Edge calculation optimization method, system and equipment based on multi-station fusion and medium

Abstract

The invention relates to the technical field of edge calculation optimization, and discloses an edge calculation optimization method, system, equipment and medium based on multi-station fusion, and the method comprises the steps: obtaining and preprocessing the first data of a charging facility of an electric vehicle in a target region, constructing a first local state evaluation model, and constructing a second global situation awareness model, and the operation situation of the station is grasped to support regional planning. And according to the second model, the power grid time-of-use electricity price and renewable energy output prediction data, a third multi-target resource optimization model is constructed, multi-factor optimization resource configuration is considered, and a win-win situation of economic and environmental benefits is realized. And constructing a fourth distributed collaborative decision-making model based on the third model, and improving decision-making efficiency and accuracy in combination with edge calculation characteristics. And calling a fourth model through an upper-layer coordination function to generate an optimal strategy, and executing a lower-layer control function to realize seamless connection of decision execution. The steps are progressive layer by layer to form a complete optimization system, efficiency and stability are improved, cost is reduced, and global resources are reasonably utilized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of edge computing optimization, and in particular to an edge computing optimization method, system, device and medium based on multi-station fusion. BACKGROUND

[0002] With the rapid popularization and wide application of electric vehicles in the market, the number of electric vehicle charging and swapping facilities that match them is also increasing continuously. This rapid growth trend makes the operation and management of charging and swapping facilities face many complex and severe challenges. The traditional centralized computing method inevitably encounters a series of problems when processing the massive data generated by such a large and dispersed charging and swapping facility. For example, there is a significant delay in the transmission process of data, the communication bandwidth is under tremendous pressure, and the computing burden of the center node is also exceptionally heavy. These problems seriously affect the real-time and efficiency of the system, making it difficult to meet the urgent needs of modern electric vehicle charging and swapping facilities for fast response and high efficiency.

[0003] Edge computing, as a new computing model that has emerged in recent years, its core idea is to place computing and data storage tasks as close to the data source as possible. This model can effectively reduce the delay of data in the transmission process, significantly improve the response speed and overall reliability of the system. Through data processing and analysis at the edge node, edge computing provides a new solution for the intelligent management of electric vehicle charging and swapping facilities.

[0004] However, although edge computing has many advantages in theory, the current edge computing optimization method for electric vehicle charging and swapping facilities still has some obvious shortcomings in practical application. For example, existing methods often lack global perception ability of multiple charging and swapping station states, making it difficult to achieve collaborative decision-making between multiple sites. This limitation makes it difficult for the system to achieve optimal allocation of resources and adaptively optimize according to real-time changes in the environment and demand, thereby affecting the overall system efficiency and user experience. Therefore, further research and improvement of edge computing optimization methods for electric vehicle charging and swapping facilities is still an important issue to be solved. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] Therefore, the present application provides an edge computing optimization method, system, device and medium based on multi-station fusion, which can solve the problem of lack of global perception and collaborative decision-making of multiple sites in existing edge computing optimization methods, and achieve optimal allocation of electric vehicle charging and swapping facility resources and adaptive optimization of the system.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] In a first aspect, the present application provides a multi-station fusion-based edge computing optimization method, comprising:

[0009] obtaining first data of a plurality of electric vehicle charging and battery swapping facilities in a target area, and preprocessing the first data;

[0010] The first data includes operation data, user demand data, and grid-side interaction data;

[0011] Based on the preprocessed first data, a first local state evaluation model is constructed on the edge computing nodes deployed at each station;

[0012] The output results of the first local state evaluation model of each station are uploaded to the regional edge coordination center through a secure communication link, and a second global situation awareness model is constructed based on multi-station state data;

[0013] According to the second global situation awareness model, the time-of-use electricity price, and the renewable energy output prediction data, a third multi-objective resource optimization model is established;

[0014] Based on the third multi-objective resource optimization model, combined with the computing power distribution and communication delay characteristics of the edge computing nodes, a fourth distributed collaborative decision-making model is constructed;

[0015] The fourth distributed collaborative decision-making model is called by a preset upper coordination function, an optimal operation strategy for each station is generated, and a corresponding lower control function is triggered to execute the optimization strategy.

[0016] As a preferred scheme of the multi-station fusion-based edge computing optimization method of the present application, wherein:

[0017] Obtaining execution result data of the optimal operation strategy at each station;

[0018] The execution result data includes actual power adjustment amount, user shunting effect, energy storage response accuracy, communication overhead, and edge node resource occupation rate;

[0019] The execution result data is fed back to the first local state evaluation model, the second global situation awareness model, the third multi-objective resource optimization model, and the fourth distributed collaborative decision-making model for online iterative updating and performance evaluation of model parameters.

[0020] As a preferred scheme of the multi-station fusion-based edge computing optimization method of the present application, wherein:

[0021] When the model prediction error is continuously lower than the preset threshold for several times;

[0022] Or when the synergistic optimization of revenue growth tends to converge, it is determined that the iteration condition is met, and the model update is suspended;

[0023] Otherwise, continue to run the above process to achieve adaptive optimization of the edge computing system.

[0024] As a preferred scheme of the edge computing optimization method based on multi-station fusion according to the application, wherein: the first local state evaluation model output result of each station is uploaded to the regional edge collaborative center through a secure communication link, and a second global situation awareness model is constructed based on multi-station state data, including:

[0025] Each station is regarded as a graph node, and the geographical distance or power grid connection relationship between stations is regarded as an edge;

[0026] The input of the second global situation awareness model is the equipment health degree, service capability, and load pressure of each station;

[0027] The output of the second global situation awareness model is a regional-level situation judgment result, including overload risk area identification, service bottleneck prediction, and resource mismatch detection.

[0028] This preferred scheme can comprehensively and accurately grasp the overall operation situation of each station in the region. By regarding each station as a graph node and regarding the geographical distance or power grid connection relationship between stations as an edge to construct the model, the spatial correlation and power grid topology between stations can be fully considered, so that the model is more suitable for the actual operation environment. Based on the analysis of the input information such as the equipment health degree, service capability, and load pressure of each station, the operation status of the station can be understood in multiple dimensions.

[0029] For overload risk area identification, the region that may be overloaded can be found in advance, so that measures can be taken in time to adjust the load or maintain the equipment, avoid equipment damage and power interruption caused by overload, and ensure the safe and stable operation of the power system. Service bottleneck prediction helps to plan resource allocation in advance, optimize service process, improve service quality and efficiency, and meet the growing demand of users. Resource mismatch detection can find unreasonable resource allocation, reallocate resources, improve resource utilization efficiency, and reduce operating costs.

[0030] As a preferred scheme of the edge computing optimization method based on multi-station fusion according to the application, wherein: further comprising:

[0031] The first local state evaluation model is used to evaluate the equipment health degree, load level, and service capability of the station in real time;

[0032] The second global situation awareness model is used to identify regional overload risk, service bottleneck, and resource mismatch problems;

[0033] The optimization strategy includes power scheduling, user guidance, energy storage charging and discharging adjustment, or dynamic pricing operation.

[0034] As a preferred scheme of the edge computing optimization method based on multi-station fusion, the third multi-objective resource optimization model comprises:

[0035] The multi-objective includes reducing the total energy consumption of the region, balancing the load of each station, minimizing the user waiting time, and improving the green electricity consumption ratio.

[0036] The third multi-objective resource optimization model includes a target function and a constraint condition, and the target function is an arbitrary function representing reducing the total energy consumption of the region, balancing the load of each station, minimizing the user waiting time, and improving the green electricity consumption ratio.

[0037] The constraint function at least includes a power upper limit constraint, an energy storage SOC constraint, and a user service quality constraint.

[0038] As a preferred scheme of the edge computing optimization method based on multi-station fusion, the upper coordination function runs in the regional edge coordination center, and the lower control function is distributed in each station edge node, forming a two-level edge computing architecture of "regional coordination-local execution".

[0039] In a second aspect, the application provides an edge computing optimization system based on multi-station fusion, comprising:

[0040] A data acquisition and processing module is configured to acquire first data of a plurality of electric vehicle charging and battery swapping facilities in a target region, and to preprocess the first data.

[0041] The first data includes operation data, user demand data, and grid-side interaction data.

[0042] A first model establishment module is configured to construct a first local state evaluation model on the edge computing nodes deployed at each station based on the preprocessed first data.

[0043] A second model establishment module is configured to upload the output results of the first local state evaluation model of each station to the regional edge coordination center through a secure communication link, and to construct a second global situation awareness model based on multi-station state data.

[0044] A third model establishment module is configured to establish a third multi-objective resource optimization model based on the second global situation awareness model and grid time-of-use electricity price and renewable energy output prediction data.

[0045] A fourth model establishing module is configured to establish a fourth distributed collaborative decision-making model based on the third multi-objective resource optimization model and in combination with the computing power distribution and communication delay characteristics of the edge computing nodes;

[0046] A performing module is configured to invoke the fourth distributed collaborative decision-making model through a preset upper-layer coordination function, generate an optimal operation strategy for each site, and trigger a corresponding lower-layer control function to execute the optimization strategy.

[0047] In a third aspect, the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method as described above when executing the computer program.

[0048] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the method as described above.

[0049] Compared with the prior art, the present application has the following beneficial effects: the present application proposes an edge computing optimization method based on multi-site fusion, and the first data of electric vehicle charging and battery swapping facilities in a target region are acquired and preprocessed, which has a significant effect. The operation data helps to master the real-time condition of the facilities, find potential problems, and ensure stable operation; the user demand data can grasp the habits and demands of users, optimize services, and improve satisfaction; the power grid side interaction data can coordinate the relationship with the power grid, reduce the pressure, and reasonably allocate power resources. The preprocessing can remove noise and outliers, improve data quality, and provide a reliable foundation for subsequent modeling.

[0050] A first local state evaluation model is constructed based on the preprocessed first data, and each site can accurately evaluate its own operation state. The model considers multiple factors in combination with the actual situation, finds problems and potential risks, provides a basis for autonomous decision-making and optimization, and enhances self-management capability.

[0051] The output results of the first local state evaluation model of each site are uploaded to a regional edge collaborative center, and a second global situation awareness model is constructed, which can grasp the operation situation of each site in the target region from a macro perspective. Integrating the state data of multiple sites, finding the correlation and influence of the sites, providing information support for regional planning and coordination, and realizing cross-site resource optimization configuration.

[0052] A third multi-objective resource optimization model is established according to the second global situation awareness model and the time-of-use electricity price and renewable energy output prediction data, which can fully consider multiple factors and realize resource optimization configuration. In combination with the time-of-use electricity price, the facilities are operated in the valley period to reduce the cost; according to the renewable energy prediction data, the renewable energy is preferentially utilized to improve the efficiency and reduce the dependence, realizing the win-win of economic and environmental benefits.

[0053] A fourth distributed collaborative decision-making model is constructed based on the third multi-objective resource optimization model, and the advantages of edge computing are given in combination with the computing power distribution and communication delay characteristics of the edge computing nodes. The model makes decisions according to the actual situation of the nodes, avoids the communication bottleneck and computing pressure of centralized decision-making, improves the decision-making efficiency and accuracy, reduces the influence of communication delay, and realizes fast and real-time decision-making.

[0054] The fourth distributed collaborative decision-making model is called by a preset upper coordination function to generate the optimal operation strategy of each site, and a lower control function is triggered to execute the optimized strategy, realizing seamless connection from decision-making to execution. The upper coordination function coordinates the whole situation, ensures the cooperation of the operation strategy, and achieves the overall optimization; the lower control function accurately executes the strategy, ensures that the facility operates according to the optimal strategy, and improves the system performance and efficiency.

[0055] In summary, the steps of the present application are interrelated and progressive, forming a complete optimization system. From data acquisition and processing, model construction and decision-making to strategy execution, each link is closely coordinated to comprehensively optimize the edge computing system based on multi-site fusion. Not only does it improve the operation efficiency and stability of the charging and swapping facility, but also reduces the cost, reasonably allocates and efficiently utilizes the power resources, enhances the system reliability and flexibility, and has significant economic and social benefits. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0057] Figure 1 A method flow chart of a multi-site fusion based edge computing optimization method provided by an embodiment of the present application.

[0058] Figure 2 An internal structure diagram of an electronic device of a multi-site fusion based edge computing optimization method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0060] Embodiment 1, refer to Figure 1For the first embodiment of the present application, the embodiment provides a multi-station fusion-based edge computing optimization method, which comprises the following steps:

[0061] In the prior related art, there are some problems, for example, in the data collection process of some edge computing systems, there may be incomplete or inaccurate data, which will affect the accuracy of subsequent model construction and decision-making. Moreover, some methods have poor adaptability to complex and variable actual environments, and it is difficult to make timely and effective adjustments according to real-time changes. In addition, in terms of communication, the secure communication link may be attacked or interfered, resulting in reduced reliability of data transmission, and thus affecting the collaborative decision-making and optimization effect of the system.

[0062] The present application provides a method that can effectively solve the above-mentioned problems. Next, how to implement the multi-station fusion-based edge computing optimization method will be described in detail in combination with multiple embodiments.

[0063] Figure 1 A method flowchart of a multi-station fusion-based edge computing optimization method is shown, which comprises the following steps:

[0064] S101, first data of a plurality of electric vehicle charging and swapping facilities in a target area is acquired, and the first data is preprocessed, wherein:

[0065] It should be noted that, in order to implement multi-station fusion-based edge computing optimization, relevant data about multi-station fusion needs to be acquired, which can be collected through various ways such as sensors, communication interfaces, etc. The running data, user demand data and grid-side interaction data of the electric vehicle charging and swapping facilities in the target area.

[0066] In the embodiment of the present application, the first data comprises running data, user demand data and grid-side interaction data.

[0067] In some specific embodiments, the running data covers the charging power, charging duration, equipment temperature, etc. of the facility, which can intuitively reflect the real-time running state of the facility; the user demand data includes the user's charging time preference, charging quantity demand, etc., which helps to grasp the user's usage habits and demand characteristics; and the grid-side interaction data contains the power exchange situation with the grid, electricity price information, etc., which can be used to coordinate the relationship between the charging and swapping facility and the grid.

[0068] It should be noted that the present application does not limit the specific content of the first data, and related technical personnel can make flexible adjustments and supplements according to actual application scenarios and needs.

[0069] It should also be noted that the first data cannot be directly used for model construction and needs to be preprocessed.

[0070] In some specific embodiments, the preprocessing process includes data cleaning, normalization, feature extraction, etc. Data cleaning can remove noise, missing values and outliers in the data, improve the quality and reliability of the data; normalization can convert data of different ranges and scales to a unified range, facilitating model processing; feature extraction can extract representative and discriminative features from the original data, reduce data dimension, and improve the training efficiency and performance of the model.

[0071] However, the above preprocessing steps cannot well meet the needs of complex and variable practical application scenarios. Therefore, according to the operation law and physical characteristics of electric vehicle charging and swapping facilities, the application sets reasonable thresholds and rules to improve the accuracy of outlier detection.

[0072] Specifically, the first data of a plurality of electric vehicle charging and swapping facilities in a target area is obtained, including operation data (such as charging power, charging time, equipment temperature, etc.), user demand data (such as user's charging time preference, charge demand, etc.) and grid side interaction data (such as power exchange with the grid, electricity price information, etc.).

[0073] For example, for each variable X i (representing a certain type of data, such as charging power in this application), a reasonable upper and lower threshold [L i , U i ] can be set, and any data point outside this range is considered as an outlier and is removed or corrected.

[0074] The outlier detection formula can be based on statistical methods (3σ principle is used in this application): if |X i -μ i |>3σ i , then X i is considered as an outlier, where μ i and σ i are the mean and standard deviation of X i , respectively.

[0075] Convert data of different ranges and scales to a unified range for model processing.

[0076] The normalization formula can be or use Z-score standardization

[0077] Extract representative and discriminative features from the original data to reduce data dimension.

[0078] It should be noted that obtaining the first data of a plurality of electric vehicle charging facilities in the target area and preprocessing the first data can make the model more accurate in evaluating the health degree, load level and service capacity of each site device, and reliable data can make the first local state evaluation model more accurate. Through the running data, it can be judged whether the device has potential failure, according to the user demand data, it can be reasonably evaluated whether the service capacity meets the demand, and combined with the grid side interaction data, it is helpful to analyze whether the load level is reasonable.

[0079] In S102, based on the preprocessed first data, a first local state evaluation model is constructed on the edge computing nodes deployed at each site, wherein:

[0080] It should be noted that after obtaining the preprocessed first data, the first local state evaluation model can be constructed on the edge computing nodes deployed at each site using these data.

[0081] In some specific embodiments, the first local state evaluation model can use machine learning or deep learning algorithms such as decision tree, neural network, etc. Taking the decision tree algorithm as an example, it can classify and evaluate the health degree, load level and service capacity of the site device according to different features in the preprocessed data, such as charging power, device temperature, user charging time preference, etc. Through learning a large amount of historical data, the decision tree can find the association rules between each feature and the device state, so as to realize the accurate judgment of the current site state.

[0082] If a neural network algorithm is used, its powerful nonlinear mapping capability can handle complex data relationships. The preprocessed first data is input into the neural network, and after calculation and processing of multiple hidden layers, the evaluation results of the health degree, load level and service capacity of the site device are finally output. In the training process, the weights and biases of the neural network are continuously adjusted to make the output results as close as possible to the actual situation, so as to improve the accuracy and reliability of the model.

[0083] In some embodiments, the first local state assessment model can also consider the uncertainty and fuzziness of the data in the real world by incorporating fuzzy logic theory. Fuzzy logic can handle fuzzy concepts such as "good", "average", and "poor" device states, and by setting fuzzy rules and membership functions, the pre-processed data can be mapped to different fuzzy sets, thus providing a more realistic assessment of the site device state. For example, for the feature of device temperature, when the temperature is in a certain interval, it can belong to both the "normal temperature" and "high temperature" fuzzy sets. Through fuzzy reasoning, the device state can be assessed to be between "good" and "average". This fuzzy logic-based approach can make the first local state assessment model more adaptable and robust in complex and variable real-world environments.

[0084] However, neither of the above two approaches can fully adapt to all real-world application scenarios. Therefore, the present application proposes an innovative method to build the first local state assessment model.

[0085] In embodiments of the present application, the first local state assessment model is used to assess the health, load level, and service capacity of the site devices in real time.

[0086] Specifically, the first local state assessment model can be established as follows:

[0087] Step 1.1: For each feature (e.g., charging power P, device temperature T, etc.), define the corresponding membership function μ A (x) to describe the degree to which it belongs to a certain fuzzy set.

[0088] For example, for the device temperature T, three fuzzy sets can be defined: "cold", "moderate", and "hot", with corresponding membership functions μ cold (T), μ moderate (T), and μ hot (T).

[0089] Step 1.2: Establish a series of fuzzy rules based on domain knowledge, such as "if the temperature is moderate and the load is light, then the device is in good health".

[0090] The rule form can be IF A AND B THEN C.

[0091] Step 1.3: Apply fuzzy rules to infer the input variables to obtain fuzzy outputs about the device health, load level, and service capacity.

[0092] For example, for a specific input X = [P, T,...] at a certain time, the fuzzy value H of the device health can be calculated through fuzzy reasoning.

[0093] Step 1.4: Convert the fuzzy output to an exact value. The commonly used method is the barycenter method. Suppose the fuzzy value of H is de-fuzzified to obtain the exact health score h.

[0094] It should be noted that based on the pre-processed first data, the first local state assessment model constructed on the edge computing nodes deployed at each site can provide accurate data support for subsequent global situation awareness and resource optimization configuration. After accurately assessing the health degree, load level and service capacity of each site device through this model, the regional edge coordination center can obtain more real and reliable running data of each site when building the second global situation awareness model. This helps to more accurately grasp the running situation of each site in the target region from a macro perspective, more clearly discover the correlation and influence between sites, and thus provide more targeted information support for regional planning and coordination, achieving more efficient cross-site resource optimization configuration.

[0095] S103, upload the output results of the first local state assessment model of each site to the regional edge coordination center through a secure communication link, and construct a second global situation awareness model based on multi-site state data; wherein:

[0096] It should be noted that after the first local state assessment model of each site is established, the output of the first local state assessment model of each site needs to be obtained and uploaded to the regional edge coordination center through a secure communication link. The use of a secure communication link ensures the security and integrity of data during transmission, preventing data from being stolen or tampered with. At the regional edge coordination center, a second global situation awareness model will be constructed based on multi-site state data.

[0097] In some specific embodiments, the secure communication link can use encryption tunnel technology such as the SSL / TLS protocol to encrypt the transmitted data. This can effectively prevent data from being eavesdropped and tampered with during network transmission, ensuring the confidentiality and integrity of the data. At the same time, the secure communication link can also set up an identity authentication mechanism, so that only authenticated sites and the regional edge coordination center can perform data transmission, avoiding illegal access and data leakage.

[0098] In other specific embodiments, in order to ensure the stability and reliability of communication, the secure communication link can use a redundant backup method. For example, multiple communication links with different paths are established at the same time, and when one of the links fails, data can be automatically switched to other links for transmission.

[0099] It is worth noting that the regional edge coordination center is the core hub of the entire edge computing optimization system, which undertakes important tasks such as the integration, analysis and decision-making of multi-site data. After receiving the first local state evaluation model output results from each site, the regional edge coordination center will use advanced data analysis techniques to deeply mine these massive data. It can identify potential correlations and mutual influences between sites in the target region, such as whether the decline in the health of some site equipment will have a chain reaction on the load level of surrounding sites.

[0100] In some specific embodiments, and based on the multi-site state data, the second global situational awareness model can be constructed in various ways. For example, a graph neural network (GNN) algorithm can be used. Graph neural networks can well handle data with graph structure, regarding each site as a node in the graph and the correlation between sites as an edge, so that a graph structure reflecting the state of multiple sites can be constructed. By learning and analyzing the features of nodes and edges in the graph, the graph neural network can capture the complex topological structure and interaction relationship between sites, thereby achieving accurate perception of the global situation in the target region.

[0101] Specifically, when constructing the graph neural network model, the feature representation of nodes and edges needs to be determined first. Node features can include evaluation results such as device health, load level and service capacity of each site, which can be obtained from the output results of the first local state evaluation model. The features of the edge can represent the physical connection relationship between the sites, the data interaction frequency, etc. Then, these features are input into the graph neural network for training. During the training process, the graph neural network will continuously adjust the parameters to optimize the perception ability of the global situation. After training is completed, the graph neural network can output the evaluation results of the global situation in the target region, such as the overall running state of each site, potential risks, etc., according to the input multi-site state data.

[0102] In other specific embodiments, a Bayesian network-based method can also be used to construct the second global situational awareness model. Bayesian networks are a kind of probabilistic graphical models that can represent the probabilistic dependence relationship between variables through nodes and directed edges. In this scenario, factors such as device health, load level, service capacity of each site can be regarded as nodes, and the conditional probability distribution between nodes can be determined according to domain knowledge and historical data, thereby constructing a Bayesian network model.

[0103] In the use of Bayesian network for global situation awareness, firstly, the first local state evaluation model output results of each site are input into the Bayesian network as evidence. Then, the joint probability distribution of each site in the target area is calculated through the Bayesian inference algorithm, and then the evaluation result of the global situation is obtained. For example, by calculating the probability of each site device health degree, load level and other factors being in a certain state at the same time, it is judged whether the operation situation of the whole region is normal. If the device health degree of a site suddenly decreases, the Bayesian network can quickly infer the influence degree of this change on the surrounding sites and the whole region according to the existing probability relationship.

[0104] The above two methods cannot well adapt to all complex and variable actual scenes with special needs. Therefore, the present application proposes a new method to construct a second global situation awareness model.

[0105] In the embodiment of the present application, the first local state evaluation model output results of each site are uploaded to the regional edge coordination center through a secure communication link, and a second global situation awareness model is constructed based on multi-site state data, which includes:

[0106] Each site is regarded as a graph node, and the geographical distance or power grid connection relationship between sites is regarded as an edge;

[0107] The input of the second global situation awareness model is the device health degree, service capability and load pressure of each site;

[0108] The output of the second global situation awareness model is the regional level situation judgment result, which includes overload risk area identification, service bottleneck prediction and resource mismatch detection.

[0109] The second global situation awareness model is used to identify regional level overload risk, service bottleneck and resource mismatch problems;

[0110] Specifically, the specific steps of constructing the second global situation awareness model can be as follows:

[0111] Step 2.1: a graph G=(V, E) needs to be constructed, wherein:

[0112] V is the node set in the graph, representing each charging and battery swapping station.

[0113] E is the edge set in the graph, representing the geographical distance or power grid connection relationship between sites.

[0114] For each node v i ∈V, its attributes include device health degree H i , service capability S i and load pressure L i .

[0115] Step 2.2: For each site i, the present invention has the following input variables:

[0116] Device health H i represents the current health status of the site device, with a value range of [0, 1], and 1 represents the best state.

[0117] Service capacity S i represents the capacity of the site to serve users, usually related to the number and type of charging stations, with a value range of [0, +∞).

[0118] Load pressure L i represents the load situation of the current site, which can be obtained by calculating the ratio of the current load to the maximum load, with a value range of [0, 1].

[0119] Step 2.3: Calculate the weight matrix W based on the geographical distance or power grid connection relationship between sites to determine the weight matrix W, where w ij represents the weight of the edge from node i to node j:

[0120]

[0121] where d ij represents the distance or power grid connection strength between sites i and j.

[0122] Step 2.4: Using the above input data and weight matrix W, the present invention can construct a second global situation awareness model to identify regional overload risk, service bottleneck and resource mismatch problems.

[0123] Step 2.4.1: Overload risk area identification By analyzing the load pressure L i of each site and the status of its adjacent sites, potential overload risk areas can be predicted. Let R i be the overload risk index of site i:

[0124]

[0125] where N(i) represents the set of all sites directly connected to site i.

[0126] Step 2.4.2: Service bottleneck prediction Service bottlenecks can be determined by comparing the service capacity S i of a site with its actual demand D i . Let B i be the bottleneck index of site i:

[0127]

[0128] It should be noted that when B i >1, it indicates that the site may have a service bottleneck.

[0129] Step 2.4.3: Resource Mismatch Detection The resource mismatch can be evaluated by comparing the device health H i , service capacity S i and load pressure L i of the sites. Let M i be the resource mismatch degree of site i:

[0130] M i = |H i - S i | + |S i - L i |

[0131] It should be noted that a higher M i value may indicate a situation of improper resource allocation.

[0132] It should be noted that uploading the first local state evaluation model output results of each site to the regional edge coordination center through a secure communication link and constructing a second global situation awareness model based on multi-site state data can provide a solid foundation for subsequent cross-site resource optimization configuration. Through the regional-level situation judgment result obtained by the second global situation awareness model, the regional edge coordination center can more accurately understand the overall operating conditions of each site in the target region. For the identified overload risk area, the regional edge coordination center can take timely measures, such as adjusting the load distribution of the surrounding sites, transferring part of the load to sites with smaller load pressure and stronger service capacity, to avoid the occurrence of overload and ensure the stable operation of devices and normal use of users.

[0133] S104, according to the second global situation awareness model and the grid time-of-use electricity price and renewable energy output prediction data, a third multi-objective resource optimization model is established, wherein:

[0134] It should be noted that after obtaining the regional-level situation judgment result output by the second global situation awareness model, combined with the grid time-of-use electricity price and renewable energy output prediction data, the resource can be more comprehensively and accurately optimized. The grid time-of-use electricity price reflects the difference in electricity cost in different time periods, with lower price in low valley period and higher price in peak period; the renewable energy output prediction data reflects the power generation situation of solar energy, wind energy and other renewable energy sources in the future period of time, which all have important influence on the optimization of resources.

[0135] The third multi-objective resource optimization model proposed by the present application aims to achieve multiple goals at the same time, including reducing operating costs, improving renewable energy utilization rate and ensuring the service quality of each site.

[0136] In some specific embodiments, the third multi-objective resource optimization model can be solved by a multi-objective optimization algorithm, such as a genetic algorithm, a particle swarm algorithm, etc.

[0137] Taking the genetic algorithm as an example, the decision variables, objective functions, and constraint conditions need to be determined first.

[0138] The decision variables can include the load distribution ratio of each site, the usage amount of renewable energy, etc. The objective functions can be set to multiple, such as minimizing the operation cost, maximizing the utilization rate of renewable energy, and optimizing the service quality. The operation cost can consider the power grid purchase cost and equipment maintenance cost, etc., and can be obtained by calculating the sum of the power purchase fee and the equipment maintenance fee in different time periods; the utilization rate of renewable energy can be measured by calculating the ratio of the actually used renewable energy power to the predicted renewable energy power generation; the service quality can be comprehensively evaluated by factors such as the equipment health degree, service ability, and load pressure of each site.

[0139] The constraint conditions need to consider the equipment capacity limit of each site, the load balance requirement, and the safe operation rules of the power grid, etc. For example, the load of each site cannot exceed its maximum load capacity, and the load distribution among the sites needs to be kept relatively balanced to avoid local overload; at the same time, the voltage, frequency, and other parameters of the power grid need to be within the safe range.

[0140] In the solving process of the genetic algorithm, a group of initial solutions are randomly generated first, which represent different resource allocation schemes. Then, each solution is evaluated according to the objective function, and its fitness value is calculated. The higher the fitness value, the better the scheme. Then, through genetic operations such as selection, crossover, and mutation, the current solutions are updated and optimized to generate new solutions. This process is iterated continuously until the stopping condition is met, such as reaching the maximum number of iterations or the fitness value no longer improving.

[0141] However, multi-objective optimization algorithms cannot well handle all complex and variable actual scenarios. In the embodiments of the present application, considering the complexity and diversity of actual situations, the present application proposes a third multi-objective resource optimization model including objective functions and constraint conditions.

[0142] In the embodiments of the present application, the third multi-objective resource optimization model is established, including:

[0143] The multi-objectives include reducing the total energy consumption of the region, balancing the load of each site, minimizing the user waiting time, and improving the green electricity consumption ratio.

[0144] The third multi-objective resource optimization model includes objective functions and constraint conditions, and the objective functions are any functions representing reducing the total energy consumption of the region, balancing the load of each site, minimizing the user waiting time, and improving the green electricity consumption ratio.

[0145] The constraint function includes at least a power upper limit constraint, an energy storage SOC constraint, and a user quality of service constraint.

[0146] Specifically, the detailed steps of establishing the third multi-objective resource optimization model can be as follows:

[0147] Step 3.1: Define several key input variables:

[0148] R i represents an overload risk index of the site i (provided by the second global situation awareness model).

[0149] B i represents a bottleneck index of the site i (provided by the second global situation awareness model).

[0150] M i represents a resource mismatch degree of the site i (provided by the second global situation awareness model).

[0151] represents a time-of-use electricity price of the power grid at time t.

[0152] represents a renewable energy output prediction value at time t.

[0153] L i (t) represents a load pressure of the site i at time t.

[0154] E i (t) represents an energy consumption of the site i at time t.

[0155] SOC i (t) represents a state of charge (SOC) of the site i at time t.

[0156] Q i represents a quality of service requirement of the site i.

[0157] Step 3.2: Establish a multi-objective function, including the following objectives:

[0158] Reduce the total energy consumption of the region, minimize the total energy consumption of all sites within a time period T.

[0159]

[0160] Balance the loads of each site, minimize the load difference between all sites.

[0161]

[0162] Minimize user waiting time, maximize service quality Q ito ensure the waiting time of each user is as short as possible.

[0163]

[0164] to increase the proportion of green power consumption and maximize the use of renewable energy.

[0165]

[0166] wherein represents the total power demand.

[0167] Step 3.3: To ensure the feasibility of the model, some constraints need to be set:

[0168] Power upper limit constraint: the power of each site cannot exceed its maximum power limit P i,max .

[0169]

[0170] Energy storage SOC constraint: the state of energy storage must be kept within a reasonable range to avoid excessive charging and discharging.

[0171]

[0172] User service quality constraint: to ensure that the service quality of each site is not lower than the minimum requirement Q i,min .

[0173]

[0174] It should be noted that the establishment of the third multi-objective resource optimization model based on the second global situation awareness model and the grid time-of-use price and renewable energy output prediction data can make the resource allocation more in line with the actual situation and improve the overall operation efficiency and economic benefits. On the one hand, combined with the grid time-of-use price, the power reserve of each site can be increased during the low price period, and the purchase of electricity from the grid can be reduced during the peak price period, and instead, the use of low-price period reserve power or renewable energy power generation is preferred, so as to reduce the operating cost. For example, during the night low-price period, each site can increase the charging capacity of the energy storage device for use during the day peak period to avoid high-price electricity purchase. On the other hand, referring to the renewable energy output prediction data, renewable energy can be better utilized. When the renewable energy generation is predicted to be sufficient, the use of renewable energy by each site is preferred, which improves the utilization rate of renewable energy, reduces the dependence on traditional power grids, and reduces carbon emissions.

[0175] S105, based on the third multi-objective resource optimization model, combined with the computing power distribution and communication delay characteristics of the edge computing node, a fourth distributed collaborative decision-making model is constructed, wherein:

[0176] It should be noted that the fourth distributed collaborative decision-making model is to further improve the flexibility and response speed of resource allocation to cope with various emergencies and dynamic changes. In the actual multi-station fusion scene, the computing power distribution and communication delay characteristics of edge computing nodes are important factors affecting decision-making efficiency and accuracy. Different edge computing nodes have different computing power and processing speed, and there is a certain delay in communication between nodes.

[0177] Therefore, the fourth distributed collaborative decision-making model proposed by the present application will fully consider these factors to achieve more efficient resource allocation.

[0178] Step 4.1: Define several key input variables to describe the characteristics of edge computing nodes, C i represents the computing power of node i (unit: number of operations per second). ij represents the communication delay between node i and node j (unit: seconds). i (t) represents the energy consumption of site i at time t (provided by the third multi-objective resource optimization model). i (t) represents the load pressure of site i at time t (provided by the third multi-objective resource optimization model). i represents the service quality requirement of site i (provided by the third multi-objective resource optimization model). i,max represents the maximum power limit of node i (provided by the third multi-objective resource optimization model). i (t) represents the State of Charge (SOC) of site i at time t (provided by the third multi-objective resource optimization model).

[0179] Step 4.2: Establish the objective function of the distributed collaborative decision-making model To fully utilize the characteristics of edge computing nodes and reduce communication delay, an objective function of a distributed collaborative decision-making model is established. The goal of this model is to maximize system efficiency and response speed while ensuring service quality and effective use of resources.

[0180] Maximize system efficiency: Improve system efficiency by minimizing total computing time and communication delay.

[0181]

[0182] Minimize communication delay: Optimize task allocation to reduce data transmission volume across nodes, thereby reducing communication delay.

[0183]

[0184] where T ij (t) represents the data transmission volume from node i to node j at time t.

[0185] Step 4.3: To ensure the feasibility of the distributed collaborative decision-making model, the following constraints are set:

[0186] Computational capability constraint: The task allocation of each node cannot exceed its computational capability.

[0187]

[0188] where T max is the maximum allowed task processing time.

[0189] Communication delay constraint: Ensure that the communication delay between all nodes does not exceed a predetermined threshold D threshold .

[0190]

[0191] Quality of service constraint: Ensure that the quality of service of each site does not fall below the minimum requirement Q i,min .

[0192]

[0193] Energy storage SOC constraint: The state of energy storage must be kept within a reasonable range to avoid excessive charging and discharging.

[0194]

[0195] Step 4.4: The model is solved using a distributed optimization algorithm (such as distributed gradient descent, alternating direction method of multipliers, etc.) to solve the above distributed collaborative decision-making model. These algorithms can find the optimal solution that minimizes the objective function while satisfying the constraints. For each node i, the distributed gradient descent method updates its task allocation strategy x i :

[0196]

[0197] where α is the step size parameter, is the gradient with respect to x i at the k-th step.

[0198] Step 4.5: By solving the optimal solution obtained, the optimal operation strategy of each site can be generated, and the corresponding lower-level control function is triggered to execute the optimization strategy.

[0199] It should be noted that based on the third multi-objective resource optimization model, combined with the computing power distribution and communication delay characteristics of the edge computing node, the fourth distributed collaborative decision model is constructed, which can further improve the accuracy and real-time performance of resource allocation. On the one hand, the model fully considers the computing power distribution of the edge computing node, and can reasonably allocate tasks according to the computing power of each node, avoiding the situation that some nodes are overloaded and some nodes are idle, thereby improving the overall computing efficiency. For example, for nodes with strong computing power, more complex and time-consuming tasks can be allocated, while for nodes with weak computing power, relatively simple tasks can be allocated. In this way, the advantages of each node can be fully utilized, and the computing resources of the entire system can be more effectively utilized. On the other hand, considering the communication delay characteristics can reduce the time loss in the data transmission process and speed up the response speed of the decision.

[0200] S106, the fourth distributed collaborative decision model is called through the preset upper coordination function, the optimal operation strategy of each site is generated, and the corresponding lower control function is triggered to execute the optimization strategy.

[0201] In an embodiment of the present application, the optimization strategy includes power scheduling, user guidance, energy storage charging and discharging adjustment, or dynamic pricing operation.

[0202] In an embodiment of the present application, the upper coordination function runs in the regional edge collaborative center, and the lower control function is distributed in the edge node of each site, forming a two-level edge computing architecture of "regional collaboration-local execution".

[0203] In an embodiment of the present application, the execution result data of each site to the optimal operation strategy is obtained;

[0204] The execution result data includes actual power adjustment amount, user shunting effect, energy storage response accuracy, communication overhead and edge node resource occupation rate.

[0205] The execution result data is fed back to the first local state evaluation model, the second global situation awareness model, the third multi-objective resource optimization model and the fourth distributed collaborative decision model, and the online iterative update and performance evaluation of the model parameters are performed.

[0206] In an embodiment of the present application, when the model prediction error is lower than the preset threshold value for a plurality of times in succession;

[0207] Or the collaborative optimization benefit growth tends to converge, it is determined that the iteration condition is met, and the model update is suspended;

[0208] Otherwise, the above process is continuously run to realize the adaptive optimization of the edge computing system.

[0209] In summary, the application discloses an edge computing optimization method based on multi-station fusion, obtains first data of electric vehicle charging and replacing facilities in a target area and pre-processes the first data, constructs a first local state evaluation model, constructs a second global situation awareness model, grasps a station operation situation to support regional planning. According to the second model, a time-of-use electricity price of a power grid and renewable energy output prediction data, a third multi-objective resource optimization model is constructed, multi-factor optimization resource allocation is considered, and economic and environmental benefits are realized. Based on the third model, a fourth distributed collaborative decision-making model is constructed, and the decision-making efficiency and accuracy are improved based on the characteristics of edge computing. An optimal strategy is generated by calling the fourth model through an upper coordination function, and a lower control function is executed, realizing seamless connection of decision-making execution. Each step is progressive to form a complete optimization system, improving efficiency and stability, reducing cost, and reasonably utilizing global resources.

[0210] In a preferred embodiment, the fourth distributed collaborative decision-making model is called by the preset upper coordination function in embodiment 2 to generate the optimal operation strategy of each station, and the operation steps of triggering the corresponding lower control function to execute the optimization strategy can be as follows:

[0211] Step 5.1: define some key variables:

[0212] S i represents the current service state of station i (such as device health, load level, etc.).

[0213] P i (t) represents the power demand of station i at time t.

[0214] T ij (t) represents the data transmission amount from node i to node j at time t.

[0215] O i represents the optimization strategy output of station i (for example: power adjustment amount, user guidance scheme, etc.).

[0216] Step 5.2: function of upper coordination function The upper coordination function is responsible for integrating information from multiple edge nodes and formulating a global optimization strategy according to the information.

[0217] The specific steps are as follows:

[0218] Step one: collect and summarize information The upper coordination function collects real-time state information S i from each edge node, including but not limited to load pressure L i (t), energy consumption E i (t) and energy storage state SOC i (t).

[0219] Step two: Determine the objective function Based on the collected information, set a comprehensive objective function aiming to minimize the overall system cost or maximize the benefit.

[0220] Assuming the goal of the invention is to minimize the total energy consumption and communication delay:

[0221]

[0222] Where λ is the weight coefficient, used to balance the impact of energy consumption and communication delay.

[0223] Step three: Use mathematical optimization methods (such as linear programming, nonlinear programming, or mixed integer programming) to find the strategy set O that minimizes the objective function F, combined with the solution provided by the fourth distributed collaborative decision-making model.

[0224] Step 5.3: The role of the lower-level control function The lower-level control function receives the optimization strategy O issued by the upper-level coordination function and executes specific control operations locally.

[0225] The specific steps are as follows:

[0226] Step one: Analyze the optimization strategy For each site i, analyze the received optimization strategy O i to get specific execution instructions, such as power adjustment amount ΔP i , user guidance plan G i , energy storage charge and discharge adjustment R i or dynamic pricing strategy Pr i .

[0227] Step two: Execute the optimization strategy According to the instructions parsed, execute the corresponding operations locally. For example:

[0228] Power adjustment, adjust the output power of the charging station according to ΔP i .

[0229] P i '(t) = P i (t) + ΔP i

[0230] User guidance, guide users to charge at sites with lighter loads according to G_i.

[0231] Energy storage charge and discharge adjustment, adjust the charge and discharge state of the energy storage system according to R_i.

[0232]

[0233] Where η is the efficiency coefficient of the energy storage system.

[0234] Dynamic pricing, according to Pr iImplement a dynamic pricing strategy to regulate user behavior.

[0235] Step three: feedback mechanism, the execution result (such as actual power adjustment amount, user shunt effect, energy storage response accuracy, etc.) is fed back to the upper coordination function, so as to further optimize and adjust the strategy.

[0236] It should be noted that through the above process, the upper coordination function can effectively integrate information from different edge nodes and develop a global optimization strategy; while the lower control function is responsible for the local execution of specific control operations to ensure that the strategy is implemented. The whole process not only improves the response speed and resource utilization of the system, but also enhances the flexibility and adaptability of the system. Through continuous feedback and adjustment, continuous performance optimization can be achieved.

[0237] Embodiment 3, refer to Figure 2 In this embodiment, a multi-station fusion based edge computing optimization system is also provided, comprising:

[0238] A data acquisition and processing module is configured to acquire first data of a plurality of electric vehicle charging and battery swapping facilities in a target area, and pre-process the first data.

[0239] The first data includes operation data, user demand data, and grid side interaction data.

[0240] A first model establishing module is configured to construct a first local state evaluation model on the edge computing nodes deployed at each site based on the pre-processed first data.

[0241] A second model establishing module is configured to upload the output results of the first local state evaluation model of each site to a regional edge coordination center through a secure communication link, and construct a second global situation awareness model based on multi-site state data.

[0242] A third model establishing module is configured to establish a third multi-objective resource optimization model based on the second global situation awareness model, the time-of-use electricity price, and the renewable energy output prediction data.

[0243] A fourth model establishing module is configured to construct a fourth distributed collaborative decision-making model based on the third multi-objective resource optimization model, in combination with the computing power distribution and communication delay characteristics of the edge computing nodes.

[0244] An execution module is configured to call the fourth distributed collaborative decision-making model through a preset upper coordination function, generate an optimal operation strategy for each site, and trigger the corresponding lower control function to execute the optimization strategy.

[0245] The above each unit module can be embedded in or independent of the processor in the electronic device in hardware form, or can be stored in the memory in the electronic device in software form, so that the processor calls and executes the corresponding operations of each module.

[0246] The embodiment also provides an electronic device, which can be a terminal, and an internal structure diagram of the terminal can be as shown in the figure. Figure 2 The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the electronic device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a multi-station fusion based edge computing optimization method. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the electronic device, or an external keyboard, touchpad or mouse, etc.

[0247] The embodiment also provides a computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the following steps:

[0248] Obtain first data of a plurality of electric vehicle charging and battery swapping facilities in a target area, and pre-process the first data;

[0249] The first data includes operation data, user demand data and grid side interaction data;

[0250] Based on the pre-processed first data, a first local state evaluation model is constructed on the edge computing nodes deployed at each station;

[0251] The output results of the first local state evaluation model of each station are uploaded to the regional edge collaborative center through a secure communication link, and a second global situation awareness model is constructed based on the multi-station state data;

[0252] According to the second global situation awareness model, the time-of-use electricity price and the renewable energy output prediction data, a third multi-objective resource optimization model is established;

[0253] Based on the third multi-objective resource optimization model, in combination with the computing power distribution and communication delay characteristics of the edge computing nodes, a fourth distributed collaborative decision-making model is constructed;

[0254] The fourth distributed collaborative decision-making model is called by the preset upper coordination function to generate the optimal operation strategy of each site and trigger the corresponding lower control function to execute the optimization strategy.

[0255] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all modifications and replacements should be covered in the scope of the claims of the present application.

[0256] Although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to the embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0257] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A multi-station fusion-based edge computing optimization method, characterized in that, The method comprises the following steps: acquiring first data of a plurality of electric vehicle charging facilities in a target area, and preprocessing the first data; the first data includes operation data, user demand data, and grid-side interaction data; based on the preprocessed first data, a first local state evaluation model is constructed on the edge computing nodes deployed at each site; the output results of the first local state evaluation model of each site are uploaded to the regional edge collaborative center through a secure communication link, and a second global situation awareness model is constructed based on the multi-site state data; a third multi-objective resource optimization model is established according to the second global situation awareness model, the time-of-use electricity price, and the renewable energy output prediction data; based on the third multi-objective resource optimization model, a fourth distributed collaborative decision-making model is constructed in combination with the computing power distribution and communication delay characteristics of the edge computing nodes; the fourth distributed collaborative decision-making model is called by a preset upper-layer coordination function to generate an optimal operation strategy for each site, and a corresponding lower-layer control function is triggered to execute the optimization strategy. 2.The multi-station fusion based edge computing optimization method of claim 1, wherein, Further comprising: acquiring execution result data of the optimal operation strategy at each site; the execution result data includes actual power adjustment amount, user shunting effect, energy storage response accuracy, communication overhead, and edge node resource occupation rate; the execution result data is fed back to the first local state evaluation model, the second global situation awareness model, the third multi-objective resource optimization model, and the fourth distributed collaborative decision-making model for online iterative updating and performance evaluation of model parameters. 3.The multi-station fusion based edge computing optimization method of claim 2, wherein, Further comprising: when the model prediction error is lower than a preset threshold for a plurality of times in succession; or when the collaborative optimization benefit growth tends to converge, it is determined that the iteration condition is met, and the model updating is suspended; otherwise, the above process is continuously run to realize adaptive optimization of the edge computing system.

4. The edge computing optimization method based on multi-station fusion according to claim 3, characterized in that, The method comprises the following steps: each site is regarded as a graph node, and the geographical distance or grid connection relationship between sites is regarded as an edge; the input of the second global situation awareness model is the device health degree, service capability, and load pressure of each site; the output of the second global situation awareness model is a regional-level situation judgment result, which includes overload risk area identification, service bottleneck prediction, and resource mismatch detection.

5. The edge computing optimization method based on multi-station fusion according to claim 4, characterized in that, Further comprising: the first local state evaluation model is used to evaluate the device health degree, load level, and service capability of the local site in real time; the second global situation awareness model is used to identify regional-level overload risk, service bottleneck, and resource mismatch problems; the optimization strategy includes power scheduling, user guidance, energy storage charging and discharging adjustment, or dynamic pricing operation.

6. The edge computing optimization method based on multi-station fusion according to claim 5, characterized in that, The method comprises the following steps: the multi-objective includes reducing the total energy consumption of the region, balancing the load of each site, minimizing the user waiting time, and improving the green electricity consumption ratio; the third multi-objective resource optimization model includes a target function and a constraint condition, and the target function is an arbitrary function representing the reduction of the total energy consumption of the region, the balancing of the load of each site, the minimization of the user waiting time, and the improvement of the green electricity consumption ratio; The constraint function includes at least a power upper limit constraint, an energy storage SOC constraint, and a user quality of service constraint.

7. The edge computing optimization method based on multi-station fusion according to claim 6, characterized in that, The upper-layer coordination function runs at a regional edge coordination center, and the lower-layer control function is distributed at each site edge node, forming a two-level edge computing architecture of "regional coordination-local execution".

8. A multi-site fusion based edge computing optimization system, applying the method according to any one of claims 1 to 7, characterized in that, The method comprises the following steps: a data acquisition and processing module configured to acquire first data of a plurality of electric vehicle charging and battery swapping facilities in a target region and to pre-process the first data; the first data comprises operation data, user demand data, and grid-side interaction data; a first model establishing module configured to construct a first local state evaluation model on an edge computing node deployed at each site based on the pre-processed first data; a second model establishing module configured to upload output results of the first local state evaluation model of each site to a regional edge coordination center through a secure communication link and to construct a second global situation awareness model based on multi-site state data; a third model establishing module configured to establish a third multi-objective resource optimization model according to the second global situation awareness model and grid time-of-use electricity price and renewable energy output prediction data; a fourth model establishing module configured to construct a fourth distributed collaborative decision-making model based on the third multi-objective resource optimization model, in combination with the computing power distribution and communication delay characteristics of the edge computing node; an execution module configured to call the fourth distributed collaborative decision-making model through a preset upper-layer coordination function, to generate an optimal operation strategy for each site, and to trigger a corresponding lower-layer control function to execute the optimization strategy. 9.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to implement the steps of the method for optimizing edge computing based on multi-site fusion according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method for optimizing edge computing based on multi-site fusion according to any one of claims 1-7.

Images