Charging pile power adjusting method based on edge calculation

Abstract

The invention discloses a charging pile power adjusting method based on edge calculation, and the method comprises the following steps: S1, collecting the operation data of a charging pile, and carrying out the preprocessing of the operation data; s2, mapping the charging piles to different areas by adopting an MAP-Elite algorithm, and calculating an efficiency evaluation value of each area; s3, generating a power adjustment strategy of the charging pile through an A3C algorithm; s4, adjusting the power output of each charging pile in real time according to a region division result in combination with the charging pile load and the power grid state; s5, issuing a power scheduling instruction to each charging pile through an edge computing node, and dynamically correcting a power regulation strategy; and S6, continuously updating a region division result, and issuing a power scheduling instruction again through the edge computing node to ensure that the power output of the charging pile realizes load balancing. According to the method, the charging pile power adjustment is optimized through edge calculation and the A3C algorithm, and the real-time performance, the accuracy and the power grid load balance are improved.

Description

Technical Field

[0001] The present invention relates to the field of charging pile control and smart grid technology, and in particular to a charging pile power regulation method based on edge computing. Background Art

[0002] With the rapid growth in the number of electric vehicles (EVs) and charging stations, power systems face enormous load pressure during peak hours. As critical EV infrastructure, charging station power regulation directly impacts grid stability and charging efficiency. To address issues such as uneven power distribution, excessive charging station loads, and grid overload, many researchers have explored various charging station power regulation methods. Most of these approaches focus on centralized scheduling through a central control system, achieving unified management of charging station power through centralized methods. However, existing technologies have limitations and are in urgent need of improvement and innovation.

[0003] Furthermore, existing power regulation methods are often based on simple load forecasting and optimization algorithms, such as linear regression based on historical data or simple control algorithms. While these methods can achieve power allocation to a certain extent, they fail to fully consider the real-time state of the power grid and the dynamic demands of individual charging stations. Because charging station loads and grid conditions are highly dynamic, a single forecasting model or optimization algorithm struggles to cope with complex power demands and real-time fluctuations. Traditional methods rely more on static scheduling, neglecting the need for rapid response and real-time optimization of power regulation in a dynamically changing environment. Consequently, existing technologies struggle to adapt to the demands of efficient and precise power regulation in the face of the rapid growth of electric vehicles and fluctuating charging demand.

[0004] With the rise of edge computing, distributed computing has become a viable solution. By processing data locally on devices at the edge of the network, edge computing reduces reliance on cloud computing, improving response speed and system efficiency. In the field of charging pile power regulation, the advantages of using edge computing for data processing and decision-making are significant. Through edge computing, charging piles can independently make real-time power regulation decisions without relying on centralized cloud computing centers. This approach can significantly reduce latency and improve the response speed of charging piles. This is especially true when there are a large number of charging piles and they are geographically dispersed. Edge computing can effectively reduce network burden and optimize overall power regulation.

[0005] However, existing edge computing-based charging pile power regulation methods still have shortcomings. First, most existing edge computing solutions use traditional rule-based or simple algorithms for power regulation. These methods typically only consider the current load of the charging pile and the grid status, lacking a comprehensive analysis of historical data and long-term trends. Furthermore, existing technologies fail to fully utilize advanced machine learning and deep learning algorithms to dynamically regulate the power of charging piles. For example, many existing solutions do not consider how to use intelligent algorithms to personalize the scheduling of charging piles and thus achieve precise control of charging pile power. Charging pile power regulation is more than just simple load distribution; it should also consider multiple factors such as grid status, electricity price fluctuations, and battery charging efficiency.

[0006] Therefore, how to provide a charging pile power regulation method based on edge computing is a problem that technicians in this field urgently need to solve. Summary of the Invention

[0007] One purpose of the present invention is to propose a charging pile power regulation method based on edge computing. The present invention makes full use of edge computing technology, A3C algorithm and multi-agent parallel learning technology, and describes in detail the process of regional division and dynamic power regulation of charging piles based on power demand characteristics. The operating data of the charging piles is collected in real time by edge nodes, and the A3C algorithm is used to optimize the scheduling of the charging pile power in combination with the grid status, electricity price and user charging demand, thereby realizing the precise allocation and dynamic adjustment of the charging pile power. This method has the advantages of low latency, high efficiency, intelligent regulation and grid load balancing. It can provide efficient and real-time power scheduling in large-scale charging networks, effectively improving charging efficiency and grid stability.

[0008] A charging pile power adjustment method based on edge computing according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect the operating data of the charging pile through the edge computing node and perform preprocessing;

[0010] S2. Use the MAP-Elite algorithm to divide the pre-processed operating data into regions, map charging piles to different regions based on power demand characteristics, and calculate the efficiency evaluation value of each region;

[0011] S3. Based on the efficiency evaluation value, generate a power regulation strategy for the charging pile using the A3C algorithm. The A3C algorithm optimizes the charging pile power scheduling through multiple parallel agents and performs joint training of global and local power regulation strategies.

[0012] S4. Based on the regional division results, combined with the charging pile load and grid status, the A3C algorithm is used to optimize the power regulation strategy, adjusting the power output of each charging pile in real time to achieve optimal power scheduling.

[0013] S5: The power scheduling instructions are sent to each charging pile through the edge computing node to control the output power of the charging pile. The power regulation strategy is dynamically modified by combining the feedback mechanism in the A3C algorithm.

[0014] S6. Based on the real-time status feedback of the charging piles, the regional division results are continuously updated, and the power scheduling instructions are re-issued through the edge computing node to ensure that the power output of the charging piles is load balanced.

[0015] Optionally, the operating data includes charging pile load, grid status, electricity price information and user charging demand data.

[0016] Optionally, the preprocessing includes data deduplication, denoising, missing value filling and data standardization.

[0017] Optionally, the S2 specifically includes:

[0018] S21, obtain the pre-processed running data D = {d1, d2, ..., d i ,...d n}, where d i =(L i ,G i ,P i ,U i ) represents the operating data of the i-th charging pile, L i is the load value of the i-th charging pile, G i is the grid status value of the i-th charging pile, P i is the electricity price information of the i-th charging pile, U i is the user charging demand data of the i-th charging pile, and n is the total number of charging piles;

[0019] S22, use MAP-Elite algorithm to divide D into regions, and according to the regional division result R, for each region r j Calculate the performance evaluation value E j :

[0020]

[0021] Among them, L ij is the load value of the i-th charging pile in the j-th area, G ij is the grid status value of the i-th charging pile in the j-th area, P ij is the electricity price information of the i-th charging pile in the j-th area, U ijis the user charging demand value of the i-th charging pile in the j-th area, L max ,G max ,P max ,U max is the maximum value of each data item, used for normalization, w1, w2, w3, w4 are the weight coefficients of each data, n j is the number of charging piles in the jth area;

[0022] S23, according to the performance evaluation value E j , the operating data of each charging pile d i Mapped to the corresponding region r j In this paper, the power demand characteristics of each charging pile are matched with the regional division, and the regional division result R and the efficiency evaluation value of each region are output to provide data support for the generation of power regulation strategy.

[0023] Optionally, the specific process of the area division is as follows:

[0024] S221. Define a feature space F = {f1, f2, ..., f i ,...f k}, the operating data of each charging pile d i is mapped into the feature space, where f i =(f i1 ,f i2 ,...,f ij ,...f ik ), represents the data point of the i-th feature dimension, k is the number of dimensions of the feature space, and each feature f ij represents the value of the i-th charging pile in the j-th dimension. All data points in the feature space constitute a k-dimensional feature space;

[0025] S222, for each data point f in the feature space F i Initialize randomly and set f i Mapped to the grid position in the feature space, and the initial performance evaluation value E of each grid position j is set to zero;

[0026] S223, according to each data point f i Position in feature space, calculate f i The fitness value Φ i , and determine f i The optimal position in feature space:

[0027] Φ i =α·L i +β·G i +γ·P i +δ·Ui ;

[0028] Among them, α, β, γ, and δ are the weight coefficients of charging pile load, grid status, electricity price information, and user charging demand respectively;

[0029] S224. By continuously optimizing the feature space and iteratively adjusting the regional boundaries, it is ensured that the power demand characteristics of the charging piles in each region are similar and the efficiency evaluation value of each region is maximized.

[0030] Optionally, the optimal position is based on the distance of each charging pile relative to each grid in the feature space, and the nearest grid is selected for mapping. For each area r j , mapping the corresponding grid position to r j The center point of the region c j , and according to the data point f i The grid position and fitness value Φ i To update:

[0031]

[0032] Among them, n j is the number of charging piles in the jth area.

[0033] Optionally, the S3 specifically includes:

[0034] S31, according to the performance evaluation value E j The region division result R defines the state space S = {s1, s2, ..., s i ,...,s n}, and design the strategy network π(a t |s t ;θ) to describe the state s t Next select action a t The probability distribution of s i is the state of the i-th charging pile, a t is the power regulation action of the charging pile, θ is the parameter of the strategy network, t represents the time step, and n is the total number of charging piles;

[0035] S32: Use the A3C algorithm to generate a power regulation strategy for the charging pile. The specific process is as follows:

[0036] Introducing multiple parallel agents into the A3C algorithm and using actor and critic networks to independently train charging piles for power regulation;

[0037] The agent is mapped to the charging pile one by one, and the agent selects a power adjustment action a at each time step t. itAs the output of the policy network, it evaluates a based on the reward value of the environment feedback it effect;

[0038] Each agent has its own state s it and the selected action a it Update the parameters θ of the policy network so that the policy tends to choose an efficient power regulation scheme;

[0039] S33, through multiple training and interaction, obtain the power adjustment strategy π of the charging pile * (a it |s it ; θ), and generate power adjustment instructions for each charging pile, adjusting the power output of each charging pile in real time to achieve optimal distribution of power scheduling.

[0040] Optionally, the Actor network updates the policy network parameters θ by optimizing the policy gradient so that the policy can maximize the reward obtained in a given state. The Critic network estimates the current state s it The value of φ is used to update the parameter φ:

[0041]

[0042] in, is the loss function of the Actor network, R t is the reward value at time step t, which represents the performance feedback of the i-th charging pile at time t, is the gradient of the policy network, which represents the action a adjusted according to the current policy it In state s it The rate of change of choice probability under it |s it ; θ) is the strategy network under the i-th charging pile, represents the expected value operation, log2(·) represents the logarithmic function, is the loss function of the Critic network, V φ (s it ) is the i-th charging pile in state s it The value function below.

[0043] Optionally, the S4 specifically includes:

[0044] S41, based on the regional division result R, combined with the charging pile load L i and the grid status G i , define the state of the i-th charging pile at time step t as s it =(L it ,G it ), where Lit is the load value of the i-th charging pile at time step t, G it is the grid state value of the i-th charging pile at time step t;

[0045] S42, using the A3C algorithm, updates the power regulation strategy π of the i-th charging pile at each time step t * (a it |s it ;θ), where a it represents the power regulation action of the i-th charging pile at time step t, θ is the parameter of the strategy network, s it is the status of the i-th charging pile;

[0046] S43, according to the optimized power regulation strategy, the power output of each charging pile is adjusted in real time, and the value function V is updated. φ (s it ), evaluate the effectiveness of the power regulation strategy and ensure real-time optimization of the power regulation of the charging pile when the grid status changes.

[0047] Optionally, the S5 specifically includes:

[0048] S51. Collect real-time status information of each charging pile through the edge computing node, including the current power output value, load value, grid status and user charging demand data of the charging pile;

[0049] S52, the power dispatch instruction a of each charging pile it The data is sent to each charging pile through the edge computing node, and the power output of the charging pile is controlled in real time;

[0050] S53. Dynamically modify the power regulation strategy through the feedback mechanism in the A3C algorithm. The feedback mechanism modifies the strategy based on the real-time power output, load conditions, and grid status of the charging pile, so that the power regulation strategy continuously adapts to the new operating environment.

[0051] S54. Generate a new scheduling instruction again based on the revised power regulation strategy and send it to the charging pile to ensure that the power regulation strategy is always in the optimal state to achieve efficient allocation of power scheduling.

[0052] The beneficial effects of the present invention are:

[0053] First, by employing edge computing technology, the present invention enables charging piles to perform data processing and decision-making locally, avoiding the delays and communication burdens associated with traditional centralized approaches that rely on cloud computing. Charging piles can collect data such as grid status, load information, electricity prices, and user demand in real time, and perform power regulation optimization locally, thereby improving response speed and overall system efficiency. This edge computing-based solution is particularly suitable for large-scale, distributed charging pile networks, significantly reducing reliance on central servers and lowering network communication costs, while ensuring the real-time and accurate power regulation of charging piles.

[0054] Secondly, the present invention combines the A3C algorithm with multi-agent parallel learning technology to dynamically optimize the charging pile power scheduling, effectively achieving global and local optimization of the charging pile power regulation strategy. The A3C algorithm can collaborate between multiple parallel agents to adjust the power output of each charging pile in real time to cope with changes in grid load in different regions and time periods. This method not only enables precise scheduling of charging piles, but also automatically adjusts the scheduling strategy based on grid demand and the real-time load of the charging piles, ensuring balanced power distribution, thereby avoiding grid overload and resource waste.

[0055] Finally, by combining the charging pile load, grid status, and power demand characteristics, the present invention achieves intelligent and refined regulation of charging pile power. By utilizing the real-time processing capabilities of edge computing nodes, it is possible to divide regions according to the power demand characteristics of different areas, and optimize the power regulation strategy based on the efficiency evaluation value of each area. This scheduling strategy based on real-time data analysis and reinforcement learning algorithms not only improves charging efficiency, but also greatly enhances the load management capabilities of the power grid, ensuring stable operation and efficient scheduling between charging piles and the power grid, and effectively solving the problems of delay and uneven power distribution faced by traditional scheduling methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0057] Figure 1 This is a flow chart of a charging pile power adjustment method based on edge computing proposed by the present invention;

[0058] Figure 2 This is a schematic diagram of the regional division and charging pile data mapping of a charging pile power adjustment method based on edge computing proposed by the present invention;

[0059] Figure 3 This is a schematic diagram of power regulation training and intelligent agent scheduling for a charging pile power regulation method based on edge computing proposed in the present invention. DETAILED DESCRIPTION

[0060] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0061] refer to Figure 1-3 , a charging pile power adjustment method based on edge computing, comprising the following steps:

[0062] S1. Collect the operating data of the charging pile through the edge computing node and perform preprocessing;

[0063] S2. Use the MAP-Elite algorithm to divide the pre-processed operating data into regions, map charging piles to different regions based on power demand characteristics, and calculate the efficiency evaluation value of each region;

[0064] S3. Based on the efficiency evaluation value, generate a power regulation strategy for the charging pile using the A3C algorithm. The A3C algorithm optimizes the charging pile power scheduling through multiple parallel agents and performs joint training of global and local power regulation strategies.

[0065] S4. Based on the regional division results, combined with the charging pile load and grid status, the A3C algorithm is used to optimize the power regulation strategy, adjusting the power output of each charging pile in real time to achieve optimal power scheduling.

[0066] S5: The power scheduling instructions are sent to each charging pile through the edge computing node to control the output power of the charging pile. The power regulation strategy is dynamically modified by combining the feedback mechanism in the A3C algorithm.

[0067] S6. Based on the real-time status feedback of the charging piles, the regional division results are continuously updated, and the power scheduling instructions are re-issued through the edge computing node to ensure that the power output of the charging piles is load balanced.

[0068] The present invention can optimize the charging pile power output in real time according to different grid states and charging pile load conditions, thereby improving the load balance of the grid and the charging efficiency.

[0069] In this embodiment, the operating data includes charging pile load, grid status, electricity price information and user charging demand data.

[0070] The present invention ensures the comprehensiveness and accuracy of the regulation strategy by taking the operating data of the charging pile including load, grid status, electricity price and user demand data as input.

[0071] In this embodiment, the preprocessing includes data deduplication, denoising, missing value filling and data standardization.

[0072] The present invention ensures the quality and consistency of input data by preprocessing the running data, including data deduplication, denoising, missing value filling and standardization.

[0073] In this embodiment, S2 specifically includes:

[0074] S21, obtain the pre-processed running data D = {d1, d2, ..., d i ,...d n}, where d i =(L i ,G i ,P i ,U i ) represents the operating data of the i-th charging pile, L i is the load value of the i-th charging pile, G i is the grid status value of the i-th charging pile, P i is the electricity price information of the i-th charging pile, U i is the user charging demand data of the i-th charging pile, and n is the total number of charging piles;

[0075] S22, use MAP-Elite algorithm to divide D into regions, and according to the regional division result R, for each region r j Calculate the performance evaluation value E j :

[0076]

[0077] Among them, L ij is the load value of the i-th charging pile in the j-th area, G ij is the grid status value of the i-th charging pile in the j-th area, P ij is the electricity price information of the i-th charging pile in the j-th area, U ij is the user charging demand value of the i-th charging pile in the j-th area, L max ,G max ,P max ,U max is the maximum value of each data item, used for normalization, w1, w2, w3, w4 are the weight coefficients of each data, n j is the number of charging piles in the jth area;

[0078] S23, according to the performance evaluation value E j , the operating data of each charging pile d i Mapped to the corresponding region r jIn this paper, the power demand characteristics of each charging pile are matched with the regional division, and the regional division result R and the efficiency evaluation value of each region are output to provide data support for the generation of power regulation strategy.

[0079] This paper optimizes the generation of charging pile power regulation strategies by using the MAP-Elite algorithm to segment charging pile operating data into regions and map power demand characteristics based on each region's performance evaluation. This method accurately assigns an appropriate power regulation strategy to each region based on factors such as the charging pile load, grid status, and electricity prices, thereby improving the accuracy and efficiency of power regulation.

[0080] In this embodiment, the specific process of the area division is as follows:

[0081] S221. Define a feature space F = {f1, f2, ..., f i ,...f k}, the operating data of each charging pile d i is mapped into the feature space, where f i =(f i1 ,f i2 ,...,f ij ,...f ik ), represents the data point of the i-th feature dimension, k is the number of dimensions of the feature space, and each feature f ij represents the value of the i-th charging pile in the j-th dimension. All data points in the feature space constitute a k-dimensional feature space;

[0082] S222, for each data point f in the feature space F i Initialize randomly and set f i Mapped to the grid position in the feature space, and the initial performance evaluation value E of each grid position j is set to zero;

[0083] S223, according to each data point f i Position in feature space, calculate f i The fitness value Φ i , and determine f i The optimal position in feature space:

[0084] Φ i =α·L i +β·G i +γ·P i +δ·U i ;

[0085] Among them, α, β, γ, and δ are the weight coefficients of charging pile load, grid status, electricity price information, and user charging demand respectively;

[0086] S224. By continuously optimizing the feature space and iteratively adjusting the regional boundaries, it is ensured that the power demand characteristics of the charging piles in each region are similar and the efficiency evaluation value of each region is maximized.

[0087] In the MAP-Elite algorithm, this invention defines a feature space and randomly initializes the data, enabling the charging pile's operating data to be properly mapped within the feature space. By calculating the fitness value of each data point and continuously optimizing region boundaries, this invention improves the accuracy of region division, ensuring more consistent power demand characteristics within each region, thereby enhancing the overall effectiveness of the power regulation strategy.

[0088] In this embodiment, the optimal position is based on the distance between each charging pile and each grid in the feature space, and the nearest grid is selected for mapping. j , mapping the corresponding grid position to r j The center point of the region c j , and according to the data point f i The grid position and fitness value Φ i To update:

[0089]

[0090] Among them, n j is the number of charging piles in the jth area.

[0091] By selecting and updating the optimal position for each area, the present invention enables each charging pile to find the optimal power adjustment position within the feature space. This method ensures the optimality and adaptability of the charging pile power adjustment strategy, effectively improves the balance of power demand within the area, and reduces unnecessary power waste and grid burden.

[0092] In this embodiment, S3 specifically includes:

[0093] S31, according to the performance evaluation value E j The region division result R defines the state space S = {s1, s2, ..., s i ,...,s n}, and design the strategy network π(a t |s t ;θ) to describe the state s t Next select action a t The probability distribution of s i is the state of the i-th charging pile, a t is the power regulation action of the charging pile, θ is the parameter of the strategy network, t represents the time step, and n is the total number of charging piles;

[0094] S32: Use the A3C algorithm to generate a power regulation strategy for the charging pile. The specific process is as follows:

[0095] Introducing multiple parallel agents into the A3C algorithm and using actor and critic networks to independently train charging piles for power regulation;

[0096] The agent is mapped to the charging pile one by one, and the agent selects a power adjustment action a at each time step t. it As the output of the policy network, it evaluates a based on the reward value of the environment feedback it effect;

[0097] Each agent has its own state s it and the selected action a it Update the parameters θ of the policy network so that the policy tends to choose an efficient power regulation scheme;

[0098] S33, through multiple training and interaction, obtain the power adjustment strategy π of the charging pile * (a it |s it ; θ), and generate power adjustment instructions for each charging pile, adjusting the power output of each charging pile in real time to achieve optimal distribution of power scheduling.

[0099] This invention uses the A3C algorithm to generate a power regulation strategy for charging piles and employs multiple parallel agents to optimize the scheduling of charging piles. Each agent continuously updates the strategy parameters based on its own state and environmental feedback, optimizing the power regulation actions of the charging piles. This achieves global and local optimization of charging pile power regulation, enhancing the adaptability and intelligence of the regulation strategy.

[0100] In this embodiment, the Actor network updates the policy network parameters θ by optimizing the policy gradient so that the policy can maximize the reward obtained in a given state. The Critic network estimates the current state s it The value of φ is used to update the parameter φ:

[0101]

[0102] in, is the loss function of the Actor network, R t is the reward value at time step t, which represents the performance feedback of the i-th charging pile at time t, is the gradient of the policy network, which represents the action a adjusted according to the current policy it In state s it The rate of change of choice probability under it|s it ; θ) is the strategy network under the i-th charging pile, represents the expected value operation, log2(·) represents the logarithmic function, is the loss function of the Critic network, V φ (s it ) is the i-th charging pile in state s it The value function below.

[0103] This paper optimizes the power regulation strategy through an actor and critic network, employing a combined training method of policy gradients and value functions. The actor network optimizes the policy gradient to maximize the reward, while the critic network estimates the value of the state, further improving the accuracy of the charging pile power regulation strategy. This method ensures efficient scheduling of charging piles in different environments and effectively optimizes power allocation.

[0104] In this embodiment, the S4 specifically includes:

[0105] S41, based on the regional division result R, combined with the charging pile load L i and the grid status G i , define the state of the i-th charging pile at time step t as s it =(L it ,G it ), where L it is the load value of the i-th charging pile at time step t, G it is the grid state value of the i-th charging pile at time step t;

[0106] S42, using the A3C algorithm, updates the power regulation strategy π of the i-th charging pile at each time step t * (a it |s it ;θ), where a it represents the power regulation action of the i-th charging pile at time step t, θ is the parameter of the strategy network, s it is the status of the i-th charging pile;

[0107] S43, according to the optimized power regulation strategy, the power output of each charging pile is adjusted in real time, and the value function V is updated. φ (s it ), evaluate the effectiveness of the power regulation strategy and ensure real-time optimization of the power regulation of the charging pile when the grid status changes.

[0108] This method uses the A3C algorithm to adjust the power regulation strategy in real time based on the status of each charging pile and grid information. By continuously updating the state value function and optimizing the scheduling strategy, it ensures optimal distribution of charging pile power output. This method can adjust power output in real time based on changes in grid status and load demand, effectively ensuring grid stability and charging efficiency.

[0109] In this embodiment, the S5 specifically includes:

[0110] S51. Collect real-time status information of each charging pile through the edge computing node, including the current power output value, load value, grid status and user charging demand data of the charging pile;

[0111] S52, the power dispatch instruction a of each charging pile it The data is sent to each charging pile through the edge computing node, and the power output of the charging pile is controlled in real time;

[0112] S53. Dynamically modify the power regulation strategy through the feedback mechanism in the A3C algorithm. The feedback mechanism modifies the strategy based on the real-time power output, load conditions, and grid status of the charging pile, so that the power regulation strategy continuously adapts to the new operating environment.

[0113] S54. Generate a new scheduling instruction again based on the revised power regulation strategy and send it to the charging pile to ensure that the power regulation strategy is always in the optimal state to achieve efficient allocation of power scheduling.

[0114] This invention uses edge computing nodes to collect real-time status information from charging piles and, combined with the feedback mechanism within the A3C algorithm, dynamically adjusts the power regulation strategy. The issuance of real-time adjustment instructions ensures the accuracy of charging pile power output, and continuously adjusts the scheduling strategy based on actual operating data to ensure optimal charging pile power regulation. This feedback mechanism effectively improves the system's adaptability and flexibility, ensuring optimal charging pile power scheduling under varying operating conditions.

[0115] Example 1:

[0116] To verify the feasibility of the present invention in practice, the present invention was applied to a large-scale charging pile network system, aiming to solve the problems of uneven charging pile power regulation, high latency, and unbalanced grid load. In traditional charging pile scheduling methods, charging pile power regulation usually relies on a central control unit, which faces problems such as communication delays, uneven loads, and untimely data processing. These problems result in charging piles being unable to respond quickly during peak hours, low charging efficiency, and even potentially burdening the grid, thereby affecting the stability of the entire power system.

[0117] In this embodiment, charging piles from multiple charging stations in a certain city were selected for experimental application. This system collects the operating data of the charging piles in real time by deploying edge computing nodes at each charging pile, including charging pile load, grid status, electricity price information, and user charging demand data. These data are processed and analyzed by the edge computing nodes, and the charging piles are divided into regions based on the MAP-Elite algorithm, and the charging piles are allocated to different areas according to power demand characteristics. Then, the power regulation strategy of each charging pile is optimized by the A3C algorithm to ensure that the power output of the charging pile matches the grid load and user demand. This method solves the delay problem and uneven grid load problem faced by the centralized scheduling system in the traditional method.

[0118] In practice, the system regulates the power of charging piles by region. Charging piles in each region are learned and scheduled in parallel using the A3C algorithm, based on factors such as the real-time load of the grid, the charging pile load, and electricity prices. Multiple parallel agents optimize the power output of the charging piles to ensure grid load balance, preventing grid overloads and charging piles from being unable to charge in a timely manner.

[0119] For our experiment, we selected 30 charging stations and conducted a week-long test. During the test, operating data from the charging stations was transmitted in real time to edge computing nodes. After data preprocessing, the MAP-Elite algorithm was used to divide the stations into zones, and power was adjusted based on the efficiency evaluation of each zone. The system dynamically adjusted the power of the charging stations based on real-time data, ensuring that each station received the optimal charging power despite fluctuations in grid load and user demand.

[0120] To verify the effectiveness of this invention, we compared the performance of charging piles before and after implementing this invention, focusing on charging efficiency, grid load fluctuations, power regulation response time, and grid stability. The following are the relevant data from the experimental results:

[0121] Table 1 Comparison of charging pile power regulation effects

[0122]

[0123]

[0124] As can be seen from the above table, the present invention significantly improves the performance of charging piles. Before implementation, the average charging time was 40 minutes, while after implementation, through real-time power regulation, the average charging time was reduced to 30 minutes, a reduction of 25%. The grid load fluctuation was also significantly reduced, with an improvement rate of 40%, which shows that the present invention effectively balanced the grid load through refined power regulation and avoided grid instability. The power regulation response time was also greatly reduced, from 10 seconds before implementation to 3 seconds, and the response speed was increased by 70%. This enables the charging pile to respond quickly to changes in the grid and user needs, ensuring charging efficiency and grid stability.

[0125] Furthermore, the frequency of grid overloads decreased from five per month to one per month, an improvement of 80%. This demonstrates that edge computing-based power regulation methods can effectively prevent grid overloads and improve grid stability and reliability. The utilization rate of charging stations also increased from 60% to 85%, meaning more users can successfully charge during peak hours.

[0126] Overall, the edge computing-based charging pile power regulation method provided by this invention demonstrates its superiority in multiple dimensions. Through real-time data processing using edge computing and dynamic regulation using the A3C algorithm, it significantly improves charging efficiency, optimizes grid load distribution, shortens power regulation response time, and ensures grid stability. These results fully demonstrate the practical value and broad application prospects of this invention in charging pile power regulation.

[0127] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A charging pile power adjustment method based on edge computing, characterized in that: The steps include: S1. Collect the operating data of the charging pile through the edge computing node and perform preprocessing; S2. Use the MAP-Elite algorithm to divide the pre-processed operating data into regions, map charging piles to different regions based on power demand characteristics, and calculate the efficiency evaluation value of each region; S3. Based on the efficiency evaluation value, generate a power regulation strategy for the charging pile using the A3C algorithm. The A3C algorithm optimizes the charging pile power scheduling through multiple parallel agents and performs joint training of global and local power regulation strategies. S4. Based on the regional division results, combined with the charging pile load and grid status, the A3C algorithm is used to optimize the power regulation strategy, adjusting the power output of each charging pile in real time to achieve optimal power scheduling. S5: The power scheduling instructions are sent to each charging pile through the edge computing node to control the output power of the charging pile. The power regulation strategy is dynamically modified by combining the feedback mechanism in the A3C algorithm. S6. Based on the real-time status feedback of the charging piles, the regional division results are continuously updated, and the power scheduling instructions are re-issued through the edge computing node to ensure that the power output of the charging piles is load balanced.

2. A charging pile power adjustment method based on edge computing according to claim 1, characterized in that: The operating data includes charging pile load, grid status, electricity price information and user charging demand data.

3. The charging pile power adjustment method based on edge computing according to claim 1 is characterized in that: The preprocessing includes data deduplication, denoising, missing value filling and data standardization.

4. The charging pile power adjustment method based on edge computing according to claim 1, characterized in that: The S2 specifically includes: S21, obtain the pre-processed running data D = {d1, d2, ..., d i ,...d n }, where d i =(L i ,G i ,P i ,U i ) represents the operating data of the i-th charging pile, L i is the load value of the i-th charging pile, G i is the grid state value of the i-th charging pile, P i is the electricity price information of the i-th charging pile, U i is the user charging demand data of the i-th charging pile, and n is the total number of charging piles; S22, use MAP-Elite algorithm to divide D into regions, and according to the regional division result R, for each region r j Calculate the performance evaluation value E j : Among them, L ij is the load value of the i-th charging pile in the j-th area, G ij is the grid status value of the i-th charging pile in the j-th area, P ij is the electricity price information of the i-th charging pile in the j-th area, U ij is the user charging demand value of the i-th charging pile in the j-th area, L max ,G max ,P max ,U max is the maximum value of each data item, used for normalization, w1, w2, w3, w4 are the weight coefficients of each data, n j is the number of charging piles in the jth area; S23, according to the performance evaluation value E j , the operating data of each charging pile d i Mapped to the corresponding region r j In this paper, the power demand characteristics of each charging pile are matched with the regional division, and the regional division result R and the efficiency evaluation value of each region are output to provide data support for the generation of power regulation strategy.

5. The charging pile power adjustment method based on edge computing according to claim 4 is characterized in that: The S22 specifically includes: S221. Define a feature space F = {f1, f2, ..., f i ,...f k }, the operating data of each charging pile d i is mapped into the feature space, where f i =(f i1 ,f i2 ,...,f ij ,...f ik ), represents the data point of the i-th feature dimension, k is the number of dimensions of the feature space, and each feature f ij represents the value of the i-th charging pile in the j-th dimension. All data points in the feature space constitute a k-dimensional feature space; S222, for each data point f in the feature space F i Initialize randomly and set f i Mapped to the grid position in the feature space, and the initial performance evaluation value E of each grid position j is set to zero; S223, according to each data point f i Position in feature space, calculate f i The fitness value Φ i , and determine f i The optimal position in feature space: F i =α·L i +β·G i +γ·P i +δ·U i ; Among them, α, β, γ, and δ are the weight coefficients of charging pile load, grid status, electricity price information, and user charging demand respectively; S224. By continuously optimizing the feature space and iteratively adjusting the regional boundaries, it is ensured that the power demand characteristics of the charging piles in each region are similar and the efficiency evaluation value of each region is maximized.

6. A charging pile power adjustment method based on edge computing according to claim 5, characterized in that: The optimal position is based on the distance of each charging pile relative to each grid in the feature space, and the nearest grid is selected for mapping. For each area r j , mapping the corresponding grid position to r j The center point of the region c j , and according to the data point f i The grid position and fitness value Φ i To update: Among them, n j is the number of charging piles in the jth area.

7. The charging pile power adjustment method based on edge computing according to claim 1, characterized in that: The S3 specifically includes: S31, according to the performance evaluation value E j The region division result R defines the state space S = {s1, s2, ..., s i ,...,s n }, and design the strategy network π(a t |s t ;θ) to describe the state s t Next select action a t The probability distribution of s i is the state of the i-th charging pile, a t is the power regulation action of the charging pile, θ is the parameter of the strategy network, t represents the time step, and n is the total number of charging piles; S32: Use the A3C algorithm to generate a power regulation strategy for the charging pile. The specific process is as follows: Introducing multiple parallel agents into the A3C algorithm and using actor and critic networks to independently train charging piles for power regulation; The agent is mapped to the charging pile one by one, and the agent selects a power adjustment action a at each time step t. it As the output of the policy network, it evaluates a based on the reward value of the environment feedback it effect; Each agent has its own state s it and the selected action a it Update the parameters θ of the policy network so that the policy tends to choose an efficient power regulation scheme; S33, through multiple training and interaction, obtain the optimal power adjustment strategy π of the charging pile * (a t |s t ; θ), and generate power adjustment instructions for each charging pile, adjusting the power output of each charging pile in real time to achieve optimal distribution of power scheduling.

8. The charging pile power adjustment method based on edge computing according to claim 7 is characterized in that: The Actor network updates the policy network parameters θ by optimizing the policy gradient so that the policy can maximize the reward obtained in a given state. The Critic network estimates the current state s it The value of φ is used to update the parameter φ: in, is the loss function of the Actor network, R t is the reward value at time step t, which represents the performance feedback of the i-th charging pile at time t, is the gradient of the policy network, indicating the action a adjusted according to the current policy it In state s it The rate of change of choice probability under it |s it ; θ) is the strategy network under the i-th charging pile, represents the expected value operation, log2(·) represents the logarithmic function, is the loss function of the Critic network, For the i-th charging pile in state s it The value below.

9. The charging pile power adjustment method based on edge computing according to claim 1, characterized in that: The S4 specifically includes: S41, based on the regional division result R, combined with the charging pile load L i and the grid status G i , define the state of the i-th charging pile at time step t as s it =(L it ,G it ), where L it is the load value of the i-th charging pile at time step t, G it is the grid state value of the i-th charging pile at time step t; S42, using the A3C algorithm, updates the power regulation strategy π of the i-th charging pile at each time step t * (a it |s it ;θ), where a it represents the power regulation action of the i-th charging pile at time step t, θ is the parameter of the strategy network, s it is the status of the i-th charging pile; S43, according to the optimized power regulation strategy, the power output of each charging pile is adjusted in real time, and the value function V is updated. φ (s it ), evaluate the effectiveness of the power regulation strategy and ensure real-time optimization of the power regulation of the charging pile when the grid status changes.

10. The charging pile power adjustment method based on edge computing according to claim 1, characterized in that: The S5 specifically includes: S51. Collect real-time status information of each charging pile through the edge computing node, including the current power output value, load value, grid status and user charging demand data of the charging pile; S52, the power dispatch instruction a of each charging pile it The data is sent to each charging pile through the edge computing node, and the power output of the charging pile is controlled in real time; S53. Dynamically modify the power regulation strategy through the feedback mechanism in the A3C algorithm. The feedback mechanism modifies the power regulation strategy based on the real-time power output, load conditions, and grid status of the charging pile, so that the power regulation strategy continuously adapts to the new operating environment. S54. Generate a new scheduling instruction again based on the revised power regulation strategy and send it to the charging pile to ensure that the power regulation strategy is always in the optimal state to achieve efficient allocation of power scheduling.

Images