Intelligent ordered charging system and method
Through the coordinated control of the central energy management controller and energy storage units, the charging distribution strategy is dynamically adjusted, which resolves the contradiction between grid load fluctuations and user charging needs, and achieves a seamless and orderly charging experience and improved grid stability.
Patent Information
- Application Number
- CN202511915740.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-18
- Publication Date
- 2026-01-20
AI Technical Summary
Existing intelligent and orderly charging systems cannot effectively balance user charging experience and grid load fluctuations in commercial charging scenarios, leading to increased grid burden and distribution safety hazards, and failing to meet real-time charging needs.
By employing a central energy management controller and energy storage units, and through dynamic adjustment of charging distribution strategies, combined with machine learning and optimization algorithms, the system predicts future load demand, coordinates the control of energy storage units and charging interfaces, optimizes power distribution, and balances grid load and user demand.
It achieves a seamless and orderly charging experience, reduces peak and valley fluctuations in grid load, lowers charging costs, improves grid stability and green energy absorption capacity, and increases the profitability potential of charging stations.
Smart Images
Figure CN121361373A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of charging infrastructure and energy management, in particular to an intelligent orderly charging system and method. BACKGROUND
[0002] At present, the distribution capacity planning of residential areas is generally planned according to the total load with a surplus of about 20% to 30%. Generally, the average charging power of a new energy vehicle is equivalent to the residential distribution capacity, which means that the demand for electricity of a new resident is increased with the charging of a new energy vehicle. The potential charging demand is seriously out of line with the distribution capacity of the power grid. On the other hand, disordered charging will lead to peak on peak of power grid load, aggravate the impact of distribution capacity, and increase the safety hazard of distribution. It is urgent to develop an intelligent orderly charging system that can adjust the charging load in real time according to the distribution capacity. However, the existing intelligent orderly system mainly considers that the charging energy is obtained from the power grid in real time, and the real-time power of charging is determined by the distribution capacity of the station, which belongs to a passive orderly charging. For commercial charging and other charging scenes with high real-time demand, such as "region 2" shown in the figure, the implementation of this function will affect the charging experience and satisfaction of users. Figure 2
[0003] Therefore, an intelligent orderly charging system and method are provided to solve the above problems. SUMMARY
[0004] In order to solve the above problems, the present application provides an intelligent orderly charging system and method, which dynamically adjusts the charging distribution strategy according to different charging scenes, meets the real-time changing charging load demand, can smooth the peak and valley power fluctuation of the power grid, helps to control the charging cost, further effectively alleviates the contradiction between supply and demand of charging stations, and realizes the non-invasive intervention of terminal user orderly charging.
[0005] To achieve the above purpose, the present application provides an intelligent orderly charging system applied to a vehicle charging station, comprising: a central energy management controller, configured to acquire real-time operation data and future demand data of the charging station, and based on the real-time operation data and future demand data, generate an optimized power distribution instruction for each charging interface and energy storage unit, with the total distribution capacity limit as a constraint condition and the overall operation efficiency of the charging station as a target; an energy storage unit, configured to store and release electric energy; a charge and discharge management module of the energy storage unit, in communication connection with the central energy management controller and the energy storage unit, configured to receive and execute the optimized power distribution instruction, and control the energy storage unit to perform charging or discharging operation; A vehicle charging and discharging management module is in communication connection with the central energy management controller and one or more charging interfaces, configured to receive and execute the optimized power distribution instructions, and control the specified charging interface to charge the electric vehicle or perform V2G discharging operation. The central energy management controller dynamically adjusts energy flow by coordinating the energy storage unit charging and discharging management module and the vehicle charging and discharging management module, and meets the user charging demand without breaking the total power distribution capacity limit.
[0006] Preferably, the real-time operation data obtained by the central energy management controller includes real-time charging load data, which is represented as: ; Wherein is the current time, is the real-time total charging power demand at the corresponding time.
[0007] Preferably, the future demand data obtained by the central energy management controller includes predicted charging load data, which is obtained by analyzing and machine learning predicting historical load data, and is represented as: ; Wherein, is the future time, is the predicted total charging power demand at the corresponding future time.
[0008] Preferably, the central energy management controller performs charging load prediction, including: Cluster charging behavior analysis, using Mini-Batch K-means algorithm to cluster charging behavior; Time series charging load prediction based on LSTM recurrent neural network, using multi-region rotation training mechanism, dividing data loader by region, realizing multi-task learning, and using dynamic learning rate, early stopping mechanism and gradient clipping for training optimization.
[0009] Preferably, the cluster charging behavior analysis specifically includes: Cleaning the load and charging duration, eliminating null values and physically impossible data, and using quantile truncation to eliminate the influence of extreme values; Constructing a feature set for cluster analysis, the feature set including charging pile density features, economic sensitivity features and periodicity features; Using Mini-Batch K-means algorithm for clustering, wherein the optimal cluster number K is selected according to the minimum distortion rate change principle, and parallel computing is used to speed up the search process of the optimal K value; by initializing multiple times and setting a fixed random seed number, the clustering result is ensured to be repeatable.
[0010] By initializing the cluster center multiple times and setting a fixed random seed number, it is ensured that the clustering experimental results are repeatable.
[0011] Preferably, the central energy management controller generates an optimized power distribution instruction based on a dynamic model of the state of charge (SOC) change of the energy storage unit, which is expressed as: wherein, is the charging and discharging power.
[0012] Preferably, the constraint conditions include global constraints and local constraints. The global constraint is expressed as: wherein, is the end time of the optimization period, is the target state of charge of the energy storage unit; The local constraint is expressed as: wherein, is the minimum state of charge of the energy storage unit, is the maximum state of charge of the energy storage unit, is the minimum charging and discharging power of the energy storage unit, is the maximum charging and discharging power of the energy storage unit.
[0013] Preferably, the target of optimizing the overall operation efficiency of the charging station includes: minimizing the total electricity cost of the station; maximizing the use of valley electricity for charging; smoothing the total load curve of the charging station; participating in grid services through discharging of the energy storage unit to obtain revenue.
[0014] Preferably, the energy storage unit adopts a modular architecture and supports capacity configuration through series, parallel or overall movement of modules. The capacity configuration strategy is dynamically optimized based on historical operation data of the charging station to ensure that the charging pile utilization rate is maintained above a preset threshold.
[0015] An intelligent and orderly charging method adopts an intelligent and orderly charging system, including the following steps: S1: obtaining real-time operation data and future demand data of the charging station; S2: based on the data, generating an optimized power distribution instruction with the total distribution capacity limit as a constraint and the optimized operation efficiency as a target. S3: execute the optimized power allocation instruction to realize the collaborative control of the charging pile and the energy storage unit, and ensure that the user charging demand is met under the distribution capacity limit.
[0016] Therefore, the intelligent orderly charging system and method can provide users with a non-susceptible and reliable charging experience, enable a charging station operator to serve more vehicles without expanding the capacity, and create additional income through V2V, V2G and peak-valley price difference; and the charging load is changed from an uncontrollable burden to a dispatchable flexible resource, which helps the power grid to shave the peak and fill the valley, improves the stability of the power grid and the green power consumption capacity.
[0017] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS
[0018] Figure 1 It is a schematic diagram of an intelligent orderly charging system in the present application; Figure 2 It is a charging load distribution diagram of a typical charging scenario studied in the present application; Figure 3 It is an electrical simulation environment diagram of the central energy management controller in the embodiment of the present application to carry out hybrid charging optimization; Figure 4 It is a charge-discharge control simulation model diagram of the energy storage module in the embodiment of the present application; Figure 5 It is a real-time power grid charging load demand diagram before / after using the energy storage device to optimize the charging load in the embodiment of the present application; Figure 6 It is a comparison diagram of the profit income of the charging station before and after charging in the embodiment of the present application; Figure 7 It is a typical charging parameter characteristic analysis diagram based on cluster analysis in the embodiment of the present application; Figure 8 It is a comparison diagram of the prediction and measurement results in the embodiment of the present application, wherein (a) is a comparison diagram of the prediction and measurement results of region 526, cluster 3, (b) is a comparison diagram of the prediction and measurement results of region 900, cluster 0, and (c) is a comparison diagram of the prediction and measurement results of region 800, cluster 1. DETAILED DESCRIPTION
[0019] The following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative labor are within the scope of protection of the present application.
[0020] Unless otherwise defined, technical terms or scientific terms used in the present application shall have the ordinary meaning of such terms for a person skilled in the art to which the present application pertains.
[0021] The terms "comprise", "comprising", "include", "including" and the like used herein are meant to be interpreted as specifying the presence of stated features or components rather than precluding the presence or addition of further features or components. The terms "inner", "outer", "upper", "lower", and the like, as used herein, refer to the orientation or position of the apparatus or element shown in the drawings, and are used only to facilitate the description of the application and the claims, and do not connote or imply that a specific orientation or position is required for the apparatus or element to be in proper operation, and therefore cannot be construed as limiting the application. In the present application, unless otherwise explicitly defined and limited, the terms "attached" and the like should be interpreted broadly, for example, can be fixedly connected, can be detachably connected, or can be integrated; can be directly connected, or can be indirectly connected through an intermediate medium; can be internal connection of two elements or interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.
[0022] Embodiments An intelligent orderly charging system applied to a car charging station, as shown in Figure 1 , comprising: A central energy management controller, configured to acquire real-time operation data and future demand data of the charging station, and generate an optimized power distribution instruction for each charging interface and energy storage unit based on the real-time operation data and the future demand data, with the total power distribution capacity limit as a constraint condition and the overall operation efficiency of the charging station as a target. The real-time operation data at least includes the total power distribution capacity limit, real-time power demand of each charging pile, and residual capacity of the energy storage unit. The future demand data at least includes future charging load predicted based on vehicle reservation information. The charging interface can be a charging pile.
[0023] The central energy management controller ensures that users of non-emergency charging vehicles have an uninterrupted charging experience without perception by dynamically adjusting charging power and discharging using energy storage units, and ensures that all vehicles complete charging before the user-set deadline through load forecasting and energy planning. The central controller considers the power distribution capacity limit and the remaining battery capacity SOC of the energy storage battery according to the real-time charging demand and future charging demand of the charging station, and optimizes the charging and discharging behavior of the individual charging pile through the energy storage unit to ensure the success rate of orderly charging without perception under the limited power distribution capacity. The central energy management controller dynamically adjusts the energy flow through the charging pile and the energy storage unit by cooperatively controlling the vehicle charging and discharging management module and the energy storage unit charging and discharging management module, so as to meet the user's charging demand without breaking the total limit of the power distribution capacity.
[0024] The central controller not only meets the user's expected charging schedule, but also prioritizes the charging terminal demand and dynamically optimizes the charging and discharging behavior of the energy storage battery with the minimum charging cost as the optimization target based on the next cycle charging load prediction.
[0025] Specifically, the charging load is represented as a discrete two-dimensional array f=Array(t,P), where t is the time, and P is the real-time total charging power demand. The real-time charging load data is represented as: ; where is the current time, is the real-time total charging power demand at the corresponding time. The predicted charging load data is obtained by analyzing and machine learning predicting the historical load data, and is represented as: ; where, is the future time, is the predicted total charging power demand at the corresponding future time.
[0026] The battery charging behavior is represented as: ; ; where, is the charging and discharging power. x=x(t)=SOC u(t)=Pbatt battery power equivalent internal resistance Rd(SOC), open circuit voltage Voc(SOC); The central energy management controller performs charging load prediction, including: Cluster charging behavior analysis: Mini-Batch K-means algorithm is used to cluster the charging behavior; the load, charging time is cleaned, and the null value and physically impossible data are removed, and the quantile truncation is used to eliminate the influence of extreme values; constructing a feature set for clustering analysis, the feature set including a charging pile density feature, an economic sensitivity feature, and a periodicity feature; The Mini-Batch K-means algorithm is used for clustering, wherein the optimal cluster number K is selected according to the minimum distortion rate change principle, and parallel computing is used to speed up the search process of the optimal K value; and the cluster center is initialized multiple times and the fixed random seed number is set to ensure the repeatability of the cluster result.
[0027] The cluster center is initialized multiple times and the fixed random seed number is set to ensure the repeatability of the cluster result.
[0028] Specifically, the clustering charging behavior analysis adopts the Mini-Batch K-means clustering algorithm, which takes the region number and timestamp as the basis, and realizes the association of data sets through the merge function of Pandas. A high-performance data processing module is designed. First, robust filtering is performed, and df.pipe(_clean_data) is executed for basic cleaning to eliminate null values and physically infeasible data. Then, the quantile truncation is used for the right-skewed distribution fields such as load and charging duration, and the 99.5% quantile truncation is used to eliminate the influence of extreme values.
[0029] In the feature construction aspect, according to the previous data spatio-temporal feature analysis, the algorithm generates four types of optimized features while retaining the basic feature values: the charging pile density (geo_density) within a 3km radius is calculated through the Haversine formula, and the logarithmic smoothing processing (geo_density_log) is performed to form a new spatio-temporal interaction feature; the price-occupancy product index (price_occupancy) is constructed to better reflect the economic sensitivity; the hour variable is converted into sine / cosine components (hour_sin / hour_cos) to enhance the model's capture of periodicity, and more accurate periodicity features are constructed. The Box-Cox transformation (λ=0.15) is performed on the right-skewed load and charging duration to make the data approach normal distribution, and the new distribution transformation feature reduces the risk of local optimization.
[0030] The optimal K value is selected according to the minimum distortion rate change of K value, which is more objective than the traditional "elbow rule" and avoids the subjectivity of manual judgment. By dynamically adjusting the batch_size (iteration sample number) parameter, the silhouette coefficient is improved, and a balance is achieved between clustering quality and computational efficiency. Through experimental verification, batch_size=8192 performs best on the data set of the present application.
[0031] To accelerate the K-value search process, the joblib library is used to implement parallel computing. By configuring n_jobs=4, the clustering tasks of different K values are run on 4 CPU cores simultaneously, reducing the search time to 1 / 4 of the original. This parallel strategy makes full use of multi-core processor resources and is suitable for clustering algorithm optimization that requires multiple iterations. By initializing multiple times, the influence of local optimal solution is reduced, and by setting a fixed random seed number (random_seed=42), the experimental results are repeatable.
[0032] Based on the LSTM recurrent neural network, the time series charging load prediction adopts a multi-region rotation training mechanism, divides the data loader by region, realizes multi-task learning, and uses dynamic learning rate, early stopping mechanism and gradient clipping for training optimization.
[0033] Specifically, the charging load prediction model based on long short-term memory network (LSTM) is implemented using the PyTorch framework. The data set is divided into training set (0.7), validation set (0.15) and prediction set (0.15) according to time sequence. Reasonable division will effectively reduce the risk of data leakage. Considering that the data contains 275 regions, the differences between regions will cause data bias, so the study adopts a region rotation training mechanism, divides the data loader by region, realizes multi-task learning, and each region is independently divided into batches (batch_size=512), realizing multi-region collaborative training. In terms of training optimization, dynamic learning rate is used to automatically adjust the learning rate based on the validation loss (initial 0.001→minimum 1e-6); early stopping mechanism is added, and training is terminated if there is no improvement for 15 consecutive epochs (PATIENCE=15); gradient clipping is performed, and the implicit adaptive learning rate is realized through the Adam optimizer. In order to improve the training rate, GPU acceleration is used first. Further, through dynamic inverse transformation mechanism, the prediction result restoration error is reduced, and the time and space alignment system is introduced to ensure the accurate matching of time stamp with actual charging events. The visualization and persistence of the results facilitate analysis and verification, and a DRL interface is designed to reduce the time consumption caused by multiple training models, facilitating subsequent tasks.
[0034] The constraints include global constraints and local constraints. The global constraint is represented as: ; Wherein, is the end time of the optimization period, is the target state of charge of the energy storage unit; The local constraint is represented as: ; ; Wherein, This represents the minimum charge of the energy storage unit. This represents the maximum charge of the energy storage unit. This represents the minimum charging and discharging power of the energy storage unit. This represents the maximum charging and discharging power of the energy storage unit.
[0035] The goals of optimizing the overall operational efficiency of charging stations include: Minimize the total electricity cost of the power station; Maximize the use of off-peak electricity for charging; Smooth the total load curve of charging stations; They generate revenue by discharging energy from energy storage devices to serve the grid.
[0036] Specifically, the optimization objective is within a certain period of time. The total profit of a charging station can be expressed as J= Maximizing, specifically, means: ; Among them, battery charging cost Battery depreciation factor , Charging revenue and battery SOC change rate; charging revenue and battery SOC change. Charging revenue, specifically represented by the input power and electricity price of the charging station, is related to the station's output power and the unit price of charging, and can be expressed as: Composed of, etc. As a penalty factor for unsatisfactory charging, p(t) should be as small as possible throughout the process, which can be expressed as... ; Based on real-time charging load demand, and taking into account the urgency of charging needs, the system estimates the required charging volume and charging cycle, as well as charging costs, and coordinates the charging power source (grid or energy storage battery) to carry out optimal charging scheduling. This can achieve charging modes such as grid power supply, grid + energy storage module power supply, and energy storage module independent power supply. Grid power supply can realize time-sharing and power-sharing charging modes in orderly charging; energy storage module power supply is beneficial to grid absorption and charging pile profitability, while energy storage module + grid joint power supply is used to improve charging success rate under limited power distribution capacity.
[0037] The energy storage unit is used to store and release electrical energy. It adopts a modular architecture, supporting capacity configuration through series, parallel, or overall module movement. Its capacity configuration strategy is dynamically optimized based on historical operational data of the charging station to ensure that the utilization rate of charging piles is continuously maintained above a preset threshold. The energy storage unit can be an energy storage battery with integrated V2V charging functionality.
[0038] The energy storage unit charge-discharge management module is in communication connection with the central energy management controller and the energy storage unit, and is configured to receive and execute the optimized power distribution instruction, and control the energy storage unit to perform charging or discharging operation. The vehicle charge-discharge management module is in communication connection with the central energy management controller and one or more charging interfaces, and is configured to receive and execute the optimized power distribution instruction, and control the specified charging interface to perform charging or V2G discharging operation on the electric vehicle. The central energy management controller dynamically adjusts energy flow by coordinating the energy storage unit charge-discharge management module and the vehicle charge-discharge management module, and meets the user charging demand without breaking the total power distribution capacity limit.
[0039] The charge-discharge action of the energy storage system is triggered by the coordinated analysis of real-time charging demand and grid state. Under the premise of meeting the state of charge, power limit and other safe operation boundaries, if the charging load causes insufficient input power in the station, the system triggers energy storage discharge. Conversely, if the forecast shows that the input power is surplus, the energy storage is charged. Further, the time-of-use price model of the power grid can be combined to select the optimal price interval for charging during the power surplus period. The objective function of this optimization problem is to minimize the operating cost or maximize the benefit, and the solution period can be set to one day or more, and iterative calculation is carried out at certain periods.
[0040] The dynamic optimization control of the energy storage system is mainly based on short-term charging load prediction. The prediction data is updated in cycles of 5-30 minutes. In the optimization algorithm level, dynamic programming (DP) is preferred for global optimization solution; further, model-based predictive control (MPC) method can also be used to effectively cope with prediction errors and system uncertainties, and improve the robustness of control by using its rolling optimization and feedback correction mechanism.
[0041] Embodiment 1 An intelligent orderly charging method adopts an intelligent orderly charging system, comprising the following steps: S1: obtaining real-time operation data and future demand data of the charging station; S2: based on the data, generating an optimized power distribution instruction with the total power distribution capacity limit as a constraint and the optimized operation efficiency as a target; S3: executing the optimized power distribution instruction to realize the coordinated control of the charging piles and the energy storage unit, and ensuring to meet the user charging demand under the power distribution capacity limit.
[0042] Embodiment 2 To verify the effectiveness of the system described in the present application, a charging station system simulation model including a power distribution network, charging piles, batteries of vehicles to be charged, and an energy storage module is constructed by using a special simulation software. Specifically, Figures 3-4The simulation optimization is carried out by setting the optimization specific goal through the program code, and the optimization result as shown in Figure 5 It can be seen that after the application of the intelligent orderly charging system of the application, the instantaneous power grid load demand of the charging station is fundamentally improved. Specifically, the peak value of the load curve is significantly reduced, the overall fluctuation tends to be smooth, and effective peak clipping and valley filling are achieved. Most importantly, the system successfully maintains the total load within the distribution capacity limit, thereby ensuring the safe and efficient operation of the charging station without expansion, and relieving the impact on the regional power grid.
[0043] Embodiment 3 This embodiment further illustrates the effectiveness of the system and method of the application from the aspects of real-time benefits of the charging station before and after optimization, typical charging parameter characteristics of clustering analysis, and comparison between prediction and measurement results of the LSTM model. The specific results are shown in Figures 6-8 .
[0044] As can be seen from Figure 6 , after introducing the dynamic optimization scheduling of the application, the income curve of the station (after optimization) shows a significant and sustained improvement compared to the income curve under the traditional charging mode (before optimization). This directly proves that the application successfully transforms the charging load from a pure cost center to a profitable resource by using the discharging of the energy storage device to participate in the power grid service strategy, thereby maximizing the income. As can be seen from Figure 7 , several typical scenarios with different charging parameter characteristics can be clearly identified in the figure. These clustering results verify the effectiveness of the clustering algorithm used in the application, providing a key basis for the central energy management controller to understand and predict diverse charging demands, so that the system can implement more targeted and refined optimization scheduling strategies. As can be seen from Figure 8 , the prediction curve and the measured curve are highly consistent, whether in overall trend or in short-term fluctuations, the model has shown excellent fitting and prediction ability. This proves that the prediction module constructed in the application can accurately predict future charging demands, providing reliable data input for the central energy management controller to develop optimal energy storage charging and discharging plans and power grid power distribution schemes in advance, thereby ensuring the control effectiveness of the entire orderly charging system.
[0045] Therefore, the application adopts the above-mentioned intelligent orderly charging system and method, dynamically adjusts the charging distribution strategy according to different charging scenarios, meets the real-time changing charging load demand, can smooth the peak and valley power fluctuations of the power grid, helps to control the charging cost, further effectively alleviates the charging supply and demand contradiction of the charging station, and realizes the non-sensitization intervention of the terminal user orderly charging.
[0046] It should be pointed out finally that the above examples are only used to illustrate the technical solutions of the present application but not to limit it, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can still be modified or replaced equivalently, and these modifications or equivalent replacements should not make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.
Claims
1. An intelligent and orderly charging system, applied to vehicle charging stations, characterized in that, include: The central energy management controller is used to acquire real-time operation data and future demand data of the charging station, and based on the real-time operation data and future demand data, with the total power distribution capacity limit as a constraint, and with the goal of optimizing the overall operation efficiency of the charging station, it generates optimized power allocation instructions for each charging interface and energy storage unit. Energy storage units are used to store and release electrical energy; The energy storage unit charging and discharging management module is communicatively connected to the central energy management controller and the energy storage unit, and is used to receive and execute the optimized power allocation command to control the energy storage unit to perform charging or discharging operations. The vehicle charging and discharging management module is communicatively connected to the central energy management controller and one or more charging interfaces. It is used to receive and execute the optimized power allocation command and control the designated charging interface to charge the electric vehicle or perform V2G discharging operation. The central energy management controller dynamically adjusts energy flow by coordinating the energy storage unit charging and discharging management module and the vehicle charging and discharging management module, so as to meet the user's charging needs without exceeding the total power distribution capacity limit.
2. The intelligent orderly charging system according to claim 1, characterized in that, The real-time operating data acquired by the central energy management controller includes real-time charging load data, which is represented as follows: ； in, For the current moment, This represents the real-time total charging power demand at the corresponding moment.
3. The intelligent orderly charging system according to claim 2, characterized in that: The future demand data acquired by the central energy management controller includes predicted charging load data, which is obtained through analysis of historical load data and machine learning prediction, and is expressed as follows: ； in, For the future moment, This is to correspond to the predicted total charging power demand at future times.
4. The intelligent orderly charging system according to claim 3, characterized in that, The central energy management controller performs charging load forecasting, including: Clustering charging behavior analysis: The Mini-Batch K-means algorithm is used to cluster charging behavior. The timing charging load prediction based on LSTM recurrent neural network adopts a multi-region rotation training mechanism, divides the data loader according to region to realize multi-task learning, and uses dynamic learning rate, early stopping mechanism and gradient clipping for training optimization.
5. The intelligent orderly charging system according to claim 4, characterized in that, Clustering charging behavior analysis, specifically including: The load and charging time are cleaned to remove null values and physically infeasible data, and quantile truncation is used to eliminate the influence of extreme values. A feature set for cluster analysis is constructed, which includes charging pile density features, economic sensitivity features, and periodic features; The Mini-Batch K-means algorithm is used for clustering. The optimal number of clusters K is selected based on the principle of minimizing the rate of change of distortion. Parallel computation is used to accelerate the search process for the optimal K value. The clustering results are repeated by initializing and setting a fixed number of random seeds multiple times.
6. The intelligent orderly charging system according to claim 3, characterized in that, The central energy management controller generates optimized power allocation commands based on a dynamic model of the state of charge (SOC) changes of the energy storage units. The SOC changes of the energy storage units are represented as follows: ； in, This refers to the charging and discharging power.
7. The intelligent orderly charging system according to claim 4, characterized in that, Constraints include: global constraints and local constraints; Global constraints are represented as: ； in, To optimize the end time of the cycle, The target state of charge for the energy storage unit; Local constraints are represented as: ； ； in, This represents the minimum charge of the energy storage unit. This represents the maximum charge of the energy storage unit. This represents the minimum charging and discharging power of the energy storage unit. This represents the maximum charging and discharging power of the energy storage unit.
8. The intelligent orderly charging system according to claim 1, characterized in that, The objectives for optimizing the overall operational efficiency of charging stations include: Minimize the total electricity cost of the power station; Maximize the use of off-peak electricity for charging; Smooth the total load curve of charging stations; They generate revenue by discharging energy from energy storage devices to serve the grid.
9. The intelligent orderly charging system according to claim 1, characterized in that, The energy storage unit adopts a modular architecture, which supports capacity configuration through series, parallel or whole-unit movement of modules. Its capacity configuration strategy is dynamically optimized based on the historical operation data of the charging station to ensure that the utilization rate of the charging pile is continuously maintained above the preset threshold.
10. A smart orderly charging method, employing the smart orderly charging system described in any one of claims 1-9, characterized in that, Includes the following steps: S1: Obtain real-time operation data and future demand data of charging stations; S2: Based on the data, with the total power distribution capacity limit as a constraint and the goal of optimizing operating efficiency, generate an optimized power allocation instruction; S3: Execute the optimized power allocation command to achieve coordinated control of the charging pile and energy storage unit, ensuring that user charging needs are met under the power distribution capacity limit.