Electric vehicle aggregation regulation and control method and system

Abstract

The invention discloses an electric vehicle aggregation regulation and control method and system, and the method comprises the steps: carrying out the association processing of vehicle feature data and aggregation resource data of different types of electric vehicles through employing an equivalence conversion algorithm, and obtaining the charging behavior feature data of each vehicle; inputting the charging behavior characteristic data of each vehicle into a vehicle charging load prediction model constructed by adopting a Monte Carlo method to obtain charging load data of each electric vehicle; inputting the charging load data into a pre-constructed integer programming model, and solving the integer programming model by adopting a branch and bound algorithm to obtain target charging load data; constructing a corresponding electric vehicle aggregation regulation and control strategy based on the target charging load data; and in the aggregation regulation and control process of the electric vehicles, the corresponding aggregation regulation and control strategy of the electric vehicles is executed according to the types of the to-be-charged electric vehicles. According to the method provided by the embodiment of the invention, the effect of improving the charging efficiency is achieved by optimizing the electric vehicle aggregation regulation and control strategy.

Description

Technical Field

[0001] The present invention relates to the field of electric vehicles, and in particular to an electric vehicle aggregation control method and system. Background Art

[0002] In recent years, with the large-scale penetration of electric vehicles, their disorderly charging will cause a lot of adverse effects on the power system. Electric vehicles can be connected to the power grid most of the time, and their loads have a high degree of operational flexibility.

[0003] With the continuous development of electric vehicles, in the existing technology, the load side lacks an effective electric vehicle charging control strategy system, which makes it difficult for electric vehicles to achieve the expected results during the charging process and the charging efficiency is low.

[0004] It can be seen that how to optimize the electric vehicle aggregation regulation strategy has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] The present invention provides an electric vehicle aggregation control method and system, which optimizes the electric vehicle aggregation control strategy to improve the charging efficiency.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides an electric vehicle aggregation control method, comprising:

[0007] The equivalence class conversion algorithm is used to associate the vehicle characteristic data and aggregated resource data of different types of electric vehicles to obtain the charging behavior characteristic data of each vehicle;

[0008] Inputting the charging behavior characteristic data of each of the vehicles into a vehicle charging load prediction model constructed by using a Monte Carlo method to obtain charging load data of each electric vehicle, wherein the construction process includes dynamically adjusting the parameters of the vehicle charging load prediction model through a grid search and a Bayesian optimization method to optimize the vehicle charging load prediction model;

[0009] Inputting the charging load data into a pre-built integer programming model, solving the integer programming model using a branch and bound algorithm, and obtaining target charging load data;

[0010] Constructing a corresponding electric vehicle aggregation control strategy based on the target charging load data;

[0011] During the electric vehicle aggregation control process, the corresponding electric vehicle aggregation control strategy is executed according to the type of electric vehicle to be charged.

[0012] As one of the preferred solutions, before using the equivalence class conversion algorithm to associate different types of electric vehicle feature data and aggregated resource data, the electric vehicle aggregation control method further includes:

[0013] Perform data cleaning on multi-source heterogeneous electric vehicle data, use ontology mapping method to solve data semantic heterogeneity, and obtain the electric vehicle vehicle characteristic data;

[0014] The static attribute data of the charging piles and the peak load demand data of the power grid are integrated to determine the available capacity of the aggregated resources, and the recursive feature elimination method is used to extract features of the available capacity of the aggregated resources to obtain the aggregated resource data.

[0015] As one of the preferred solutions, the method of solving the integer programming model by using a branch and bound algorithm includes:

[0016] The integer programming model is decomposed into a branch model by the branch and bound algorithm, and the branch model is solved by using the relaxed integer constraints and the breadth-first search strategy.

[0017] As one of the preferred solutions, after executing the corresponding electric vehicle aggregation control strategy according to the type of electric vehicle to be charged, the electric vehicle aggregation control method further includes:

[0018] The aggregation control effect is evaluated using an incremental learning method based on a self-organizing incremental learning neural network, and the aggregation control strategy is dynamically optimized using a deep Q-network algorithm.

[0019] As one of the preferred solutions, the aggregated control effect includes vehicle status data and charging power data;

[0020] The method of evaluating the aggregation control effect by using an incremental learning method based on a self-organizing incremental learning neural network and dynamically optimizing the aggregation control strategy by using a deep Q-network algorithm includes:

[0021] Preprocessing the vehicle status data and the charging power data using an incremental decision tree algorithm to obtain a data set;

[0022] Using Spark Streaming to perform real-time analysis on the data set to obtain evaluation indicators, wherein the evaluation indicators include a fit between a charging load curve and an expected curve and a response rate of a charging device;

[0023] According to the evaluation index, the aggregation control strategy is optimized using a deep Q-network algorithm to obtain an optimized strategy.

[0024] Another embodiment of the present invention provides an electric vehicle aggregation control system, comprising:

[0025] An association module is used to associate different types of electric vehicle feature data and aggregated resource data using an equivalence class conversion algorithm to obtain charging behavior feature data of each vehicle;

[0026] A processing module, used for inputting the charging behavior characteristic data of each vehicle into a vehicle charging load prediction model constructed by using a Monte Carlo method to obtain charging load data of each electric vehicle, wherein the construction process includes dynamically adjusting the parameters of the vehicle charging load prediction model by using a grid search and a Bayesian optimization method to optimize the vehicle charging load prediction model;

[0027] A solution module, used for inputting the charging load data into a pre-built integer programming model, solving the integer programming model by using a branch and bound algorithm, and obtaining target charging load data;

[0028] A construction module, used to construct a corresponding electric vehicle aggregation control strategy based on the target charging load data;

[0029] The execution module is used to execute the corresponding electric vehicle aggregation control strategy according to the type of electric vehicle to be charged during the electric vehicle aggregation control process.

[0030] As one of the preferred solutions, before using the equivalence class conversion algorithm to associate different types of electric vehicle feature data and aggregated resource data, the electric vehicle aggregation control system further includes:

[0031] A mapping module is used to clean the multi-source heterogeneous electric vehicle data, adopt an ontology mapping method to solve the data semantic heterogeneity, and obtain the electric vehicle feature data;

[0032] The extraction module is used to fuse the static attribute data of the charging pile and the peak load demand data of the power grid, determine the available capacity of the aggregated resources, and use the recursive feature elimination method to extract the features of the available capacity of the aggregated resources to obtain the aggregated resource data.

[0033] As one of the preferred solutions, the method of solving the integer programming model by using a branch and bound algorithm includes:

[0034] The integer programming model is decomposed into a branch model by using the branch and bound algorithm, and the branch model is solved by using relaxed integer constraints and a breadth-first search strategy.

[0035] As one of the preferred solutions, after executing the corresponding electric vehicle aggregation control strategy according to the type of electric vehicle to be charged, the electric vehicle aggregation control system further includes:

[0036] The aggregation control effect is evaluated using an incremental learning method based on a self-organizing incremental learning neural network, and the aggregation control strategy is dynamically optimized using a deep Q-network algorithm.

[0037] As one of the preferred solutions, the aggregated control effect includes vehicle status data and charging power data;

[0038] The method of evaluating the aggregation control effect by using an incremental learning method based on a self-organizing incremental learning neural network and dynamically optimizing the aggregation control strategy by using a deep Q-network algorithm includes:

[0039] Preprocessing the vehicle status data and the charging power data using an incremental decision tree algorithm to obtain a data set;

[0040] Using Spark Streaming to perform real-time analysis on the data set to obtain evaluation indicators, wherein the evaluation indicators include a fit between a charging load curve and an expected curve and a response rate of a charging device;

[0041] According to the evaluation index, the aggregation control strategy is optimized using a deep Q-network algorithm to obtain an optimized strategy.

[0042] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0043] (1) The integer programming model established by charging load data can comprehensively consider multiple factors such as grid load, electric vehicle charging demand, and charging facility capacity to formulate an aggregate control strategy; the branch and bound algorithm is used to solve the model to ensure that the global optimal solution is found within a limited computing time, thereby ensuring the effectiveness and feasibility of the control strategy;

[0044] (2) The electric vehicle aggregation control strategy can intelligently arrange charging time according to the grid load and the needs of electric vehicle users. For example, by analyzing the grid off-peak hours and vehicle idle time, vehicle charging can be concentrated in time periods with low grid load, thus avoiding grid congestion caused by a large number of vehicles charging at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic flow chart of an electric vehicle polymerization control method in one embodiment of the present invention;

[0046] Figure 2 It is a schematic diagram of the structure of an electric vehicle aggregation control system in one embodiment of the present invention;

[0047] Reference numerals:

[0048] Among them, 11, association module; 12, processing module; 13, solution module; 14, construction module; 15, execution module. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0051] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the term "and / or" used herein includes any and all combinations of one or more related listed items. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0052] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0053] An embodiment of the present invention provides an electric vehicle aggregation control method. For details, see Figure 1 , Figure 1 The figure shows a schematic flow chart of an electric vehicle aggregation control method in one embodiment of the present invention, the method comprising:

[0054] S1: Use the equivalence class conversion algorithm to associate different types of electric vehicle feature data and aggregated resource data to obtain charging behavior feature data of each vehicle;

[0055] S2: Inputting the charging behavior characteristic data of each vehicle into the vehicle charging load prediction model constructed by the Monte Carlo method to obtain the charging load data of each electric vehicle, wherein the construction process includes dynamically adjusting the parameters of the vehicle charging load prediction model through grid search and Bayesian optimization methods to optimize the vehicle charging load prediction model;

[0056] S3: input the charging load data into a pre-built integer programming model, and use a branch and bound algorithm to solve the integer programming model to obtain the target charging load data;

[0057] S4: Construct corresponding electric vehicle aggregation control strategy based on target charging load data;

[0058] S5: During the electric vehicle aggregation control process, a corresponding electric vehicle aggregation control strategy is executed according to the type of electric vehicle to be charged.

[0059] In step S1, before using the equivalence class conversion algorithm to associate different types of electric vehicle feature data and aggregated resource data, it is necessary to perform data preprocessing on them first, and different data preprocessing processes are different.

[0060] For multi-source heterogeneous electric vehicle data, data cleaning is required, and the ontology mapping method is used to solve the data semantic heterogeneity and obtain the electric vehicle vehicle characteristic data. The specific mapping process is as follows:

[0061] 1. Data source semantic analysis: Analyze the concepts and semantics involved in the data for each different data source. For example, in the data of the charging pile operation platform, the concept of "charging start time" mentioned needs to find the corresponding concept in the ontology model. Through manual annotation, natural language processing technology and other methods, identify the ontology concepts corresponding to the data items in each data source.

[0062] 2. Mapping rule formulation: According to the analysis results, formulate the ontology mapping rules. The rules can be based on vocabulary similarity (e.g., "parking" and "parking" have high vocabulary similarity and can be mapped to the same ontology concept "vehicle stop status"), concept attribute matching (if data items in two different data sources have attributes that describe the battery power range and the attribute value domains are similar, they can be mapped to the "battery power" concept in the ontology), and domain knowledge and business logic (e.g., based on the general knowledge of the electric vehicle industry on the classification of vehicle driving status, similar expressions in different data sources are mapped to the correct ontology concept), etc.

[0063] 3. Mapping execution and verification: Use the established mapping rules to map data items from different data sources to the corresponding concepts in the constructed ontology model. After the mapping is completed, verification is required to check whether there are mapping errors or omissions. The accuracy of the mapping can be ensured by extracting some sample data for manual verification or using some verification tools (such as semantic rule-based checking tools).

[0064] The static attribute data of charging piles and the peak-shaving demand data of the power grid must be integrated. After the integration, the available capacity of the aggregated resources is determined. The recursive feature elimination method is used to extract the features of the available capacity of the aggregated resources to obtain the aggregated resource data.

[0065] Specifically, the peak load demand data of the power grid mainly reflects the demand for load adjustment of the power grid in different periods in order to maintain stable operation and balance load. It includes the peak and valley time division of each period, the expected peak load capacity target, the current carrying capacity of the power grid, and the safety margin and other related information.

[0066] The static attribute data of charging piles and the peak-shaving demand data of the power grid need to be integrated. The fusion process can adopt a fusion method based on time-space correlation. After the fusion, the available capacity of the aggregated resources is determined.

[0067] Recursive feature elimination is a feature selection method based on machine learning. Its core idea is to repeatedly build the model, and remove the features that have the least impact on the model performance in each iteration until the predetermined number of features or other stopping conditions are reached. When extracting features from the available capacity of aggregated resources, many factors related to the available capacity of aggregated resources (such as the distribution location of charging piles, the proportion of different types of charging piles, the difference in peak load targets in different periods of the power grid, etc.) are regarded as feature variables, and the available capacity itself is used as the target variable.

[0068] After preprocessing the data, the equivalence class conversion algorithm is used to associate the vehicle characteristic data and aggregate resource data of different types of electric vehicles to obtain the charging behavior characteristic data of each vehicle.

[0069] Specifically, the equivalence class transformation algorithm is an algorithm commonly used in the field of data processing. It classifies and transforms data to mine the association between data. Its core is to divide data into different equivalence classes according to specific rules or attributes, so that the data in the same equivalence class have similarities or equivalence in certain key aspects, thereby facilitating the discovery of associations between different data sets.

[0070] The specific steps of using the equivalence class conversion algorithm for association processing are:

[0071] (1) Determine the equivalence classification standard. There are two types of standards: one based on vehicle attributes and one based on aggregated resource attributes. The specific standard to be used depends on the specific situation.

[0072] (2) Equivalence class conversion and mapping: convert the electric vehicle feature data and aggregated resource data according to the above-mentioned standards to classify them into corresponding equivalence classes; after completing the data conversion, establish a mapping relationship between the equivalence classes of different data sets;

[0073] (3) Association analysis and feature extraction: By analyzing the established equivalence class mapping relationship, the association rules between different equivalence classes are mined. For example, it is found that the vehicle equivalence class of high-power on-board charging equipment is frequently associated with the aggregated resource equivalence class in areas with dense DC charging piles, which means that such vehicles are more likely to choose to charge in places where DC charging piles are concentrated. Based on the mined association rules, the charging behavior feature data of each vehicle is extracted. For example, based on the association between the vehicle and the charging pile, the common charging locations, preferred charging time (combined with the usage of charging piles in different areas and at different times), and approximate charging frequency (determined by the vehicle's usage pattern and the availability of charging piles) of a certain vehicle can be determined.

[0074] By using the equivalence class conversion algorithm to perform the above-mentioned association processing on the electric vehicle characteristic data and the aggregated resource data, different types of data resources can be effectively integrated to accurately obtain the charging behavior characteristic data of each vehicle.

[0075] In step S2, the Monte Carlo method is a random simulation method based on probability statistics theory. It simulates the uncertainty factors in a complex system through a large number of repeated random sampling experiments, and then estimates and analyzes the behavior and results of the system.

[0076] When constructing a vehicle charging load prediction model, the Monte Carlo method is very applicable because the charging behavior of electric vehicles itself has many uncertainties, such as the charging start time, charging duration, and charging power, which are affected by many factors such as the owner's travel habits and vehicle usage scenarios, and are random.

[0077] The basic idea of ​​using the Monte Carlo method to build a vehicle charging load prediction model is as follows:

[0078] First, it is necessary to determine the key random variables related to vehicle charging load, which usually include the charging start time, charging duration, charging power, etc. For example, based on the statistical analysis of a large number of electric vehicle user travel and charging data, it is assumed that the charging start time follows a normal distribution (the mean and standard deviation may be different in different regions and different user groups, and the charging power has a corresponding discrete value probability distribution according to factors such as vehicle model and charging pile type;

[0079] Then, multiple random samplings are performed based on these set random variables and their probability distributions. Each sampling is equivalent to simulating an electric vehicle charging scenario, and based on the basic principles of electricity (such as charging load equals charging power multiplied by charging time), the charging load value in this scenario is calculated. After random sampling simulation, the charging load data obtained from each simulation is collected and sorted, and finally the overall vehicle charging load situation is described through statistical analysis, so as to build a vehicle charging load prediction model.

[0080] During the construction process, grid search and Bayesian optimization methods are usually used in combination and dynamically adjusted according to actual conditions. Grid search can be used to conduct preliminary exploration of the parameter space at a coarse granularity to quickly locate a roughly optimal parameter area; then, in this initially determined optimal parameter area, Bayesian optimization is used for a more refined search, and its iterative optimization characteristics are used to deeply explore the optimal parameters in this area. Moreover, throughout the process, with the continuous accumulation of new vehicle charging behavior feature data and changes in the charging environment (such as changes in the distribution of charging piles, changes in user travel habits, etc.), the parameter range, step size of the grid search, and the initial sample point selection and stop conditions of the Bayesian optimization can be timely readjusted to achieve dynamic and flexible adjustment of the parameters of the vehicle charging load prediction model, ensuring that the model can always maintain good prediction performance.

[0081] After completing the construction of the vehicle charging load prediction model and parameter optimization, the collected and organized vehicle charging behavior characteristic data can be input into the model. The model will perform a large number of random sampling simulation calculations based on these input data according to the previously set Monte Carlo simulation method. Each sampling simulates a charging scenario for the vehicle and calculates the corresponding charging load value.

[0082] The integer programming model in step S3 aims to achieve specific goals, such as minimizing the peak-to-valley difference of the grid load and reducing the total charging cost, by reasonably arranging and optimizing certain discrete decision variables (such as the charging status of the electric vehicle, the charging power level, etc.) while satisfying a series of constraints (such as grid capacity limitations, electric vehicle battery capacity limitations, user charging needs, etc.).

[0083] Inputting the charging load data obtained by the vehicle charging load prediction model constructed by the Monte Carlo method in step S2 into the pre-constructed integer programming model means incorporating the actual charging load situation into the optimization decision-making system, providing a basis for finding target charging load data that better meets specific goals and constraints through integer programming; decomposing the integer programming model into a branch model through the branch and bound algorithm, and solving the branch model using relaxed integer constraints and a breadth-first search strategy.

[0084] In other words, if the goal of the integer programming model is to distribute the charging load as evenly as possible in different time periods while ensuring the stable operation of the power grid, then these charging load data can help the model analyze the load conditions in each current time period and find the direction of optimization.

[0085] The target charging load data is obtained by inputting the charging load data into a pre-built integer programming model and solving it using a branch-and-bound algorithm. The integer programming model can optimize the charging load on the basis of satisfying various constraints (such as grid capacity limitations, electric vehicle battery capacity limitations, etc.).

[0086] The charging load data is obtained by building a vehicle charging load prediction model using the Monte Carlo method. It is mainly based on the random simulation of vehicle charging behavior characteristic data, reflecting the possible charging load of electric vehicles in a natural state. However, the charging behavior in this natural state is often disordered, without considering the carrying capacity of the power grid, energy efficiency and other systematic factors. For example, without any regulation, a large number of electric vehicles may charge in the evening, which will cause a peak load on the power grid during this period, increasing the risk and cost of power grid operation.

[0087] Since the Monte Carlo simulation process involves many random variables (such as charging start time, charging duration, etc.), the charging load data will have large volatility and uncertainty. Even under the same vehicle charging behavior characteristic input, the charging load data obtained in each simulation may be different due to differences in random sampling. This uncertainty makes it difficult to formulate accurate and reliable plans based on charging load data when making decisions such as power grid planning and charging facility layout. For example, for a new charging pile construction plan, if it is based only on charging load data, it is difficult to determine the appropriate number and location of charging piles, because the uncertainty of the data may cause the plan to be either too conservative (insufficient number of charging piles) or too aggressive (too many charging piles resulting in waste of resources).

[0088] Compared with charging load data, target charging load data has been optimized by integer programming model and accurately solved by branch and bound algorithm, so the data is more certain. It provides more valuable reference for power grid planning, charging facility layout, charging strategy formulation, etc. For example, based on target charging load data, it can accurately determine how many charging piles need to be added in a certain area, what charging control strategy needs to be adopted at what time, etc., making relevant decisions more scientific and reasonable, and can effectively improve the operating efficiency and economic benefits of the entire electric vehicle charging system.

[0089] The corresponding electric vehicle aggregation control strategy is constructed based on the target charging load data. In the electric vehicle aggregation control process, the corresponding electric vehicle aggregation control strategy is executed according to the type of electric vehicle to be charged.

[0090] After executing the corresponding electric vehicle aggregation control strategy according to the type of electric vehicles to be charged, the strategy needs to be evaluated in order to optimize it.

[0091] The evaluation process is as follows: the vehicle status data and charging power data are preprocessed using the incremental decision tree algorithm to obtain a data set; Spark Streaming is used to perform real-time analysis on the data set to obtain evaluation indicators, which include the fit between the charging load curve and the expected curve and the response rate of the charging equipment; based on the evaluation indicators, the deep Q-network algorithm is used to optimize the aggregation control strategy to obtain the optimization strategy.

[0092] The incremental decision tree algorithm is an algorithm that can process continuously flowing data and gradually update and build a decision tree model. Unlike the traditional decision tree algorithm that processes all data at once, it is suitable for scenarios where data changes dynamically over time and the model needs to be updated in real time or frequently, such as in the processing of electric vehicle-related data, where vehicle status data and charging power data are continuously generated and updated.

[0093] Spark Streaming is an open source framework for real-time data processing. It can process continuously flowing data in small batches (micro-batches). It has the advantages of high throughput and fault tolerance. It is very suitable for real-time analysis of dynamic data sets related to electric vehicles.

[0094] The deep Q-network algorithm is an algorithm based on reinforcement learning. It learns the optimal strategy by having the agent continuously perform actions in the environment and based on the rewards fed back by the environment.

[0095] In this scenario, the intelligent agent can be regarded as the executor of the aggregate control strategy. In the actual operating environment of the electric vehicle charging system, it continuously adjusts the charging time, changes the charging power and other control strategy-related operations according to the evaluation indicators to learn the optimal strategy.

[0096] Evaluation indicators such as the fit between the charging load curve and the expected curve and the response rate of the charging device are used as reward signals in the deep Q-network algorithm. The aggregate control strategy will initially adjust the control strategy, and then observe the reward of environmental feedback, that is, the numerical change of the evaluation indicator. For example, if the fit between the charging load curve and the expected curve is improved after a strategy adjustment, and the response rate of the charging device is also improved, then the intelligent agent will get a higher reward. Based on this reward, it will learn that this action is beneficial and will be more inclined to take similar actions in the future; conversely, if the evaluation indicator deteriorates and the reward is low, the intelligent agent will try other different actions to explore better strategies.

[0097] That is to say, the incremental learning method based on the self-organizing incremental learning neural network is used to evaluate the aggregation control effect, and the deep Q-network algorithm is used to dynamically optimize the aggregation control strategy, where the aggregation control effect includes vehicle status data and charging power data.

[0098] An embodiment of the present invention provides an electric vehicle aggregation control system. For details, see Figure 2 , Figure 2 The figure shows a schematic diagram of the structure of an electric vehicle aggregation control system in one embodiment of the present invention, the system includes:

[0099] The association module 11 is used to associate different types of electric vehicle feature data and aggregated resource data using an equivalence class conversion algorithm to obtain charging behavior feature data of each vehicle;

[0100] The processing module 12 is used to input the charging behavior characteristic data of each vehicle into the vehicle charging load prediction model constructed by the Monte Carlo method to obtain the charging load data of each electric vehicle, wherein the construction process includes dynamically adjusting the parameters of the vehicle charging load prediction model through grid search and Bayesian optimization method to optimize the vehicle charging load prediction model;

[0101] A solving module 13 is used to input the charging load data into a pre-built integer programming model, and solve the integer programming model using a branch and bound algorithm to obtain target charging load data;

[0102] A construction module 14 is used to construct a corresponding electric vehicle aggregation control strategy based on the target charging load data;

[0103] The execution module 15 is used to execute corresponding electric vehicle aggregation control strategies according to the types of electric vehicles to be charged during the electric vehicle aggregation control process.

[0104] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0105] (1) The integer programming model established by charging load data can comprehensively consider multiple factors such as grid load, electric vehicle charging demand, and charging facility capacity to formulate an aggregate control strategy; the branch and bound algorithm is used to solve the model to ensure that the global optimal solution is found within a limited computing time, thereby ensuring the effectiveness and feasibility of the control strategy;

[0106] (2) The electric vehicle aggregation control strategy can intelligently arrange charging time according to the grid load and the needs of electric vehicle users. For example, by analyzing the grid off-peak hours and vehicle idle time, vehicle charging can be concentrated in time periods with low grid load, thus avoiding grid congestion caused by a large number of vehicles charging at the same time.

[0107] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for regulating the polymerization of electric vehicles, characterized in that: include: The equivalence class conversion algorithm is used to associate the vehicle characteristic data and aggregated resource data of different types of electric vehicles to obtain the charging behavior characteristic data of each vehicle; Inputting the charging behavior characteristic data of each of the vehicles into a vehicle charging load prediction model constructed by using a Monte Carlo method to obtain charging load data of each electric vehicle, wherein the construction process includes dynamically adjusting the parameters of the vehicle charging load prediction model through a grid search and a Bayesian optimization method to optimize the vehicle charging load prediction model; Inputting the charging load data into a pre-built integer programming model, solving the integer programming model using a branch and bound algorithm, and obtaining target charging load data; Constructing a corresponding electric vehicle aggregation control strategy based on the target charging load data; During the electric vehicle aggregation control process, the corresponding electric vehicle aggregation control strategy is executed according to the type of electric vehicle to be charged.

2. The electric vehicle aggregation control method according to claim 1, characterized in that: Before using the equivalence class conversion algorithm to associate different types of electric vehicle feature data and aggregated resource data, the electric vehicle aggregation control method further includes: Perform data cleaning on multi-source heterogeneous electric vehicle data, use ontology mapping method to solve data semantic heterogeneity, and obtain the electric vehicle vehicle characteristic data; The static attribute data of the charging piles and the peak load demand data of the power grid are integrated to determine the available capacity of the aggregated resources, and the recursive feature elimination method is used to extract features of the available capacity of the aggregated resources to obtain the aggregated resource data.

3. The electric vehicle aggregation control method according to claim 1, characterized in that: The step of using a branch and bound algorithm to solve the integer programming model comprises: The integer programming model is decomposed into a branch model by using the branch and bound algorithm, and the branch model is solved by using relaxed integer constraints and a breadth-first search strategy.

4. The electric vehicle aggregation control method according to claim 1, characterized in that: After executing the corresponding electric vehicle aggregation control strategy according to the type of electric vehicle to be charged, the electric vehicle aggregation control method further includes: The aggregation control effect is evaluated using an incremental learning method based on a self-organizing incremental learning neural network, and the aggregation control strategy is dynamically optimized using a deep Q-network algorithm.

5. The electric vehicle aggregation control method according to claim 4, characterized in that: The aggregated control effect includes vehicle status data and charging power data; The method of evaluating the aggregation control effect by using an incremental learning method based on a self-organizing incremental learning neural network and dynamically optimizing the aggregation control strategy by using a deep Q-network algorithm includes: Preprocessing the vehicle status data and the charging power data using an incremental decision tree algorithm to obtain a data set; Using Spark Streaming to perform real-time analysis on the data set to obtain evaluation indicators, wherein the evaluation indicators include a fit between a charging load curve and an expected curve and a response rate of a charging device; According to the evaluation index, the aggregation control strategy is optimized using a deep Q-network algorithm to obtain an optimized strategy.

6. An electric vehicle aggregation control system, characterized in that: include: An association module is used to associate different types of electric vehicle feature data and aggregated resource data using an equivalence class conversion algorithm to obtain charging behavior feature data of each vehicle; A processing module, used for inputting the charging behavior characteristic data of each vehicle into a vehicle charging load prediction model constructed by using a Monte Carlo method to obtain charging load data of each electric vehicle, wherein the construction process includes dynamically adjusting the parameters of the vehicle charging load prediction model by using a grid search and a Bayesian optimization method to optimize the vehicle charging load prediction model; A solution module, used for inputting the charging load data into a pre-built integer programming model, solving the integer programming model by using a branch and bound algorithm, and obtaining target charging load data; A construction module, used to construct a corresponding electric vehicle aggregation control strategy based on the target charging load data; The execution module is used to execute the corresponding electric vehicle aggregation control strategy according to the type of electric vehicle to be charged during the electric vehicle aggregation control process.

7. The electric vehicle aggregation control system according to claim 6, characterized in that: Before using the equivalence class conversion algorithm to associate different types of electric vehicle feature data and aggregated resource data, the electric vehicle aggregation control system further includes: A mapping module is used to clean the multi-source heterogeneous electric vehicle data, adopt an ontology mapping method to solve the data semantic heterogeneity, and obtain the electric vehicle feature data; The extraction module is used to fuse the static attribute data of the charging pile and the peak load demand data of the power grid, determine the available capacity of the aggregated resources, and use the recursive feature elimination method to extract the features of the available capacity of the aggregated resources to obtain the aggregated resource data.

8. The electric vehicle aggregation control system according to claim 6, characterized in that: The step of using a branch and bound algorithm to solve the integer programming model comprises: The integer programming model is decomposed into a branch model by using the branch and bound algorithm, and the branch model is solved by using relaxed integer constraints and a breadth-first search strategy.

9. The electric vehicle aggregation control system according to claim 6, characterized in that: After executing the corresponding electric vehicle aggregation control strategy according to the type of electric vehicle to be charged, the electric vehicle aggregation control system further includes: The aggregation control effect is evaluated using an incremental learning method based on a self-organizing incremental learning neural network, and the aggregation control strategy is dynamically optimized using a deep Q-network algorithm.

10. The electric vehicle aggregation control system according to claim 9, characterized in that: The aggregated control effect includes vehicle status data and charging power data; The method of evaluating the aggregation control effect by using an incremental learning method based on a self-organizing incremental learning neural network and dynamically optimizing the aggregation control strategy by using a deep Q-network algorithm includes: Preprocessing the vehicle status data and the charging power data using an incremental decision tree algorithm to obtain a data set; Using Spark Streaming to perform real-time analysis on the data set to obtain evaluation indicators, wherein the evaluation indicators include a fit between a charging load curve and an expected curve and a response rate of a charging device; According to the evaluation index, the aggregation control strategy is optimized using a deep Q-network algorithm to obtain an optimized strategy.

Images