A community orderly intelligent charging method and device

Abstract

The present application discloses a community orderly intelligent charging method and device, which are used to optimize the allocation and management of power resources in the community. The method includes the following steps: first, the data to be processed is acquired in real time through the intelligent acquisition terminal of the Internet of Things and the power space available for vehicle charging is determined. Subsequently, the data to be processed is jointly clustered, key features are extracted and feature weights are adjusted to generate a charging priority sequence. An orderly judgment is made based on the total power of the vehicle to be charged and the power space available for charging. If the orderly control state is entered, the output power and charging time of the charging pile are dynamically adjusted, and the power space data is continuously monitored through an adaptive exit control model to ensure that the normal working state of the charging pile is gradually restored after the power space is restored to the preset threshold. This effectively solves the problems of overload of the power system and uneven distribution of charging resources during the peak period of charging demand, and realizes the allocation and efficient utilization of charging resources.

Description

Technical Field

[0001] The present invention relates to the technical field of charging piles, and in particular to a community orderly intelligent charging method and device. Background Art

[0002] With the continuous growth of sales of new energy vehicles, charging piles, as the most basic and most critical supporting facilities, have been widely used and continuously developed. Traditional charging piles mainly refer to high-power DC charging piles built in fast charging stations. However, with the establishment of fast charging stations in major cities and the saturation of the market, small and medium-power AC charging piles in communities have become a very important supplement to the charging and energy replenishment network. However, in some old communities or communities with high occupancy rates, the problem of too many charging piles has become increasingly prominent. This has led to the total power demand of the community far exceeding the total power supply power of the community, resulting in insufficient charging and even overloading and damage to the power supply equipment. This phenomenon not only affects the normal electricity consumption of residents, but also poses a threat to the safety and stability of the community power system. At the same time, there are the following problems in the community intelligent charging system: insufficient real-time data collection and processing, limited dynamic management capabilities of charging resources, lack of intelligent charging priority management based on feature analysis, lack of effective orderly control mechanism and adaptive exit control mechanism, and inability to achieve optimal allocation and efficient utilization of charging resources during peak charging demand periods. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a community orderly intelligent charging method and device thereof to solve the overload problem of the community power system under high charging demand, insufficient real-time data collection and processing, limited dynamic management capability of charging resources, lack of intelligent charging priority management based on feature analysis, lack of effective orderly control mechanism and adaptive exit control mechanism, and inability to achieve optimal allocation and efficient utilization of charging resources during peak charging demand periods.

[0004] A community orderly intelligent charging method, characterized by comprising the following steps:

[0005] S1, obtaining the data to be processed, wherein the data to be processed is connected to the community main transformer through the IoT intelligent collection terminal to collect the current community power load data, real-time load data and the total power demand of the vehicles to be charged in real time, and obtain the historical charging behavior data;

[0006] S2, performing surplus space processing on the data to be processed, so as to obtain power space data available for vehicle charging;

[0007] S3, jointly clustering the power load data and the charging demand data to form a feature subset; calling the feature optimization model to extract key features from the feature subset; adjusting the feature weight according to the key features to obtain the feature weight;

[0008] S4. Based on the feature weights in each feature subset and combined with the real-time load and historical charging behavior, a charging priority sequence is dynamically generated through dynamic priority processing;

[0009] S5, making an orderly judgment based on the total power of the vehicles to be charged and the power space available for charging the vehicles; if the total power of the vehicles to be charged is not less than the power space available for charging the vehicles, entering an orderly control state, which is used to dynamically adjust the output power and charging time of the charging pile;

[0010] S6, in the orderly control state, calls the adaptive exit ordered state control model, which continuously monitors the power space data available for vehicle charging. When it is monitored that the power space available for vehicle charging continues to increase and reaches a preset threshold, it triggers the adaptive recovery control, thereby controlling the recovery order and power output adjustment of each charging pile, and gradually restoring the normal working state of the charging pile.

[0011] Furthermore, the power space P available for vehicle charging available (t), the formula is P available (t) = P total -P load (t)-P reserve , where P reserve It is the power space reserved to ensure the stability of the system. total is the total transformer capacity, P load (t) is the total electrical load.

[0012] Furthermore, the S3 step includes the following sub-steps:

[0013] S31. Performing a joint cluster analysis on the power load data and the charging demand data to form different feature categories and obtain feature subsets;

[0014] S32. Calling the feature optimization model, inputting the feature subset into the feature optimization model, thereby extracting key features;

[0015] S33. Adjust the feature weight according to the key feature to obtain the feature weight; the feature weight adjustment determines the weight of each key feature by analyzing the impact of the key feature on charging demand prediction and load balancing.

[0016] Furthermore, the S4 step includes the following sub-steps:

[0017] S41, calling the feature weight aggregation model, performing weighted aggregation on the feature weights in each feature subset, and obtaining a comprehensive feature weight;

[0018] S42, constructing a feature contribution matrix based on the comprehensive feature weights to quantify the contribution of each feature to the charging demand priority; the feature contribution matrix quantifies the importance of each feature in the priority decision by analyzing the performance of each feature under different charging demand scenarios;

[0019] S43, dynamically adjusting the feature contribution and optimizing the feature contribution matrix in combination with the real-time load and historical charging behavior; the dynamic adjustment adjusts the weight of the feature contribution by real-time monitoring and analyzing the changes in the power load and charging behavior to adapt to the current power demand and load balance;

[0020] S44, calling the priority decision model to dynamically generate a charging priority sequence based on the optimized feature contribution matrix; the priority decision model dynamically generates a charging priority sequence by comprehensively considering the weights of each feature in the feature contribution matrix to achieve allocation and utilization of charging resources.

[0021] Furthermore, in the step S5, the orderly control state dynamically adjusts the output power and charging time of the charging pile in real time through the following steps:

[0022] S51, acquiring real-time output power data and charging time data of the charging pile, generating a real-time data sequence, and performing prediction processing in combination with historical data to obtain a predicted data sequence;

[0023] S52, obtaining an error sequence by predicting the data sequence and combining it with the real-time data sequence;

[0024] S53: Based on the error sequence and the priority sequence, a differential evolution algorithm is used to dynamically adjust the output power and charging time of the charging pile.

[0025] Furthermore, the prediction processing in step 51 predicts the future output power and charging time of each charging pile through a spatiotemporal graph convolutional network.

[0026] Furthermore, the differential evolution algorithm in step S53 dynamically adjusts the output power and charging time of the charging pile through three main operations: mutation, crossover and selection through a fitness function. The calculation formula of the fitness function is as follows: Among them, α, β and γ are weight coefficients, S i represents the priority score of the i-th vehicle, and the error sequence E(t)=(P predicted (t)-P out (t),T predicted (t)-T charge (t)) reflects the difference between the predicted value and the actual value.

[0027] Furthermore, the triggering adaptive recovery control judgment condition in step S6 is as follows: Then it enters the adaptive recovery control state, where P threshold is the recovery threshold, P available (t) is the power space available for vehicle charging.

[0028] Furthermore, the adaptive recovery control in the adaptive exit ordered state control model in step S6 includes priority control of the recovery order and power output adjustment control, wherein the power output adjustment control is performed by setting the target power output P of each charging pile. target , according to the current power space P available for vehicle charging available (t) and charging demand P demand , gradually adjust the power output, the power output adjustment formula in each time period t is: Where i represents the number of the charging pile currently restored, P out (t) is the actual output power of the i-th charging pile at time t, P target is the target power output of the i-th charging pile, P available (t) is the total power space available for charging at the current moment, It is the total power that has been allocated to the previous charging pile.

[0029] In addition, the present application also discloses a community orderly intelligent charging device, characterized in that the device comprises:

[0030] A data acquisition unit is used to acquire the data to be processed. The data to be processed is connected to the community main transformer through the Internet of Things intelligent acquisition terminal to collect the current community power load data, real-time load data and the total power demand of the vehicles to be charged in real time, and acquire historical charging behavior data;

[0031] A power space processing unit, used for performing surplus space processing on the data to be processed, so as to obtain power space data available for charging the vehicle;

[0032] A feature clustering optimization unit is used to jointly cluster the power load data and the charging demand data to form a feature subset; call the feature optimization model to extract key features; adjust the feature weight according to the key features to obtain the feature weight;

[0033] A priority processing unit, configured to dynamically generate a charging priority sequence through dynamic priority processing based on feature weights in each feature subset and in combination with real-time load and historical charging behavior;

[0034] An orderly judgment control unit is used to make an orderly judgment based on the total power of the vehicles to be charged and the power space available for charging the vehicles; if the total power of the vehicles to be charged is not less than the power space available for charging the vehicles, an orderly control state is entered, and the orderly control state is used to dynamically adjust the output power and charging time of the charging pile;

[0035] The adaptive recovery control unit is used to call the adaptive exit ordered state control model in the orderly control state. The adaptive exit ordered state control model continuously monitors the power space data available for vehicle charging. When it is monitored that the power space available for vehicle charging continues to increase and reaches a preset threshold, the adaptive recovery control is triggered, thereby controlling the recovery order and power output adjustment of each charging pile, and gradually restoring the normal working state of the charging pile.

[0036] This application provides a community orderly intelligent charging method and device, which effectively prevents power system overload by real-time monitoring and dynamic adjustment of the output power and charging time of the charging pile. At the same time, the dynamic priority processing and adaptive control mechanism are adopted to enhance the dynamic management capability of charging resources, ensure that high-priority vehicles have priority in obtaining charging services, and improve resource utilization efficiency. The introduced orderly control mechanism can reasonably limit or suspend charging to ensure the overall balance and safety of the system, while the adaptive exit control mechanism can quickly restore the normal operation of the system after the peak demand, optimizing the allocation of charging resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A method flow chart of a method for orderly intelligent charging of a community;

[0038] Figure 2 A schematic diagram of the feature clustering optimization process in a community orderly smart charging method;

[0039] Figure 3 A schematic diagram of a process of dynamically generating a charging priority sequence in a community orderly smart charging method;

[0040] Figure 4 A schematic flow chart of a method for dynamically adjusting the output power and charging time of a charging pile in real time in a sequence control state in a community orderly intelligent charging method;

[0041] Figure 5 A schematic diagram of the structure of a community orderly intelligent charging device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] The following will be combined with the accompanying drawings of the present invention to clearly and completely describe the simulation technology route in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] With the growth in sales of new energy vehicles, charging piles have become increasingly important as an infrastructure to support their development. Traditional charging piles mainly refer to DC fast charging piles, which are usually set up in fast charging stations and have high charging power. However, with the establishment of fast charging stations in first- and second-tier cities and the gradual saturation of the market, facing the continued growth in the number of new energy vehicles, AC slow charging piles established in communities have become increasingly important. However, the establishment of AC slow charging piles in communities faces some challenges. One of the main bottlenecks is the limitation of the total load power of the community transformer. As the number of new energy vehicles in each community increases, the demand for charging has also gradually increased. Especially in old communities or communities with high occupancy rates, the power required for vehicle charging has begun to gradually affect the normal electricity demand of residents. Faced with this situation, some community properties have taken measures to prevent users from installing charging piles, but this approach cannot completely solve the problem of community charging, and may even affect the vital interests of the owners. Therefore, the problem of new energy charging piles has become a prominent problem in the field of property services complained by owners. The present invention proposes a method for orderly intelligent charging in a community, please refer to the attached Figure 1 As shown, the following steps are included:

[0045] S1, obtaining the data to be processed, wherein the data to be processed is connected to the community main transformer through the IoT intelligent collection terminal to collect the current community power load data, real-time load data and the total power demand of the vehicles to be charged in real time, and obtain the historical charging behavior data;

[0046] Specifically, by connecting the IoT intelligent data collection terminal to the community main transformer, the real-time collection of community power load data, real-time load data, and the total power demand of vehicles to be charged can be realized, and historical charging behavior data can be obtained. The IoT intelligent data collection terminal includes multiple sensors and communication modules. These sensors are installed at the key nodes of the community main transformer and can monitor and collect the current community power load data in real time. Power load data refers to the total power consumption of the entire community within a specific time period, including the power consumption of all residents and public facilities. At the same time, real-time load data refers to the instantaneous power consumption of the community at a specific moment. Through the high-frequency sampling of the collection terminal, the real-time load status of the community can be accurately obtained.

[0047] The total power demand of the vehicles to be charged is collected in real time through the communication module with the charging pile management system. When each vehicle to be charged is connected to the charging pile, its charging demand data will be uploaded to the management system and obtained in real time by the collection terminal. The historical charging behavior data is retrieved and analyzed by calling and analyzing the historical records of the charging pile management system, including detailed information such as the charging start time, duration, charging power and power of all charging vehicles in the past period of time, so as to obtain the historical average load P avg_load and the historical average demand P avg_demand , used to analyze past charging behavior patterns and load changes.

[0048] The real-time collection and processing of all the above data are realized through the built-in computing unit and communication module of the IoT intelligent collection terminal. The collected data are uploaded to the central control system in real time through wireless or wired networks. The central control system will perform preliminary processing and storage on these data for use in subsequent steps.

[0049] In this way, the system can obtain the current community's power load status, real-time load status and detailed requirements of each vehicle to be charged in a short time, and combine historical charging behavior data to provide a reliable data basis for subsequent power space calculation, feature extraction and priority generation. This combination of real-time and historical data can not only improve the system's dynamic response capability to power load, but also provide accurate data support for the formulation of intelligent charging strategies, ensuring the efficient and orderly charging process.

[0050] S2, performing surplus space processing on the data to be processed, so as to obtain power space data available for vehicle charging;

[0051] Specifically, the power load data, real-time load data and total power demand of the vehicles to be charged are processed to determine the power space that can be currently allocated to the vehicles to be charged. The data obtained from the IoT intelligent acquisition terminal is preprocessed to ensure its accuracy and consistency. The preprocessing includes data cleaning, denoising and format conversion operations. Then the total power load P of the cell is calculated based on the current power load data and real-time load data. load (t), where N is the number of all electrical equipment in the community, P i (t) is the power consumption of the ith device at time t. After obtaining the current total power load, the system will combine the capacity of the total transformer in the community P total and real-time load data P real (t) is further calculated. Real-time load data is obtained through high-frequency sampling, which represents the instantaneous power consumption at the current moment. Next, the power space P available for vehicle charging is calculated.available (t), the formula is P available (t) = P total -P load (t)-P reserve , where P reserve It is the power space reserved to ensure system stability and to cope with sudden power demand or load fluctuations.

[0052] Through the above steps, on the basis of real-time monitoring of power load and real-time load data, the power space available for vehicle charging can be obtained in real time. This lays the foundation for the dynamic management of the subsequent charging process. This mechanism can not only maximize the power space available for vehicle charging, but also ensure the stable operation of the power system. It should be noted that if trend analysis and prediction are combined with historical charging behavior data, the reserved power can be adjusted more accurately to cope with future charging demand peaks, further improving charging efficiency and user satisfaction. I will not go into too much detail here.

[0053] S3, jointly clustering the power load data and the charging demand data to form a feature subset; calling the feature optimization model to extract key features; adjusting the feature weights according to the key features to obtain feature weights;

[0054] For details, please refer to the attached Figure 2 As shown, the S3 step includes the following sub-steps:

[0055] S31. Performing a joint cluster analysis on the power load data and the charging demand data to form different feature categories and obtain feature subsets;

[0056] In this step, the system performs a joint cluster analysis on the power load data and the charging demand data to form different feature categories and obtain feature subsets. First, the power load data and charging demand data obtained by the acquisition terminal constitute a multidimensional data set, in which each data point represents the load and demand situation within a time period. In order to perform cluster analysis, the K-means clustering algorithm is used in this embodiment. The algorithm distributes data points into k clusters through iterative optimization, so that the similarity between data points in each cluster is maximized, and the difference between different clusters is maximized. Specifically, the data set X = {x1, x2, ..., x n}, where x i represents the i-th data point, including the power load P load,i and the charging demand P demand,i The goal of the K-means algorithm is to minimize the intra-cluster square error, and its objective function is: Among them C j represents the jth cluster, μ j It is cluster C jThe algorithm steps include initializing the centroids of k clusters, and then iterating two steps: (1) assigning each data point to the cluster to which the nearest centroid belongs; (2) recalculating the centroid of each cluster until the centroid position no longer changes significantly or the maximum number of iterations is reached. In this way, the power load and charging demand data are assigned to k different clusters, each cluster representing a feature category. The data points in each feature category have similar load and demand patterns, and these clusters are feature subsets. For example, in this embodiment, a cluster analysis is performed on the community data, and the results divide the data into three feature categories, each category representing a feature subset:

[0057] Feature subset 1: High load-high demand period. Specifically, the data points of this feature subset represent the time period when the electricity load in the community is high and the charging demand is also high. For example, during the evening peak period (18:00-21:00), many residents go home after get off work, the household electricity load increases, and a large number of electric vehicles need to be charged.

[0058] Feature subset 2: low load-high demand period. Specifically, the data points of this feature subset represent the time period when the electricity load in the community is low but the charging demand is high. For example, at night (21:00-23:00), the household electricity load is reduced, but there are still many electric vehicles charging.

[0059] Feature subset 3: Medium load-low demand period. Specifically, the data points of this feature subset represent the time period when the electricity load in the community is medium, but the charging demand is low. For example, during the daytime working hours (10:00-16:00), the household electricity load is moderate and most electric vehicles are not charging.

[0060] This cluster analysis can identify the electricity consumption and charging demand patterns in different time periods within the community, thereby forming different feature subsets. These feature subsets provide a basis for subsequent feature extraction and optimization, enabling the system to dynamically adjust and optimize charging priorities according to actual conditions, ensuring efficient use of power resources within the community and orderly charging processes.

[0061] S32. Calling the feature optimization model, inputting the feature subset into the feature optimization model, thereby extracting key features;

[0062] Specifically, each feature subset obtained from step S31 is used as input data and input into the feature optimization model. Each feature subset contains a set of multidimensional data points, which are composed of various related features, including but not limited to current power load, charging demand, time period, historical average load and historical average demand. In one or more embodiments, the feature optimization model is one of principal component analysis, linear discriminant analysis, and random forest. The feature optimization model extracts the features that best explain the data differences by performing dimensionality reduction processing on the input feature subset. In this embodiment, principal component analysis is used for processing, and the original feature space is converted to a new feature space through linear transformation, so that the converted new features are independent of each other and sorted by variance size. The specific process is as follows: First, the feature subset X is input to calculate the covariance matrix Σ, and the formula is: Among them, x i represents the i-th data point, μ is the mean vector of the data. Then the covariance matrix Σ is decomposed to obtain the eigenvalue λ and eigenvector v: ∑ v =λ v , the eigenvalue λ represents the variance of the new feature, and the eigenvector v represents the direction of the new feature. The eigenvectors with the largest eigenvalues ​​are selected as new features, and the features corresponding to these eigenvectors are the key features. The new feature matrix Y is obtained by projecting the original data X onto the eigenvector v: Y = XV k , where V k is a matrix composed of the first k feature vectors. For example, in this embodiment, there is a feature subset, in which each data point includes the following features: current power load (P load ), charging demand (P demand ), time period (Time), historical average load (P avg_load ), historical average demand (P avg_demand ). The PCA model is used to extract the key features that best explain the data changes. The input data is the feature subset X, specifically X = [P load , P demand ,Time,P avg_load , P avg_demand], calculate the covariance matrix Σ, then perform eigenvalue decomposition to obtain eigenvalues ​​λ1,λ2,…,λ5 and corresponding eigenvectors v1,v2,…,v5, select the eigenvectors v1,v2,v3 with the largest eigenvalues ​​as new key features, and finally project the original data into the new feature space to obtain a new feature matrix Y=[Y1,Y2,Y3]. The key features Y1,Y2,Y3 obtained after processing by the feature optimization model correspond to the fluctuation characteristics of the power load, the change pattern of charging demand, and the influence of historical data in this embodiment. The key features will be used for further weight adjustment and dynamic generation of charging priority in subsequent steps to ensure that the system can operate efficiently under complex load and demand environments.

[0063] S33. Adjust the feature weight according to the key feature to obtain the feature weight; the feature weight adjustment determines the weight of each key feature by analyzing the impact of the key feature on charging demand prediction and load balancing.

[0064] Specifically, the key feature Y obtained from step S32 is used as input data and input into the feature weight adjustment model. In this embodiment, the key features include important features extracted from the original features, including the current power load fluctuation characteristics, the charging demand change pattern, and the impact of historical data. By analyzing the contribution of each key feature to the system goal, its importance is evaluated. In this example, a multivariate linear regression model is used to analyze the impact of each key feature on the charging demand prediction and calculate the regression coefficient of the feature. The form of the regression model is: Y demand =β0+β1Y1+β2Y2+...+β n Y n +ε, where Y demand is the predicted value of charging demand, Y i is the key feature, β i is the feature coefficient, and ε is the error term. The importance of each feature is evaluated by analyzing the size and significance of the regression coefficient. Next, the weight of each key feature is assigned according to the result of the feature importance evaluation. In this example, the weight assignment uses the normalization method so that the sum of the weights of all features is 1. Specifically, the importance score of each feature is I i , then the weight w of each feature i By normalization calculation, we get: Among them I iis the importance score of the i-th feature, and n is the total number of key features. Finally, the assigned weights are verified and optimized through experiments and simulations. In this embodiment, a cross-validation method is used to evaluate the performance of feature weights in actual predictions through training and validation data sets. According to the prediction results, the feature weights are further adjusted and optimized to maximize the accuracy of charging demand prediction and load balancing. The weight of each key feature is determined through the above process to ensure that these weights can accurately reflect the importance and impact of each feature in charging demand prediction and load balancing. The final feature weights will be used for subsequent dynamic priority processing and the generation of charging priority sequences to achieve the optimization goal of orderly intelligent charging in the community.

[0065] S4. Based on the feature weights in each feature subset and combined with the real-time load and historical charging behavior, a charging priority sequence is dynamically generated through dynamic priority processing;

[0066] For details, please refer to the attached Figure 3 As shown, the S4 step includes the following sub-steps:

[0067] S41, calling a feature weight aggregation model to perform weighted aggregation on the feature weights in each feature subset to obtain a comprehensive feature weight;

[0068] In this embodiment, characteristic weights are extracted from different feature subsets. The feature subsets are high load-high demand period, low load-high demand period, and medium load-low demand period. Each feature subset contains key features Y i Its key characteristics include the fluctuation characteristics of power load, the changing pattern of charging demand, and the influence of historical data. First, determine the weighting coefficient, weighting coefficient α i Used to reflect the importance of each feature subset in the comprehensive feature weight. In one or more embodiments, the weighting coefficient can be automatically determined based on actual needs, data analysis results, or through a machine learning algorithm. In this embodiment, the weighting coefficients are α1, α2, and α3, and meet the following conditions: α i ≥0. Next, the weights in each feature subset are integrated through layer-by-layer weighted aggregation to ensure that the comprehensive feature weight can fully reflect the importance of each feature. The first layer of aggregation performs weighted aggregation within each feature subset to generate the initial weight of the subset. Assume that the feature subset S i The key feature weight contained in is Y i1 ,Y i2 ,…,Y in , and its weighted aggregation formula is as follows: Where W i is the feature subset X i The initial weight, β ijis the weight coefficient of each key feature in the feature subset, satisfying The second layer of aggregation is to perform weighted aggregation of the preliminary weights across subsets to generate a comprehensive feature weight W. The formula is as follows: Where W is the final comprehensive feature weight, W i is the feature subset X i The initial weight, α i is the weight coefficient of the feature subset. It is worth mentioning that in this embodiment, in fact, X i The key features included in include the fluctuation characteristics of power load, the change pattern of charging demand, and the impact of historical data, that is, it is calculated according to the three. At the same time, in this embodiment, a dynamic weight adjustment mechanism is introduced to dynamically adjust the weight coefficient α according to the changes in real-time data and historical data. i and β ij To ensure that the comprehensive feature weights can reflect the current actual situation in real time. Specifically, through the IoT intelligent collection terminal, the community power load data, charging demand data and other related data are monitored in real time, and the feature weights in the feature subset are dynamically updated. Adopting an adaptive adjustment algorithm, the weighting coefficient α is dynamically adjusted according to the changes in real-time monitoring data and historical data. i and β ij In one or more embodiments, the adaptive adjustment algorithm can be implemented based on gradient descent, Bayesian optimization or other optimization algorithms to minimize prediction error and improve model accuracy.

[0069] In summary, through the multi-layer weighted aggregation strategy and dynamic weight adjustment mechanism, the comprehensive feature weight W is finally generated. It is worth mentioning that the key feature weight is used for feature extraction and optimization, while the comprehensive feature weight is used for comprehensive analysis and decision-making. The comprehensive feature weight can fully reflect the importance of each feature in different time periods, different loads and demand conditions, and provide a reliable basis for subsequent priority decisions.

[0070] S42, constructing a feature contribution matrix based on the comprehensive feature weights to quantify the contribution of each feature to the charging demand priority; the feature contribution matrix quantifies the importance of each feature in the priority decision by analyzing the performance of each feature under different charging demand scenarios;

[0071] Specifically, based on the comprehensive feature weights, a feature contribution matrix is ​​constructed to quantify the contribution of each feature to the charging demand priority; the feature contribution matrix quantifies the importance of each feature in the priority decision by analyzing the performance of each feature in different charging demand scenarios. In this embodiment, the feature contribution matrix C is first constructed using the comprehensive feature weight W. Specifically, the above steps in this embodiment indicate the key features extracted from the feature subset. The key features include the fluctuation characteristics of the power load, the changing pattern of the charging demand, and the influence of historical data. The feature contribution matrix is ​​defined for this purpose. The feature contribution matrix C is a two-dimensional matrix in which each row represents a feature subset and each column represents a key feature. Matrix element C ij Represents feature subset X i Medium feature Y j Contribution to charging demand priority. For each feature subset X i The feature contribution is calculated using the comprehensive feature weight W and the key feature weights extracted from each feature subset. The specific formula is as follows: ij =α i β ij ·W j , where α i is the feature subset X i The weighting coefficient of reflects the importance of the feature subset in the comprehensive feature weight; β ij is the weighted coefficient of each key feature in the feature subset, reflecting the importance of the key feature in the feature subset; W j is a specific key feature Y j The comprehensive weight of . Through the above calculation, the characteristic contribution matrix C can be obtained, which is as follows: In which, each matrix element C ij Represents feature subset X i Medium feature Y j Finally, the characteristic contribution matrix is ​​adjusted dynamically based on the real-time load and historical charging behavior. Specifically, by real-time monitoring and analyzing the changes in power load and charging behavior, the weights of characteristic contributions are adjusted to adapt to current power demand and load balance. The dynamic adjustment mechanism includes introducing an adaptive adjustment algorithm to dynamically adjust the weighting coefficient α according to the changes in real-time monitoring data and historical data. i and β ij In this embodiment, the adaptive adjustment algorithm is based on gradient descent, Bayesian optimization or other optimization algorithms to minimize the prediction error and improve the model accuracy.

[0072] In summary, through the above steps, we can construct and dynamically adjust the feature contribution matrix C so that it can fully reflect the importance of each feature in the priority decision. This provides a reliable basis for subsequent priority decisions and ensures that the system can operate efficiently under complex load and demand environments.

[0073] S43, dynamically adjusting the characteristic contribution and optimizing the characteristic contribution matrix by combining the real-time load and historical charging behavior; the dynamic adjustment adjusts the weight of the characteristic contribution by real-time monitoring and analyzing the changes in the power load and charging behavior to adapt to the current power demand and load balance;

[0074] Specifically, the community's power load data and charging demand data are monitored in real time through the IoT intelligent data collection terminal. These real-time data include the current power load P load 、Real-time charging demand P demand In addition to real-time data, historical charging behavior data, including historical average load P avg_load and the historical average demand P avg_demand , which is used to analyze past charging behavior patterns and load changes. The real-time monitored power load and charging demand data are combined with historical data to construct a real-time feature matrix X real-time The feature matrix is ​​represented as X real-time =[P load ,P demand ,P avg_load ,P avg_demand ]. Next, a dynamic weight adjustment mechanism is introduced to dynamically adjust the weight coefficients in the characteristic contribution matrix according to the changes in real-time data and historical data. Let the initial characteristic contribution matrix be M, which contains the contribution of each characteristic in different charging demand scenarios. Through the input of real-time data, the weight coefficient α is dynamically adjusted. i and β ij , optimize the characteristic contribution matrix M. The adjustment formula is as follows: α i (t) = a i (t-1)+η·Δα i , β ij =β ij (t-1)+η·Δβ ij , where η is the learning rate, Δα i and Δβ ij They are α i and β ij The adjusted value is calculated based on the current real-time data and historical data. The adjusted weighting coefficient is applied to the characteristic contribution matrix to optimize the contribution of each characteristic in the matrix. The optimized characteristic contribution matrix M optimized It is expressed as: By optimizing the characteristic contribution matrix, the contribution of each characteristic in different charging demand scenarios is quantified to ensure that the comprehensive characteristic weight can reflect the current actual situation in real time. At the same time, a feedback adjustment mechanism is introduced to continuously optimize the characteristic contribution matrix by real-time monitoring and analysis of changes in power load and charging behavior. The feedback mechanism dynamically adjusts the weighting coefficient and characteristic contribution based on the comparison of real-time data and historical data to adapt to the current power demand and load balance. For example, when there is a significant difference between real-time data and historical data, the system will automatically adjust the characteristic contribution matrix to ensure the accuracy and reliability of priority decisions.

[0075] In summary, through the above steps, combined with real-time load and historical charging behavior, the characteristic contribution is dynamically adjusted and the characteristic contribution matrix is ​​optimized to ensure that the system can operate efficiently under complex load and demand environments. This dynamic adjustment mechanism makes charging priority decisions more flexible and intelligent, effectively improving the allocation and utilization efficiency of charging resources.

[0076] S44, calling the priority decision model to dynamically generate a charging priority sequence based on the optimized feature contribution matrix; the priority decision model dynamically generates an optimal charging priority sequence by comprehensively considering the feature weights in the feature contribution matrix to achieve optimal allocation and utilization of charging resources.

[0077] Specifically, the optimized characteristic contribution matrix M optimized It is obtained by comparing, analyzing and dynamically adjusting real-time data and historical data. The matrix reflects the contribution of different characteristics (such as the fluctuation characteristics of power load and the changing pattern of charging demand) in the current power load and charging demand scenarios. The priority decision model is a decision support system based on the optimized characteristic contribution matrix, which aims to generate the optimal charging priority sequence. The model comprehensively considers the weights of each characteristic in the characteristic contribution matrix and dynamically adjusts it according to the current state of charging demand and power load. Specifically, the input data first includes the optimized characteristic contribution matrix M optimized And real-time load data, charging demand data. These data constitute the characteristic characteristics and demand background at the current moment. Next, the characteristic score is calculated. For each vehicle to be charged or charging pile, the characteristic score is calculated using the optimized characteristic contribution matrix according to its current power load and charging demand. Characteristic score S i It is expressed by the following formula: Among them, W ij is the weight in the optimized feature contribution matrix, F ijis the jth characteristic value of the i-th charging demand object, and n is the number of characteristics. Finally, a charging priority sequence is generated based on the calculated characteristic scores. The priority sequence can be obtained by sorting the scores of all charging objects, and the objects with the highest scores are given priority for charging. And the dynamically generated charging priority sequence is used to optimize the allocation of charging resources. The model adjusts the allocation of charging piles according to the priority sequence to ensure the optimal utilization of charging resources while meeting the current power load and charging demand. It is worth mentioning that this embodiment also introduces a feedback mechanism to continuously monitor the effect of charging priority decisions. Monitor the changes in power load and charging demand in real time, and adjust the parameters of the priority decision model regularly to ensure the accuracy and real-time nature of the decision. The feedback mechanism includes monitoring data analysis and dynamic adjustment, and the monitoring data analysis is used to analyze the actual charging situation and load changes and compare them with the expected results. The dynamic adjustment is used to adjust the input parameters and characteristic contribution matrix of the decision model according to the feedback results to improve the generation strategy of the charging priority sequence.

[0078] S5, making an orderly judgment based on the total power of the vehicles to be charged and the power space available for charging the vehicles; if the total power of the vehicles to be charged is not less than the power space available for charging the vehicles, entering an orderly control state, which is used to dynamically adjust the output power and charging time of the charging pile;

[0079] Specifically, in this step, the required data include: the total power P of the vehicle to be charged total Represents the total power demand of all vehicles to be charged. The power space P available for vehicle charging available (t) represents the total power space available for charging at the current moment. This data has been obtained in step S2 and will be dynamically updated according to real-time data. The system will compare the total power of the vehicle to be charged with the power space available for charging. The judgment conditions are as follows: If P total ≥P available (t), the system enters the orderly control state. Otherwise, the system proceeds according to the normal charging plan and does not need to enter the orderly control state. In the orderly control state, the system needs to dynamically adjust the output power and charging time of the charging pile to achieve the optimal allocation of charging resources. The specific steps are as follows:

[0080] S51, acquiring real-time output power data and charging time data of the charging pile, generating a real-time data sequence, and performing prediction processing in combination with historical data to obtain a predicted data sequence, wherein the prediction processing is to predict the future output power and charging time of each charging pile through a spatiotemporal graph convolutional network;

[0081] Specifically, the output power P of the charging pile is monitored in real time through the IoT intelligent data collection terminal. out(t). These data reflect the power output of each charging pile at the current moment; similarly, the charging time T of the charging pile is monitored through the intelligent data collection terminal. charge (t), records the cumulative charging time of each vehicle from the start of charging to the current moment. The collected real-time output power data and charging time data will form a real-time data sequence X real-time For example, for a charging pile, its real-time data sequence can be expressed as: real-time ={(P out (t1),T charge (t1)),(P out (t2),T charge (t2)),...(P out (t n ),T charge (t n ))}, the real-time data sequence will be used for real-time prediction and dynamic adjustment. And collect the historical output power data and charging time data of each charging pile to form a historical data set. The historical data set contains the output power and charging time in multiple time periods, for example: X history ={(P out (t h1 ),T charge (t h1 )),(P out (t h2 ),T charge (t h2 )),...,(P out (t hn ),T charge (t hn ))}. Next, we use the spatiotemporal graph convolutional network (ST-GCN) for prediction. Specifically, we combine the real-time data sequence and the historical data sequence into an input matrix X as input: X = {X real-time ,X history}, the input matrix contains continuous data in the time dimension, including both past data and current real-time data. Graph convolution is performed on the input matrix X to capture spatial dependencies. The relationship between charging piles is represented by the adjacency matrix A, and the input data is converted into a node feature matrix: H = σ(D -1 / 2 AD -1 / 2 XW), perform temporal convolution on the feature matrix H after the graph convolution operation to capture the temporal dependency: Y = σ(W t *H), after the temporal convolution, a pooling operation is performed to reduce the dimension of the feature and capture the main features. In one or more embodiments, a maximum pooling (MaxPooling) or an average pooling (Average Pooling) can be used: Y pooled=Pooling(Y), the pooled feature matrix passes through one or more fully connected layers (also called fully connected neural network layers) to further extract features and generate the final prediction result: Z = σ(W f Y pooled +b f ), W f is the weight matrix of the fully connected layer, b f is the bias vector, σ is the activation function. Finally, the output layer generates the predicted data sequence X at the future moment predicted :X predicted ={(P predicted (t n+1 ),T predicted (t n+1 )),(P predicted (t n+2 ),T predicted (t n+2 )),....,(P predicted (t n+k ),T predicted (t n+k ))}, Through the above steps, the ST-GCN model can use real-time data and historical data to accurately predict the future output power and charging time of each charging pile, thereby generating a prediction data sequence. This process not only takes into account the spatial dependency between charging piles, but also captures the temporal dependency of the output power and charging time of the charging piles, ensuring the accuracy and reliability of the prediction results.

[0082] S52, obtaining an error sequence by predicting the data sequence and combining it with the real-time data sequence;

[0083] Specifically, first align the predicted data sequence and the real-time data sequence to ensure that the predicted data sequence X predicted and real-time data series X real-time The timestamps of the prediction data are aligned to make an accurate comparison. Next, the error sequence E(t) is calculated, which represents the difference between the predicted data and the real-time data. For each time t, the error calculation formula is as follows: E(t) = (P predicted (t)-P out (t),T predicted (t)-T charge (t)), where P predicted (t) and T predicted (t) represent the predicted output power and charging time at time t, respectively, out (t) and T charge (t) represent the real-time output power and charging time at time t respectively. The specific form of the error sequence E(t) is: E(t) = {(e P (t),eT (t))}, where e P (t) = P predictrd (t)-P out (t), e T (t) = T predictrd (t)-T charge (t). Construct an error function, expressed as E = {E(t n+1 ),E(t n+2 ),...,E(t n+k )}. In this step, by comparing the real-time data series and the predicted data series, an error series E is generated to evaluate and optimize the performance of the prediction model. This process ensures that the system can dynamically adapt to real-time changes and improve the efficiency and accuracy of charging resource allocation.

[0084] S53: Based on the error sequence and the priority sequence, a differential evolution algorithm is used to dynamically adjust the output power and charging time of the charging pile.

[0085] The differential evolution algorithm dynamically adjusts the output power and charging time of the charging pile through the fitness function and three main operations: mutation, crossover and selection. Specifically, the differential evolution algorithm parameters include:

[0086] Population size NP: The number of individuals in the differential evolution algorithm.

[0087] Coefficient of variation F: a parameter that controls the range of variation. In this embodiment, it is in the range of [0, 2].

[0088] Crossover probability CR: controls the probability of the crossover operation. In this embodiment, it is in the range of [0, 1].

[0089] Iteration number G: The maximum number of generations that the algorithm runs.

[0090] First, an initial population is generated, and each individual represents a charging scheme, including the output power and charging time of the charging pile. In one or more embodiments, the initial population can be generated randomly or based on historical data and current status. Next, a fitness function f is defined to evaluate the pros and cons of each individual's charging scheme. In this embodiment, the fitness function considers the following factors:

[0091] Error minimization: Minimize the error between predicted data and actual data.

[0092] Priority is met: Vehicles with higher priority are charged first.

[0093] The fitness function can be expressed as: Among them, α, β and γ are weight coefficients, S irepresents the priority score of the i-th vehicle, and the error sequence E(t)=(P predicted (t)-P out (t),T predicted (t)-T charge (t)) reflects the difference between the predicted value and the actual value. Then the differential evolution operation is performed. The differential evolution algorithm includes three main operations: mutation, crossover and selection:

[0094] The mutation is for each individual x in the population i (i-th individual), select three different individuals x r1 、x r2 、x r3 , generate mutant individuals: v i =x r1 +F·(x r2 -x r3 );

[0095] The crossover is performed on each variant individual v i and the original individual x i , perform crossover operation according to the crossover probability CR to generate the test individual u i :

[0096] The selection is for each test individual u i and the original individual x i , according to the fitness function, individuals with higher fitness are selected to enter the next generation:

[0097] Next, we will integrate the adjustment process of error sequence and priority sequence. Specifically, in the differential evolution algorithm, the use of error sequence and priority sequence is mainly reflected in the definition of fitness function and the selection of solution. The fitness function needs to consider the error sequence E(t) and the priority sequence S, where error sequence E(t) = (P predicted (t)-P out (t),T predicted (t)-T charge (t)), which reflects the difference between the predicted data and the actual data. It is worth mentioning that the priority sequence S is used to ensure that high-priority vehicles have priority in obtaining charging resources when charging resources are tight. In the differential evolution process, the priority part ∑ i S i It can ensure that the charging needs of high-priority vehicles are met first.

[0098] Through the above differential evolution algorithm, the output power and charging time of the charging pile can be dynamically adjusted to limit or suspend charging of some or all charging piles:

[0099] Limit charging power: Adjust the output power of the charging pile according to the priority and load of the charging pile to ensure that the charging needs of high-priority vehicles are met while avoiding system overload.

[0100] Adjust charging time: Dynamically adjust the charging time of the charging pile according to the priority sequence to give priority to the charging needs of high-priority vehicles.

[0101] Suspend charging: When the power load is too high or the charging demand is too large, the system temporarily stops charging operations for low-priority vehicles to ensure the balance and safety of the overall power system.

[0102] Through the above process, the differential evolution algorithm can effectively realize charging optimization under orderly control state and ensure the rational allocation and efficient utilization of charging resources.

[0103] S6, in the orderly control state, calls the adaptive exit ordered state control model, which continuously monitors the power space data available for vehicle charging. When it is monitored that the power space available for vehicle charging continues to increase and reaches a preset threshold, it triggers the adaptive recovery control, thereby controlling the recovery order and power output adjustment of each charging pile, and gradually restoring the normal working state of the charging pile.

[0104] Specifically, first set the relevant threshold parameters, including the recovery threshold P threshold and detection window T window . Set the parameters of the recovery strategy. In this embodiment, the parameters of the recovery strategy are the priority strategy of the recovery order and the adjustment strategy of the power output. Through the IoT intelligent collection terminal, continuously monitor the power space P available for vehicle charging at the current moment available (t). Record within a period of time (detection window T window ) to form the historical data sequence X history To further determine the response conditions, specifically, for each time period t, the power space data P available (t), analyze and judge. If in the detection window T window Internal power space data P available (t) continues to increase and reaches the preset threshold value P threshold , then the adaptive recovery control is triggered. Then enter the adaptive recovery control state. Specifically, first determine the recovery order: determine the recovery order of the charging piles according to the priority strategy. The priority strategy is based on the vehicle's priority sequence S and charging requirements. Prioritize the recovery of the charging piles of high-priority vehicles. Then adjust the power output: dynamically adjust the power output of the charging piles to ensure that the normal working state of the charging piles is gradually restored. According to the current power space P available(t) and charging demand, gradually increase the output power of the charging pile. Finally, gradually restore the normal working state of the charging pile: the system gradually restores the normal charging operation of each charging pile to ensure that the power load is within a controllable range. Realize the staged recovery of the charging pile to avoid system instability caused by sudden load increase. In summary, the adaptive recovery control in the adaptive exit ordered state control model includes priority control of the recovery order and power output adjustment control. The priority control of the recovery order is controlled according to the priority sequence of the charging pile, and the recovery order of the charging pile is determined according to the priority score; the power output adjustment control is achieved by setting the target power output P of each charging pile. target , according to the current power space P available for vehicle charging available (t) and charging demand P demand , gradually adjust the power output. The power output adjustment formula in each time period t is: Where i represents the number of the charging pile currently restored, P out (t) is the actual output power of the i-th charging pile at time t, P target is the target power output of the i-th charging pile, P available (t) is the total power space available for charging at the current moment, It is the total power that has been allocated to the previous charging pile.

[0105] Based on the description of the above-mentioned community orderly intelligent charging method embodiment, the present application embodiment also discloses a community orderly intelligent charging device, which can be a computer program (including program code) running the above-mentioned community orderly intelligent charging method. Figure 5 As shown, the community orderly intelligent charging device can operate the following units:

[0106] The data acquisition unit 110 is used to acquire the data to be processed. The data to be processed is connected to the community main transformer through the Internet of Things intelligent acquisition terminal to collect the current community power load data, real-time load data and the total power demand of the vehicles to be charged in real time, and acquire historical charging behavior data;

[0107] The power space processing unit 120 is used to perform surplus space processing on the data to be processed, so as to obtain power space data that can be used for charging the vehicle;

[0108] The feature clustering optimization unit 130 is used to jointly cluster the power load data and the charging demand data to form a feature subset; call the feature optimization model to extract key features; adjust the feature weight according to the key features to obtain the feature weight;

[0109] A priority processing unit 140 is used to dynamically generate a charging priority sequence through dynamic priority processing based on the feature weights in each feature subset and in combination with the real-time load and historical charging behavior;

[0110] An orderly judgment control unit 150 is used to make an orderly judgment based on the total power of the vehicles to be charged and the power space available for charging the vehicles; if the total power of the vehicles to be charged is not less than the power space available for charging the vehicles, an orderly control state is entered, and the orderly control state is used to dynamically adjust the output power and charging time of the charging pile;

[0111] The adaptive recovery control unit 160 is used to call an adaptive exit-ordered state control model in an orderly control state. The adaptive exit-ordered state control model continuously monitors the power space data available for vehicle charging. When it is monitored that the power space available for vehicle charging continues to increase and reaches a preset threshold, the adaptive recovery control is triggered, thereby controlling the recovery order and power output adjustment of each charging pile, and gradually restoring the normal working state of the charging pile.

[0112] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A community orderly intelligent charging method, characterized in that: The following steps are involved: S1, obtaining the data to be processed, wherein the data to be processed is connected to the community main transformer through the IoT intelligent collection terminal to collect the current community power load data, real-time load data and the total power demand of the vehicles to be charged in real time, and obtain the historical charging behavior data; S2, performing surplus space processing on the data to be processed, so as to obtain power space data available for vehicle charging; S3, jointly clustering the power load data and the charging demand data to form a feature subset; Calling a feature optimization model to extract key features from a feature subset; adjusting feature weights according to the key features to obtain feature weights; S4, based on the feature weights in each feature subset and combined with the real-time load and historical charging behavior, a charging priority sequence is dynamically generated through dynamic priority processing; S5, making an orderly judgment based on the total power of the vehicles to be charged and the power space available for charging the vehicles; if the total power of the vehicles to be charged is not less than the power space available for charging the vehicles, entering an orderly control state, which is used to dynamically adjust the output power and charging time of the charging pile; S6, in the orderly control state, calls the adaptive exit ordered state control model, which continuously monitors the power space data available for vehicle charging. When it is monitored that the power space available for vehicle charging continues to increase and reaches a preset threshold, it triggers the adaptive recovery control, thereby controlling the recovery order and power output adjustment of each charging pile, and gradually restoring the normal working state of the charging pile.

2. A community orderly intelligent charging method according to claim 1, characterized in that: The power space P available for vehicle charging available (t), the formula is P available (t) = P total -P load (t)-P reserve , where P reserve It is the power space reserved to ensure the stability of the system. total is the total transformer capacity, P load (t) is the total electrical load.

3. A community orderly intelligent charging method according to claim 1, characterized in that: The S3 step includes the following sub-steps: S31. Performing a joint cluster analysis on the power load data and the charging demand data to form different feature categories and obtain feature subsets; S32. Calling the feature optimization model, inputting the feature subset into the feature optimization model, thereby extracting key features; S33. Adjust the feature weight according to the key feature to obtain the feature weight; the feature weight adjustment determines the weight of each key feature by analyzing the impact of the key feature on charging demand prediction and load balancing.

4. A community orderly intelligent charging method according to any one of claims 1 to 3, characterized in that: The S4 step includes the following sub-steps: S41, calling a feature weight aggregation model to perform weighted aggregation on the feature weights in each feature subset to obtain a comprehensive feature weight; S42, constructing a feature contribution matrix based on the comprehensive feature weights to quantify the contribution of each feature to the charging demand priority; The feature contribution matrix quantifies the importance of each feature in the priority decision by analyzing the performance of each feature under different charging demand scenarios; S43, combining the real-time load and historical charging behavior, dynamically adjusting the feature contribution and optimizing the feature contribution matrix; The dynamic adjustment adjusts the weight of the characteristic contribution by real-time monitoring and analyzing the changes in power load and charging behavior to adapt to the current power demand and load balance; S44, calling the priority decision model to dynamically generate a charging priority sequence based on the optimized feature contribution matrix; the priority decision model dynamically generates a charging priority sequence by comprehensively considering the weights of each feature in the feature contribution matrix to achieve allocation and utilization of charging resources.

5. A community orderly intelligent charging method according to claim 4, characterized in that: In step S5, the orderly control state dynamically adjusts the output power and charging time of the charging pile in real time through the following steps: S51, acquiring real-time output power data and charging time data of the charging pile, generating a real-time data sequence, and performing prediction processing in combination with historical data to obtain a predicted data sequence; S52, obtaining an error sequence by predicting the data sequence and combining it with the real-time data sequence; S53: Based on the error sequence and the priority sequence, a differential evolution algorithm is used to dynamically adjust the output power and charging time of the charging pile.

6. A community orderly intelligent charging method according to claim 5, characterized in that: The prediction process in step S51 is to predict the future output power and charging time of each charging pile through a spatiotemporal graph convolutional network.

7. A community orderly intelligent charging method according to claim 5, characterized in that: The differential evolution algorithm in step S53 dynamically adjusts the output power and charging time of the charging pile through three main operations: mutation, crossover and selection through the fitness function. The calculation formula of the fitness function is as follows: Among them, α, β and γ are weight coefficients, S i represents the priority score of the i-th vehicle, and the error sequence E(t)=(P predicted (t)-P out (t),T predicted (t)-T charge (t)) reflects the difference between the predicted value and the actual value, where P predicted (t) and T predicted (t) represent the predicted output power and charging time at time t, respectively, out (t) and T charge (t) represent the real-time output power and charging time at time t respectively.

8. A community orderly intelligent charging method according to claim 5, characterized in that: The triggering adaptive recovery control judgment condition in step S6 is as follows: Then it enters the adaptive recovery control state, where P threshold is the recovery threshold, P available (t) is the power space available for vehicle charging, T window For the detection window.

9. A community orderly intelligent charging method according to claim 8, characterized in that: The adaptive recovery control in the adaptive exit ordered state control model in step S6 includes priority control of the recovery order and power output adjustment control. The power output adjustment control is achieved by setting the target power output P of each charging pile. target , according to the current power space P available for vehicle charging available (t) and charging demand P demand , gradually adjust the power output, the power output adjustment formula in each time period t is: Where i represents the number of the charging pile currently restored, P out (t) is the actual output power of the i-th charging pile at time t, P target is the target power output of the i-th charging pile, P available (t) is the total power space available for charging at the current moment, It is the total power that has been allocated to the previous charging pile.

10. A community orderly intelligent charging device, characterized in that: The device comprises: A data acquisition unit is used to acquire the data to be processed. The data to be processed is connected to the community main transformer through the Internet of Things intelligent acquisition terminal to collect the current community power load data, real-time load data and the total power demand of the vehicles to be charged in real time, and acquire historical charging behavior data; A power space processing unit, used for performing surplus space processing on the data to be processed, so as to obtain power space data available for charging the vehicle; A feature clustering optimization unit is used to jointly cluster the power load data and the charging demand data to form a feature subset; call the feature optimization model to extract key features; adjust the feature weight according to the key features to obtain the feature weight; A priority processing unit, configured to dynamically generate a charging priority sequence through dynamic priority processing based on feature weights in each feature subset and in combination with real-time load and historical charging behavior; An orderly judgment control unit is used to make an orderly judgment based on the total power of the vehicles to be charged and the power space available for charging the vehicles; if the total power of the vehicles to be charged is not less than the power space available for charging the vehicles, an orderly control state is entered, and the orderly control state is used to dynamically adjust the output power and charging time of the charging pile; The adaptive recovery control unit is used to call the adaptive exit ordered state control model in the orderly control state. The adaptive exit ordered state control model continuously monitors the power space data available for vehicle charging. When it is monitored that the power space available for vehicle charging continues to increase and reaches a preset threshold, the adaptive recovery control is triggered, thereby controlling the recovery order and power output adjustment of each charging pile, and gradually restoring the normal working state of the charging pile.

Images