Electric vehicle battery swapping network state sensing method and system based on edge computing

Abstract

This invention provides a state perception method and system for electric vehicle battery swapping networks based on edge computing, belonging to the field of electric vehicle battery swapping network technology. First, a state interaction topology for the battery swapping network is constructed, and dynamic data on battery storage at swapping stations, operation of swapping equipment, and swapping requests are collected to generate a multi-dimensional state perception stream. This stream is then transmitted to an edge computing state processing link for distributed feature mapping, resulting in a set of feature vectors. Next, a state association evolution model containing multiple state evolution relationships is constructed to trace the origin and diffusion path of operational bottlenecks and determine control priorities. Finally, a dynamic control stream is generated based on the priorities and the set of feature vectors, converted into a standardized instruction sequence, and transmitted to the corresponding link. This achieves real-time perception and dynamic control of the battery swapping network state, improving the operating efficiency and service quality of the battery swapping network.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle battery swapping network technology, and more specifically, to a method and system for state perception of electric vehicle battery swapping networks based on edge computing. Background Technology

[0002] In the current booming development of the electric vehicle industry, the stability and efficiency of battery swapping networks are crucial as a key infrastructure ensuring the efficient operation of electric vehicles. Currently, traditional methods for sensing the status of battery swapping networks mainly rely on a centralized data processing model. Under this model, battery swapping stations need to transmit large amounts of battery storage data, battery swapping equipment operation data, and battery swapping request data to a central server for processing.

[0003] However, with the continuous expansion of the battery swapping network, the amount of data is growing explosively, and centralized processing faces many problems. On the one hand, data transmission is easily affected by factors such as network bandwidth and latency, leading to untimely and inaccurate data transmission, which in turn affects the real-time performance and accuracy of state awareness. On the other hand, the central server needs to process massive amounts of data, resulting in enormous computational pressure and making it prone to processing delays or even system crashes, unable to quickly respond to the dynamic changes of the battery swapping network. In addition, traditional methods are unable to deeply explore the complex relationships between the various state dimensions of the battery swapping network, and it is difficult to accurately locate the origin and spread path of the operational bottlenecks of the battery swapping network, thus failing to achieve effective dynamic control and limiting the operational efficiency and service quality of the battery swapping network. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for state perception of electric vehicle battery swapping networks based on edge computing, the method comprising: Construct a state interaction topology for the battery swapping network, and collect dynamic data on battery storage at the battery swapping station, dynamic data on the operation of the battery swapping equipment, and dynamic data on battery swapping requests based on the state interaction topology to generate a multi-dimensional state perception stream for the battery swapping network. The multi-dimensional state perception stream of the battery swapping network is transmitted to the state processing link corresponding to edge computing, and distributed feature mapping processing is performed to obtain the feature vector set of each state dimension of the battery swapping network. A state correlation evolution model for the battery swapping network is constructed based on a set of feature vectors. The state correlation evolution model for the battery swapping network includes the evolutionary correlation between the battery storage state and the operating state of the battery swapping equipment, the evolutionary correlation between the operating state of the battery swapping equipment and the battery swapping request state, and the evolutionary correlation between the battery storage state and the battery swapping request state. By using the state-related evolution model of the battery swapping network, the origin state dimension and diffusion path of the operational bottleneck of the battery swapping network are traced, and the priority of bottleneck control is determined. Based on the bottleneck control priority and feature vector set, a dynamic control flow for the battery swapping network is generated. This dynamic control flow is then converted into a standardized control instruction sequence and transmitted to the state execution link corresponding to the battery swapping station and the resource adjustment link corresponding to the edge computing, thereby realizing real-time perception and dynamic control of the battery swapping network status.

[0005] Furthermore, embodiments of the present invention also provide a state perception system for electric vehicle battery swapping networks based on edge computing, characterized in that it includes: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described edge computing-based electric vehicle battery swapping network state perception method by executing the machine-executable instructions.

[0006] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, the processor of the edge computing-based electric vehicle battery swapping network state perception system reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the edge computing-based electric vehicle battery swapping network state perception system to execute the above-described edge computing-based electric vehicle battery swapping network state perception method.

[0007] Based on the above, by constructing a state interaction topology for the battery swapping network, dynamic data from various aspects such as battery storage, battery swapping equipment operation, and battery swapping requests can be collected comprehensively and accurately. This generates a multi-dimensional state perception stream, which is then transmitted to the corresponding state processing link in edge computing for distributed feature mapping processing. This fully leverages the low latency and high bandwidth advantages of edge computing, avoiding the data transmission bottlenecks and computational pressure of centralized processing, greatly improving the speed and efficiency of data processing, and achieving real-time perception of the battery swapping network status. A state correlation evolution model for the battery swapping network, constructed based on a set of feature vectors, deeply explores the complex evolutionary relationships between battery storage status, battery swapping equipment operation status, and battery swapping request status. This model can accurately trace the origin state dimension and diffusion path of bottlenecks in the battery swapping network operation, and determine reasonable bottleneck control priorities. The dynamic control stream of the battery swapping network generated according to the bottleneck control priorities and the set of feature vectors is converted into a standardized control command sequence and transmitted to the corresponding link, achieving precise and dynamic control of the battery swapping network status and effectively improving the operational efficiency, stability, and service quality of the battery swapping network. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the execution flow of the electric vehicle battery swapping network state perception method based on edge computing provided in an embodiment of the present invention.

[0009] Figure 2 This is a schematic diagram of exemplary hardware and software components of an electric vehicle battery swapping network state perception system based on edge computing provided in an embodiment of the present invention. Detailed Implementation

[0010] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating an edge computing-based electric vehicle battery swapping network state perception method according to an embodiment of the present invention. The following is a detailed description of this edge computing-based electric vehicle battery swapping network state perception method.

[0011] Step S110: Construct the battery swapping network status interaction topology, collect dynamic data of battery storage at battery swapping stations, dynamic data of battery swapping equipment operation, and dynamic data of battery swapping requests based on the battery swapping network status interaction topology, and generate a multi-dimensional status perception stream for the battery swapping network.

[0012] In this embodiment, a regional battery swapping network containing multiple battery swapping stations is used as an example to illustrate the implementation process of this step. First, it is necessary to construct a state interaction topology that can fully cover the battery swapping stations in the region and meet the real-time requirements of edge computing. Through this topology, various dynamic data are collected and integrated into a multi-dimensional state awareness stream.

[0013] Step S111: Based on the geographical distribution of battery swapping stations and the coverage area of ​​edge computing, plan the node layout of the battery swapping network status interaction topology, and determine the number of battery swapping stations covered and the data transmission range corresponding to each topology node.

[0014] When planning the node layout, the geographical coordinates of all battery swapping stations in the area are first collected and organized. Simultaneously, the coverage parameters of the edge computing nodes are obtained, including signal strength attenuation models and data transmission rate versus distance models. Based on the distribution density of battery swapping stations, topology nodes are evenly distributed within the edge computing coverage area, ensuring that the straight-line distance from each topology node to the battery swapping stations it covers is within a range that guarantees data transmission quality. For areas with densely distributed battery swapping stations, the number of topology nodes is appropriately increased to reduce the number of battery swapping stations each topology node needs to cover, avoiding data transmission congestion. For areas with sparsely distributed battery swapping stations, the coverage range of a single topology node can be appropriately expanded to reduce the total number of topology nodes and lower construction costs. When determining the number of battery swapping stations covered by each topology node, the average daily battery swapping volume, data generation volume, and processing capacity of the edge computing nodes are comprehensively considered, ensuring that the computing and storage resources of each topology node can meet the data processing needs of the battery swapping stations it covers. The determination of the data transmission range requires on-site testing of data transmission packet loss rate, latency, and other indicators at different distances. Combined with the real-time requirements of the battery swapping station data, a maximum data transmission distance threshold is set, and the coverage range of the topology node does not exceed this threshold.

[0015] Step S112: Configure the link transmission parameters of the battery swapping network status interaction topology. The link transmission parameters include the data frame transmission period, data field length and link transmission protocol type. Set the data frame transmission period to not exceed the preset timeliness threshold, the data field length to match the transmission capacity of the data output port of the battery swapping station, and select a protocol that meets the real-time transmission requirements.

[0016] The data frame transmission period needs to be determined based on the frequency of change of various data types at the battery swapping station. For example, the temperature change trend in the dynamic data of battery swapping equipment operation is relatively slow, so the data frame transmission period can be appropriately longer; while the movement trajectory changes of the battery swapping robotic arm are rapid and require a higher sampling frequency, so the data frame transmission period needs to be set shorter. The preset timeliness threshold is determined based on the real-time requirements of the battery swapping network status control. For example, if a response is required within a few seconds after an abnormal state occurs in the battery swapping equipment, the data frame transmission period needs to be set to no more than a fraction of that response time to ensure that abnormal data can be transmitted to the processing node in a timely manner. The configuration of the data field length needs to match the hardware parameters of the data output port of the battery swapping station. By consulting the technical manual of the data output port of the battery swapping station, parameters such as the maximum supported data field length and data transmission rate can be obtained to ensure that the configured data field length does not exceed the processing capacity of the port and to avoid truncation or errors during data transmission. When choosing a link transmission protocol, considering the requirements of battery swapping networks for real-time and reliable data transmission, real-time transmission protocols are usually selected. These protocols can provide lower latency and higher transmission efficiency during data transmission, ensuring the timely delivery of battery swapping network status data.

[0017] Step S113: Deploy a status acquisition interface on each topology node. The status acquisition interface includes a battery storage dynamic acquisition interface, a battery swapping equipment operation dynamic data acquisition interface, and a battery swapping request dynamic data acquisition interface. Each interface is connected to the corresponding data output port in the battery swapping station.

[0018] The hardware selection for the status acquisition interface must match the interface type of the corresponding data output port within the battery swapping station. For example, if the data output port of the battery management system in the swapping station is an Ethernet interface, then the battery storage dynamic acquisition interface should also be an Ethernet interface; if the controller of the swapping equipment uses serial communication, then the dynamic data acquisition interface for the swapping equipment operation should be a corresponding serial interface. During interface deployment, strict electrical characteristic matching is required, including voltage level and signal transmission method, to ensure the accuracy and stability of data acquisition. After physical connection, each interface also requires software configuration to set parameters such as the interface's IP address, port number, and communication rate, enabling it to correctly interact with the corresponding data output port within the swapping station. Simultaneously, a data verification mechanism, such as cyclic redundancy check, should be configured for each interface to detect potential errors during data transmission and request retransmission upon error detection, ensuring the integrity of the acquired data.

[0019] Step S114: Collect dynamic data of battery storage at the battery swapping station, dynamic data of battery swapping equipment operation, and dynamic data of battery swapping requests. The dynamic data of battery storage at the battery swapping station includes the remaining power change trend of each battery, the health status fluctuation, and the storage location transfer record. The dynamic data of battery swapping equipment operation includes the movement trajectory change of the battery swapping robotic arm, the positioning deviation change of the battery swapping platform, and the temperature change trend of the battery swapping equipment. The dynamic data of battery swapping requests includes the distribution of battery swapping request initiation time, the distribution of requested battery types, and the change of battery swapping waiting time for requested vehicles.

[0020] For collecting dynamic data on battery storage, a data acquisition module is deployed in the battery management system to read parameters such as voltage, current, and temperature of each battery in real time. Based on these parameters, the remaining capacity is calculated, and the trend of remaining capacity changes is obtained through continuous sampling. Health status fluctuations are assessed by analyzing indicators such as the number of charge-discharge cycles, internal resistance changes, and capacity decay rate. The acquisition module periodically detects and records these indicators, forming health status fluctuation data. Storage location transfer records are generated by installing position sensors on the battery storage racks. When batteries move between racks or are removed or placed in, the sensors detect changes in battery position and record the position information and timestamp, forming a storage location transfer record.

[0021] The collection of dynamic operational data for battery swapping equipment requires the installation of corresponding sensors on devices such as the battery swapping robotic arm and the battery swapping platform. For example, angle sensors and displacement sensors are installed at various joints of the battery swapping robotic arm to collect the arm's motion angle and displacement data in real time, and the changes in the robotic arm's motion trajectory are obtained through data fusion; high-precision positioning sensors, such as GPS or laser positioning sensors, are installed on the battery swapping platform to collect the real-time position information of the battery swapping platform and compare it with the preset standard position to obtain the changes in positioning deviation; temperature sensors are installed on key components of the battery swapping equipment, such as motors and controllers, to monitor their temperature changes in real time and generate temperature change trend data.

[0022] The collection of dynamic data on battery swapping requests is achieved through the user interaction system and backend server of the battery swapping network. The distribution of request initiation times is obtained by recording the initiation time of each request and statistically analyzing it by hour or minute to determine the distribution of request numbers across different time periods. The distribution of requested battery types is analyzed by examining the battery type information selected by users when initiating battery swapping requests, and statistically analyzing the proportion of requests for each battery type. The change in battery swapping waiting time for requested vehicles is monitored by recording the time interval from when a user initiates a request to when the swap is completed, updating the waiting time data in real time, and analyzing its trend over time.

[0023] Step S115: The collected battery storage dynamic data, battery swapping equipment operation dynamic data, and battery swapping request dynamic data of the battery swapping station are aligned with the time axis according to a unified time base, and then integrated sequentially according to the preset perception stream structure to obtain the integrated perception stream. The perception stream structure includes a state dimension identifier, a data acquisition time marker, and a dynamic data content field.

[0024] Because different types of data come from different acquisition devices, their internal clocks may deviate, leading to inconsistent timestamps. Therefore, a unified time base needs to be established, such as using the system time of edge computing nodes as the standard time. During data acquisition, each acquisition device needs to convert its local time to a timestamp under the unified time base after acquiring data. For acquisition devices without network time synchronization capabilities, they can periodically synchronize with edge computing nodes to ensure the accuracy of timestamps. The specific implementation of time axis alignment is to map the timestamps of different types of data onto the unified time base as the horizontal axis, and combine data within the same time point or time interval. The preset perception stream structure is designed according to the needs of subsequent data processing and analysis. The state dimension identifier is used to distinguish data of different state dimensions such as battery storage, battery swapping equipment operation, and battery swapping requests; the data acquisition time stamp records the data acquisition time, accurate to the millisecond level, for time series analysis; the dynamic data content field stores the specific dynamic data values, and its data format is defined according to different types of data, such as numeric, string, and array types. During the integration process, various types of data are sequentially filled into the perception stream structure according to the order of the state dimension identifiers to form the integrated perception stream.

[0025] Step S116: Add topology node identifiers to the integrated sensing streams, and summarize the sensing streams after adding topology node identifiers to form a multi-dimensional state sensing stream of the battery swapping network covering all topology nodes. The topology node identifiers are used to distinguish the sensing streams generated by different topology nodes.

[0026] A topology node identifier is a unique identifier for each topology node, which can be a string or a numeric code. After each topology node integrates the collected data to form a sensing stream, its identifier is added to the header of the sensing stream. This allows the central processing node to quickly distinguish the data source when sensing streams from different topology nodes are transmitted. The aggregation process is achieved through communication links established between the central processing node and each topology node. The central processing node receives and stores the data according to a certain period or after receiving sensing stream data from the topology nodes. During aggregation, the sensing stream data undergoes an integrity check to ensure no data is lost or corrupted. If sensing stream data from a topology node is found to be missing, the central processing node sends a data retransmission request to that topology node to ensure the integrity and accuracy of the multi-dimensional status sensing streams of the battery swapping network.

[0027] Step S120: Transmit the multi-dimensional state perception stream of the battery swapping network to the state processing link corresponding to the edge computing, perform distributed feature mapping processing, and obtain the feature vector set of each state dimension of the battery swapping network.

[0028] After the multi-dimensional state perception stream of the battery swapping network is generated, it needs to be transmitted to edge computing nodes for processing. Edge computing nodes are characterized by low latency and high bandwidth, enabling them to quickly process and analyze data. The state processing link is a series of data processing modules and processes designed within the edge computing node to process the state data of the battery swapping network. Through distributed feature mapping processing, the raw state data is transformed into a set of vectors that can reflect the characteristics of each state dimension.

[0029] Step S121: The multi-dimensional state perception flow of the battery swapping network is divided into battery storage dynamic flow, battery swapping equipment operation dynamic flow and battery swapping request dynamic flow according to the state dimension. Each dynamic flow is matched with a state processing sub-link of edge computing.

[0030] After receiving the multi-dimensional state perception stream of the battery swapping network, the edge computing node first parses it to extract the state dimension identifiers from the perception stream structure. Based on the different state dimension identifiers, the perception stream is split into battery storage dynamic streams, battery swapping equipment operation dynamic streams, and battery swapping request dynamic streams. For example, data with the state dimension identifier "battery storage" is assigned to the battery storage dynamic stream, data with the identifier "battery swapping equipment operation" is assigned to the battery swapping equipment operation dynamic stream, and so on. The edge computing node internally sets up multiple independent state processing sub-links, each dedicated to processing one type of dynamic stream data. When establishing matching relationships, by configuring the routing rules of the edge computing node, the split battery storage dynamic stream is routed to the battery storage state processing sub-link, the battery swapping equipment operation dynamic stream is routed to the battery swapping equipment operation state processing sub-link, and the battery swapping request dynamic stream is routed to the battery swapping request state processing sub-link. This enables parallel data processing and improves data processing efficiency.

[0031] Step S122: In the battery storage dynamic processing sub-link, the remaining power change trend in the battery storage dynamic stream is extracted to obtain the power change trend feature; the health status fluctuation is extracted to obtain the health fluctuation feature; and the storage location transfer record is extracted to obtain the location transfer feature.

[0032] Step S1221: Extract the remaining power change trend data of each battery from the battery storage dynamic stream, and arrange them in chronological order to form a remaining power time series.

[0033] The battery storage dynamic stream contains remaining power trend data for multiple batteries, each with a unique battery identifier. During processing, the remaining power trend data for different batteries is separated based on their identifiers. For each battery, its remaining power data at different time points is arranged in chronological order according to its timestamps, forming a time series of its remaining power. The length of the time series is determined based on the data acquisition time span and the data frame transmission period, ensuring that the time series reflects the complete change in the battery's remaining power over a period of time.

[0034] Step S1222: Segment the remaining power time series, dividing the continuous time series into multiple continuous time periods, with the remaining power changes in each time period having similar trends.

[0035] A sliding window method is used to segment the remaining power time series. First, a fixed-length time window is set, its length determined by the characteristics of remaining power changes. For example, if the remaining power typically shows a clear trend within a few minutes, the window length is set to a few minutes. Then, the time window slides from left to right across the time series, each slide taking a fixed step. The step size can be equal to or less than the window length to achieve overlapping or non-overlapping segmentation of the data. For the remaining power data within each window, statistics such as the slope and variance of the changes are calculated to determine whether the trends of remaining power changes within that window are similar. If the trend statistics of adjacent windows are small in difference, the two windows are merged into one time period; if the difference is large, the current window is taken as the start of a new time period. In this way, the entire remaining power time series is divided into multiple time periods with similar trends.

[0036] Step S1223: Calculate the change range of the remaining power in each time period. The change range value of the time period is obtained by the difference between the remaining power at the start and end of the time period. Calculate the change frequency of the remaining power in each time period. The change frequency value is obtained by dividing the number of times the remaining power changes in the time period by the duration of the time period. Integrate the change range value and change frequency value of each time period in the order of the time periods to form the power change trend characteristics.

[0037] For each defined time period, the remaining battery power at the start and end times of that period is obtained. The remaining battery power at the end time is subtracted from the remaining battery power at the start time to obtain the change in remaining battery power within that time period. If the change in magnitude is positive, it indicates that the remaining battery power has increased during that time period; if it is negative, it indicates that the remaining battery power has decreased; if it is zero, it indicates that the remaining battery power has not changed. The calculation of the change frequency requires counting the number of times the remaining battery power changes within that time period. Here, "change" refers to a significant difference in remaining battery power compared to the previous sampling point (exceeding a preset threshold). The number of changes is divided by the duration of that time period (in seconds or minutes) to obtain the change frequency value of the remaining battery power. The change in magnitude and change frequency values ​​for each time period are arranged in chronological order to form a two-dimensional feature array, which represents the battery power change trend feature.

[0038] Step S1224: Separate the health status fluctuation data of each battery from the battery storage dynamic stream, and record the initial value, peak value and duration of each health status fluctuation.

[0039] Similar to the separation of remaining power trend data, health status fluctuation data for each battery is extracted from the battery storage dynamic stream based on battery identifiers. Health status fluctuation data typically exists as periodically monitored health status values, with one value recorded for each monitoring. By analyzing the sequence of these health status values, the start and end of each health status fluctuation can be identified. The fluctuation start value is the health status value at the beginning of the fluctuation; the fluctuation peak value is the maximum or minimum health status value reached during the fluctuation (depending on the direction of the fluctuation); and the fluctuation duration is the time elapsed from the start to the end of the fluctuation. When recording these parameters, a unique fluctuation identifier needs to be assigned to each fluctuation event for subsequent feature extraction and analysis.

[0040] Step S1225: Calculate the fluctuation range of each health status fluctuation, and obtain the fluctuation range value by the difference between the fluctuation peak value and the fluctuation start value; calculate the number of health status fluctuations per unit time, and obtain the fluctuation frequency value by dividing the total number of fluctuations by the total observation duration; integrate the fluctuation range value, fluctuation duration and fluctuation frequency value per unit time of each fluctuation to form the health fluctuation characteristics.

[0041] For each health status fluctuation event, the fluctuation range is obtained by subtracting the initial value from the peak value. If the fluctuation is positive (health status value increases), the fluctuation range is positive; if it is negative (health status value decreases), the fluctuation range is negative. Calculating the fluctuation frequency requires determining a total observation period, such as one day or one week, counting the total number of health status fluctuations within that period, and then dividing the total number by the total observation period to obtain the fluctuation frequency per unit time. The fluctuation range, duration, and frequency of each fluctuation, along with the fluctuation range for the entire observation period, are integrated to form a feature vector containing multiple elements; this feature vector is the health fluctuation feature. The fluctuation range and duration of each fluctuation event are included as part of the feature vector, while the fluctuation frequency is added as a separate feature element.

[0042] Step S1226: Separate the storage location transfer record data of each battery from the battery storage dynamic stream, and record the starting position identifier, target position identifier and transfer time for each transfer.

[0043] The storage location transfer record data for each battery is extracted from the battery storage dynamic stream based on the battery identifier. This data contains detailed information for each transfer event, including the time of the transfer, the starting position, and the destination position. When recording the starting and destination position identifiers, an encoding method corresponding to the battery storage rack position number is used to ensure the uniqueness and accuracy of the position identifiers. The transfer time is obtained by calculating the difference between the start and end times of the transfer event; the start time is the time the battery leaves the starting position, and the end time is the time the battery arrives at the destination position.

[0044] Step S1227: Count the number of transfers corresponding to each starting position identifier to obtain the starting position transfer frequency; count the number of receptions corresponding to each target position identifier to obtain the target position reception frequency; calculate the average time consumption for each transfer, and obtain the average time consumption value by dividing the total transfer time consumption by the total number of transfers; integrate the starting position transfer frequency, the target position reception frequency, and the average time consumption value to form the position transfer feature.

[0045] The starting location identifiers in all transfer records are categorized and statistically analyzed. The frequency of each starting location identifier is counted, which is the starting location transfer frequency, indicating how frequently the battery is transferred from that starting location. Similarly, the target location identifiers are categorized and statistically analyzed to obtain the number of receptions for each target location identifier, i.e., the target location reception frequency, indicating how frequently the battery is received at that target location. The total transfer time is the sum of the transfer times of all transfer events, and the total number of transfer events is the total number of transfer events. Dividing the total transfer time by the total number of transfers yields the average transfer time for each transfer. The starting location transfer frequency, target location reception frequency, and average time are integrated into a single feature vector. The starting location transfer frequency and target location reception frequency can be represented as two arrays (each array has a length equal to the total number of location identifiers, and the array elements are the frequency values ​​of the corresponding location identifiers). The average time is a separate value, which together constitute the location transfer feature.

[0046] Step S123: In the dynamic data processing sub-link of the battery swapping equipment, the trajectory feature of the robotic arm movement trajectory change in the dynamic flow of the battery swapping equipment is extracted to obtain the trajectory change feature; the deviation feature of the positioning deviation change of the battery swapping platform is extracted to obtain the positioning deviation feature; and the temperature feature of the temperature change trend of the battery swapping equipment is extracted to obtain the temperature change feature.

[0047] Step S1231: Extract the motion trajectory change data of the battery swapping robot arm from the dynamic flow of the battery swapping equipment operation. The motion trajectory change data includes the sequence of changes in the position coordinates of the robot arm in three-dimensional space over time.

[0048] The dynamic flow of battery swapping equipment operation includes data on the movement trajectory changes of the battery swapping robotic arm. This data is collected by sensors installed on the robotic arm and stored in the form of timestamps and corresponding three-dimensional spatial coordinates (X, Y, Z). When extracting data, data belonging to the movement trajectory of the battery swapping robotic arm is filtered out based on the equipment identifier and parameter identifier. The above data is arranged in chronological order according to the timestamps to form a time series of the robotic arm's movement trajectory, with each time point corresponding to a three-dimensional spatial coordinate, reflecting the position of the robotic arm at that moment.

[0049] Step S1232: Smooth the time series data of the robotic arm's motion trajectory to remove noise interference, then calculate trajectory parameters such as curvature, velocity, and acceleration, and arrange the above parameters in time order to form trajectory change characteristics.

[0050] Since sensor-collected data may be affected by noise, causing fluctuations or outliers in the trajectory data, smoothing is necessary. Smoothing methods such as moving averages and low-pass filtering can be used to remove high-frequency noise by weighted averaging or filtering adjacent data points, resulting in a smoother trajectory curve. After smoothing, for each point on the trajectory, its curvature is calculated, reflecting the degree of curvature of the trajectory; the robot's speed at that point is calculated by dividing the distance between two adjacent points by the time interval; and acceleration is calculated by dividing the change in speed by the time interval. These trajectory parameters (curvature, speed, and acceleration) are arranged in chronological order to form the trajectory variation characteristics.

[0051] Step S1233: Extract the positioning deviation change data of the battery swapping platform from the dynamic flow of battery swapping equipment operation. The deviation change data includes the time sequence of the positioning deviation value of the battery swapping platform in the X-axis and Y-axis directions.

[0052] The positioning deviation data of the battery swapping platform also comes from the dynamic flow of battery swapping equipment operation. The positioning deviation data of the battery swapping platform is filtered out by equipment identification and parameter identification. The positioning deviation is usually divided into the X-axis direction and the Y-axis direction (for planar positioning), and the deviation value in each direction changes over time. The positioning deviation values ​​in the X-axis and Y-axis directions are arranged in chronological order according to timestamps to form two independent time series, which reflect the changes in the positioning deviation of the battery swapping platform in the two directions respectively.

[0053] Step S1234: Calculate the root mean square error, maximum deviation value, and deviation change rate of the battery swapping platform positioning deviation, and integrate the above parameters into positioning deviation characteristics.

[0054] The root mean square error (RMSE) is an indicator that measures the overall magnitude of positioning deviation. It is obtained by calculating the square root of the average of the sum of the squares of all positioning deviation values. The maximum deviation value is the largest positioning deviation in the entire time series, reflecting the most severe positioning deviation that the battery swapping platform may experience. The deviation change rate refers to the amount of change in positioning deviation per unit time, obtained by dividing the difference in deviation values ​​between two adjacent time points by the time interval, reflecting how quickly the positioning deviation changes. Integrating parameters such as the RMSE, the maximum deviation values ​​in the X and Y axes, and the deviation change rates in the X and Y axes forms a positioning deviation feature that includes multiple elements.

[0055] Step S1235: Extract the temperature change trend data of the battery swapping equipment from the dynamic flow of the battery swapping equipment operation. The temperature change trend data of the battery swapping equipment includes the temperature value change sequence of the key components of the battery swapping equipment over time.

[0056] The temperature change trend data of the battery swapping equipment comes from temperature sensors installed on key components of the equipment. This data is extracted from the dynamic flow of the equipment's operation using equipment and parameter identifiers. Key components include motors, controllers, and battery interfaces, each corresponding to one or more temperature sensors. The collected temperature values ​​change over time. The temperature values ​​of each key component are arranged in chronological order according to timestamps to form their respective temperature time series.

[0057] Step S1236: Perform trend analysis on the temperature time series, calculate temperature parameters such as the average rate of change of temperature, the variance of temperature fluctuation, and the highest temperature value, and integrate the above parameters into temperature change characteristics.

[0058] Trend analysis can use methods such as linear regression to fit the changing trend of temperature time series and obtain the average rate of temperature change. The average rate of change reflects the overall direction and speed of temperature change over time. The variance of temperature fluctuation is obtained by averaging the squares of the deviations between the temperature values ​​and the average temperature values, reflecting the degree of temperature fluctuation. The maximum temperature value is the maximum value in the temperature time series and is an important indicator of whether the equipment may overheat. The average rate of change, temperature fluctuation variance, and maximum temperature value of each key component are integrated into a feature vector. The parameters of different key components occupy different positions in the feature vector, forming the temperature change characteristics.

[0059] Step S124: In the battery swapping request dynamic data processing sub-link, extract time period features from the request initiation time period distribution in the battery swapping request dynamic flow to obtain time period distribution features; extract type features from the request battery type distribution to obtain type distribution features; and extract duration features from the changes in battery swapping waiting time of requesting vehicles to obtain waiting time features.

[0060] Step S1241: Extract the initiation time data of battery swapping requests from the dynamic stream of battery swapping requests, group the initiation times according to preset time intervals (such as hours or minutes), count the number of requests in each time interval, and obtain the distribution data of request initiation time periods.

[0061] The dynamic stream of battery swapping requests contains the initiation timestamp of each request. After extracting these timestamps, the request initiation times within a day or a period of time are grouped according to a preset time interval. For example, if the preset time interval is 1 hour, the day is divided into 24 time periods, and the number of requests in each time period is counted. During the statistics, a histogram can be used, with the horizontal axis representing the time period and the vertical axis representing the number of requests, to obtain histogram data on the distribution of request initiation time periods.

[0062] Step S1242: Calculate the proportion of requests in each time period to the total number of requests to obtain the time period distribution ratio; analyze the peak and trough periods of the time period distribution, and integrate the time period distribution ratio, the start and end times of the peak period and the start and end times of the trough period to form the time period distribution characteristics.

[0063] The total number of requests is the sum of requests across all time periods. Dividing the number of requests in each time period by the total number of requests yields the time period distribution ratio, reflecting the relative concentration of requests within that time period. Peak periods are those with a high number of requests and a large time period distribution ratio, determined by comparing the number of requests or distribution ratios across different time periods; typically, the time periods with the highest number of requests are selected as peak periods. Low periods are those with a low number of requests and a small time period distribution ratio. The start and end times of peak and low periods are recorded, and the time period distribution ratios (represented as an array, with array elements corresponding to the proportion of each time period), peak period information, and low period information are integrated to form the time period distribution characteristics.

[0064] Step S1243: Extract the request battery type distribution data from the battery swap request dynamic stream. This request battery type distribution data includes the request count statistics for different battery types.

[0065] The battery type distribution data is obtained by statistically analyzing the battery type parameter in battery swapping requests. In the dynamic flow of battery swapping requests, each request record contains a battery type field, indicating the type of battery requested by the user. By classifying and statistically analyzing the battery type field of all request records, the number of requests corresponding to each battery type is obtained, forming the battery type distribution data.

[0066] Step S1244: Calculate the proportion of requests for each battery type to the total number of requests to obtain the type distribution ratio; analyze the request frequency of different battery types, and integrate the type distribution ratio and request frequency to form the type distribution characteristics.

[0067] The total number of requests is the sum of the number of requests for all battery types. Dividing the number of requests for each battery type by the total number of requests yields the type distribution ratio for that battery type. Request frequency refers to the number of requests for a particular battery type per unit time, obtained by dividing the total number of requests for that battery type by the total observation duration. The type distribution ratios of all battery types (represented as an array, with array elements corresponding to the ratio value for each battery type) and the request frequency (also represented as an array) are combined to form the type distribution characteristics.

[0068] Step S1245: Extract the battery swapping waiting time change data of the requested vehicle from the battery swapping request dynamic stream. The battery swapping waiting time change data of the requested vehicle contains the sequence of changes in the waiting time value of each battery swapping request over time.

[0069] The data on changes in battery swapping waiting times comes from the processing records of battery swapping requests. Each battery swapping request has a waiting time, which is the time interval from when the user initiates the request to when the battery swap is completed. Arranging these waiting time values ​​in order of the timestamp of the request initiation forms a time series of waiting times, reflecting how the waiting time changes over time.

[0070] Step S1246: Calculate the average, variance, maximum and minimum values ​​of the waiting time, analyze the trend of the waiting time, and integrate the above parameters and trend characteristics to form the waiting time characteristics.

[0071] The average waiting time reflects the overall waiting level; the variance reflects the dispersion of waiting time, with a larger variance indicating greater fluctuations; the maximum and minimum values ​​reflect waiting times under extreme conditions. Trend analysis can be performed by calculating the moving average of waiting time or using methods such as linear regression to obtain the overall direction of change in waiting time over time (increasing, decreasing, or remaining stable). Integrating the above duration parameters (average, variance, maximum, and minimum values) with trend characteristics (such as trend slope) forms the waiting time feature.

[0072] Step S125: Perform vector transformation on the power change trend features, health fluctuation features, and location transfer features obtained from the battery storage dynamic processing sub-link to obtain the battery storage state feature vector.

[0073] Battery level change trend features, health fluctuation features, and location transfer features are all feature data represented in different forms, and they need to be converted into a unified vector form. For the battery level change trend feature, it is already a two-dimensional feature array, which can be directly flattened into a one-dimensional vector. The health fluctuation feature includes the parameters and frequency of each fluctuation. Arranging the fluctuation range, duration, and frequency of each fluctuation in a specific order forms a one-dimensional vector. The starting location transfer frequency and target location reception frequency in the location transfer feature are in array form. These two arrays can be concatenated, and the average latency value added to form a one-dimensional vector. Then, these three one-dimensional vectors are concatenated in a specific order (e.g., first the battery level change trend feature vector, then the health fluctuation feature vector, and finally the location transfer feature vector) to form a longer one-dimensional vector, which serves as the battery storage state feature vector. During the vector conversion process, it is necessary to ensure that the dimensions and order of each feature are fixed for subsequent model training and feature comparison.

[0074] Step S126: Perform vector transformation on the trajectory change features, positioning deviation features, and temperature change features obtained from the dynamic data processing sub-link of the battery swapping equipment to obtain the operating status feature vector of the battery swapping equipment.

[0075] The trajectory change feature is a multi-dimensional time series containing parameters such as curvature, velocity, and acceleration. The parameter values ​​at each time point can be flattened into a one-dimensional vector in chronological order. The positioning deviation feature includes parameters such as root mean square error, maximum deviation value, and deviation change rate. These parameters are arranged in a specific order to form a one-dimensional vector. The temperature change feature includes parameters such as the average change rate, variance, and highest temperature value of each key component. These parameters are also arranged into a one-dimensional vector. Then, the trajectory change feature vector, positioning deviation feature vector, and temperature change feature vector are concatenated to form the operating status feature vector of the battery swapping equipment. The concatenation order can be determined based on the importance of the features or the source of the data and remains fixed.

[0076] Step S127: Perform vector transformation on the time period distribution characteristics, type distribution characteristics, and waiting time characteristics obtained from the battery swapping request dynamic data processing sub-link to obtain the battery swapping request status feature vector.

[0077] The time-period distribution feature includes information on time-period distribution ratios, peak periods, and off-peak periods. The time-period distribution ratio array is flattened, and the start and end times of peak and off-peak periods are converted into numerical parameters (e.g., converting time into minutes of a day), then arranged into a one-dimensional vector. The type distribution feature includes an array of type distribution ratios and an array of request frequencies; these two arrays are concatenated and flattened into a one-dimensional vector. The waiting time feature includes parameters such as average, variance, maximum, minimum, and trend characteristics; these parameters are arranged into a one-dimensional vector. Finally, these three one-dimensional vectors are concatenated together to form the battery swapping request status feature vector.

[0078] Step S128: Perform dimension unification processing on the battery storage state feature vector, the battery swapping equipment operation state feature vector, and the battery swapping request state feature vector, and classify and organize them according to the state dimension to form a feature vector set for each state dimension of the battery swapping network. The feature vector set includes the battery storage state feature vector group, the battery swapping equipment operation state feature vector group, and the battery swapping request state feature vector group.

[0079] Since feature vectors of different state dimensions may have different dimensions (lengths), dimension unification is necessary to facilitate subsequent model building and data processing. Dimension unification can be achieved through methods such as feature selection, feature reduction, or feature padding. For example, if a feature vector has a high dimension, dimensionality reduction methods such as principal component analysis can be used to reduce its dimension to the same level as other feature vectors; if a feature vector has a low dimension, its dimension can be increased to the target dimension by adding zero vectors or other padding values. After dimension unification, feature vectors belonging to the same state dimension are grouped together to form feature vector sets. For example, all battery storage state feature vectors constitute the battery storage state feature vector set, all battery swapping equipment operation state feature vectors constitute the battery swapping equipment operation state feature vector set, and all battery swapping request state feature vectors constitute the battery swapping request state feature vector set. These feature vector sets together form the feature vector set for each state dimension of the battery swapping network. The vectors in each feature vector set have the same dimension and correspond to different time points or different samples of the same state dimension.

[0080] Step S130: Construct a battery swapping network state correlation evolution model based on the feature vector set. The battery swapping network state correlation evolution model includes the evolution correlation between battery storage state and battery swapping equipment operation state, the evolution correlation between battery swapping equipment operation state and battery swapping request state, and the evolution correlation between battery storage state and battery swapping request state.

[0081] The feature vector set contains characteristic information of each state dimension of the battery swapping network. Based on this feature information, a state correlation evolution model is constructed to reveal the intrinsic connections and evolutionary patterns between different state dimensions. This battery swapping network state correlation evolution model can help analyze the changing trends of the battery swapping network state and predict potential problems.

[0082] Step S131: Extract the battery storage status feature vector group, the battery swapping equipment operation status feature vector group, and the battery swapping request status feature vector group from the feature vector set, and sort each vector group in chronological order.

[0083] Each feature vector group in the feature vector set contains multiple feature vectors, each corresponding to a specific point in time or time period. After extracting the vector groups, the battery storage state feature vector group, the battery swapping equipment operation state feature vector group, and the battery swapping request state feature vector group need to be sorted chronologically according to the timestamp information corresponding to each feature vector. In the sorted vector groups, the order of the feature vectors is consistent with the chronological order, which facilitates subsequent analysis of the evolution of states over time and the temporal correlation between different state dimensions.

[0084] Step S132: Select the battery storage state feature vector and the battery swapping equipment operation state feature vector at adjacent time nodes, and calculate the correlation between these two vectors in the time dimension to obtain the battery-equipment correlation degree; select the battery swapping equipment operation state feature vector and the battery swapping request state feature vector at adjacent time nodes, and calculate the correlation between these two vectors in the time dimension to obtain the equipment request correlation degree; select the battery storage state feature vector and the battery swapping request state feature vector at adjacent time nodes, and calculate the correlation between these two vectors in the time dimension to obtain the battery request correlation degree.

[0085] Step S1321: Determine the time interval between adjacent time nodes so that the interval between all adjacent time nodes remains the same.

[0086] The time interval needs to be determined based on the sampling frequency and rate of change of the battery swapping network status data. If the status data changes rapidly, the time interval should be set smaller to capture subtle changes; if the changes are slow, the time interval can be appropriately increased. By examining the timestamps of the feature vectors, the time difference between adjacent feature vectors is calculated, and then a suitable fixed time interval is selected so that the time difference of most adjacent feature vectors is close to this interval. For cases where the time difference does not match the fixed interval, interpolation or resampling methods can be used to generate a sequence of feature vectors with equal time intervals. It is crucial to ensure that the intervals between all adjacent time nodes are the same for subsequent calculations of status changes and correlations.

[0087] Step S1322: Standardize the battery storage state feature vector, the battery swapping equipment operation state feature vector, and the battery swapping request state feature vector respectively to eliminate the dimensional differences of each feature value and make all feature values ​​within a comparable order of magnitude.

[0088] The purpose of standardization is to eliminate the influence of different dimensions between feature values. Common standardization methods include Z-score standardization (converting feature values ​​to a normal distribution with a mean of 0 and a standard deviation of 1) or min-max standardization (scaling feature values ​​to the interval [0,1] or [-1,1]). For each feature vector in each feature vector group, standardization is performed on each feature dimension. For example, for the battery storage state feature vector group, the mean and standard deviation of all vectors in each feature dimension are calculated (Z-score standardization). Then, for each feature value of each vector, the mean of that dimension is subtracted, and then divided by the standard deviation to obtain the standardized feature value. Through standardization, the feature values ​​in feature vectors of different state dimensions have the same order of magnitude, which facilitates comparison and calculation of correlation.

[0089] Step S1323: Extract the standardized battery storage state feature vector at the Nth time node and the standardized battery storage state feature vector at the N+1th time node, calculate the degree of difference between the two vectors, and obtain the change in battery state; extract the standardized battery swapping equipment operation state feature vector at the Nth time node and the standardized battery swapping equipment operation state feature vector at the N+1th time node, calculate the degree of difference between the two vectors, and obtain the change in equipment state.

[0090] For the battery storage state feature vector, the vectors at the Nth and (N+1)th time nodes are selected from the standardized vector group. The degree of difference between these two vectors can be calculated using methods such as Euclidean distance and Manhattan distance. Euclidean distance is the square root of the sum of the squares of the differences between corresponding elements of two vectors, and it can effectively reflect the overall difference between the vectors. The calculated Euclidean distance is used as the battery state change measure; the larger the value, the greater the change in battery storage state between adjacent time nodes. Similarly, for the battery swapping equipment operating state feature vector, the same method is used to calculate the Euclidean distance between the vectors at the Nth and (N+1)th time nodes to obtain the equipment state change measure.

[0091] Step S1324: Calculate the degree of coordinated change between battery state change and device state change. Obtain the coordinated change value by the overlap ratio and the consistency of the direction of change between battery state change and device state change. Convert the coordinated change value into battery-device correlation degree. The range of correlation degree values ​​corresponds to the range of coordinated change value.

[0092] The overlap ratio refers to the proportion of feature dimensions in the changes in battery state and device state that share the same direction of change. For example, if both the battery state feature vector and the device state feature vector have multiple feature dimensions, for each feature dimension, its direction of change (increase or decrease) at adjacent time points is determined. If the two vectors change in the same direction on a certain feature dimension, then that dimension is considered to overlap. The overlap ratio is the number of overlapping feature dimensions divided by the total number of feature dimensions. Consistency of change direction refers to whether the overall trends of the two changes are consistent. For example, if the change in battery state increases and the change in device state also increases, the change direction is consistent; if one increases while the other decreases, the change direction is inconsistent. The consistency of change direction can be measured by calculating the correlation coefficient between the two changes; the closer the correlation coefficient is to 1, the better the consistency. The co-change value is a comprehensive indicator of the overlap ratio and the consistency of change direction. It can be obtained by weighted summation of the overlap ratio and the correlation coefficient, with the weights determined according to their importance. Then, the collaborative change values ​​are transformed into a preset range of values ​​(such as [0,1]) through linear transformation or normalization to obtain the battery device correlation degree. The higher the correlation degree, the stronger the correlation between the battery storage state and the battery swapping equipment operation state in the time dimension.

[0093] Step S1325: Calculate the degree of difference between the standardized battery swapping equipment operating state feature vector at the Nth time node and the standardized battery swapping equipment operating state feature vector at the N+1th time node to obtain the equipment state change amount; calculate the degree of difference between the standardized battery swapping request state feature vector at the Nth time node and the standardized battery swapping request state feature vector at the N+1th time node to obtain the request state change amount.

[0094] The calculation method for the change in equipment status is the same as in step S1323, that is, calculating the Euclidean distance between the standardized battery swapping equipment operating status feature vectors at the Nth and N+1th time nodes. The calculation of the change in request status is also the same, calculating the Euclidean distance between the standardized battery swapping request status feature vectors at the Nth and N+1th time nodes to reflect the magnitude of the change in battery swapping request status between adjacent time nodes.

[0095] Step S1326: Calculate the degree of coordinated change between the device state change and the request state change. Obtain the coordinated change value by the overlap ratio and the consistency of the change direction between the device state change and the request state change. Convert the coordinated change value into device request correlation degree. The correlation degree value range corresponds to the change range of the coordinated change value.

[0096] The calculation method for this step is similar to that of step S1324. First, the overlap ratio between the equipment status change and the request status change is calculated, that is, the proportion of feature dimensions with the same direction of change in the two changes. Then, the consistency of the change direction of the two is calculated and measured by the correlation coefficient. The overlap ratio and the correlation coefficient are weighted and summed to obtain the collaborative change value. The collaborative change value is then converted into the equipment request correlation degree. The correlation degree range corresponds to the range of the collaborative change value. The higher the correlation degree, the stronger the correlation between the changes in the operating status of the battery swapping equipment and the battery swapping request status.

[0097] Step S1327: Calculate the degree of difference between the standardized battery storage state feature vector at the Nth time node and the standardized battery storage state feature vector at the N+1th time node to obtain the battery state change amount; calculate the degree of difference between the standardized battery swap request state feature vector at the Nth time node and the standardized battery swap request state feature vector at the N+1th time node to obtain the request state change amount.

[0098] The calculation methods for battery state change and request state change are the same as before. The Euclidean distance between the battery storage state feature vector and the battery swap request state feature vector at adjacent time nodes is calculated respectively.

[0099] Step S1328: Calculate the degree of coordinated change between the battery state change and the requested state change. Obtain the coordinated change value by the overlap ratio and the consistency of the change direction between the battery state change and the requested state change. Convert the coordinated change value into a battery request correlation degree. The correlation degree value range corresponds to the change range of the coordinated change value.

[0100] Similarly, first calculate the overlap ratio and direction consistency (correlation coefficient) between the changes in battery state and the changes in the request state. Then, weight and sum the two to obtain the collaborative change value, and finally convert it into the battery request correlation degree to reflect the degree of correlation between changes in battery storage state and battery swap request state.

[0101] Step S1329: Record the battery device correlation degree, device request correlation degree and battery request correlation degree calculated for each adjacent time node, so that each time node corresponds to a set of correlation degree data.

[0102] For each pair of adjacent time nodes (N and N+1), a set of battery device correlation, device request correlation, and battery request correlation are calculated. These correlation data are then associated with the corresponding time nodes (which can be timestamps of N or N+1) and stored in a database or file to form correlation time series data. Each time node (except the last time node) corresponds to a set of correlation data.

[0103] Step S133: Record the battery device correlation degree, device request correlation degree, and battery request correlation degree at different time points to form a correlation degree time series.

[0104] The correlation data recorded in step S1329 is arranged in chronological order to form a correlation time series. The correlation time series includes three independent series: battery device correlation time series, device request correlation time series, and battery request correlation time series. Each series contains elements representing the correlation value at different time points, reflecting how the correlation changes over time.

[0105] Step S134: Analyze the changing trend of the correlation time series, determine the evolution law of correlation over time, and divide the correlation into periods of increasing correlation, decreasing correlation, and stable correlation.

[0106] Trend analysis of correlation time series can be performed using methods such as sliding window and trend line fitting. The sliding window method divides the time series into multiple windows, calculates the average or median correlation degree within each window, and judges trend changes by comparing the statistical values ​​of adjacent windows. Trend line fitting uses methods such as linear regression and multinomial regression to fit a trend curve of the correlation time series, and judges whether the correlation degree is increasing, decreasing, or stable based on the slope of the trend curve. Based on the results of trend analysis, the correlation time series is divided into different periods: when the trend slope is positive, it is an increasing period; when the trend slope is negative, it is a decreasing period; when the trend slope is close to zero and the correlation degree value fluctuates within a small range, it is a stable period. When dividing the period, a threshold for trend judgment needs to be determined; only when the absolute value of the trend slope is greater than this threshold is the trend considered to be significantly increasing or decreasing.

[0107] Step S135: Based on the evolutionary law of correlation, construct the evolutionary correlation between battery storage state and battery swapping equipment operation state, and describe the mutual influence between the two states at different time stages; and construct the evolutionary correlation between battery swapping equipment operation state and battery swapping request state, and describe the mutual influence between the two states at different time stages; and construct the evolutionary correlation between battery storage state and battery swapping request state, and describe the mutual influence between the two states at different time stages.

[0108] The construction of evolutionary correlations requires combining the evolutionary patterns of correlation degree with the actual operating mechanism of the battery swapping network. During periods of increased correlation degree, it indicates that the mutual influence between the two states is enhanced, and changes in one state are more likely to cause changes in the other. During periods of decreased correlation degree, the mutual influence weakens. During stable periods, the mutual influence remains at a relatively stable level. For the evolutionary correlation between battery storage state and battery swapping equipment operating state, it is necessary to analyze how changes in battery storage state (such as changes in charge level and health status) affect the operating state of the battery swapping equipment (such as robotic arm movements and platform positioning) at different times, and how changes in the operating state of the battery swapping equipment, in turn, affect the battery storage state (such as equipment failure preventing normal battery storage or retrieval). Similarly, for the other two evolutionary correlations, it is also necessary to analyze the direction, intensity, and mechanism of the mutual influence between the two states at different times, and describe them in the form of text or mathematical models.

[0109] Step S136: Integrate the evolutionary relationships between battery storage status and battery swapping equipment operation status, battery swapping equipment operation status and battery swapping request status, and battery storage status and battery swapping request status according to the time dimension, and set the weight of each relationship at different time stages. The weight is determined based on the degree of correlation.

[0110] The three evolutionary relationships are integrated according to time stages (i.e., the previously defined rising period, falling period, and stable period). For each time stage, weights are assigned to the three relationships based on the degree of correlation within that stage. A higher correlation indicates a more important relationship within that stage, resulting in a higher weight; conversely, a lower correlation results in a lower weight. The weights can be set using a normalization method: the value of each of the three correlations within each stage is divided by the sum of the three correlations within that stage, resulting in a weight for each relationship such that the sum of the weights is 1. In this way, the model can adjust the emphasis on different relationships based on the magnitude of the weights at different time stages.

[0111] Step S137: Integrate the three evolutionary relationships after weight setting to form a battery swapping network state correlation evolution model. The battery swapping network state correlation evolution model includes time stage division, correlation relationship of each stage and corresponding weight.

[0112] This model integrates time-stage segmentation information (start and end times of each period), descriptions of three evolutionary relationships within each time stage, and corresponding weight values ​​into a single model structure. The battery swapping network state correlation evolution model can be designed using an object-oriented approach, including time-stage classes, relationship classes, and weight classes. The complete model is constructed through the combination of these classes. The model's input is the battery swapping network's state data (such as feature vectors or raw sensing flow data), and the output is the state assessment result or correlation prediction result based on the relationships and weights. The model's parameters include time-stage segmentation parameters, relationship description parameters, and weight parameters, which can be adjusted and optimized based on actual data.

[0113] Step S140: Trace the origin state dimension and diffusion path of the battery swapping network operation bottleneck through the battery swapping network state correlation evolution model, and determine the bottleneck control priority.

[0114] After the state-related evolution model of the battery swapping network is constructed, it is used to analyze the operational status of the network, identify potential operational bottlenecks, and trace their origins and spread paths. Determining the bottleneck control priority ensures that control resources are used to solve key problems first, thereby improving control efficiency.

[0115] Step S141: Input the multi-dimensional state perception flow of the battery swapping network into the state association evolution model of the battery swapping network. The state association evolution model of the battery swapping network outputs the operation status evaluation results of each state dimension at different time stages based on the internal evolutionary association relationship and weight.

[0116] The multi-dimensional state perception stream of the battery swapping network contains dynamic data of each state dimension of the network, which is input into the state correlation evolution model of the battery swapping network. The model first parses and preprocesses the input data, extracting feature information for each state dimension. Then, based on the model's internal time stage division, it determines the time stage to which the current input data belongs. For that time stage, it calls upon the corresponding evolutionary correlation relationships and weight parameters to calculate and evaluate the mutual influence between each state dimension. The operational status evaluation result can be a comprehensive score or a combination of multiple evaluation indicators, reflecting the operational status of each state dimension within that time stage (e.g., whether it is normal, its efficiency, its stability, etc.). During the evaluation process, the model considers the weights of different correlation relationships, incorporating the mutual influence between state dimensions with high correlation into the evaluation calculation.

[0117] Step S142: Analyze the operational status assessment results of each state dimension, identify the state dimensions whose operational status assessment results are in the abnormal range, and determine the initial abnormal dimension; based on the evolutionary correlation relationship in the battery swapping network state correlation evolution model, find other state dimensions that are correlated with the initial abnormal dimension, and determine the correlation dimension.

[0118] Step S1421: Determine the normal operating range of each state dimension based on the parameter value range in the historical normal operation data of the battery swapping network. Compare the operating status evaluation results of each state dimension at different time stages with the corresponding normal range. If the evaluation result exceeds the normal range, mark the state dimension as abnormal at the corresponding time stage.

[0119] Historical normal operation data of the battery swapping network refers to the status data collected and the corresponding operational status assessment results collected when the network is operating normally. Statistical analysis of this historical data determines the normal range (normal interval) for the operational status assessment results of each status dimension. For example, the mean and standard deviation of the historical assessment results can be calculated, and the normal interval can be set as the range of the mean plus or minus a certain number of standard deviations; or the percentile method can be used to set the normal interval as the range where 95% or 99% of the historical data falls. For each status dimension, its operational status assessment result is compared with the corresponding normal interval at each time period. If the assessment result is less than the lower limit or greater than the upper limit of the normal interval, the status dimension is considered abnormal at that time period and is marked as abnormal.

[0120] Step S1422: Calculate the percentage of abnormal time for each state dimension. The percentage of abnormal time is obtained by dividing the total duration of abnormal time by the total observation duration. The state dimension with the largest percentage of abnormal time is selected as the initial abnormal dimension. If there are multiple state dimensions with the same and largest percentage of abnormal time, the severity of abnormality for each dimension is further compared, and the state dimension with the largest severity of abnormality is selected as the initial abnormal dimension.

[0121] For each state dimension, the total time it was marked as an anomaly within the total observation period is calculated. The percentage of anomaly time is the total anomaly time divided by the total observation period. For example, if a state dimension was marked as an anomaly for 20 hours out of a 100-hour observation period, its percentage of anomaly time is 0.2. Compare the percentages of anomaly time for all state dimensions and select the state dimension with the highest percentage as the initial anomaly dimension. If multiple state dimensions have the same and the highest percentage of anomaly time, the severity of the anomaly for these dimensions needs to be further calculated. The severity of the anomaly can take into account factors such as the degree to which the assessment result exceeds the normal range and the duration of the anomaly. For example, the severity of the anomaly can be obtained by multiplying the degree of deviation (the difference between the assessment result and the boundary of the normal range) by the duration of the anomaly and then summing the results. The state dimension with the highest severity of anomaly is selected as the initial anomaly dimension.

[0122] Step S1423: Retrieve all evolutionary relationships in the state correlation evolution model of the battery swapping network, filter out the relationships that contain the initial anomaly dimension, extract other state dimensions associated with the initial anomaly dimension from the relationships that contain the initial anomaly dimension, and use the extracted dimensions as correlation dimensions.

[0123] The battery swapping network state correlation evolution model stores three types of evolutionary correlations: the correlation between battery storage state and battery swapping equipment operating state, the correlation between battery swapping equipment operating state and battery swapping request state, and the correlation between battery storage state and battery swapping request state. The descriptive information of these correlations is retrieved, and correlations containing the initial anomaly dimension are filtered out. For example, if the initial anomaly dimension is battery storage state, then correlations containing battery storage state are filtered out (i.e., the correlation between battery storage state and battery swapping equipment operating state, and the correlation between battery storage state and battery swapping request state). Then, from these filtered correlations, other state dimensions besides the initial anomaly dimension are extracted; these dimensions are the correlation dimensions associated with the initial anomaly dimension.

[0124] Step S1424: Classify the correlation dimensions. Based on the type of correlation, the correlation dimensions are divided into direct correlation dimensions and indirect correlation dimensions. Direct correlation dimensions are dimensions that are directly correlated with the initial abnormal dimension, while indirect correlation dimensions are dimensions that are correlated with the initial abnormal dimension through other dimensions.

[0125] Directly related dimensions are state dimensions that are directly connected to the initial anomaly dimension in their evolutionary relationship. For example, if the initial anomaly dimension is the battery storage state, then the battery swapping equipment operating state and the battery swapping request state are directly related dimensions because they are directly related to the battery storage state in their evolutionary relationships. Indirectly related dimensions, on the other hand, are dimensions that are not directly related to the initial anomaly dimension, but are related to it through other directly or indirectly related dimensions. For example, if the battery swapping equipment operating state is a directly related dimension, and the battery swapping equipment operating state is also related to the battery swapping request state (but the battery swapping request state is already considered a directly related dimension), then in this case, there might not be any indirectly related dimensions; or, if a more complex relationship network exists, there might be more levels of indirectly related dimensions.

[0126] Step S1425: Analyze the correlation strength between the directly related dimensions and the initial abnormal dimensions. Determine the correlation strength value by the weight in the correlation relationship. Sort the directly related dimensions in descending order of correlation strength value to form a sorting table of directly related dimensions.

[0127] In the battery swapping network state correlation evolution model, weights are assigned to each correlation at different time stages, and the magnitude of the weight reflects the strength of the correlation. For the correlation between a directly correlated dimension and an initial anomaly dimension, the weight value at the current time stage is the correlation strength value. For example, if the correlation between battery storage state and battery swapping equipment operating state has a weight of 0.6 at the current stage, and the correlation between battery storage state and battery swapping request state has a weight of 0.4, then the correlation strength value between battery swapping equipment operating state and the initial anomaly dimension (battery storage state) is 0.6, and the correlation strength value between battery swapping request state is 0.4. The directly correlated dimensions are arranged in descending order of correlation strength value to form a directly correlated dimension ranking table.

[0128] Step S1426: Analyze the association path length between the indirect association dimension and the initial anomaly dimension. The association path length is the number of intermediate dimensions that the indirect association dimension is associated with the initial anomaly dimension through intermediate dimensions. The smaller the path length value, the tighter the association. Sort the indirect association dimensions in ascending order of association path length value to form an indirect association dimension sorting table.

[0129] For indirectly related dimensions, it's necessary to find the association path between them and the initial anomaly dimension. An association path refers to the chain that starts from the initial anomaly dimension, passes through a series of intermediate related dimensions, and ultimately reaches the indirect related dimension. The length of the association path is the number of intermediate dimensions in the path. For example, if the initial anomaly dimension A is related to the indirect related dimension C through intermediate dimension B, then the association path length is 1 (the number of intermediate dimensions B is 1); if the initial anomaly dimension A is related to the indirect related dimension D through intermediate dimensions B and C, then the association path length is 2. If multiple association paths exist, the path with the shortest length is selected as the association path between the indirect related dimension and the initial anomaly dimension. Then, the indirect related dimensions are sorted in ascending order of their association path lengths to form an indirect related dimension sorting table.

[0130] Step S1427: Integrate the direct correlation dimension sorting table and the indirect correlation dimension sorting table to form a correlation dimension list. The correlation dimension list records the correlation method and correlation tightness of each correlation dimension with the initial abnormal dimension.

[0131] The related dimensions from the directly related dimensions sorting table and the indirectly related dimensions sorting table are merged into a single list, which is then deduplicated to form a list of related dimensions. This list records the relationship between each related dimension and the initial abnormal dimension (direct or indirect), as well as data related to the strength of the relationship (e.g., the relationship strength value for directly related dimensions, and the relationship path length for indirectly related dimensions). This clearly displays all dimensions associated with the initial abnormal dimension and their relationship characteristics.

[0132] Step S143: Analyze the evolutionary relationship between the initial abnormal dimension and the associated dimension, and determine the order of influence of the initial abnormal dimension on the associated dimension. The associated dimension first affected by the initial abnormal dimension is the first-level associated dimension, the associated dimension affected by the first-level associated dimension is the second-level associated dimension, and so on.

[0133] Based on the evolutionary relationships in the battery swapping network state correlation evolution model, this paper analyzes how the initial anomaly dimension affects the correlated dimensions. Directly correlated dimensions are usually the first objects affected by the initial anomaly dimension; therefore, they are identified as first-level correlated dimensions. Then, it is examined whether these first-level correlated dimensions have evolutionary relationships with other correlated dimensions. If so, these correlated dimensions affected by the first-level correlated dimensions are identified as second-level correlated dimensions. This process continues to determine third-level, fourth-level, and so on, correlated dimensions. The order of influence reflects the hierarchical relationship of anomaly propagation from the initial anomaly dimension to the correlated dimensions.

[0134] Step S144: Track the anomaly start time of the initial anomaly dimension, compare it with the time when the anomaly appears in the associated dimension, and determine the time sequence of the anomaly spreading from the initial anomaly dimension to the associated dimension.

[0135] By examining the anomaly marker records of the initial anomaly dimension, its anomaly initiation time is determined, i.e., the timestamp of the first time it was marked as an anomaly. Then, the anomaly marker records of each associated dimension are examined to find their respective anomaly initiation times. Comparing the anomaly initiation times of the initial anomaly dimension with those of each associated dimension, and arranging them in chronological order, yields the time sequence of anomaly propagation. Typically, the anomaly initiation time of the first-level associated dimension is later than that of the initial anomaly dimension, the anomaly initiation time of the second-level associated dimension is later than that of the first-level associated dimension, and so on. However, there may be cases where some associated dimensions exhibit anomalies earlier due to other factors, requiring a comprehensive judgment based on the order of influence.

[0136] Step S145: Based on the order of impact and time sequence, construct the diffusion path of the bottleneck in the battery swapping network operation. The path includes the initial anomaly dimension, the correlation dimensions at each level, and the corresponding diffusion time.

[0137] The diffusion path is a directed graph structure, starting with the initial anomaly dimension, with each level of related dimensions arranged sequentially according to their influence. Each node (dimension) is marked with its anomaly initiation time (diffusion time). For example, the path can be represented as: initial anomaly dimension (diffusion time T0) → first-level related dimension A (diffusion time T1) → second-level related dimension B (diffusion time T2), and simultaneously initial anomaly dimension → first-level related dimension C (diffusion time T3) → second-level related dimension D (diffusion time T4), and so on. The diffusion path clearly demonstrates how an anomaly, starting from the originating state dimension, spreads to each level of related dimensions at different times.

[0138] Step S146: Determine the initial anomaly dimension as the origin state dimension of the battery swapping network operation bottleneck.

[0139] Since the initial anomaly dimension is the state dimension where anomalies first appear, have the largest proportion of anomaly time, or have the highest anomaly severity, and is the starting point for anomaly propagation, it is identified as the originating state dimension of the bottleneck in the battery swapping network operation.

[0140] Step S147: Analyze the degree of abnormality in each dimension of the diffusion path, standardize the duration of the abnormality and the degree of deviation from the normal range to obtain dimensionless duration score and deviation score, and then combine the duration score and deviation score to determine the numerical value of the severity of the abnormality in each dimension.

[0141] The duration of an anomaly refers to the total time the abnormal state persists within the observation period for each dimension; the degree of deviation from the normal range refers to the extent to which the operational status assessment result deviates from the normal range during each anomaly (such as the absolute value or square of the excess value). The duration of anomalies is standardized, for example, by dividing the duration of anomalies for each dimension by the maximum duration of anomalies to obtain a duration score (range [0,1]); the degree of deviation from anomalies is standardized, for example, by dividing the average deviation of each dimension by the maximum average deviation to obtain a deviation score (range [0,1]). Then, a weighted summation (such as the duration score and deviation score each having a certain weight) is used to obtain the severity value of the anomaly. The larger the severity value, the more severe the anomaly.

[0142] Step S148: Based on the numerical severity of the anomaly and the hierarchy in the diffusion path, determine the bottleneck control priority. The control priority of the origin state dimension is higher than that of the first-level correlation dimension, the control priority of the first-level correlation dimension is higher than that of the second-level correlation dimension, and the control priority of the dimension with the larger anomaly severity value within the same level is higher.

[0143] The determination of bottleneck control priorities follows a hierarchical priority principle: the origin state dimension, as the source of the anomaly, has the highest control priority; followed by the first-level related dimensions, then the second-level related dimensions, and so on. Within the same level, the severity values ​​of the anomalies in each dimension are compared, with dimensions having higher control priority for those with larger values. This ensures that problems with a wide impact and severe anomalies are addressed first.

[0144] Step S149: Record the origin state dimension, diffusion path and bottleneck control priority to form a bottleneck tracing report. The bottleneck tracing report includes details of anomalies in each dimension, diffusion order and control priority ranking.

[0145] The bottleneck tracing report summarizes the bottleneck tracing process and results, including details such as the identification of the origin status dimension, the anomaly start time, and the severity of the anomaly; a graphical or textual description of the propagation path, including the various levels of correlation dimensions and their propagation time; and a priority ranking table for bottleneck control at each dimension. This bottleneck tracing report can provide clear guidance for the operation and maintenance personnel of the battery swapping network, helping them to carry out targeted control and maintenance.

[0146] Step S150: Generate a dynamic control flow for the battery swapping network based on the bottleneck control priority and feature vector set. Convert the dynamic control flow for the battery swapping network into a standardized control instruction sequence and transmit it to the state execution link corresponding to the battery swapping station and the resource adjustment link corresponding to the edge computing, so as to realize real-time perception and dynamic control of the battery swapping network status.

[0147] Based on bottleneck control priorities, the state dimensions requiring priority control are determined. Combining the feature information of these dimensions from the feature vector set, specific control strategies are formulated, generating a dynamic control flow. This control flow is then converted into a standardized sequence of control instructions that can be recognized and executed by the battery swapping station and edge computing nodes. This sequence is transmitted to the corresponding execution links via communication links to execute the control operations, thereby achieving real-time perception and dynamic control of the battery swapping network status and resolving operational bottlenecks.

[0148] Step S151: Based on the bottleneck control priority, select the state dimensions with control priorities from high to low in sequence, and determine the current feature vector of each dimension to be controlled.

[0149] From the bottleneck control priority ranking table, state dimensions are selected as controllable dimensions in descending order of priority. For each controllable dimension, the latest feature vector (corresponding to the latest time point) is extracted from the feature vector set, which reflects the current state characteristics of that dimension.

[0150] Step S152: Based on the current feature vector of the dimension to be adjusted and the normal feature vector range of the dimension to be adjusted, determine the feature parameters that need to be adjusted and the target parameter values. The target parameter values ​​are the parameter values ​​that make the feature parameters return to the normal feature vector range.

[0151] The normal eigenvector range for the dimension to be adjusted is determined based on historical normal data, similar to the normal range of the operational status assessment results, but more specific to each dimension of the eigenvector. Each feature parameter of the current eigenvector is compared with the normal eigenvector range to identify those parameters that exceed the normal range; these are the eigenparameters that need adjustment. The target parameter value refers to the desired value that these eigenparameters should achieve. This target value should be within the normal eigenvector range, typically the middle value of the normal range or a value determined based on the optimization objective.

[0152] Step S153: Calculate the adjustment range required to adjust the feature parameter of the dimension to be adjusted from the current value to the target parameter value. The adjustment range is obtained by the difference between the target parameter value and the current value.

[0153] The adjustment magnitude is the target parameter value minus the current feature parameter value. If the current value is lower than the target value, the adjustment magnitude is positive, indicating that the parameter value needs to be increased; if the current value is higher than the target value, the adjustment magnitude is negative, indicating that the parameter value needs to be decreased. The magnitude of the adjustment magnitude reflects the degree of adjustment required.

[0154] Step S154: Based on the evolutionary relationship in the battery swapping network state correlation evolution model, determine the correlation dimension affected by the adjustment of the dimension to be regulated, calculate the characteristic parameters and adjustment magnitude of the correlation dimension that need to be adjusted in a coordinated manner, so that the characteristic parameters of the correlation dimension remain within the normal range after the adjustment of the dimension to be regulated.

[0155] Adjustments to the dimension to be regulated may affect the state of its associated dimensions through evolutionary relationships. Therefore, it is necessary to identify the associated dimensions (mainly directly associated dimensions) of the dimension to be regulated based on the state-related evolution model of the battery swapping network. For these associated dimensions, the potential impact of adjustments to the dimension to be regulated on their characteristic parameters should be analyzed. If the aforementioned impact may cause the characteristic parameters of the associated dimensions to exceed the normal range, it is necessary to calculate the coordinated adjustment magnitude of the corresponding characteristic parameters of the associated dimensions. The calculation of the coordinated adjustment magnitude needs to consider factors such as the adjustment magnitude of the dimension to be regulated, the strength and direction of the association relationship, to ensure that the characteristic parameters of the associated dimensions remain within the normal range after the dimension to be regulated is adjusted.

[0156] Step S155: Configure the adjustment sequence for the adjustment parameters of each dimension to be adjusted and related dimensions. The adjustment sequence includes the adjustment start time, adjustment duration and adjustment step interval. The adjustment of related dimensions is carried out synchronously with the adjustment of the dimension to be adjusted.

[0157] The start time for adjustments is typically set to the current time or a short delay to ensure timely control. The adjustment duration is determined based on the adjustment magnitude and the parameter's response speed; parameters with large adjustment magnitudes or slow responses require a longer duration. The adjustment step interval refers to the time interval between two adjacent steps when adjusting in steps. Step-by-step adjustments can prevent sudden parameter changes from impacting the system. The adjustment sequence of related dimensions should be synchronized with the adjustment sequence of the dimension to be controlled, meaning adjustments should begin simultaneously, and the adjustment duration and step interval should also be consistent to achieve coordinated control.

[0158] Step S156: The adjustment parameters, adjustment magnitude values, and adjustment timing of each dimension to be controlled are classified and integrated according to the state dimension to form a dynamic control flow for the battery swapping network. The dynamic control flow for the battery swapping network includes dimension identifiers, a list of adjustment parameters, adjustment magnitude values, and adjustment timing arrangements.

[0159] The dynamic control flow of the battery swapping network integrates adjustment information from all controllable and related dimensions. Adjustment parameters, adjustment magnitude values, and adjustment timing are categorized according to state dimensions, with each state dimension corresponding to a set of adjustment information. Dimension identifiers distinguish different state dimensions; the adjustment parameter list lists the names of the characteristic parameters that need adjustment; the adjustment magnitude values ​​correspond to the adjustment amount for each parameter; and the adjustment timing arrangement includes information such as the adjustment start time, duration, and step interval. The format of the dynamic control flow should facilitate subsequent conversion into standardized control commands.

[0160] Step S157: Read the dimension identifier, adjustment parameter, adjustment magnitude value and adjustment timing in the dynamic control flow of the battery swapping network, and determine the execution link corresponding to each adjustment parameter. The execution link includes the battery adjustment link of the battery swapping station status execution link, the equipment adjustment link of the battery swapping station status execution link, the request adjustment link of the battery swapping station status execution link, the resource allocation link of the edge computing resource adjustment link, and the task scheduling link of the edge computing resource adjustment link.

[0161] Different state dimensions and adjustment parameters correspond to different execution chains. For example, the adjustment parameters for battery storage status typically correspond to the battery adjustment chain of a battery swapping station, the adjustment parameters for the operation status of battery swapping equipment correspond to the equipment adjustment chain, and the adjustment parameters for the battery swapping request status correspond to the request adjustment chain; the adjustment parameters for edge computing resources correspond to the resource allocation chain and the task scheduling chain. Based on the dimension identifier and the type of adjustment parameter, the corresponding execution chain is retrieved from a pre-defined mapping table.

[0162] Step S158: Select the corresponding instruction format template according to the type of execution link. The instruction format template of different execution links contains the parameter fields and instruction structure that can be recognized by the execution link.

[0163] Each execution chain has its specific instruction format requirements, determined by the hardware or software system of the execution chain. The instruction format template is predefined and includes parameter fields (such as parameter name, adjustment range, start time, etc.) and instruction structure (such as instruction header, parameter area, checksum, etc.) recognizable by that execution chain. The appropriate instruction format template is selected based on the type of execution chain.

[0164] Step S159: Fill the adjustment parameters, adjustment magnitude values, and adjustment timing into the corresponding instruction format template fields to generate a preliminary control instruction. The preliminary control instruction includes the execution link identifier, parameter name, adjustment magnitude value, start time, and duration.

[0165] Information such as adjustment parameters, adjustment magnitude, adjustment start time, and duration related to the execution link is extracted from the dynamic control flow of the battery swapping network and filled into the corresponding fields according to the instruction format template. The execution link identifier indicates which execution link the instruction should be sent to. After filling, a preliminary control instruction is formed, which contains all the key information required for the execution link to perform the control operation.

[0166] Step S1510: Sort the preliminary control instructions according to the order of adjustment timing to form a standardized control instruction sequence. The order of instructions in the standardized control instruction sequence is the same as the execution order of the adjustment timing.

[0167] Different initial control commands may have different start times, and these commands need to be sorted according to their start times. If multiple commands have the same start time, they can be sorted according to their importance or other priority rules. The sorted command sequence is the standardized control command sequence, and the order of the commands in the sequence is consistent with their actual execution order, ensuring that the control operation is carried out according to the predetermined timing.

[0168] Step S1511: Add a sequence identifier to the standardized control instruction sequence. The sequence identifier includes the sequence generation time, control batch identifier, and total number of instructions.

[0169] Sequence identifiers are used to uniquely identify and manage standardized control command sequences. The sequence generation time refers to the exact timestamp of the generation of the command sequence; the control batch identifier is a unique number assigned to the control operation of that batch; the total number of commands refers to the total number of preliminary control commands contained in the sequence. The above information helps the execution link to receive and confirm the command sequence, retransmit errors, and track the execution status.

[0170] Step S1512: Based on the location and transmission requirements of the execution link, the standardized control command sequence is transmitted to the corresponding execution link, and the command transmission completion time is recorded. This ensures that after receiving the standardized control command sequence, the execution link executes the adjustment operation in the order of the commands. During the operation, the current value of the adjustment parameters is collected in real time to form status feedback data. The status feedback data is transmitted to the battery swapping network status perception link for real-time monitoring of the adjustment effect. If the adjustment effect does not meet expectations, the adjusted battery swapping network dynamic control flow and the adjusted standardized control command sequence are regenerated based on the status feedback data and feature vector set, and then transmitted to the execution link for execution again.

[0171] For example, in step S15121: after the execution link receives the standardized control instruction sequence, it parses the instruction order, execution link identifier, adjustment parameters, adjustment amplitude values ​​and adjustment timing in the standardized control instruction sequence.

[0172] The execution link receives standardized control command sequences through a communication interface, verifies the received data to ensure its integrity and correctness. After verification, the command sequence is parsed to extract information such as command order, execution link identifier (used to confirm whether it is a command of this link), adjustment parameters corresponding to each command, adjustment magnitude value, and adjustment timing (start time, duration, step interval), and the above information is stored in a local cache for subsequent execution.

[0173] Step S15122: Adjust the start time according to the order of instructions, and start the corresponding adjustment operation when the start time is reached. If the adjustment operation contains multiple steps, execute each step in sequence according to the adjustment step interval in the instructions.

[0174] The execution chain contains a timer to track the current time. When the timer reaches the start time of an instruction's adjustment, the execution chain retrieves the instruction's adjustment parameters, adjustment magnitude, and step interval from the cache, and initiates the adjustment operation. If the adjustment magnitude is large and requires multiple steps, each step is executed sequentially according to the step interval, with each step involving a small adjustment magnitude, gradually approaching the target parameter value. For example, if the adjustment magnitude is 10, the step interval is 1 second, and the execution is divided into 5 steps, then each step adjusts by 2, and the operation is performed once every 1 second.

[0175] Step S15123: During the adjustment operation, the current values ​​of the adjustment parameters are collected in real time through the parameter acquisition component of the execution link. The acquisition frequency is kept the same as the interval between adjustment steps, so that each adjustment step corresponds to a set of current parameter values.

[0176] The parameter acquisition component of the execution link is connected to the status acquisition interface of the battery swapping network, enabling real-time acquisition of the current values ​​of adjustment parameters. To accurately track the effect of each adjustment step, the acquisition frequency is set to the same interval as the adjustment step; that is, the current value of the adjustment parameter is acquired immediately after each adjustment step is executed. This way, each adjustment step has a corresponding record of the current parameter value, facilitating analysis of parameter changes during the adjustment process.

[0177] Step S15124: Record the current value of the adjustment parameter, the collection time, and the corresponding adjustment steps for each collection, forming a single status feedback record.

[0178] For each collected value of the adjustment parameter, it is associated with the corresponding collection time (accurate to the millisecond level) and the adjustment step number to form a status feedback record. The status feedback record may also include information such as execution link identifier and parameter name to ensure the uniqueness and traceability of the record.

[0179] Step S15125: Integrate multiple status feedback records in the order of collection time to form status feedback data. The status feedback data includes the execution link identifier, the name of the adjustment parameter, the collection time sequence, and the corresponding current value sequence of the parameter.

[0180] Multiple status feedback records from the same execution link and the same adjustment parameter are arranged in chronological order of collection time to form a collection time sequence and a corresponding parameter current value sequence. This sequence, along with the execution link identifier and adjustment parameter name, constitutes the status feedback data. The status feedback data can be organized using formats such as JSON or XML for easy transmission and parsing.

[0181] Step S15126: Establish a connection with the battery swapping network status sensing link, transmit status feedback data to the battery swapping network status sensing link through this connection, and record the data transmission completion time; after receiving the status feedback data, the battery swapping network status sensing link extracts the current value sequence of adjustment parameters, compares it with the adjustment target parameter value, calculates the difference between the current parameter value and the target parameter value at each time point, and obtains the adjustment deviation sequence.

[0182] The execution link establishes a connection with the battery swapping network status sensing link through a preset communication protocol and port. After the connection is established, it sends out status feedback data and records the data transmission completion time. Upon receiving the status feedback data, the battery swapping network status sensing link parses the data and extracts the current value sequence of adjustment parameters. Each value in this sequence is compared with the corresponding target parameter value, and the difference between the two (current value minus target value) is calculated to obtain the adjustment deviation sequence. The adjustment deviation sequence reflects the difference between the current parameter value and the target value at each acquisition time point.

[0183] Step S15127: Analyze the changing trend of the adjustment deviation sequence. If the deviation sequence gradually decreases and the final deviation is within the allowable range, it is determined that the adjustment effect has reached the expected level. If the deviation sequence does not decrease or the final deviation exceeds the allowable range, it is determined that the adjustment effect has not reached the expected level.

[0184] The trend of the deviation sequence can be analyzed by calculating the moving average of the deviation and plotting the deviation curve. If the overall trend of the deviation sequence is a gradual decrease, and the absolute value of the final deviation is less than the preset allowable range (which is determined according to the importance of the parameter and the control precision requirements) at the end of the adjustment operation (or after a period of time), the adjustment effect is considered to have met expectations. If the deviation sequence does not show a significant decreasing trend, or the final deviation exceeds the allowable range, the adjustment effect is considered to have failed to meet expectations.

[0185] Step S15128: If the adjustment effect does not meet expectations, combine the current parameter value sequence in the status feedback data and the feature vector of that dimension in the feature vector set to recalculate the adjustment magnitude and adjustment timing, so that the new adjustment magnitude can enable the parameter to quickly return to the target parameter value.

[0186] When the adjustment effect does not meet expectations, the reasons need to be analyzed. It may be that the initial adjustment magnitude was insufficient, or the adjustment timing was set improperly, causing the parameters to fail to quickly return to the target value. By combining the current parameter value sequence in the state feedback data (reflecting the actual changes in the parameters) and the feature vector of that dimension in the feature vector set (reflecting the relationship between the parameters and other features), the parameter adjustment needs should be reassessed. The new adjustment magnitude can be calculated based on factors such as the current deviation and the parameter's response characteristics, and may be larger or smaller than the initial adjustment magnitude. The adjustment timing can also be optimized, such as shortening the interval between adjustment steps or extending the adjustment duration, to accelerate the speed at which the parameters return to the target value.

[0187] Step S15129: Based on the recalculated adjustment magnitude value and adjustment timing, generate the adjusted dynamic control flow of the battery swapping network. The adjusted dynamic control flow of the battery swapping network includes the updated adjustment parameters, the updated adjustment magnitude value, and the updated adjustment timing.

[0188] Following a method similar to steps S151 to S156, an adjusted dynamic control flow for the battery swapping network is generated based on the recalculated adjustment magnitude and timing. The adjustment parameters in this dynamic control flow may be the same as before (if only the adjustment magnitude and timing have changed), but the adjustment magnitude and timing (start time, duration, step interval) are updated.

[0189] Step S15130: Following the aforementioned instruction conversion process, the adjusted dynamic control flow of the battery swapping network is converted into an adjusted standardized control instruction sequence. The adjusted standardized control instruction sequence is then transmitted to the execution link via the corresponding transmission method. After receiving the adjusted standardized control instruction sequence, the execution link executes the adjustment operation according to the new instructions, while continuing to collect status feedback data and transmit it to the battery swapping network status perception link. The above process is repeated until the adjustment deviation sequence is finally within the allowable range and the adjustment effect reaches the expected level.

[0190] The adjusted dynamic control flow of the battery swapping network is converted into an adjusted standardized control command sequence according to steps S157 to S1511. After adding a new sequence identifier, it is transmitted to the corresponding execution link. Upon receiving the command, the execution link executes the adjustment operation according to the new instructions and continues to collect and transmit status feedback data. The battery swapping network status awareness link re-evaluates the adjustment effect. If the expected result is still not achieved, the above adjustment process is repeated until the current parameter value returns to the target parameter value and the deviation is within the allowable range, at which point the adjustment effect meets expectations. Through the above closed-loop control mechanism, the status of the battery swapping network can be effectively controlled and restored to normal operating level.

[0191] In one exemplary embodiment, an edge computing-based electric vehicle battery swapping network state awareness system is provided. This edge computing-based electric vehicle battery swapping network state awareness system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2 As shown, the edge computing-based electric vehicle battery swapping network status perception system includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for information exchange between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an edge computing-based electric vehicle battery swapping network status perception method. The display unit of this edge computing-based electric vehicle battery swapping network status perception system is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of this edge computing-based electric vehicle battery swapping network status perception system can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad set on the shell of the edge computing-based electric vehicle battery swapping network status perception system, or an external keyboard, touchpad, or mouse, etc.

[0192] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A state perception method for electric vehicle battery swapping networks based on edge computing, characterized in that, The method includes: Construct a state interaction topology for the battery swapping network, and collect dynamic data on battery storage at the battery swapping station, dynamic data on the operation of the battery swapping equipment, and dynamic data on battery swapping requests based on the state interaction topology to generate a multi-dimensional state perception stream for the battery swapping network. The multi-dimensional state perception stream of the battery swapping network is transmitted to the state processing link corresponding to edge computing, and distributed feature mapping processing is performed to obtain the feature vector set of each state dimension of the battery swapping network. A state correlation evolution model for the battery swapping network is constructed based on a set of feature vectors. The state correlation evolution model for the battery swapping network includes the evolutionary correlation between the battery storage state and the operating state of the battery swapping equipment, the evolutionary correlation between the operating state of the battery swapping equipment and the battery swapping request state, and the evolutionary correlation between the battery storage state and the battery swapping request state. By using the state-related evolution model of the battery swapping network, the origin state dimension and diffusion path of the operational bottleneck of the battery swapping network are traced, and the priority of bottleneck control is determined. Based on the bottleneck control priority and feature vector set, a dynamic control flow for the battery swapping network is generated. This dynamic control flow is then converted into a standardized control instruction sequence and transmitted to the state execution link corresponding to the battery swapping station and the resource adjustment link corresponding to the edge computing, thereby realizing real-time perception and dynamic control of the battery swapping network status.

2. The electric vehicle battery swapping network state perception method based on edge computing according to claim 1, characterized in that, The construction of the battery swapping network status interaction topology involves collecting dynamic data on battery storage at battery swapping stations, dynamic data on the operation of battery swapping equipment, and dynamic data on battery swapping requests based on this topology. This generates a multi-dimensional status awareness stream for the battery swapping network, including: Based on the geographical distribution of battery swapping stations and the coverage area of ​​edge computing, the node layout of the battery swapping network status interaction topology is planned, and the number of battery swapping stations covered and the data transmission range corresponding to each topology node are determined. Configure the link transmission parameters of the battery swapping network status interaction topology. The link transmission parameters include the data frame transmission period, data field length and link transmission protocol type. Set the data frame transmission period to not exceed the preset timeliness threshold, match the data field length to the transmission capacity of the data output port of the battery swapping station, and select a link transmission protocol type that meets the real-time transmission requirements. A status acquisition interface is deployed on each topology node. The status acquisition interface includes a battery storage dynamic acquisition interface, a battery swapping equipment operation dynamic data acquisition interface, and a battery swapping request dynamic data acquisition interface. Each interface is connected to the corresponding data output port in the battery swapping station. The system collects dynamic data on battery storage at battery swapping stations, dynamic data on the operation of battery swapping equipment, and dynamic data on battery swapping requests. The dynamic data on battery storage at battery swapping stations includes the remaining power change trend of each battery, the fluctuation of its health status, and the storage location transfer record. The dynamic data on the operation of battery swapping equipment includes the movement trajectory changes of the battery swapping robotic arm, the positioning deviation changes of the battery swapping platform, and the temperature change trend of the battery swapping equipment. The dynamic data on battery swapping requests includes the distribution of battery swapping request initiation time, the distribution of requested battery types, and the changes in the battery swapping waiting time of the requested vehicles. The collected dynamic data of battery storage at the battery swapping station, dynamic data of battery swapping equipment operation, and dynamic data of battery swapping requests are aligned with a unified time base and integrated sequentially according to a preset perception stream structure to obtain an integrated perception stream. The perception stream structure includes a state dimension identifier, a data acquisition time marker, and a dynamic data content field. Add topology node identifiers to the integrated sensing streams, and then aggregate the sensing streams with the added topology node identifiers to form a multi-dimensional state sensing stream of the battery swapping network covering all topology nodes. The topology node identifiers are used to distinguish sensing streams generated by different topology nodes.

3. The electric vehicle battery swapping network state perception method based on edge computing according to claim 1, characterized in that, The process involves transmitting the multi-dimensional state perception stream of the battery swapping network to the state processing link corresponding to edge computing, performing distributed feature mapping processing, and obtaining a set of feature vectors for each state dimension of the battery swapping network, including: The multi-dimensional state perception flow of the battery swapping network is divided into battery storage dynamic flow, battery swapping equipment operation dynamic flow and battery swapping request dynamic flow according to the state dimension. Each dynamic flow is matched with a state processing sub-link of edge computing. In the battery storage dynamic processing sub-link, trend features are extracted from the remaining power change trend in the battery storage dynamic stream to obtain power change trend features; fluctuation features are extracted from the health status fluctuations to obtain health fluctuation features; and location features are extracted from the storage location transfer records to obtain location transfer features. In the dynamic data processing sub-link of the battery swapping equipment operation, trajectory features are extracted from the changes in the robotic arm's motion trajectory in the dynamic flow of the battery swapping equipment operation to obtain trajectory change features; deviation features are extracted from the changes in the positioning deviation of the battery swapping platform to obtain positioning deviation features; and temperature features are extracted from the temperature change trend of the battery swapping equipment to obtain temperature change features. In the dynamic data processing sub-link of battery swapping requests, time period features are extracted from the request initiation time distribution in the dynamic flow of battery swapping requests to obtain time period distribution features; type features are extracted from the request battery type distribution to obtain type distribution features; and time features are extracted from the changes in the waiting time for battery swapping of requesting vehicles to obtain waiting time features. The battery storage state feature vector is obtained by vectorizing the power change trend features, health fluctuation features, and location transfer features obtained from the battery storage dynamic processing sub-link. The trajectory change features, positioning deviation features, and temperature change features obtained from the dynamic data processing sub-link of the battery swapping equipment are vectorized to obtain the operating status feature vector of the battery swapping equipment. The time period distribution characteristics, type distribution characteristics, and waiting time characteristics obtained from the dynamic data processing sub-link of battery swapping requests are transformed into vectors to obtain the battery swapping request status feature vector. The feature vectors of battery storage status, battery swapping equipment operation status, and battery swapping request status are processed to unify their dimensions and then classified and organized according to the status dimensions to form a feature vector set for each status dimension of the battery swapping network. The feature vector set includes the battery storage status feature vector group, the battery swapping equipment operation status feature vector group, and the battery swapping request status feature vector group.

4. The electric vehicle battery swapping network state perception method based on edge computing according to claim 3, characterized in that, In the battery storage dynamic processing sub-link, the remaining power change trend in the battery storage dynamic stream is extracted to obtain the power change trend feature. Fluctuation features are extracted from the fluctuations in health status to obtain health fluctuation features; Location features are extracted from the storage location transfer records to obtain location transfer features, including: The remaining power trend data of each battery is separated from the battery storage dynamic stream and arranged in chronological order to form a remaining power time series; The remaining power time series is segmented, dividing the continuous time series into multiple consecutive time periods, and the remaining power changes in each time period have similar trends. Calculate the change range of remaining power in each time period. The change range value for that time period is obtained by the difference between the remaining power at the start and end of the time period. Calculate the change frequency of remaining power in each time period. The change frequency value is obtained by dividing the number of times the remaining power changes in the time period by the duration of the time period. Integrate the change range value and change frequency value of each time period in the order of the time periods to form the power change trend characteristics. Extract the health status fluctuation data of each battery from the battery storage dynamic stream, and record the starting value, peak value and duration of each health status fluctuation. Calculate the fluctuation range of each health status fluctuation, and obtain the fluctuation range value by the difference between the fluctuation peak value and the fluctuation start value; calculate the number of health status fluctuations per unit time, and obtain the fluctuation frequency value by dividing the total number of fluctuations by the total observation duration; integrate the fluctuation range value, fluctuation duration and fluctuation frequency value per unit time of each fluctuation to form the health fluctuation characteristics; Extract the storage location transfer record data of each battery from the battery storage dynamic stream, and record the starting location identifier, target location identifier and transfer time for each transfer; The number of transfers corresponding to each starting position identifier is counted to obtain the starting position transfer frequency; the number of receptions corresponding to each target position identifier is counted to obtain the target position reception frequency. Calculate the average time for each transfer by dividing the total transfer time by the total number of transfers. The frequency of starting position transfer, the frequency of receiving at the target position, and the average time consumption are integrated to form the position transfer characteristics.

5. The electric vehicle battery swapping network state perception method based on edge computing according to claim 1, characterized in that, The method for constructing a state-related evolution model for the battery swapping network based on a set of feature vectors includes: Extract the battery storage status feature vector group, the battery swapping equipment operation status feature vector group, and the battery swapping request status feature vector group from the feature vector set, and sort each vector group in chronological order; The battery storage state feature vector and the battery swapping equipment operation state feature vector at adjacent time points are selected. The correlation between these two vectors in the time dimension is calculated to obtain the battery-equipment correlation degree. The battery swapping equipment operation state feature vector and the battery swapping request state feature vector at adjacent time points are selected. The correlation between these two vectors in the time dimension is calculated to obtain the equipment-request correlation degree. The battery storage state feature vector and the battery swapping request state feature vector at adjacent time points are selected. The correlation between these two vectors in the time dimension is calculated to obtain the battery-request correlation degree. Record the battery device correlation, device request correlation, and battery request correlation at different time points to form a correlation time series; Analyze the changing trend of correlation over time, determine the evolution law of correlation over time, and divide the correlation into periods of increasing correlation, decreasing correlation, and stable correlation. Based on the evolutionary law of correlation, this study constructs an evolutionary correlation between battery storage state and battery swapping equipment operation state, describing the mutual influence between these two states at different time stages; and also constructs an evolutionary correlation between battery swapping equipment operation state and battery swapping request state, describing the mutual influence between these two states at different time stages; and also constructs an evolutionary correlation between battery storage state and battery swapping request state, describing the mutual influence between these two states at different time stages. The evolutionary relationships between battery storage status and battery swapping equipment operation status, battery swapping equipment operation status and battery swapping request status, and battery storage status and battery swapping request status are integrated according to the time dimension. The weights of each relationship are set at different time stages, and the weights are determined based on the degree of correlation. The three evolutionary relationships after integrating the weight settings are used to form a battery swapping network state correlation evolution model, which includes time stage division, correlation relationship of each stage and corresponding weight.

6. The electric vehicle battery swapping network state perception method based on edge computing according to claim 5, characterized in that, The process involves selecting battery storage state feature vectors and battery swapping equipment operation state feature vectors at adjacent time points, calculating the correlation between these two vectors over time, and obtaining the battery-equipment correlation degree. It also involves selecting battery swapping equipment operation state feature vectors and battery swapping request state feature vectors at adjacent time points, calculating the correlation between these two vectors over time, and obtaining the equipment request correlation degree. Finally, it involves selecting battery storage state feature vectors and battery swapping request state feature vectors at adjacent time points, calculating the correlation between these two vectors over time, and obtaining the battery request correlation degree. This process includes: Determine the time interval between adjacent time nodes so that the interval between all adjacent time nodes remains the same; The feature vectors of battery storage state, battery swapping equipment operation state, and battery swapping request state are standardized to eliminate the dimensional differences of each feature value and make all feature values ​​comparable to each other. Extract the standardized battery storage state feature vector at the Nth time node and the standardized battery storage state feature vector at the N+1th time node, calculate the degree of difference between the two vectors, and obtain the change in battery state; extract the standardized battery swapping equipment operation state feature vector at the Nth time node and the standardized battery swapping equipment operation state feature vector at the N+1th time node, calculate the degree of difference between the two vectors, and obtain the change in equipment state. Calculate the degree of coordinated change between battery state change and device state change. Obtain the coordinated change value by the overlap ratio and the consistency of the direction of change between battery state change and device state change. Convert the coordinated change value into battery-device correlation degree. The range of correlation degree values ​​corresponds to the range of changes in the coordinated change value. Calculate the degree of difference between the standardized power swapping equipment operating state feature vector at the Nth time node and the standardized power swapping equipment operating state feature vector at the N+1th time node to obtain the change in equipment state. Calculate the degree of difference between the standardized battery swapping request state feature vector at the Nth time node and the standardized battery swapping request state feature vector at the (N+1)th time node to obtain the change in request state. Calculate the degree of coordinated change between the change in device state and the change in requested state, and obtain the coordinated change value by the overlap ratio and the consistency of the direction of change between the change in device state and the change in requested state. The collaborative change value is converted into a device request correlation degree, and the range of correlation degree values ​​corresponds to the range of change of the collaborative change value. Calculate the degree of difference between the standardized battery storage state feature vector at the Nth time node and the standardized battery storage state feature vector at the N+1th time node to obtain the battery state change; calculate the degree of difference between the standardized battery swap request state feature vector at the Nth time node and the standardized battery swap request state feature vector at the N+1th time node to obtain the request state change. Calculate the degree of coordinated change between the battery state change and the requested state change, and obtain the coordinated change value by the overlap ratio and the consistency of the direction of change between the battery state change and the requested state change. The collaborative change value is converted into a battery request correlation degree, and the range of correlation degree values ​​corresponds to the range of change of the collaborative change value. Record the battery device correlation, device request correlation, and battery request correlation calculated for each adjacent time node, so that each time node corresponds to a set of correlation data.

7. The electric vehicle battery swapping network state perception method based on edge computing according to claim 1, characterized in that, The process of tracing the origin and spread path of operational bottlenecks in the battery swapping network through a state-related evolution model, and determining bottleneck control priorities, includes: The multi-dimensional state perception flow of the battery swapping network is input into the state association evolution model of the battery swapping network. Based on the internal evolutionary association relationship and weight, the state association evolution model of the battery swapping network outputs the operation status evaluation results of each state dimension at different time stages. Analyze the operational status assessment results of each state dimension, identify the state dimensions whose operational status assessment results are in the abnormal range, and determine the initial abnormal dimension; based on the evolutionary correlation relationship in the battery swapping network state correlation evolution model, find other state dimensions that are correlated with the initial abnormal dimension, and determine the correlated dimension. Analyze the evolutionary relationship between the initial anomaly dimension and the correlation dimension, determine the order of influence of the initial anomaly dimension on the correlation dimension, the correlation dimension first affected by the initial anomaly dimension is the first-level correlation dimension, the correlation dimension affected by the first-level correlation dimension is the second-level correlation dimension, and so on. Track the start time of the anomaly in the initial anomaly dimension, compare it with the time when the anomaly appeared in the associated dimensions, and determine the time sequence in which the anomaly spread from the initial anomaly dimension to the associated dimensions. Based on the order of impact and the time sequence, the diffusion path of the bottleneck in the battery swapping network is constructed. The path includes the initial anomaly dimension, the correlation dimensions at each level, and the corresponding diffusion time. The initial anomaly dimension is determined as the origin state dimension of the bottleneck in the battery swapping network operation; The degree of abnormality in each dimension of the diffusion path is analyzed. The duration of the abnormality and the degree of deviation from the normal range are standardized to obtain dimensionless duration scores and deviation scores. Then, the severity of the abnormality in each dimension is determined by combining the duration scores and deviation scores. Based on the severity of anomalies and the hierarchy in the diffusion path, the priority of bottleneck control is determined. The control priority of the origin state dimension is higher than that of the first-level associated dimension, the control priority of the first-level associated dimension is higher than that of the second-level associated dimension, and the control priority of the dimension with the larger severity of anomalies within the same level is higher. Record the origin state dimension, diffusion path and bottleneck control priority to form a bottleneck tracing report. The bottleneck tracing report includes anomaly details, diffusion order and control priority ranking for each dimension.

8. The electric vehicle battery swapping network state perception method based on edge computing according to claim 7, characterized in that, The analysis of the operational status assessment results for each state dimension identifies the state dimensions whose operational status assessment results fall within the abnormal range, thus determining the initial abnormal dimension. Based on the evolutionary correlation relationship in the battery swapping network state correlation evolution model, other state dimensions that are correlated with the initial abnormal dimension are identified, thus determining the correlated dimensions, including: Based on the parameter value range in the historical normal operation data of the battery swapping network, the normal operating range of each state dimension is determined. The operating status evaluation results of each state dimension at different time stages are compared with the corresponding normal range. If the evaluation result exceeds the normal range, the state dimension is marked as abnormal at the corresponding time stage. The percentage of abnormal time for each state dimension is calculated by dividing the total abnormal time duration by the total observation duration. The state dimension with the largest percentage of abnormal time is selected as the initial abnormal dimension. If there are multiple state dimensions with the same and largest percentage of abnormal time, the severity of the abnormality of each dimension is further compared, and the state dimension with the largest severity of the abnormality is selected as the initial abnormal dimension. Retrieve all evolutionary relationships in the state correlation evolution model of the battery swapping network, filter out the relationships that contain the initial anomaly dimension, extract other state dimensions associated with the initial anomaly dimension from the relationships that contain the initial anomaly dimension, and use the extracted dimensions as correlation dimensions. The correlation dimensions are classified into direct correlation dimensions and indirect correlation dimensions based on the type of correlation relationship. Direct correlation dimensions are those that are directly correlated with the initial anomaly dimension, while indirect correlation dimensions are those that are correlated with the initial anomaly dimension through other dimensions. Analyze the correlation strength between directly related dimensions and the initial abnormal dimensions, determine the correlation strength value by the weight in the correlation relationship, and sort the directly related dimensions in descending order of correlation strength value to form a sorting table of directly related dimensions; Analyze the association path length between the indirect association dimension and the initial anomaly dimension. The association path length is the number of intermediate dimensions that the indirect association dimension is associated with the initial anomaly dimension through intermediate dimensions. The smaller the path length value, the stronger the association. The indirect association dimensions are sorted in ascending order of their associated path length values ​​to form an indirect association dimension sorting table. Integrate the sorting tables of directly related dimensions and indirectly related dimensions to form a list of related dimensions. The list of related dimensions records the relationship between each related dimension and the initial abnormal dimension, as well as the degree of relationship.

9. The electric vehicle battery swapping network state perception method based on edge computing according to claim 1, characterized in that, The process of generating a dynamic control flow for the battery swapping network based on bottleneck control priorities and feature vector sets, converting the dynamic control flow into a standardized control command sequence, and transmitting it to the state execution link corresponding to the battery swapping station and the resource adjustment link corresponding to edge computing, enables real-time perception and dynamic control of the battery swapping network status, including: Based on the bottleneck control priority, state dimensions with control priorities from high to low are selected in sequence to determine the current feature vector of each dimension to be controlled. Based on the current feature vector of the dimension to be adjusted and the normal feature vector range of the dimension to be adjusted, determine the feature parameters that need to be adjusted and the target parameter values. The target parameter values ​​are the parameter values ​​that make the feature parameters return to the normal feature vector range. Calculate the adjustment magnitude required to adjust the feature parameter of the dimension to be adjusted from its current value to the target parameter value, and obtain the adjustment magnitude value by the difference between the target parameter value and the current value; Based on the evolutionary relationship in the state-related evolution model of the battery swapping network, the related dimensions affected by the adjustment of the dimension to be regulated are determined, and the characteristic parameters and adjustment magnitude of the related dimensions that need to be adjusted in coordination are calculated so that the characteristic parameters of the related dimensions remain within the normal range after the adjustment of the dimension to be regulated. Configure adjustment timing for the adjustment parameters of each dimension to be adjusted and related dimensions. The adjustment timing includes the adjustment start time, adjustment duration and adjustment step interval. The adjustment of related dimensions is carried out synchronously with the adjustment of the dimension to be adjusted. The adjustment parameters, adjustment magnitude values, and adjustment timing of each dimension to be regulated are classified and integrated according to the state dimension to form a dynamic control flow for the battery swapping network. The dynamic control flow for the battery swapping network includes dimension identifiers, a list of adjustment parameters, adjustment magnitude values, and adjustment timing arrangements. Read the dimension identifier, adjustment parameters, adjustment magnitude values ​​and adjustment timing in the dynamic control flow of the battery swapping network, and determine the execution link corresponding to each adjustment parameter. The execution link includes the battery adjustment link of the battery swapping station status execution link, the equipment adjustment link of the battery swapping station status execution link, the request adjustment link of the battery swapping station status execution link, the resource allocation link of the edge computing resource adjustment link, and the task scheduling link of the edge computing resource adjustment link. Select the corresponding instruction format template according to the type of execution link. The instruction format templates of different execution links contain the parameter fields and instruction structures that can be recognized by that execution link. Fill the adjustment parameters, adjustment magnitude values, and adjustment timing into the corresponding instruction format template fields to generate a preliminary control instruction. The preliminary control instruction includes the execution link identifier, parameter name, adjustment magnitude value, start time, and duration. The initial control instructions are sorted according to the order of adjustment timing to form a standardized control instruction sequence. The order of instructions in the standardized control instruction sequence is the same as the execution order of the adjustment timing. Add a sequence identifier to the standardized control instruction sequence. The sequence identifier includes the sequence generation time, control batch identifier, and total number of instructions. Based on the location and transmission requirements of the execution link, the standardized control command sequence is transmitted to the corresponding execution link, and the command transmission completion time is recorded. This ensures that after receiving the standardized control command sequence, the execution link executes the adjustment operation in the order of the commands. During the operation, the current value of the adjustment parameters is collected in real time to form status feedback data. The status feedback data is transmitted to the battery swapping network status perception link for real-time monitoring of the adjustment effect. If the adjustment effect does not meet expectations, the adjusted battery swapping network dynamic control flow and the adjusted standardized control command sequence are regenerated based on the status feedback data and feature vector set, and then transmitted to the execution link for execution again.

10. A state perception system for electric vehicle battery swapping networks based on edge computing, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the electric vehicle battery swapping network state awareness method based on edge computing according to any one of claims 1 to 9 by executing the machine-executable instructions.

Images