A charging pile cluster power coordination method, device, equipment and storage medium

Abstract

The application provides a charging pile cluster power coordination method and device, equipment and storage medium, relates to the charging pile optimization technical field, and is applied to a charging pile power distribution system, wherein the charging pile power distribution system comprises a plurality of charging piles; the method comprises the following steps: acquiring current state data, historical use data and position data of each charging pile in different time periods, determining corresponding priority weight data according to the position data and the historical use data; obtaining use probability data of the corresponding charging pile based on a preset prediction model according to each historical use data and corresponding current state data; obtaining corresponding power reference data according to power grid dispatching plan data, current state data and use probability data; and obtaining power adjustment strategies of each charging pile based on a deep reinforcement learning model according to current state data, priority weight data and power reference data. The application combines the prediction model and the deep reinforcement learning technology, and realizes intelligent and efficient power coordination.

Description

Technical Field

[0001] This invention relates to the field of charging pile optimization technology, and more specifically, to a method, apparatus, equipment, and storage medium for coordinating the power of a charging pile cluster. Background Technology

[0002] With the widespread application of electric vehicles, the construction and management of charging stations have gradually become important issues. In recent years, charging station technology has continued to develop, and related power regulation schemes have emerged one after another. However, existing charging station power regulation methods have some limitations: for example, traditional charging station power regulation methods are mostly based on simple rules or real-time grid load information to allocate power. Moreover, these methods coordinate and adjust the power of charging station clusters in a unified and real-time manner, which results in power waste or lag; this lag may lead to grid load fluctuations, and even cause grid stability problems in some cases. Summary of the Invention

[0003] The problem solved by this invention is one or more of the aforementioned related technical problems.

[0004] To address the aforementioned problems, this invention provides a method, apparatus, device, and storage medium for coordinating the power of a charging pile cluster.

[0005] In a first aspect, the present invention provides a power coordination method for a charging pile cluster, applied to a charging pile power distribution system, wherein the charging pile power distribution system includes multiple charging piles; the coordination method includes:

[0006] Obtain the current status data, historical usage data and location data of each charging pile for different time periods, and determine the corresponding priority weight data based on the location data and the historical usage data;

[0007] Based on a preset prediction model, the usage probability data of the charging pile is obtained according to the historical usage data and the corresponding current status data.

[0008] Obtain power grid dispatch plan data, and obtain corresponding power reference data based on the power grid dispatch plan data, the current status data, and the usage probability data;

[0009] Based on a deep reinforcement learning model, a power adjustment strategy for each charging pile is obtained according to the current state data, the priority weight data, and the power reference data, and the power adjustment of each charging pile is performed according to the power adjustment strategy.

[0010] Optionally, the historical usage data includes charging time distribution data and charging duration data, the current status data includes the current operating status, and the preset prediction model includes an LSTM module and a Transformer module; the process of obtaining corresponding usage probability data based on the preset prediction model, according to each of the historical usage data and the corresponding current status data, includes:

[0011] The LSTM module is used to extract features from the charging time distribution data, the charging duration data, and the current operating status to obtain temporary feature data.

[0012] The temporary feature data is processed by the Transformer module to obtain the corresponding usage probability data.

[0013] Optionally, the power grid dispatch plan data includes the power grid load forecast data and power supply capacity data, the current status data includes real-time electricity price data and rated power, and the step of obtaining the corresponding power reference data based on the power grid dispatch plan data, the current status data, and the usage probability data includes:

[0014] The maximum power limit data for each charging pile is determined based on the power grid load forecast data and the power supply capacity data.

[0015] The maximum power limit data, the real-time electricity price data, the rated power, and the grid load prediction data are input into a preset power prediction model. The operator splitting QP algorithm is used to solve the preset power prediction model and output the power reference data corresponding to each charging pile.

[0016] The preset power prediction model is constructed based on an objective function and multiple constraints, with the goal of minimizing the optimal charging pile power output.

[0017] Optionally, the objective function specifically includes:

[0018] The formula for the objective function is:

[0019] ;

[0020] in, The total number of time periods. This represents the total number of charging stations. Time period The cost per unit of energy is determined by the real-time electricity price. For the charging pile In time period The actual power output within; Peak power cost factor; For the charging pile In time period The penalty value for waiting time or charging interruption within the time limit; For the charging pile In time period The probability of use within, For the charging pile The rated power.

[0021] Optionally, the step of obtaining the power adjustment strategy for each charging pile based on the deep reinforcement learning model, according to the current state data, the priority weight data, and the power reference data, includes:

[0022] Construct the deep reinforcement learning model and initialize it.

[0023] The reward value is determined based on the current status data, the corresponding priority weight data, the power reference data, the real-time electricity price data, and the rated power.

[0024] The policy gradient is determined based on the reward value, and the policy network parameters of the deep reinforcement learning model are updated according to the policy gradient to maximize the expected cumulative reward.

[0025] The power adjustment strategy for each charging pile is determined based on the updated strategy network parameters.

[0026] Optionally, the charging pile cluster power coordination method further includes:

[0027] The real-time voltage data of each charging pile is acquired, and the real-time voltage data is compared with a preset voltage threshold.

[0028] When the real-time voltage data is less than the preset voltage threshold, the corresponding charging pile is designated as a faulty pile, and the current actual power of the adjacent charging piles of the faulty pile is obtained. Based on the current actual power of each adjacent charging pile, the maximum reverse power data output by the corresponding adjacent charging pile is calculated, and all the maximum reverse power data are transmitted to the faulty pile.

[0029] Optionally, the historical usage data includes usage frequency data, charging duration data, and charging volume data; determining the corresponding priority weight data based on the location data and the historical usage data includes:

[0030] Determine the corresponding initial weight data based on the location data;

[0031] Based on the principal component analysis algorithm, the location data, usage frequency data, charging duration data, and charging power data are analyzed to obtain temporary weight data;

[0032] The priority weight data corresponding to the charging pile is determined based on the initial weight data and the corresponding temporary weight data.

[0033] Secondly, the present invention provides a charging pile cluster power coordination device, applied to a charging pile power distribution system, wherein the charging pile power distribution system includes multiple charging piles; the coordination device includes:

[0034] The acquisition unit is used to acquire the current status data, historical usage data and location data of each charging pile, and determine the corresponding priority weight data based on the location data and the historical usage data.

[0035] The calculation unit is used to obtain the usage probability data of the charging pile based on the preset prediction model, according to the historical usage data and the corresponding current state data; to obtain the power grid dispatch plan data, and to obtain the corresponding power reference data according to the power grid dispatch plan data, the current state data and the usage probability data;

[0036] The adjustment unit is used to obtain the power adjustment strategy of each charging pile based on the current state data, the priority weight data and the power reference data according to the deep reinforcement learning model, and to adjust the power of each charging pile according to the power adjustment strategy.

[0037] Thirdly, the present invention provides a charging pile cluster power coordination device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the charging pile cluster power coordination method as described in the first aspect.

[0038] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the charging pile cluster power coordination method as described in the first aspect.

[0039] The beneficial effects of the charging pile cluster power coordination method, device, system, and storage medium of the present invention are:

[0040] By combining grid dispatch plan data with the current status and usage probability of charging piles, power reference data is determined. Adjusting the power of charging pile clusters based on this reference data allows for better coordination with the overall grid operation. For example, when grid load is high, the power output of charging piles can be appropriately reduced to avoid excessive impact on the grid; conversely, when grid load is low, the power of charging piles can be appropriately increased to fully utilize the grid's remaining capacity. This dynamic power adjustment method helps balance grid load, reduce peak-to-valley differences, improve grid stability and reliability, and reduce the risk of grid failures.

[0041] A power regulation strategy based on a deep reinforcement learning model comprehensively considers the priority weights of charging piles, their current state, and power reference data to achieve fine-grained power adjustment. For example, high-priority charging piles (such as those located in transportation hubs or commercial centers) receive sufficient power supply to meet their higher usage demands; other charging piles are allocated power rationally based on power reference data. This strategy not only ensures the normal use of charging piles but also maximizes their utilization rate within the limits allowed by the power grid, reduces idle time, and improves overall efficiency.

[0042] Furthermore, the power adjustment strategy fully considers the usage probability data of charging stations and predicts their usage in advance. When the usage probability is high, sufficient power is allocated in advance to ensure that users can quickly obtain stable charging services and reduce waiting time. At the same time, by dynamically adjusting the power, it avoids situations where the charging station has insufficient or excessive power during use, improving the stability and safety of the charging process and optimizing the user charging experience.

[0043] Therefore, this charging pile cluster power coordination method achieves coordinated optimization between the charging pile cluster and the power grid by closely integrating real-time power adjustment with the grid's operating status and charging demand. This method dynamically adjusts the power allocation of each charging pile based on the grid's real-time operating status and charging pile usage, with each charging pile's power allocation determined by priority. Traditional methods often lead to power waste or delayed adjustments due to uniform power allocation or similar strategies for charging piles. In contrast, this method effectively avoids these problems, significantly improving the performance of the charging pile cluster and the utilization rate of grid resources.

[0044] From the perspective of the power grid, reasonable power regulation reduces grid losses and equipment wear, extends equipment lifespan, and lowers operating costs. From the perspective of charging pile operation, this method improves the utilization rate of charging piles and increases operational revenue. For users, they can obtain more efficient and stable charging services, saving time and costs. Therefore, this method improves the overall efficiency of charging pile clusters and the power grid system, achieving a win-win situation for multiple parties. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating a power coordination method for a charging pile cluster according to an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram of the structure of a charging pile cluster power coordination device according to an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of a charging pile cluster power coordination device according to an embodiment of the present invention. Detailed Implementation

[0048] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0049] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0050] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0051] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0052] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0053] To address the problems existing in the aforementioned related technologies, embodiments of the present invention provide a method, apparatus, device, and storage medium for coordinating the power of a charging pile cluster.

[0054] like Figure 1 As shown in the figure, an embodiment of the present invention provides a power coordination method for a charging pile cluster, which is applied to a charging pile power distribution system, the charging pile power distribution system including multiple charging piles; the coordination method includes:

[0055] Step S100: Obtain the current status data, historical usage data and location data of each charging pile for different time periods, and determine the corresponding priority weight data based on the location data and the historical usage data.

[0056] Specifically, the system acquires current status data for each charging station, including real-time power, availability, and whether it is currently charging. If charging, this includes the vehicle's charge level, reflecting the station's operational status in real time. Simultaneously, it collects historical usage data for charging stations across different time periods (weekdays, weekends, daytime, and nighttime), covering usage frequency, charging duration, and charge level, to analyze usage patterns and predict future usage. Furthermore, it acquires geographical location information for charging stations, including their distribution and distance from surrounding stations, to assess the mutual influence between them. For example, charging stations located close together may share user groups, and their usage patterns may be interconnected.

[0057] Based on the location data and historical usage data of charging stations, priority weights are determined for each charging station. Location data reflects the rationality of the layout; for example, charging stations in transportation hubs or commercial centers have higher priority. Historical usage data reflects usage frequency and importance; charging stations with high usage frequency need to have their power supply guaranteed first. Taking into account both location and historical usage, priority weights are assigned to each charging station for subsequent power adjustment strategy formulation.

[0058] Step S200: Based on a preset prediction model, obtain the usage probability data of the corresponding charging pile according to the historical usage data and the corresponding current status data.

[0059] Specifically, historical usage data: collect usage records of charging stations over a period of time (such as several months), including usage time, usage duration, charging amount, etc.

[0060] Current status data, such as the real-time status data of a charging station when it is in use, typically includes: Is it currently charging? Yes, the charging station is charging the electric vehicle. Current charging power: For example, the current charging power is 30kW. Remaining battery capacity (SOC): The electric vehicle's battery has 40% remaining energy. Current charging current and voltage: The current charging current is 100A, and the voltage is 380V.

[0061] Real-time status data when the charging station is not in use, such as: Idle Status: Yes, the charging station is idle and waiting for vehicles to connect. Rated Power: The rated power of the charging station is 50kW. Available Charging Power: The maximum charging power that the charging station can currently provide is 50kW. Connection Status: The charging station is normally connected to the power grid and can start charging at any time. Last Use Time: For example, if the last time the charging station ended was 9:00 AM today.

[0062] Preprocessing of the data includes data cleaning (removing missing values, outliers, and erroneous data), data transformation (converting timestamps to hours, days, weeks, etc., to suit the time series analysis requirements of the model), and feature engineering (extracting useful features from the raw data, such as usage frequency, average usage duration, and charging distribution).

[0063] Analyze the correlation between features and the target variable (using probability), and select features with high correlation. You can also use models (such as random forests) to evaluate the importance of features and select those with high importance.

[0064] Choose an appropriate prediction model, such as time series analysis (ARIMA, SARIMA), regression models (linear regression, ridge regression), machine learning models (random forest, gradient boosting tree, support vector machine), or deep learning models (LSTM, Transformer). Use methods such as cross-validation to evaluate the performance of different models and select the optimal model.

[0065] Based on a pre-defined prediction model, and combining historical usage data and current status data of charging stations, the probability of use for each charging station is calculated. The prediction model can employ machine learning algorithms such as time series analysis, regression models, or LSTM modules. By learning from and analyzing historical data, it predicts the probability of a charging station's use within a future timeframe. This usage probability data reflects the likelihood of a charging station being used in the future, providing a forward-looking basis for power adjustment.

[0066] Step S300: Obtain power grid dispatch plan data, and obtain corresponding power reference data based on the power grid dispatch plan data, the current status data, and the usage probability data.

[0067] Specifically, the main purpose of step S300 is to obtain power grid dispatch plan data and, in conjunction with the current status data and usage probability data of the charging piles, calculate the power reference data for each charging pile. This step is to ensure that the power output of the charging piles can both meet the charging needs of the vehicles and coordinate with the power grid's operation plan, thereby maximizing the stable operation of the power grid and the charging efficiency.

[0068] First, grid dispatching plan data can be obtained from relevant departments of grid operation. This data may include grid power constraints, electricity price information, and renewable energy output forecasts.

[0069] The power grid dispatch plan data (such as load data) is integrated with the current status data of charging piles (such as current power output) and usage probability data (the likelihood that the charging piles may be used in the future).

[0070] Based on the integrated data, power reference data for each charging station is calculated. This process may utilize advanced data analysis tools and algorithms to process and analyze the data, such as machine learning models and optimization algorithms. This may involve considering grid power limitations and the charging demand of the charging stations to determine the ideal power output for each charging station at different times.

[0071] Acquire grid dispatch plan data, including grid load, generation plans, and power supply capacity. Combine this data with the current status and usage probability data of each charging station to calculate power reference data. This power reference data represents a reasonable power allocation target determined based on the overall grid operation and charging station usage demand, considering both the safe and stable operation of the grid and the actual needs of the charging stations.

[0072] Step S400: Based on the deep reinforcement learning model, obtain the power adjustment strategy for each charging pile according to the current state data, the priority weight data and the power reference data, and adjust the power of each charging pile according to the power adjustment strategy.

[0073] Specifically, the purpose of step S400 is to use a deep reinforcement learning model to formulate a power adjustment strategy for each charging pile based on the current state data, corresponding priority weight data, and power reference data of the charging pile, and to execute these strategies to achieve intelligent power coordination of the charging pile cluster.

[0074] Using a trained model, a power adjustment strategy is generated for each charging station based on current charging station status data, priority weights, and power reference data. The power output of each charging station is then adjusted according to this strategy to meet the vehicle's charging needs, while also considering grid stability and electricity costs. Deep learning frameworks (such as TensorFlow or PyTorch) can be used to implement and train the deep reinforcement learning model.

[0075] Based on a deep reinforcement learning model, a power regulation strategy for each charging station is formulated by comprehensively considering its current state data, priority weight data, and power reference data. The deep reinforcement learning model can continuously learn and optimize to find the optimal power regulation strategy. During power regulation, priority is given to ensuring the power supply to high-priority charging stations, while the power of each charging station is dynamically adjusted based on the power reference data to ensure reasonable and efficient power allocation. The output power is also adjusted in real time by the charging station's control system to meet the requirements of the power regulation strategy. Real-time monitoring of the charging station's operating status and the grid load is necessary during power regulation to ensure the safety and stability of the power regulation.

[0076] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties. The collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0077] In this embodiment, by combining grid dispatch plan data with the current status and usage probability of charging piles, power reference data is determined, enabling the power adjustment of the charging pile cluster to be coordinated with the overall operation of the grid. When the grid load is high, the power output of the charging piles is reasonably reduced to avoid excessive impact on the grid; when the grid load is low, the power of the charging piles is appropriately increased to fully utilize the grid's remaining capacity. This dynamic power adjustment method helps balance the grid load, reduce peak-valley differences, improve the stability and reliability of grid operation, and reduce the risk of grid failures.

[0078] A power regulation strategy based on a deep reinforcement learning model comprehensively considers the priority weights of charging piles, their current state, and power reference data to achieve fine-grained power adjustment. High-priority charging piles (such as those located in transportation hubs or commercial centers) receive sufficient power to meet their higher usage demands; other charging piles are allocated power rationally based on the power reference data. This strategy not only ensures the normal operation of charging piles but also maximizes their utilization rate within the limits allowed by the power grid, reducing idle time and improving overall efficiency.

[0079] Furthermore, the power adjustment strategy fully considers the usage probability data of charging stations and predicts their usage in advance. When the usage probability is high, sufficient power is allocated in advance to ensure that users can quickly obtain stable charging services and reduce waiting time. At the same time, by dynamically adjusting the power, it avoids situations where the charging station has insufficient or excessive power during use, improving the stability and safety of the charging process and optimizing the user charging experience.

[0080] Therefore, this charging pile cluster power coordination method organically combines the power adjustment of charging piles with grid operation and charging demand, achieving coordinated optimization between the charging pile cluster and the power grid. This method can dynamically adjust the power allocation of each charging pile based on the real-time status of the power grid and the usage of the charging piles, ensuring the consistency and optimization of the power allocation strategy. This approach avoids the uneven power allocation and conflicts caused by the independent operation of charging piles in traditional methods, effectively improving the overall performance of the charging pile cluster and the utilization rate of grid resources.

[0081] From the perspective of the power grid, reasonable power regulation reduces grid losses and equipment wear, extends equipment lifespan, and lowers operating costs. From the perspective of charging pile operation, this method improves the utilization rate of charging piles and increases operational revenue. For users, they can obtain more efficient and stable charging services, saving time and costs. Therefore, this method improves the overall efficiency of charging pile clusters and the power grid system, achieving a win-win situation for multiple parties.

[0082] Optionally, the historical usage data includes charging time distribution data and charging duration data, the current status data includes the current operating status, and the preset prediction model includes an LSTM module and a Transformer module; the process of obtaining corresponding usage probability data based on the preset prediction model, according to each of the historical usage data and the corresponding current status data, includes:

[0083] The LSTM module is used to extract features from the charging time distribution data, the charging duration data, and the current operating status to obtain temporary feature data.

[0084] The temporary feature data is processed by the Transformer module to obtain the corresponding usage probability data.

[0085] Specifically, first, the following data was collected and organized:

[0086] Charging time distribution data: the frequency of use of charging stations at different times of the day.

[0087] Charging time data: The duration for which a vehicle is charging at a charging station.

[0088] Current operating status: Is the charging station currently idle, charging, or experiencing a malfunction?

[0089] The LSTM (Long Short-Term Memory) module is used to process time series data, enabling it to capture long-term dependencies within the data. In this example, LSTM is used for:

[0090] Processing charging time distribution data: Learning the time patterns of charging station usage, such as peak hours and off-peak hours.

[0091] Processing charging duration data: Analyze the distribution of charging duration and identify typical charging session lengths.

[0092] Integrate the current operating status (which may also include the expected charging completion time, i.e., when charging is expected to end if the charging station is in use): use this information as additional input features to improve the accuracy of predictions.

[0093] The temporary feature data output by the LSTM contains key information from the input data, which is encoded into a format suitable for further processing.

[0094] The Transformer module utilizes a self-attention mechanism to process the temporary feature data output by the LSTM. The purpose of this step is:

[0095] Capturing complex relationships between features: Through a self-attention mechanism, Transformer can identify the relationships between different features (such as charging time distribution and charging duration).

[0096] Enhance the model's predictive power: Through multi-head attention and feedforward networks, the Transformer further extracts and integrates features to generate more accurate probability predictions.

[0097] Finally, the Transformer module outputs usage probability data for each charging station at different time periods. This data reflects the model's predictions about the future likelihood of the charging station being used.

[0098] By combining a hybrid model of LSTM and Transformer, historical and current state data can be used more effectively to predict the probability of charging station usage. LSTM is responsible for capturing time-series features, while Transformer focuses on capturing global features; the combination of the two provides more accurate predictions. This model can handle diverse data types, including time-series and state information, thus better adapting to different charging stations and scenarios.

[0099] Accurate usage probability prediction is crucial for optimizing power allocation at charging stations. It not only improves charging efficiency but also reduces user waiting time. Furthermore, the model supports real-time updates and adjustments to predictions to adapt to changes in charging station usage patterns, enabling more flexible and dynamic grid management. This provides robust data support for the operation and management of charging stations.

[0100] Optionally, the power grid dispatch plan data includes the power grid load forecast data and power supply capacity data, the current status data includes real-time electricity price data and rated power, and the step of obtaining the corresponding power reference data based on the power grid dispatch plan data, the current status data, and the usage probability data includes:

[0101] The maximum power limit data for each charging pile is determined based on the power grid load forecast data and the power supply capacity data.

[0102] The maximum power limit data, the real-time electricity price data, the rated power, and the grid load prediction data are input into a preset power prediction model. The operator splitting QP algorithm is used to solve the preset power prediction model and output the power reference data corresponding to each charging pile.

[0103] The preset power prediction model is constructed based on an objective function and multiple constraints, with the goal of minimizing the optimal charging pile power output.

[0104] Optionally, the objective function specifically includes:

[0105] The formula for the objective function is:

[0106] ;

[0107] in, Minimize means to minimize the value of the objective function; The summation symbol indicates a summation over a range (such as all time periods). or all charging stations Summing up the terms within the range. It is a time period The length. This represents the total number of time periods. This represents the total number of charging stations. Time period The cost per unit of energy can be the price of electricity or other energy-related costs. For charging piles In time period The actual power output within. This is the peak power cost factor, used to penalize power usage during peak hours in order to reduce grid load fluctuations. For charging piles In time period The penalty value for waiting time or charging interruption. For charging piles In time period The probability of use within. For charging piles The rated power is the maximum power that the charging pile can continuously output under normal operating conditions.

[0108] Specifically, the data includes: Grid load forecast data: Predicted grid load demand over different time periods. Power supply capacity data: The maximum power the grid can provide over different time periods. Real-time electricity price data: Current electricity market prices, typically related to grid load and power supply capacity. Rated power: The maximum output power of the charging station.

[0109] Based on grid load forecast data and power supply capacity data, the maximum power limit for each charging pile is calculated to ensure that the total power does not exceed the grid's power supply capacity.

[0110] The data includes: Grid load forecast data: Predicted total grid load demand over a specific time period; Power supply capacity data: The maximum power the grid can provide during a specific time period; Number and rated power of charging stations: The number of charging stations and the rated power of each charging station.

[0111] Calculate the total power supply capacity of the power grid within a specific time period, i.e., power supply capacity data.

[0112] The available power supply capacity that can be used for charging piles is obtained by subtracting the grid load forecast data from the power supply capacity data.

[0113] Calculate the maximum power limit for each charging station based on available power supply capacity, the number of charging stations, and their rated power. This can be achieved using the following formula:

[0114] ;

[0115] in: It is the first The maximum power limit of a single charging station. It refers to the available power supply capacity of the power grid. This is the total number of charging stations. It is the first The rated power of each charging station.

[0116] Maximum power limit data, real-time electricity price data, rated power, and grid load forecast data are input into a pre-defined power prediction model. The operator-split quadratic programming (OSQP) algorithm is used to solve the power prediction model to find the optimal power output of the charging pile. The model outputs power reference data for each charging pile at different time periods, taking into account the grid's operating status and economic factors. Specifically, the OSQP algorithm is used to solve the pre-defined power prediction model. OSQP is an efficient method for solving quadratic programming problems, suitable for handling optimization problems with linear constraints. Based on the OSQP solution, power reference data for each charging pile is output, which will guide actual power regulation. The parameters of the power prediction model are dynamically adjusted based on real-time data and environmental changes to adapt to changes in grid load and fluctuations in charging demand.

[0117] The constraints include:

[0118] 1. Power grid security constraints:

[0119] Grid load limit constraint: Ensure that the sum of the power output of all charging piles does not exceed the power supply capacity of the grid during that time period, i.e. ,in It is the power grid during the time period Internal power supply capacity.

[0120] Voltage and frequency stability constraints: Power distribution must ensure that the grid voltage and frequency are within safe limits, i.e. and ,in and They are time periods The current grid voltage and frequency within the system. This is the minimum safe lower limit value of the grid voltage. This is the maximum safe upper limit of the grid voltage. This is the minimum safe lower limit for the power grid frequency. This represents the maximum safe upper limit for the power grid frequency.

[0121] 2. Charging pile characteristic constraints:

[0122] Charging pile rated power constraint: The power output of each charging pile cannot exceed its rated power, i.e. ,in It is a charging station Rated power.

[0123] Charging station status constraints: If the charging station is idle, the power output is 0; if it is charging, the power output must be within the allowable range and meet the charging requirements.

[0124] 3. User requirement constraints:

[0125] Charging demand meets constraints: Power allocation should meet the vehicle's basic charging needs, i.e. ,in It is the sum of the charging requirements of vehicle i at all times. The charging duration within each time period.

[0126] Waiting time minimization constraint: Minimize user waiting time, i.e. It should be as small as possible.

[0127] 4. Economic constraints:

[0128] Cost minimization constraint: Power allocation should minimize total cost, including energy cost, peak power cost, etc. minimize.

[0129] Due to the cost minimization constraint, power output needs to adapt to real-time electricity prices, and charging should be performed during periods of low electricity prices as much as possible. Should be with It is inversely proportional.

[0130] Limiting the maximum power output of charging stations can effectively prevent excessive pressure on the power grid, thereby improving grid stability and reliability. Utilizing real-time electricity price data to guide charging stations to operate during periods of lower electricity prices not only reduces charging costs but also improves economic efficiency. Furthermore, by rationally allocating charging station power based on grid load forecasts and power supply capacity, the use of electricity resources can be optimized.

[0131] Accurate power reference data can improve the efficiency of charging stations, reduce user waiting time, and thus enhance the user experience. The model can dynamically adjust power output based on real-time data, adapting to changes in grid load and electricity prices, enabling flexible and dynamic grid management. Furthermore, by optimizing charging time, it is possible to better utilize renewable energy sources such as wind and solar power, supporting the integration and use of green energy. Improving charging efficiency and optimizing power resource allocation helps reduce unnecessary energy waste, thereby lowering carbon emissions.

[0132] In summary, this process, by comprehensively considering the power grid dispatch plan, current state, and usage probability, and utilizing advanced algorithms and models, achieves precise control and optimization of charging pile power, resulting in significant economic, environmental, and social benefits.

[0133] Optionally, the step of obtaining the power adjustment strategy for each charging pile based on the deep reinforcement learning model, according to the current state data, the priority weight data, and the power reference data, includes:

[0134] Construct the deep reinforcement learning model and initialize it.

[0135] The reward value is determined based on the current status data, the corresponding priority weight data, the power reference data, the real-time electricity price data, and the rated power.

[0136] The policy gradient is determined based on the reward value, and the policy network parameters of the deep reinforcement learning model are updated according to the policy gradient to maximize the expected cumulative reward.

[0137] The power adjustment strategy for each charging pile is determined based on the updated strategy network parameters.

[0138] Specifically, a deep reinforcement learning model is constructed and initialized. This model consists of a policy network and a value network. The policy network's input includes the charging pile's current state data, priority weight data, and power reference data, and its output is the charging pile's power adjustment policy. The value network's input is the same as the policy network, and its output is the estimated value of the current state. Initialization includes randomly initializing the network parameters (weights and biases).

[0139] The reward value is calculated based on factors such as the current status data of the charging pile (e.g., current operating status, remaining power), priority weight data, power reference data, real-time electricity price data, and rated power.

[0140] The formula for calculating the reward value is: Reward value = Charging efficiency reward + Grid load balance reward + Charging cost reward + System stability reward.

[0141] The rewards are categorized into three types: charging efficiency reward, charging pile power output, and charging duration; grid load balancing reward, charging cost reward, and system stability reward, all related to grid voltage and frequency fluctuations. The policy gradient is calculated based on the reward values, representing the direction and extent of the policy network parameters' influence on the expected cumulative reward. The policy gradient is used to update the policy network parameters using the formula: new parameters = old parameters + learning rate × policy gradient. Through multiple iterations, the policy network gradually learns the parameters that maximize the expected cumulative reward. Based on the updated policy network parameters and current state data, the power adjustment strategy for each charging pile is obtained. The power adjustment strategy includes the power output value of each charging pile at the current moment, which should meet the charging pile's power constraints (e.g., not exceeding the rated maximum power).

[0142] By employing a deep reinforcement learning model and comprehensively considering multiple factors, an intelligent power regulation strategy is implemented. This strategy can be dynamically adjusted based on real-time data, improving system flexibility and adaptability. Simultaneously, the learning process of the policy network continuously optimizes the power regulation strategy, enhancing system performance and efficiency. This power regulation strategy can balance grid load, reduce charging costs, and maintain system stability.

[0143] In some embodiments, a deep reinforcement learning model is constructed and initialized:

[0144] 1. Policy Network: Inputs include current state data (such as current power output) Remaining battery power (etc.), priority weight Power reference data Real-time electricity price The output is the power regulation strategy (power output value). ).

[0145] Value Network: The input is the same as the policy network, and the output is the estimated value of the current state. .

[0146] Initialization: Randomly initialize the parameters of the policy network and the value network. and .

[0147] 2. Determine the reward value: Reward value function Taking multiple factors into consideration, the formula is as follows:

[0148] ;

[0149] in: (Charging efficiency reward) Based on the current power output With rated power (ratio) (Power grid load balancing reward) Based on total power output Deviation from grid load forecast ), (Charging cost incentive) Based on real-time electricity prices and power output). (System stability reward) (Based on fluctuations in grid voltage and frequency). This represents the grid voltage value during the current time period t. This refers to the rated voltage value of the power grid, i.e., the target voltage when the power grid is operating normally. This represents the power grid frequency value for the current time period t. This refers to the rated frequency value of the power grid, i.e., the target frequency during normal operation of the power grid. It is the weighting coefficient.

[0150] 3. Policy Gradient and Network Parameter Update: Policy Gradient: Calculating the policy and network parameters. Gradient of expected cumulative reward:

[0151] ;

[0152] in It represents a trajectory (a set of state-action sequences). Expected value, representing the expected value for all possible trajectories. On average, It is in state Take action below The probability is determined by the policy network. Provided. Charging pile In time period The reward value obtained, Charging pile In time period The state. Charging pile In time period Actions taken (such as power output values).

[0153] Update policy network parameters:

[0154] ;

[0155] in It is the learning rate, used to control the step size for updating the policy network parameters. : The objective function of the policy network, used to evaluate the performance of the policy network.

[0156] 4. Determine the power regulation strategy: Based on the updated strategy network parameters In the current state Down-output power regulation strategy That is, power output value This indicates a charging station. In time period The power output value.

[0157] Example numerical explanation: Assume that during a certain time period Charging piles The relevant data is: Current power output kW, rated power kW, total power output of the power grid kW, power grid load forecast kW, real-time electricity price Yuan / kWh, current grid voltage V (rated voltage) V), current power grid frequency Hz (rated frequency) Hz), let the weighting coefficient be , , , , .

[0158] Calculate each reward item: ;

[0159] ;

[0160] ;

[0161] ;

[0162] Total reward value:

[0163]

[0164] .

[0165] Based on this reward value, the policy gradient is calculated and the policy network parameters are updated to ultimately determine the charging station. In time period The power regulation strategy.

[0166] Optionally, the charging pile cluster power coordination method further includes:

[0167] The real-time voltage data of each charging pile is obtained and compared with a preset voltage threshold.

[0168] When the real-time voltage data is less than the preset voltage threshold, the corresponding charging pile is designated as a faulty pile, and the current actual power of the adjacent charging piles of the faulty pile is obtained. Based on the current actual power of each adjacent charging pile, the maximum reverse power data output by the corresponding adjacent charging pile is calculated, and all the maximum reverse power data are transmitted to the faulty pile.

[0169] Specifically, real-time voltage data is collected from each charging station. This data reflects the current voltage level provided by the power grid. The collected real-time voltage data is compared with a preset voltage threshold. The preset voltage threshold is a minimum safe voltage value set according to power grid specifications and charging station technical requirements.

[0170] If the real-time voltage data is lower than the preset voltage threshold, the charging station will be identified as a faulty station. This could be due to insufficient power or a faulty voltage.

[0171] Once a faulty charging station is identified, the current actual power output of adjacent charging stations will be automatically acquired. For each adjacent charging station, its maximum reverse power output will be calculated based on its current actual power. This refers to the maximum power that the charging station can feed back to the grid without violating charging station power limits and battery state-of-charge constraints. All calculated maximum reverse power data will be transmitted to the faulty charging station for power regulation and fault recovery.

[0172] The faulty pile adjusts its power based on the maximum reverse power data received to restore normal operation or at least reduce the burden on the power grid.

[0173] By monitoring voltage in real time and addressing faults promptly, we can improve the reliability and stability of charging pile clusters. Utilizing the reverse power of adjacent charging piles to support faulty piles allows for more efficient use of existing resources and reduces energy waste. When grid power is insufficient, power regulation between charging piles can alleviate the burden on the grid and enhance grid stability. Automated fault detection and handling reduce the need for manual intervention, thereby lowering maintenance costs. Rapid identification and handling of faulty piles reduces user wait times and improves user satisfaction. This method supports the operation of smart grids, better integrating renewable energy sources and promoting sustainable development by dynamically adjusting the power output of charging piles.

[0174] In summary, the process of acquiring real-time voltage data of charging piles and comparing it with preset voltage thresholds, as well as intelligently adjusting power when insufficient voltage is detected, significantly improves the operating efficiency of charging pile clusters and the stability of the power grid, while also enhancing the user's charging experience.

[0175] Optionally, the historical usage data includes usage frequency data, charging duration data, and charging volume data; determining the corresponding priority weight data based on the location data and the historical usage data includes:

[0176] Determine the corresponding initial weight data based on the location data;

[0177] Based on the principal component analysis algorithm, the location data, usage frequency data, charging duration data, and charging power data are analyzed to obtain temporary weight data;

[0178] The priority weight data corresponding to the charging pile is determined based on the initial weight data and the corresponding temporary weight data.

[0179] Specifically, analyze the geographical location of charging stations to determine whether they are located in high-priority areas (such as city centers, transportation hubs, high-demand areas, etc.).

[0180] Assign higher initial weights to charging piles located in high-priority areas and lower initial weights to charging piles located in low-priority areas. For example: commercial center: initial weight = 1.0, transportation hub: initial weight = 0.8, residential area: initial weight = 0.5, other areas: initial weight = 0.3.

[0181] Principal component analysis (PCA) processes historical usage data, collecting information on charging station usage frequency, charging duration, and charging volume. The data is then standardized to eliminate the influence of different units of measurement.

[0182] Principal component analysis (PCA) process: Calculate the correlation coefficient matrix to analyze the correlation between variables. Then, calculate eigenvalues ​​and eigenvectors to extract the principal components. Principal components are linear combinations of the original variables and can explain most of the variability in the data.

[0183] Based on the principal component analysis results, a temporary weight is calculated for each charging station. The temporary weight reflects the performance of the charging station on different usage indicators, such as charging stations with high usage frequency, long charging time, and large charging capacity receiving higher temporary weights.

[0184] The initial and temporary weights are combined to determine the final priority weight of the charging piles. The fusion method can be a simple weighted average.

[0185] ;

[0186] in, It is a charging station The final priority weight (priority weight data). These are the initial weights. It's a temporary weight. Weighting coefficients (e.g.) ).

[0187] Determining the priority weights of charging stations optimizes grid resource allocation, prioritizing the needs of charging stations with high usage frequency, long charging times, and large charging capacities, thereby improving overall utilization efficiency. Simultaneously, it ensures that charging stations located in high-priority areas receive better service, reducing user wait times and enhancing user experience. Furthermore, optimizing power allocation reduces electricity waste and lowers operating costs. Priority weights can be dynamically adjusted based on real-time data to adapt to changes in grid load and user demand, enabling flexible and dynamic grid management.

[0188] like Figure 2 As shown in the figure, an embodiment of the present invention provides a charging pile cluster power coordination device, which is applied to a charging pile power distribution system, the charging pile power distribution system including multiple charging piles; the coordination device includes:

[0189] The acquisition unit is used to acquire the current status data, historical usage data and location data of each charging pile, and determine the corresponding priority weight data based on the location data and the historical usage data.

[0190] The calculation unit is used to obtain the usage probability data of the charging pile based on the preset prediction model, according to the historical usage data and the corresponding current state data; to obtain the power grid dispatch plan data, and to obtain the corresponding power reference data according to the power grid dispatch plan data, the current state data and the usage probability data;

[0191] The adjustment unit is used to obtain the power adjustment strategy of each charging pile based on the current state data, the priority weight data and the power reference data according to the deep reinforcement learning model, and to adjust the power of each charging pile according to the power adjustment strategy.

[0192] like Figure 3 As shown in the figure, an embodiment of the present invention provides a charging pile cluster power coordination device, including a memory and a processor; the memory is used to store a computer program; the processor is used to implement the charging pile cluster power coordination method as described above when the computer program is executed.

[0193] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the charging pile cluster power coordination method described above.

[0194] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A charging pile cluster power coordination method, characterized in that, The application is applied to a charging pile power distribution system, the charging pile power distribution system comprises a plurality of charging piles; the coordination method comprises: obtaining current state data, historical use data of different time periods and location data of each charging pile, and determining corresponding priority weight data according to the location data and the historical use data; based on a preset prediction model, according to each historical use data, the corresponding current state data, the use probability data of the corresponding charging pile is obtained; obtaining power grid scheduling plan data, the power grid scheduling plan data includes power grid load prediction data and power supply capacity data, the current state data includes real-time electricity price data and rated power; according to the power grid scheduling plan data, the current state data and the use probability data, the corresponding power reference data is obtained, including: determining the maximum power upper limit data of each charging pile according to the power grid load prediction data and the power supply capacity data; inputting each maximum power upper limit data, the real-time electricity price data, the rated power, the power grid load prediction data into a preset power prediction model, using operator splitting QP algorithm to solve the preset power prediction model, and outputting the corresponding power reference data of each charging pile; wherein, the preset power prediction model is constructed based on objective function and multiple constraint conditions, and the optimal charging pile power output with minimized objective function is taken as the target; based on a deep reinforcement learning model, according to the current state data, the priority weight data and the power reference data, the power adjustment strategy of each charging pile is obtained, and the power of each charging pile is adjusted according to the power adjustment strategy.

2. The charging pile cluster power coordination method according to claim 1, characterized in that, The historical use data includes charging time distribution data and charging time length data, the current state data includes current running state, and the preset prediction model includes LSTM module and Transformer module; the preset prediction model is used to obtain corresponding use probability data according to each historical use data and corresponding current state data, including: the charging time distribution data, the charging time length data and the current running state are extracted by the LSTM module to obtain temporary feature data; the temporary feature data is processed by the Transformer module to obtain the corresponding use probability data.

3. The charging pile cluster power coordination method according to claim 2, characterized in that, The objective function specifically includes: the formula of the objective function is: ； wherein, is the total number of time periods, is the total number of charging piles, is a time period a cost per unit energy determined by the real-time electricity price; is the charging pile a real power output in a time period ; is a peak power cost coefficient; is the charging pile a penalty value for waiting time or charging interruption in a time period ; is the charging pile a usage probability in a time period ; is the rated power of the charging pile ; is the length of a time period .

4. The charging pile cluster power coordination method according to claim 2, characterized in that, based on a deep reinforcement learning model, according to the current state data, the priority weight data and the power reference data, the power adjustment strategy of each charging pile is obtained, including: the deep reinforcement learning model is constructed and initialized; determining the reward value according to each current state data, corresponding priority weight data, power reference data, real-time electricity price data and rated power; determining the strategy gradient based on the reward value, updating the policy network parameters of the deep reinforcement learning model according to the strategy gradient to maximize the expected cumulative reward; Determine the power adjustment strategy of each charging pile according to the updated policy network parameters.

5. The charging pile cluster power coordination method according to claim 1, characterized in that, The charging pile cluster power coordination method further comprises: Obtain real-time voltage data of each charging pile, and compare the real-time voltage data with a preset voltage threshold, When the real-time voltage data is less than the preset voltage threshold, the corresponding charging pile is regarded as a fault pile, the current actual power of the adjacent charging pile of the fault pile is obtained, the maximum reverse power data output by the corresponding adjacent charging pile is calculated according to each current actual power, and all the maximum reverse power data is transmitted to the fault pile.

6. The charging pile cluster power coordination method of claim 1, wherein, The historical use data includes use frequency data, charging duration data and charging power data; the corresponding priority weight data is determined according to the location data and the historical use data, comprising: Determine the initial weight data corresponding to the location data; Based on the principal component analysis algorithm, analyze the location data, use frequency data, charging duration data and charging power data to obtain temporary weight data; Determine the priority weight data of the corresponding charging pile according to the initial weight data and the corresponding temporary weight data.

7. A charging pile cluster power coordination device, characterized in that, The charging pile power distribution system comprises a plurality of charging piles; the coordination device comprises: An acquisition unit is configured to acquire current state data, historical use data and location data of each charging pile, and determine corresponding priority weight data according to the location data and the historical use data; A calculation unit is configured to obtain use probability data of the corresponding charging pile based on a preset prediction model according to the historical use data and the corresponding current state data; obtain power grid scheduling plan data, wherein the power grid scheduling plan data comprises power grid load prediction data and power supply capacity data, and the current state data comprises real-time electricity price data and rated power; obtain corresponding power reference data according to the power grid scheduling plan data, the current state data and the use probability data, comprising: determining maximum power upper limit data of each charging pile according to the power grid load prediction data and the power supply capacity data; inputting the maximum power upper limit data, the real-time electricity price data, the rated power and the power grid load prediction data into a preset power prediction model, and solving the preset power prediction model by using an operator splitting QP algorithm, and outputting the power reference data corresponding to each charging pile; wherein the preset power prediction model is constructed based on an objective function and multiple constraint conditions, and the optimal charging pile power output with minimized objective function is taken as the target; An adjustment unit is configured to obtain power adjustment strategies of each charging pile based on a deep reinforcement learning model according to the current state data, the priority weight data and the power reference data, and adjust the power of each charging pile according to the power adjustment strategies.

8. A charging pile cluster power coordination device, comprising a memory, a processor and a computer program stored in the memory and capable of running on the processor, characterized in that, The processor executes the computer program to implement the charging pile cluster power coordination method of any one of claims 1-6. The processor executes the computer program to implement the charging pile cluster power coordination method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and when the computer program is executed by a processor, the charging pile cluster power coordination method in any one of claims 1 to 6 is implemented.

Images