first j The minimum value of each label. For the first j The maximum value of each label.

[0094] In this embodiment, to improve the readability of electric vehicle user feature tags, a scoring system is adopted, with a score of 0-1 as the baseline, to measure the scores of different types of electric vehicle users across all feature dimensions. The optimal charging feature set is semantically represented, and the calculation method is as follows: (34) In the formula, For the first i Class User j Individual tag scores, For all belonging to the first i Class User j The average value of each label, and The first j The minimum and maximum values ​​of each label.

[0095] To enable business personnel to more intuitively understand the charging characteristics of various electric vehicle users, the obtained user charging tags are visualized to create user behavior profiles for different user types. The visualization consists of two parts: intra-category user charging behavior profiles and inter-category charging characteristic comparisons. Intra-category user charging behavior profiles use methods such as radar charts to display the charging characteristics of various users, while inter-category charging characteristic comparisons are presented using bar charts to highlight the comparison of the frequency of the same tag across different user groups. Combining these two methods allows business personnel to more accurately and conveniently understand the commonalities and individual characteristics of electric vehicle user charging behavior.

[0096] In one possible implementation, before calculating the approximation between the charging behavior characteristics of the electric vehicle users corresponding to that user profile and the ideal solution using the TOPSIS method in the optimal feature set, and thus obtaining the demand response potential of the electric vehicle users corresponding to that user profile, the following steps are also included: The weights of each evaluation index in the optimal feature set are determined using the entropy weight method.

[0097] In this embodiment, to quantitatively assess the potential of electric vehicles to participate in grid demand response, a comprehensive evaluation model based on the Entropy Weight Method (EWM) and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) is constructed. This method first uses the Entropy Weight Method to determine the objective weights of each evaluation indicator, and then uses the TOPSIS method to calculate the degree of similarity between various user groups and the ideal solution, thereby achieving a quantitative assessment of demand response potential. The final result is presented in the form of a four-quadrant bubble chart, with the horizontal axis representing "power adjustability" and the vertical axis representing "time flexibility," thus creating a profile of the response potential of the electric vehicle user group through two-dimensional features.

[0098] Entropy weighting is a typical objective weighting method. Its basic idea is to use the dispersion of each indicator's values ​​to reflect the amount of information it contains. The greater the difference between indicators, the more information they provide, and the higher their weight. According to the definition of information entropy, the first... j The formula for calculating the entropy value of each indicator is: (35) in, i and j The meaning of reference formula (34) is as follows. The number of evaluation objects, i.e. the specific number of categories, is determined by the result of the feature set optimization in Section 3.3.

[0099] (36) Therefore, the first j Entropy weight of item label for: (37) In one possible implementation, for each user profile, the TOPSIS method is used to calculate the degree of closeness between the charging behavior characteristics of the electric vehicle users corresponding to that user profile in the optimal feature set and the ideal solution, thereby obtaining the demand response potential of the electric vehicle users corresponding to that user profile, including: The original data matrix is ​​established based on all charging behavior features in the optimal feature set, and then dimensionless processing is performed to obtain the normalized matrix. Weights are assigned to the indices in the normalized matrix, and a weighted normalized matrix is ​​constructed. The maximum value of each column element in the weighted normalized matrix is ​​taken as the positive ideal solution, and the minimum value of each column element is taken as the negative ideal solution. For each user profile, calculate the distance from the charging behavior features of the electric vehicle user corresponding to that user profile in the optimal feature set to the positive and negative ideal solutions, and calculate the relative proximity to the positive and negative ideal solutions to obtain the demand response potential of the electric vehicle user corresponding to that user profile.

[0100] In this embodiment, the TOPSIS method is a multi-attribute comprehensive evaluation method that ranks objects based on their approximation to an ideal solution. It ranks the objects according to their relative closeness to the ideal target, thereby determining the relative merits of the objects. Its basic principle is to construct positive and negative ideal systems among a finite number of objects, and compare the object to be evaluated with each of the positive and negative ideal objects to obtain their relative closeness. Objects closer to the positive ideal object and farther from the negative ideal object are considered superior, and each object is then evaluated. Its basic steps are as follows: ①Establish the original data matrix R .

[0101] (38) ② The initial matrix is ​​dimensionless to obtain the normalized matrix. .

[0102] ③ Assign weights to the indicators according to the entropy weight method.

[0103] ④ Construct a weighted normalized matrix.

[0104] Weighted combination of various indicators Constructing a weight vector Multiplying the weighted normalized matrix by the normalized matrix yields the weighted normalized matrix. .

[0105] ⑤ Determine the positive and negative ideal solutions.

[0106] In reality, there is no absolutely optimal or worst solution; therefore, ideal objects... By weighted normalization matrix The negative ideal object is composed of the maximum value of each column element. By weighted normalization matrix It is composed of the minimum value of each column element.

[0107] ⑥ Calculate the distance scale.

[0108] Calculate each object to be evaluated to iThe distance between the positive and negative ideal solutions can be measured using the following scale: The distance to the ideal solution is calculated using Euclidean distance. The distance to the negative ideal solution is : (40) (41) ⑦ Calculate the relative proximity.

[0109] The relative closeness of each evaluation object to the positive and negative ideal solutions (42) In the formula, To evaluate how close the proposed solution is to the ideal solution, the value is between 0 and 1. The closer the value is to 1, the higher the score of the proposed solution, and the closer the value is to 0, the lower the score.

[0110] ⑧ Based on the ideal closeness Sort by size.

[0111] In one specific embodiment, electric vehicle (EV) charging load data was collected for a certain region from January 1, 2020 to December 31, 2021. This region has a high coverage rate of charging facilities and diverse user types, providing a relatively comprehensive reflection of the actual charging behavior characteristics of residential, public, and commercial EVs. The raw data comes from the local public charging pile operation platform and includes multi-dimensional information such as charging start and end times, charging amount, charging duration, charging location, order creation time, transaction amount, geographical location, and weather conditions. To eliminate interference from outliers and invalid records, the raw data underwent data cleaning and preprocessing, including removing abnormal orders with charging durations less than 5 minutes or greater than 24 hours, removing records with missing key fields, and standardizing the time format, resulting in 400,000 valid data entries. Furthermore, the peak-valley time periods for users with unified large industrial electricity prices and general industrial and commercial electricity prices in this region are shown in Table 1, and the time-of-use electricity prices for EV charging are shown in Table 2.

[0112] Table 1 Peak and off-peak periods for users of the unified industrial electricity price and general industrial and commercial electricity price

[0113] Table 2 Electricity Time-of-Use Pricing for Electric Vehicle Charging

[0114] Based on the original feature set constructed from temporal behavioral characteristics and demand response potential characteristics, this embodiment uses the SOM-GMM (Self-Organizing Map-Gaussian Mixture Model) joint clustering algorithm to perform cluster analysis on electric vehicle users. The SOM network first realizes the self-organizing mapping of user features in a low-dimensional space, initially capturing the topological relationship of the feature space; then, the GMM algorithm is used to perform probabilistic clustering of users on this basis to obtain smoother and more interpretable category boundaries.

[0115] To further characterize the charging behavior of electric vehicle users, this embodiment first performs information gain analysis on the original feature set to evaluate the contribution of each feature to the user clustering results.

[0116] Secondly, feature redundancy was quantitatively evaluated to select features that contribute the most and have the least redundancy in the charging behavior profile. The results showed the average redundancy among features, indicating that there is almost no linear redundancy among features and the overall feature set has good information complementarity.

[0117] Based on the principles of high information content and low redundancy, and considering both information gain and feature redundancy, this embodiment ultimately selects six indicators covering user charging time distribution (ActiveWindow, StartTimeConcentration), adjustment behavior (AdjustRate, AdjustRange), and charging power characteristics (ValleyRatio, PeakRatio) to comprehensively reflect the multidimensional characteristics of user charging behavior. Further calculation of the feature set adaptability evaluation coefficient yields a value of 0.9468, indicating that this feature subset possesses high information carrying capacity and good adaptability in distinguishing different user groups and constructing charging behavior profiles, providing a reliable data foundation for subsequent profile analysis and group behavior pattern recognition based on quantitative features.

[0118] Furthermore, based on the scores of different types of electric vehicle users across all feature dimensions, the charging behavior of the aforementioned eight types of electric vehicle users was profiled, resulting in... Figure 2 The charging behavior profile radar chart shown below and Figure 3 The diagram shown is a four-quadrant representation of demand response potential.

[0119] Depend on Figure 2 , Figure 3It is evident that different user groups exhibit significant behavioral differences in key characteristics. The first group of users has very short charging times but possesses high power regulation capabilities. Combined with moderate off-peak and peak charging rates, this indicates that they are typical short-duration, high-power, interruptible charging users with high load regulation potential. The second group of users are strict price followers, charging almost exclusively during off-peak hours, exhibiting regular charging behavior but showing low willingness to actively participate in real-time regulation. The third group of users primarily charges during peak hours but retains some power regulation capabilities, possessing certain controllable potential. The fourth group of users is similar to the first group, belonging to the short-duration, high-regulation type. Their key difference lies in the fact that they almost never charge during peak hours; their controllable potential is mainly reflected in flat and off-peak hours, making their charging behavior more "friendly." The fifth group of users shows a moderate tendency in time selection and power regulation, representing the most balanced and least extreme group in charging behavior. The sixth group of users is even more extreme than the second group, consisting entirely of off-peak electricity users. Their charging behavior patterns are highly consistent, making them the foundational users for ensuring off-peak electricity consumption. Category 7 users have long charging times and frequently adjust their charging power. They are "active managers" in the charging process, and their high AdjustRate indicates they are well-suited for demand response projects requiring real-time, frequent responses. Category 8 users are typical "rigid" peak-hour chargers, insensitive to electricity prices, and prefer to charge during peak hours. Their high AdjustRange means that with strong incentive policies, it is possible to shift this portion of the load that puts the most pressure on the grid.

[0120] To quantify the adjustable space for demand response by eight types of electric vehicle users, this embodiment calculates two types of potential indicators: the first is the maximum DR potential (an ideal adjustable upper limit ignoring time flexibility); the second is the DR potential considering time flexibility (Time-Flex DR), which incorporates movable time windows and sustainable response duration into the constraints, making it closer to the engineering feasibility potential.

[0121] In terms of maximum DR potential, the upper limits of each cluster are in the same order of magnitude (56–69 kW), with only clu6 slightly lower (40.9 kW), indicating that when the equipment and station-end capacity conditions are similar, the theoretical adjustable upper limits of each cluster do not differ significantly. However, once time flexibility is introduced, the differences between clusters are significantly amplified: clu7 = 54.22 kW. clu5=29.82 kW>clu4=20.17 kW>clu1=14.00 kW ≈ clu6=13.56 kW ≈ clu3=12.59 kW>clu2=8.75 kW>clu8=7.65 kW. This ranking is highly consistent with the characteristics of charging start time: the wider the time window, the more dispersed the distribution, and the longer the duration of the cluster, the higher its Time-Flex DR; conversely, rigid peak charging or strict valley charging leads to limited time-shifting space, and the Time-Flex DR is significantly reduced.

[0122] The differences can be further explained by combining charging behavior characteristics: (1) clu7 users are long-term and high-frequency adjustment type: the start time coverage is wide and the daily load curve remains at a medium-high level for a long period of time. Time-Flex DR is almost equal to its upper limit (54.22 vs. 69.02 kW), indicating that it can reduce power and shift time, and is suitable for fast / frequent real-time DR and frequency modulation auxiliary services.

[0123] (2) The users of clu5 are balanced: the usage is relatively balanced throughout the day with moderate fluctuations. Although the maximum limit is average (56.68 kW), the Time-Flex DR reaches 29.82 kW, ranking second. This cluster has the best engineering feasibility and is suitable for price-based and incentive-based peak shifting synergy.

[0124] (3) clu4 users are short-term high-regulation but peak-avoidance type: their activities are more "friendly" during flat and valley periods. Time-FlexDR=20.17 kW, which is significantly higher than clu1 (14.00 kW), which is also short-term high-regulation. This shows that under similar power adjustability, the peak-avoidance strategy significantly improves the feasibility of implementation.

[0125] (4) Users of clu1 and 3 are short-term high-regulation or peak-segment dominant: the maximum upper limit is relatively high (60.79 / 65.50kW), but the time flexibility is generally low (14.00 / 12.59 kW), which is more suitable for event-type peak shaving or power limiting, while the cross-time period transfer capability is limited.

[0126] (5) Users of clu2 and 6 are strictly / extremely off-peak electricity types: their time selection is highly concentrated in the off-peak period at night, and their Time-Flex DR is low (8.75 / 13.56 kW). These two types of users are beneficial for off-peak absorption and capacity protection, but their contribution to peak shifting is limited.

[0127] (6) clu8 users are rigid peak charging type: they have a strong peak preference and are not sensitive to electricity prices. Although the maximum upper limit is high (64.61 kW) and the Time-Flex DR is the lowest (7.65 kW), they need a combination of "strong incentive + scheduled scheduling + capacity constraint" to shift the load to the flat / valley.

[0128] In summary, the dominant factor determining the feasibility of DR is not "maximum adjustable power," but rather "movable time window and sustainable response duration." From this perspective, clu7 significantly outperforms other clusters, clu5 and clu4 exhibit good engineering flexibility, clu1 / 3 / 6 are suitable for event-driven and power-limited scenarios, while clu2 / 8 require differentiated pricing and strong incentive mechanisms to guide behavior reconfiguration. Therefore, the following recommendations are made: (i) Project matching: prioritize real-time / frequent DR for clu7, followed by clu4 / 1; prioritize intraday peak shifting and arbitrage for clu5 / 4; prioritize valley absorption for clu2 / 6; and adopt a combined mandatory and subsidy strategy for rigid peak load (clu8); (ii) Incentive design: incorporate the time flexibility coefficient into the settlement (time shift amount × flexibility coefficient × incentive unit price) to increase the revenue priority of clu7 / 5; (iii) Rolling scheduling: implement multi-period rolling optimization for clu4 / 5 / 7, adopt event-based power limiting for clu1 / 3, guarantee valley absorption for clu2 / 6, and implement reservation and capacity constraints for clu8. This integrated clustering-matching-incentive framework can significantly improve the "effective potential" of demand-side schedulable resources without changing the upper limit of equipment.

[0129] Conclusion: This study first constructs two major categories of user behavior characteristic analysis indexes from a demand response perspective, targeting the charging load characteristics of electric vehicles: time-series behavior indicators and demand response potential indicators, to comprehensively characterize users' charging behavior features. Based on this, the SOM-GMM clustering algorithm is used to classify and identify electric vehicle users, ultimately forming eight representative user groups. This reveals significant differences among different user types in charging time preferences, load adjustment capabilities, and response characteristics. Subsequently, to improve the clustering accuracy and feature interpretability of the model, information gain analysis was performed on the original feature set. Through feature redundancy quantification, highly correlated feature pairs (such as ActiveWindow and MeanDuration, LoadFluctuation and AdjustRange) were identified. In the feature selection process, the best features were retained based on the principle of information contribution, thus constructing a non-redundant feature subset with high information carrying capacity. Combining the information gain and redundancy analysis results, six key features were finally determined: ActiveWindow, AdjustRate, StartTimeConcentration, ValleyRatio, PeakRatio, and AdjustRange. The feature subset achieved an adaptability evaluation coefficient of 0.9468, indicating that it has high adaptability and explanatory power in distinguishing different charging behavior patterns and supporting the construction of charging behavior profiles, providing solid data support and theoretical basis for the subsequent formulation of user profiles and differentiated demand response strategies.

[0130] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0131] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0132] Figure 4 The diagram shows a schematic of the user demand response potential assessment device based on GMM clustering and information entropy provided in an embodiment of the present invention. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown, and are described in detail below: like Figure 4 As shown, the user demand response potential assessment device 4 based on GMM clustering and information entropy includes: Feature construction module 41 is used to construct the original feature set of electric vehicle users' charging behavior based on time-flexible and power-adjustable indicators; User profiling module 42 is used to cluster the original feature set of charging behavior based on SOM model and GMM clustering algorithm to obtain multiple user profiles; each user profile corresponds to a type of electric vehicle user. The potential assessment module 43 is used to calculate the degree of closeness between the charging behavior characteristics of electric vehicle users corresponding to each user profile and the ideal solution using the TOPSIS method, thereby obtaining the demand response potential of electric vehicle users corresponding to that user profile.

[0133] In one possible implementation, the user profiling module 42 is specifically used for: Based on the SOM model, the data in the original feature set of charging behavior is mapped to a low-dimensional space to obtain the mapped charging behavior feature set. The target number of clusters is determined using the sum of squared errors as the clustering evaluation index. The GMM clustering algorithm is used to cluster charging behavior feature sets after mapping the target cluster number pairs, resulting in multiple clusters; each cluster corresponds to a user profile.

[0134] In one possible implementation, the user profiling module 42 is also used for: Before calculating the approximation of the charging behavior characteristics of electric vehicle users corresponding to each user profile to the ideal solution using the TOPSIS method, and obtaining the demand response potential of electric vehicle users corresponding to the user profile, the time-flexible indexes and power-adjustable indexes in the original charging behavior feature set are combined to obtain multiple candidate behavior feature sets. The feature set adaptability evaluation coefficient of each candidate behavior feature set is calculated based on the preset feature set adaptability function, and the optimal feature set is selected based on the feature set adaptability evaluation coefficient. Accordingly, the potential assessment module 43 is specifically used for: For each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle users corresponding to that user profile in the optimal feature set and the ideal solution, thereby obtaining the demand response potential of the electric vehicle users corresponding to that user profile.

[0135] In one possible implementation, the feature set fitness function is:

[0136] Where L is the feature set fitness evaluation coefficient of candidate feature set S, G(S) is the information gain index of all features in candidate feature set S, and R(S) is the redundancy index of all features in candidate feature set S.

[0137] In one possible implementation, the user profiling module 42 is also used for: After using the GMM clustering algorithm to cluster the charging behavior feature set mapped based on the target cluster number, resulting in multiple clusters, the scores of electric vehicle users in each cluster are calculated across all feature dimensions using a preset formula. Then, the optimal charging feature set is semantically represented based on each score. The preset formula is as follows:

[0138] in, For the first i Class User j Individual tag scores, For all belonging to the first i Class User j The average value of each label, For the first j The minimum value of each label. For the first j The maximum value of each label.

[0139] In one possible implementation, the potential assessment module 43 is also used for: Before calculating the approximation of the charging behavior characteristics of electric vehicle users corresponding to that user profile with the ideal solution using the TOPSIS method, and thus obtaining the demand response potential of electric vehicle users corresponding to that user profile, the entropy weight method is used to determine the weight of each evaluation index in the optimal feature set.

[0140] In one possible implementation, the potential assessment module 43 is specifically used for: The original data matrix is ​​established based on all charging behavior features in the optimal feature set, and then dimensionless processing is performed to obtain the normalized matrix. Weights are assigned to the indices in the normalized matrix, and a weighted normalized matrix is ​​constructed. The maximum value of each column element in the weighted normalized matrix is ​​taken as the positive ideal solution, and the minimum value of each column element is taken as the negative ideal solution. For each user profile, calculate the distance from the charging behavior features of the electric vehicle user corresponding to that user profile in the optimal feature set to the positive and negative ideal solutions, and calculate the relative proximity to the positive and negative ideal solutions to obtain the demand response potential of the electric vehicle user corresponding to that user profile.

[0141] By employing the SOM-GMM clustering method, we conduct multi-level and multi-dimensional analysis of electric vehicle users' charging behavior, overcoming the limitations of traditional clustering algorithms in adapting to complex behavioral patterns and improving clustering accuracy and flexibility. By profiling electric vehicle users' charging behavior and analyzing demand response potential as a crucial dimension, we reveal significant differences among different user groups in terms of active periods, load regulation capabilities, and peak-valley distribution, providing data support for subsequent personalized demand response strategy design. Finally, we use the TOPSIS method to calculate the degree of closeness between various user groups and the ideal solution, thereby achieving a quantitative assessment of demand response potential and providing a basis for demand response scheduling.

[0142] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 5 As shown, the electronic device 5 of this embodiment includes a processor 50 and a memory 51. The memory 51 stores a computer program 52. When the processor 50 executes the computer program 52, it implements the steps in the various method embodiments described above. Alternatively, when the processor 50 executes the computer program 52, it implements the functions of each module / unit in the various device embodiments described above.

[0143] For example, computer program 52 may be divided into one or more modules / units, which are stored in memory 51 and executed by processor 50 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 52 in electronic device 5.

[0144] Electronic device 5 may include, but is not limited to, processor 50 and memory 51. Those skilled in the art will understand that... Figure 5 This is merely an example of electronic device 5 and does not constitute a limitation on electronic device 5. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 5 may also include input / output devices, network access devices, buses, etc.

[0145] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0146] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0147] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for evaluating user demand response potential based on GMM clustering and information entropy, characterized in that, include: Based on time-flexible and power-adjustable indicators, we construct the original feature set of electric vehicle users' charging behavior. Based on the SOM model and GMM clustering algorithm, the original feature set of charging behavior is clustered to obtain multiple user profiles; each user profile corresponds to a type of electric vehicle user. For each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle user corresponding to that user profile and the ideal solution, thereby obtaining the demand response potential of the electric vehicle user corresponding to that user profile.

2. The user demand response potential assessment method based on GMM clustering and information entropy according to claim 1, characterized in that, The SOM model and GMM clustering algorithm are used to cluster the original feature set of charging behavior to obtain various user profiles, including: Based on the SOM model, the data in the original feature set of the charging behavior is mapped to a low-dimensional space to obtain the mapped charging behavior feature set. The target number of clusters is determined using the sum of squared errors as the clustering evaluation index. The GMM clustering algorithm is used to cluster the charging behavior feature set mapped from the target cluster number, resulting in multiple clusters; each cluster corresponds to a user profile.

3. The user demand response potential assessment method based on GMM clustering and information entropy according to claim 2, characterized in that, Before calculating the approximation of the charging behavior characteristics of electric vehicle users corresponding to each user profile to the ideal solution using the TOPSIS method, and obtaining the demand response potential of electric vehicle users corresponding to that user profile, the process further includes: By combining time-flexible and power-adjustable indicators from the original feature set of charging behavior, multiple candidate behavior feature sets are obtained. The feature set adaptability evaluation coefficient of each candidate behavior feature set is calculated based on the preset feature set adaptability function, and the optimal feature set is selected based on the feature set adaptability evaluation coefficient. Accordingly, for each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle user corresponding to that user profile and the ideal solution, thereby obtaining the demand response potential of the electric vehicle user corresponding to that user profile, including: For each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle users corresponding to that user profile in the optimal feature set and the ideal solution, thereby obtaining the demand response potential of the electric vehicle users corresponding to that user profile.

4. The user demand response potential assessment method based on GMM clustering and information entropy according to claim 3, characterized in that, The feature set fitness function is: Where L is the feature set fitness evaluation coefficient of candidate feature set S, G(S) is the information gain index of all features in candidate feature set S, and R(S) is the redundancy index of all features in candidate feature set S.

5. The user demand response potential assessment method based on GMM clustering and information entropy according to claim 4, characterized in that, After using the GMM clustering algorithm to cluster the charging behavior feature set based on the target cluster number pairs to obtain multiple clusters, the method further includes: The scores of electric vehicle users in each cluster across all feature dimensions are calculated based on a preset formula, and the optimal charging feature set is semantically represented based on each score; wherein the preset formula is: in, For the first i Class User j Individual tag scores, For all belonging to the first i Class User j The average value of each label, For the first j The minimum value of each label. For the first j The maximum value of each label.

6. The user demand response potential assessment method based on GMM clustering and information entropy according to claim 3, characterized in that, Before calculating the approximation between the charging behavior characteristics of electric vehicle users corresponding to that user profile and the ideal solution using the TOPSIS method in the optimal feature set, and obtaining the demand response potential of the electric vehicle users corresponding to that user profile, the method further includes: The weights of each evaluation index in the optimal feature set are determined using the entropy weight method.

7. The user demand response potential assessment method based on GMM clustering and information entropy according to claim 6, characterized in that, For each user profile, the TOPSIS method is used to calculate the degree of similarity between the charging behavior characteristics of the electric vehicle user corresponding to that user profile in the optimal feature set and the ideal solution, thereby obtaining the demand response potential of the electric vehicle user corresponding to that user profile, including: An original data matrix is ​​established based on all charging behavior features in the optimal feature set, and then dimensionless processing is performed to obtain a normalized matrix. Weights are assigned to the indices in the normalized matrix, and a weighted normalized matrix is ​​constructed. The maximum value of each column element in the weighted normalized matrix is ​​taken as the positive ideal solution, and the minimum value of each column element is taken as the negative ideal solution. For each user profile, the distance from the charging behavior features of the electric vehicle user corresponding to that user profile in the optimal feature set to the positive ideal solution and the negative ideal solution is calculated, and the relative proximity to the positive ideal solution and the negative ideal solution is calculated to obtain the demand response potential of the electric vehicle user corresponding to that user profile.

8. A user demand response potential assessment device based on GMM clustering and information entropy, characterized in that, include: The feature construction module is used to construct the original feature set of electric vehicle users' charging behavior based on time-flexible and power-adjustable indicators. The user profiling module is used to cluster the original feature set of charging behavior based on the SOM model and GMM clustering algorithm to obtain multiple user profiles; wherein each user profile corresponds to a type of electric vehicle user. The potential assessment module is used to calculate the degree of similarity between the charging behavior characteristics of electric vehicle users corresponding to each user profile and the ideal solution using the TOPSIS method, thereby obtaining the demand response potential of electric vehicle users corresponding to that user profile.

9. The user demand response potential assessment device based on GMM clustering and information entropy according to claim 8, characterized in that, The user profiling module is specifically used for: Based on the SOM model, the data in the original feature set of the charging behavior is mapped to a low-dimensional space to obtain the mapped charging behavior feature set. The target number of clusters is determined using the sum of squared errors as the clustering evaluation index. The GMM clustering algorithm is used to cluster the charging behavior feature set mapped from the target cluster number, resulting in multiple clusters; each cluster corresponds to a user profile.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

Images