A mobile charging pile intelligent management system and method

Abstract

The present application relates to the technical field of charging pile management, and particularly relates to a mobile charging pile intelligent management system and method, comprising: collecting data, fusing mobile charging pile operation data, user data and environment data to obtain a normalized data set; inputting the normalized data set into a graph neural network to construct a graph structure to obtain an initial graph structure, fusing a feature matrix and an adjacency matrix in the initial graph structure to obtain a fused multi-dimensional data set; inputting the fused multi-dimensional data set into a demand prediction model to obtain a three-dimensional demand matrix; determining a high-demand area, inputting the three-dimensional demand matrix and the area code of the high-demand area into a scheduling cost formula to obtain a scheduling cost, determining a charging pile calling priority, and generating a scheduling instruction according to the charging pile calling priority and the area code of the high-demand area. The scheme constructs a hybrid prediction model through GCN and LSTM, improves prediction accuracy, and optimizes scheduling efficiency.

Description

Technical Field

[0001] This invention relates to the field of charging pile management technology, and in particular to an intelligent management system and method for mobile charging piles. Background Technology

[0002] Mobile charging stations face numerous challenges in data collection during actual operation. Positioning is susceptible to obstruction by tall buildings and electromagnetic interference, leading to insufficient trajectory recording accuracy. Current and voltage sensors lack redundancy, making data acquisition prone to interruption in case of malfunction, and transient pulse interference affects the accuracy of power curves. User data collection lacks a clear authorization mechanism, resulting in insufficient protection of sensitive information, low efficiency in evaluation and feedback processing, and difficulty in accurately extracting user preferences. Environmental data acquisition frequency is low, traffic and meteorological information is not updated in a timely manner, parking space positioning errors are large, and data from different areas lacks refined statistical analysis, failing to provide effective support for demand analysis.

[0003] In terms of data processing, the integration of multi-source heterogeneous data is low, and the methods for handling outliers and missing values ​​are simple, which can easily lead to data distortion. During scheduling, the differences in regional demand and the actual status of charging piles are not fully considered, resulting in low resource matching efficiency and a lack of dynamic response to grid load, which can easily affect power supply stability. Summary of the Invention

[0004] This invention constructs a hybrid prediction model using GCN and LSTM to improve prediction accuracy and optimize scheduling efficiency.

[0005] The technical solution proposed in this invention is: an intelligent management method for mobile charging piles, the method comprising:

[0006] Collect mobile charging pile operation data, user data, and environmental data, and fuse these data through edge computing nodes to obtain a normalized dataset;

[0007] The normalized dataset is input into a graph neural network to construct the graph structure and obtain the initial graph structure. The feature matrix and adjacency matrix in the initial graph structure are fused using a graph convolutional network to obtain the fused multidimensional dataset.

[0008] The fused multi-dimensional dataset is input into the demand prediction model to predict the probability distribution of charging demand in different regions and at different times, thus obtaining a three-dimensional demand matrix.

[0009] High-demand areas are identified based on the three-dimensional demand matrix. The three-dimensional demand matrix and the area codes of the high-demand areas are then input into the scheduling cost formula to obtain the scheduling cost. The priority of charging piles is determined based on the scheduling cost. Scheduling instructions are then generated based on the priority of charging piles and the area codes of the high-demand areas.

[0010] Preferably, the specific process for collecting mobile charging pile operation data, user data, and environmental data is as follows:

[0011] The positioning module adjusts the acquisition frequency according to the moving speed of the mobile charging pile to collect and record the real-time time and timestamp. It collects the real-time power through current and voltage sensors. The BMS dynamically adjusts the monitoring frequency according to the charging stage to monitor the remaining power of the mobile charging pile and sets a breakpoint resume mechanism.

[0012] After user authorization, reservation information and evaluation feedback are collected, and sensitive information is processed using differential privacy technology;

[0013] It connects to the traffic management department's API to collect traffic flow, average speed, and congestion index for each road segment, introduces road topology data, establishes a road network distance matrix, sets up a data caching mechanism, and collects parking space data through a geographic information system.

[0014] Preferably, the process of obtaining the normalized dataset is as follows:

[0015] Adopt 3 The criteria are as follows: an outlier threshold is set, and if an outlier exceeds the threshold twice consecutively, it is considered an outlier and is removed. Missing values ​​are filled using linear interpolation to obtain a cleaned dataset. The cleaned data is normalized using the min-max standardization formula, and the location data is converted into planar coordinates using Gaussian projection to obtain a normalized dataset.

[0016] Preferably, the process of obtaining the initial graph structure is as follows:

[0017] Define charging pile nodes, traffic segment nodes, and parking lot nodes, obtain the node feature matrix, calculate the shortest path time of the road network using Dijkstra's algorithm, input the shortest path time into the edge weight calculation formula to calculate the edge weight, construct the edge set and adjacency matrix based on the edge weight, and construct the initial graph structure based on the node feature matrix, edge set, and adjacency matrix.

[0018] Preferably, the process for obtaining the fused multi-dimensional dataset is as follows:

[0019] The adjacency matrix is ​​input into the first layer of the graph convolutional network to obtain the first-layer feature matrix. The first-layer feature matrix, user preference data, and meteorological data are weighted and summed using an attention mechanism to calculate the attention weights, resulting in a weighted feature matrix. The weighted feature matrix is ​​then concatenated with the user preference data and meteorological data to obtain a fused feature matrix. The concatenation dimension is determined using 5-fold cross-validation. The fused feature matrix is ​​then integrated with the original data timestamps to obtain a spatiotemporal structured dataset. The spatiotemporal structured dataset is then standardized and encapsulated to obtain a fused multi-dimensional dataset.

[0020] Preferably, the process of obtaining the three-dimensional demand matrix is ​​as follows:

[0021] The fused multi-dimensional dataset is processed for temporal serialization. The fused feature matrix is ​​sorted by geographic region and time to construct a regional temporal feature set. Feature extraction is performed on the regional temporal feature set and adjacency matrix. The extracted features are input into a GNN spatial model. For the feature matrix at each time step, graph convolution with consistent GNN feature fusion is used to calculate the regional spatial feature sequence. The regional spatial feature sequence is input into an LSTM time series model. Residual connections are added between LSTM layers, and the sequence is mapped to the charging demand probability distribution through a fully connected layer to obtain the demand probability of a single region at a single time step. A three-dimensional demand matrix is ​​obtained based on the demand probability of a single region at a single time step.

[0022] Preferably, the process for determining the priority of charging pile activation is as follows:

[0023] The total probability of high power level demand in a certain region at a certain time step is calculated based on the three-dimensional demand matrix. If the total probability of high power level demand is greater than or equal to the regional demand threshold, it is determined to be a high-demand region. The region code corresponding to the total demand probability of the high-demand region is input into the scheduling cost formula to calculate the scheduling cost of each mobile charging pile. All mobile charging piles are prioritized according to the call cost. The time required for each mobile charging pile to reach the high-demand region is calculated. If the time required for a high-priority mobile charging pile is greater than 30 minutes, the next priority mobile charging pile is selected.

[0024] Preferably, the scheduling instruction generation process is as follows:

[0025] Using elliptic curve cryptography, public and private keys for charging piles are generated based on the charging pile device number, area code, and the public key of the power grid dispatch center. These keys are then submitted to the blockchain network for registration, and identity authentication is completed after consensus among all network nodes. Based on the Solidity language, a smart contract is written, incorporating regional demand thresholds from a three-dimensional demand matrix, idle charging pile criteria, power grid load adjustment coefficients, and a blockchain node list. The smart contract calls the QueryIdle() function to filter a list of charging piles with idle status and battery levels ≥70% according to the idle criteria, and determines the call priority using a scheduling cost formula. For high-demand areas, scheduling instructions are generated based on the area code and the estimated arrival time of the charging piles.

[0026] The present invention also provides a mobile charging pile intelligent management system, the system being used to execute the aforementioned mobile charging pile intelligent management method.

[0027] The present invention also provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the aforementioned intelligent management method for mobile charging piles.

[0028] The beneficial effects of this invention are:

[0029] 1. Through multi-dimensional data collection from BeiDou positioning, BMS system, and environmental data, combined with 3σ criterion cleaning and min-max standardization, a comprehensive improvement in data quality was achieved. In graph structure construction, road network distance was introduced to replace straight-line distance in edge weight calculation. Then, a hybrid GCN and LSTM model was used to capture regional spatial correlations and temporal patterns, reducing charging demand prediction errors to within 8%. This provides accurate data support for subsequent scheduling, avoiding resource misallocation or supply-demand imbalance.

[0030] 2. A blockchain network built on ECC encryption and PBFT consensus algorithm enables trusted storage of charging pile identity authentication and scheduling commands. The smart contract incorporates a scheduling cost function (combining distance and demand urgency) and tiered power adjustment rules, reducing response time in high-demand areas by 30% and grid load fluctuations by 35%. Simultaneously, the introduction of a private key multi-signature mechanism and third-party audit nodes effectively prevents data tampering and single-point-of-failure risks, ensuring that the entire scheduling latency remains stable within 12 seconds, meeting both real-time and security requirements.

[0031] 3. From dynamic frequency adjustment of data collection (such as reducing positioning frequency to reduce redundancy when stationary), to the strategy of marking and retaining abnormal data, and then to blockchain notarization and model iteration after scheduling execution, a complete closed-loop management system has been formed. At the user level, differential privacy technology protects sensitive information while improving the interactive experience through voice feedback and a rating system, increasing satisfaction from 72 to 89 points. At the system level, the breakpoint resume mechanism reduces the data loss rate from 5.3% to 0.8% and increases resource utilization by 37.8%, ensuring the stable operation of the charging network in complex scenarios. Attached Figure Description

[0032] Figure 1 This is a flowchart of an intelligent management method for mobile charging piles according to the present invention;

[0033] Figure 2 This is a flowchart illustrating the prediction process of an intelligent management method for mobile charging piles according to the present invention. Detailed Implementation

[0034] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0035] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0036] like Figure 1 and Figure 2 As shown, mobile charging pile operation data is collected through its built-in basic sensors, battery management system (BMS), and positioning module. The positioning module uses a BeiDou dual-mode positioning module and is equipped with an anti-interference antenna. This antenna employs a multi-band design, effectively resisting interference from urban tall buildings and electromagnetic interference, improving positioning accuracy to 0.5 meters. The acquisition frequency is dynamically adjusted according to the moving speed; when the speed is ≥30km / h, it is 0.5 seconds / time. At 30km / h, the system automatically switches to a sampling frequency of 5 seconds per cycle, recording not only real-time location but also generating detailed movement trajectory curves for subsequent analysis of the charging pile's movement path and activity range. The current and voltage sensors employ a redundant design, with the primary and backup sensors operating simultaneously, increasing the sampling frequency to 500Hz. If the primary sensor malfunctions (e.g., abnormal data fluctuations exceeding ±10%), the system automatically switches to the backup sensor within 0.1 seconds, ensuring uninterrupted power data acquisition. Simultaneously, a digital filtering algorithm is introduced to automatically filter out transient pulse interference, resulting in a smooth real-time power curve. The BMS dynamically adjusts the monitoring frequency based on the charging stage to monitor the remaining battery power: 20 seconds per cycle during the initial charging phase (0-80% SOC) and 60 seconds per cycle during the later charging phase (80%-100% SOC). Every 20 seconds, the temperature of different parts of the battery (including cell surface, battery pack casing, heat sink, etc.) is monitored, and the temperature difference between these parts is calculated. When the temperature difference exceeds 5℃, a high-temperature warning is immediately triggered, and detailed temperature distribution data is uploaded. In addition, the voltage difference of individual battery cells is monitored. When it exceeds 0.1V, it is marked as an abnormal state, and the data upload frequency is increased to once every 5 seconds to promptly grasp the battery balance. A breakpoint resume mechanism is set up, and data is cached for up to 15 minutes when the network is interrupted, and automatically retransmitted after the network is restored. An abnormal data is marked and retained using a strategy that dynamically adjusts the 3σ criterion threshold by using a sliding window standard deviation (window size set to 5 minutes) to avoid accidentally deleting normal fluctuation data at the moment of startup.

[0037] Data is stored directly in a local database. All operational data is encrypted using the AES-256 encryption algorithm before being stored in the encrypted database on the local edge node. The encryption key is automatically updated every 24 hours, and the key is stored and managed through a hardware security module (HSM) to prevent key leakage. In addition to regular checks, data integrity verification incorporates hash value comparison. The stored data is hashed hourly and compared with the original hash value to ensure data integrity. After the charging pile is started, each sensor and module collects data at a set frequency, which is then processed, encrypted, and stored in the local encrypted database. The system monitors the equipment's operating status in real time. When an anomaly occurs (such as sensor failure or data anomaly), it automatically triggers corresponding early warning mechanisms and data re-collection strategies. Data integrity is verified hourly to ensure data accuracy and reliability. This results in interference-resistant location trajectory data with millisecond-level timestamps, filtered real-time power curves from redundant acquisitions, and battery status datasets with temperature gradient monitoring and early warning indicators.

[0038] When collecting user data, a data collection authorization agreement must be clearly displayed when a user first uses the app. The agreement should detail the types of data collected, their purpose, and the storage period. Unnecessary preference data should only be collected after user authorization. When a user initiates a reservation, in addition to recording the reservation time, desired charging amount, and specified charging location, the app simultaneously retrieves the user's historical charging data for the past three months. Big data analysis is used to derive characteristics such as fast / slow charging preferences (e.g., users who choose fast charging more than 70% are marked as "fast charging preference users"), average charging time, and commonly used payment methods, which are then stored in association with the reservation information. The app includes a voice feedback function, allowing users to submit reviews via voice. The system uses deep learning-based speech recognition technology to accurately convert speech into text. Along with the text reviews, natural language processing technology is used to extract keywords (e.g., "slow charging," "loose interface," "good service"), which are then categorized and stored according to categories such as "facilities issues," "service quality," and "charging experience." Additionally, a 1-5 star rating option is provided on the charging completion page for convenient user feedback on the charging experience.

[0039] When analyzing users' historical charging data, differential privacy technology is used to process sensitive information, such as blurring the average charging cost by ±10% and dividing specific charging times into time slots (e.g., "8:00-10:00") to avoid leaking users' precise consumption information and travel patterns. A data usage report is pushed to users regularly (monthly), detailing the data usage scenarios and scope. Users can choose to modify the data collection scope or delete some historical data at any time. After users register for the app and authorize data collection, the system collects reservation information, order data, and feedback information in real time during user usage. The collected data is processed for privacy and categorized before being stored in a user preference database. A monthly data usage report is generated and pushed to users, and the data collection strategy is adjusted based on user feedback. This results in user order data containing reservation information (authorized by the user), user profiles with preference tags and differential privacy processing, and user feedback datasets categorized from multiple channels.

[0040] When collecting environmental data, the system connects to the real-time traffic data API of the traffic management department, increasing the data acquisition frequency to once every 3 minutes. The acquired information includes traffic flow, average vehicle speed, and congestion index (0-10) for each road segment. Traffic conditions within a 1-kilometer radius of charging stations are highlighted; areas with a congestion index exceeding 7 are automatically marked as congested. Road topology data is incorporated to create a road network distance matrix instead of straight-line distances. Real-time traffic data is obtained through OpenStreetMap, and the actual travel time between nodes is calculated as the basis for edge weight calculation. A data caching mechanism is also implemented. When the network is interrupted, cached data is automatically activated (caching time not exceeding 15 minutes), and the latest data is quickly synchronized after the network is restored to ensure data continuity.

[0041] In addition to collecting parameters such as temperature, humidity, precipitation, and wind force, the real-time data interface connected to the meteorological service platform now includes Air Quality Index (AQI) and visibility parameters. When severe weather or environmental conditions such as heavy rain (daily precipitation ≥ 50 mm), strong winds (wind force ≥ level 6), or AQI ≥ 200 are detected, the data refresh frequency is automatically increased to once per minute, and a severe weather warning is triggered.

[0042] Parking space data is linked to Geographic Information System (GIS) basic data to pinpoint the location of available parking spaces to specific coordinates (error ≤ 3 meters). Combined with the parking lot's aisle layout and charging pile dimensions, the system analyzes and determines the optimal parking route, facilitating navigation of mobile charging piles to available spaces. Simultaneously, the system synchronizes the total number of parking spaces and the number of occupied spaces every 5 minutes, calculates the parking space occupancy rate, and generates a heat map to visually display the parking space shortage in different areas of the parking lot.

[0043] During the environmental data spatial matching process, buffer analysis is introduced. Centered on the current location of the charging pile, data statistics are performed within a 500-meter, 1-kilometer, and 2-kilometer radius, including traffic flow distribution, differences in meteorological parameters, and changes in parking space occupancy rates. This provides refined data support for analyzing charging demand in different radii. The system periodically connects to various data interfaces to acquire traffic, meteorological, and parking data. After processing, correlation, and analysis, the data is categorized and stored according to region and buffer. When severe weather or traffic congestion is detected, corresponding data processing and early warning mechanisms are activated to ensure the timeliness and effectiveness of the data. This results in a cached traffic condition dataset with regional identifiers, meteorological data including air quality and visibility with severe weather warnings, and parking space occupancy data with heatmaps and associated geographic coordinates.

[0044] Edge computing nodes are deployed at charging station terminals to fuse the collected multi-source heterogeneous data. The collected data includes location trajectory data. ,in , For the first The longitude (unit: degrees) of each record. For the first The latitude of each record Timestamps (in milliseconds); power curve data ,in For the first The charging power (unit: kW) of each record; battery status data. ,in , For the first The remaining battery power (in %) recorded for each item. For timestamps (in milliseconds). Battery temperature (unit: °C); User order data ,in , The reservation time is in seconds. Charging time (in seconds). Charging capacity (unit: kWh); traffic data ,in , For the first The congestion index of each record (range 0-10). Average vehicle speed (unit: km / h); meteorological data ,in , Temperature (unit: °C) Humidity (unit: %) Air Quality Index (range 0-500); Parking data ,in , Parking space occupancy rate (unit: %) The coordinates of the parking space are in degrees.

[0045] Adopt 3 Criteria, defining outlier thresholds For data sequences Calculate the mean ( (sequence length) and standard deviation ,when Furthermore, values ​​appearing more than twice consecutively are considered outliers and removed to avoid misjudgment based on a single fluctuation. Missing values ​​are filled using linear interpolation: if... and ( There are missing values ​​between them. ( ),but The dataset is obtained after cleaning. , where the superscript " "" represents the cleaned data. The cleaned data is normalized using the min-max normalization formula, which is: ,in, The minimum value of the data sequence. The maximum value of the data sequence after normalization Location data Transform into planar coordinates using Gaussian projection (Unit: meters) Obtain the normalized dataset .

[0046] Construct a graph structure based on a normalized dataset. Define a set of nodes. ,in, For charging pile nodes (number) ), feature vector , To normalize the electrical quantity, For normalized power, Normalized temperature; Traffic segment nodes (number) ), feature vector , To normalize the congestion index, To normalize vehicle speed; Parking lot nodes (number) ), feature vector , To normalize the occupancy rate, the node feature matrix ,in For feature dimension ( (Eigenvectors are padded with zeros to 3D). This yields the node feature matrix. Edge weights are calculated using road network travel time; the formula for edge weight calculation is as follows: ,in, The time decay coefficient, For nodes arrive The actual travel time (calculated using Dijkstra's algorithm to find the shortest path in the road network), when hour ,otherwise Thus, the adjacency matrix is ​​obtained. ,in Based on the node feature matrix Adjacency matrix , build edge set Obtain the initial graph structure .

[0047] The initial graph structure is constructed using a graph convolutional network (GCN). The characteristic matrix in Adjacency matrix Fusion was performed. The hyperparameters of the GCN were determined through grid search. The number of hidden units in the first layer was set to 128, and in the second layer to 64, with a dropout rate of 0.3 to avoid overfitting. The first layer of the Graph Convolutional Network (GCN) was used for computation. ,in, ( (For the identity matrix, add self-loops). for The degree matrix satisfies , This is the weight matrix. For bias terms, It is the ReLU activation function, i.e. Obtain the first layer feature matrix The first layer feature matrix is ​​processed through an attention mechanism. User preference data Meteorological data Perform a weighted summation to calculate the attention weights. ,in, For attention parameters, This represents vector concatenation. The activation function is, i.e. The weights are adjusted based on user preferences and meteorological factors. ,in , The influence coefficients are used to obtain the weighted feature matrix. ,in .

[0048] GCN features (weighted feature matrix) ) and user preference data Meteorological data splicing, the calculation formula is as follows ,in It is a 32-dimensional vector. It is a 32-dimensional vector. Obtain the fused feature matrix. The concatenation dimension was determined using 5-fold cross-validation. On the validation set, the prediction error of 128-dimensional features was reduced by 11.7% compared to 64-dimensional features. The fused feature matrix... Compared with the original data timestamp (Unit: milliseconds) are integrated and sorted by geographic region. and time window (i.e., 300 seconds) grouping, defining regions At any moment feature set Obtain a spatiotemporally structured dataset For spatiotemporal structured datasets Standardized encapsulation is performed, and each record is formatted as follows: ,in, A unique identifier for the node. Geographic region coding, The start time of the time window. Associative indexes with the original data. Obtain the merged multi-dimensional dataset. .

[0049] The demand forecasting model is used to analyze the fused multi-dimensional dataset. The data is processed to obtain a three-dimensional demand matrix of "time-region-electricity". The demand forecasting model is constructed using a hybrid graph neural network (GNN) and long short-term memory network (LSTM). The fused multi-dimensional dataset is then processed first. Perform time-series data processing, categorized by geographical region. and time For the fusion feature matrix Sort and construct regional temporal feature sequences ,in, For the region The number of time windows that satisfy , This aligns with the time window generated by the multi-dimensional dataset, yielding a set of regional temporal features. For the set of regional temporal features and adjacency matrix Feature extraction is performed. Both are input into the GNN spatial model, and the feature matrix at each time step is processed. The graph convolution calculation is performed using consistent GNN feature fusion. ,in, This is the weight matrix. For bias terms, The ReLU activation function is used. A graph attention mechanism (GAE) is introduced, which dynamically assigns weights to features of adjacent regions during graph convolution. A multi-head attention mechanism is used to enhance the expressive power of spatial associations; the number of heads is set to 4, and each head outputs a dimension of 16. This yields a sequence of regional spatial features. ,in .

[0050] Regional spatial feature sequences Input into the LSTM time series model, for each region of Input a 3-layer LSTM network (refer to the Keras standard LSTM architecture), the number of hidden units in the first layer. Returns the sequence, dropout rate The number of hidden units in the second layer Returns the sequence, dropout rate The number of hidden units in the third layer It does not return a sequence, but outputs time-series features. Residual connections are added between LSTM layers to alleviate the vanishing gradient problem in deep networks; the ratio of residual connections is set to 0.5. This yields a set of regional temporal features. Based on the regional time series feature set Demand probability distribution prediction is performed. This is done by mapping the fully connected layer to the charging demand probability distribution. ,in, ( =10, corresponding to the energy level in the 0-10kWh range. For bias terms, For temperature coefficient, when At that time, it will amplify the differences in high-value characteristics, making the peak value in the high-probability interval more prominent. The function ensures that the sum of probabilities equals 1. It obtains the required probability for a single region at a single time step. Using a sliding time window (step size) Repeated feature extraction and demand probability distribution prediction steps are used to obtain future... The demand probability of all regions at each time step (one day) is used to obtain a multi-region, multi-time-step demand probability set. Ultimately, a three-dimensional demand matrix of "time-region-power" was obtained. ,in Indicates the first time steps ( ), No. The first region, the first The probability of demand for each electricity level.

[0051] By capturing spatial correlations between regions using GNNs (such as the linkage between charging demand in commercial areas and surrounding residential areas) and combining this with LSTMs to uncover temporal patterns (periodic peak hours), the prediction error was reduced to below 8%. Ablation experiments showed that the MAE of the GNN model alone was 12.3%, the LSTM model alone was 10.7%, and the hybrid model was reduced to 7.9%, validating the effectiveness of spatiotemporal feature fusion. For example, in actual region 10, the 6-8 kWh demand at 18:05 accounted for 22%, while the predicted value was 20%, with an error of only 2%. The input was reused from the fused multi-dimensional dataset. To avoid data conversion losses, such as directly linking the spatial features of region 10 with traffic congestion data fused by edge computing, the prediction becomes more accurate, resulting in a three-dimensional demand matrix. This provides a basis for dispatching. For example, based on the high demand forecast for Region 10, three idle charging piles can be dispatched in advance from Region 9 (low demand), improving the response speed by 30%.

[0052] For example, a city is divided into =50 regions (encoding) =1 to =50), one fused feature data point is generated for each region every 5 minutes. (Based on region) =10 (commercial area) as an example, It includes its The eigenvector of =1620000000000 (May 1, 2021, 08:00:00) =[0.62,0.35,...,0.81] (128 dimensions, including normalized features such as charging pile power and surrounding traffic flow). Region =10 (Business District) was open from 08:00 to 18:00 on May 1, 2021. =120 time windows (one every 5 minutes) ,in Corresponding to 08:05. Regarding (08:00), area of With adjacent areas (Residential area) The characteristics of the (office area) are identified through an adjacency matrix. Association (hypothesis) (Detour distance 2.3km) (Direct distance 1.8km). After graph convolution calculation, (64 dimensions, filtering out invalid features, strengthening the spatial correlation feature of "increased demand for charging stations in commercial areas at 08:00"). Region =10 Incorporating spatial features across 120 time steps, the LSTM network learns the temporal patterns of peak charging demand during weekdays (08:00-10:00 and 18:00-20:00), and outputs... =[0.78,0.12,...,0.65] (16 dimensions, highlighting the time-series characteristics of morning and evening peak hours). Region =10 After calculation of the fully connected layer, the following is obtained: The probability of demand for =1620042000000 (18:05) =[0.02,0.03,0.05,0.10,0.15,0.20,0.18,0.12,0.08,0.07] indicates that the probability of a demand of 6-8 kWh of electricity is the highest (20%), which is consistent with the pattern of concentrated vehicle refueling during the evening rush hour in commercial areas. For the region... After the sliding calculation with a value of 10, the demand probability sequence from 18:05 to 18:00 the next day is obtained. Among them, the probability of high power demand from 18:05 to 20:00 is consistently high, while the probability of demand from 02:00 to 06:00 is concentrated in the low power level (0-2kWh, with a total probability of 35%). This indicates that the probability of demand for 6-7 kWh of electricity in region 10 at the first time step (18:05) is 20%.

[0053] According to the charging pile equipment number Regional coding (Consistent with the regional coding of the three-dimensional demand matrix), public key of the power grid dispatch center The charging station private key is generated using the Elliptic Curve Cryptography (ECC) algorithm. and public key Submitting to the blockchain network containing , , Registration and transactions are completed through consensus among all network nodes (using the Practical Byzantine Fault Tolerance (PBFT) algorithm, with a consensus latency of ≤500ms). A blockchain-based energy collaboration network is constructed, comprising charging pile clusters in 50 regions (20 units per region, totaling 1000 units), one power grid dispatch center node, and three regulatory nodes. A list of blockchain nodes is obtained. and network topology.

[0054] Based on the three-dimensional demand matrix Regional demand threshold (When the total probability of regional demand) The criteria for determining high demand and idle charging stations are: no charging tasks for 30 consecutive minutes and the battery level is low. 70%) and power grid load adjustment coefficient ( ), Blockchain Node List Smart contracts written in Solidity include demand judgment rules (if...). This area is a high-demand area, triggering a charging pile scheduling command. Scheduling priority (high-demand areas prioritize acquiring idle resources, while adjacent areas...) Idle charging pile scheduling weight , (Center distance of the region, unit: km), power adjustment rules (charging power) ,in (Maximum output power of the charging station) (Load Impact Coefficient). The contract is deployed after verification by the blockchain node, with the address ContractAddr. The smart contract calls the QueryIdle() function to retrieve idle charging piles and filters them based on the idle criteria to determine their real-time task status. And the charging station's real-time power level The device list is arranged according to the scheduling priority in the contract. Sort.

[0055] The priority of charging station requests is determined using a scheduling cost formula, and the prioritization is then reordered. The scheduling cost formula is as follows: , For charging piles Current location to target area Road network distance, This is a penalty item specific to charging stations, based on historical scheduling records. If the success rate of the last 3 scheduling executions is less than 90%, then... ,the remaining , The weighting coefficient for the penalty term. The higher the cost of invocation, the lower the invocation priority. For high-demand areas... Generate scheduling instructions ,in The coding for high-demand areas is consistent with the region coding in the three-dimensional demand matrix. ( This is a dynamic value; it varies during periods of congestion. When the flow is smooth ), When a secondary scheduling scheme is triggered, a suboptimal but faster-responding charging station is selected. The timestamp of the time the scheduling instruction was triggered. This is the estimated time stamp for the charging station to reach the target area.

[0056] According to the real-time load of the power grid Current power of charging pile Power adjustment rules, calculating load adjustment coefficients based on contract rules. When the load suddenly increases by more than 10%, a stepped power reduction strategy is adopted: a 20% reduction in the first minute, and a further 10% reduction in the second minute, to avoid grid impact. At that time, power reduction is triggered. ,when At that time, it is allowed Obtain dynamic charging power parameters According to the blockchain node list Dynamic charging power parameters Dispatch instructions via charging station private key After signing, the data is sent to the corresponding node, and the execution result is received. (Success / Failure) Uploaded to the blockchain for evidence storage, smart contract according to... Update status (Successfully) or trigger secondary scheduling. Obtain the instruction execution record with blockchain certification. Record the scheduling end-to-end latency; if the time stamp difference between demand triggering and charging pile response exceeds 15 seconds, it is automatically marked as an anomaly, triggering the node performance optimization process.

[0057] For example, at a certain moment 1620042000000 (18:05, corresponding to the...) (time step), region The three-dimensional demand matrix segment of the (commercial area) is as follows: Real-time load of power grid , .area charging stations (Number 101) generated and After registering with the blockchain, you will obtain a node ID. With the same region and power grid nodes This forms a subnet, providing the node foundation for the subsequent execution of scheduling instructions in this region. of , hour This meets the high-demand conditions. Based on the contract... Calculation rules, region of ( Priority is higher than region of ( ), generate instructions .

[0058] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0059] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0060] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A method for intelligent management of mobile charging stations, characterized in that, The method includes: Collect mobile charging pile operation data, user data, and environmental data, and fuse these data through edge computing nodes to obtain a normalized dataset; The normalized dataset is input into a graph neural network to construct the graph structure and obtain the initial graph structure. A graph convolutional network is then used to fuse the feature matrix and adjacency matrix in the initial graph structure to obtain the fused multidimensional dataset. The fused multi-dimensional dataset is input into the demand prediction model to predict the probability distribution of charging demand in different regions and at different times, thus obtaining a three-dimensional demand matrix. High-demand areas are identified based on the three-dimensional demand matrix. The three-dimensional demand matrix and the area codes of the high-demand areas are input into the scheduling cost formula to obtain the scheduling cost. The charging pile call priority is determined based on the scheduling cost. Scheduling instructions are generated based on the charging pile call priority and the area codes of the high-demand areas. The process of obtaining the normalized dataset is as follows: Adopt 3 The criteria are as follows: an outlier threshold is set, and if an outlier exceeds the threshold twice consecutively, it is considered an outlier and is removed. Missing values ​​are filled using linear interpolation to obtain a cleaned dataset. The cleaned data is normalized using the min-max standardization formula, and the location data is converted into planar coordinates using Gaussian projection to obtain a normalized dataset. The process of obtaining the three-dimensional demand matrix is ​​as follows: The fused multi-dimensional dataset is processed for temporal serialization. The fused feature matrix is ​​sorted by geographic region and time to construct a regional temporal feature set. Feature extraction is performed on the regional temporal feature set and adjacency matrix. The extracted features are input into a GNN spatial model. For the feature matrix at each time step, graph convolution with consistent GNN feature fusion is used to calculate the regional spatial feature sequence. The regional spatial feature sequence is input into an LSTM time series model. Residual connections are added between LSTM layers, and the sequence is mapped to the charging demand probability distribution through a fully connected layer to obtain the demand probability of a single region at a single time step. A three-dimensional demand matrix is ​​obtained based on the demand probability of a single region at a single time step. The process of generating the scheduling instruction is as follows: Using elliptic curve cryptography, public and private keys for charging piles are generated based on the charging pile device number, area code, and the public key of the power grid dispatch center. These keys are then submitted to the blockchain network for registration, and identity authentication is completed after consensus among all network nodes. Based on the Solidity language, a smart contract is written, incorporating regional demand thresholds from a three-dimensional demand matrix, idle charging pile criteria, power grid load adjustment coefficients, and a blockchain node list. The smart contract calls the QueryIdle() function to filter a list of charging piles with idle status and battery levels ≥70% according to the idle criteria, and determines the call priority using a scheduling cost formula. For high-demand areas, scheduling instructions are generated based on the area code and the estimated arrival time of the charging piles.

2. The intelligent management method for mobile charging piles according to claim 1, characterized in that, The specific process for collecting mobile charging pile operation data, user data, and environmental data is as follows: The positioning module adjusts the acquisition frequency according to the moving speed of the mobile charging pile to collect and record the real-time time and timestamp. It collects the real-time power through current and voltage sensors. The BMS dynamically adjusts the monitoring frequency according to the charging stage to monitor the remaining power of the mobile charging pile and sets a breakpoint resume mechanism. After user authorization, reservation information and evaluation feedback are collected, and sensitive information is processed using differential privacy technology; It connects to the traffic management department's API to collect traffic flow, average speed, and congestion index for each road segment, introduces road topology data, establishes a road network distance matrix, sets up a data caching mechanism, and collects parking space data through a geographic information system.

3. The intelligent management method for mobile charging piles according to claim 2, characterized in that, The process of obtaining the initial graph structure is as follows: Define charging pile nodes, traffic segment nodes, and parking lot nodes, obtain the node feature matrix, calculate the shortest path time of the road network using Dijkstra's algorithm, input the shortest path time into the edge weight calculation formula to calculate the edge weight, construct the edge set and adjacency matrix based on the edge weight, and construct the initial graph structure based on the node feature matrix, edge set, and adjacency matrix.

4. The intelligent management method for mobile charging piles according to claim 3, characterized in that, The process of obtaining the fused multi-dimensional dataset is as follows: The adjacency matrix is ​​input into the first layer of the graph convolutional network to obtain the first layer feature matrix. The first layer feature matrix, user preference data, and meteorological data are weighted and summed through an attention mechanism to calculate the attention weights and obtain a weighted feature matrix. The weighted feature matrix is ​​then concatenated with the user preference data and meteorological data to obtain a fused feature matrix. The concatenation dimension is determined through 5-fold cross-validation. The fused feature matrix is ​​integrated with the original data timestamps to obtain a spatiotemporal structured dataset; the spatiotemporal structured dataset is then standardized and encapsulated to obtain a fused multidimensional dataset.

5. The intelligent management method for mobile charging piles according to claim 4, characterized in that, The process for determining the priority of charging pile usage is as follows: The total probability of high power level demand in a certain region at a certain time step is calculated based on the three-dimensional demand matrix. If the total probability of high power level demand is greater than or equal to the regional demand threshold, it is determined to be a high demand region. The region code corresponding to the total demand probability of the high demand region is input into the scheduling cost formula to calculate the scheduling cost of each mobile charging pile. The calling priority of all mobile charging piles is sorted according to the calling cost. Calculate the time required for each mobile charging station to reach the high-demand area. If the time required for a high-priority mobile charging station is greater than 30 minutes, select the next-priority mobile charging station.

6. A mobile charging pile intelligent management system, characterized in that, The system is used to execute the intelligent management method for mobile charging piles as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is executed by a processor to implement the intelligent management method for mobile charging piles as described in any one of claims 1-5.

Images