An intelligent charging scheduling system and resource allocation method

Abstract

This invention belongs to the field of charging pile technology and provides an intelligent charging scheduling system and resource allocation method. The method includes: data collection, energy distribution model prediction, charging pile transportation, energy supply and demand ratio calculation, charging pile movement plan determination, and charging pile recovery. The present invention generates an energy scheduling plan based on energy demand data, weather data, road condition data, and an energy distribution model. This allows energy distribution to be designed based on historical and real-time data, improving the rationality and accuracy of the distribution. By calculating the energy supply and demand ratio and determining the charging pile movement plan, energy transfer in emergency situations is achieved, improving the system's fault tolerance.

Description

Technical Field

[0001] The present invention relates to the technical field of charging piles, and in particular to an intelligent charging scheduling system and a resource allocation method. Background Art

[0002] Mobile charging station technology has experienced rapid development in recent years, primarily in terms of intelligence, modularity, and convenience. With the growing demand for electric vehicles, mobile charging stations have gradually implemented features such as fast battery charging, remote monitoring, and data analysis. Through cloud platforms and apps, users can view charging status in real time and make reservations. New mobile charging stations also feature more flexible power scheduling, automatic vehicle recognition, and support for multiple charging interfaces. Furthermore, based on connected vehicle technology, some charging stations can intelligently interact with electric vehicles to optimize the charging process and manage energy. Overall, mobile charging stations demonstrate significant potential for application in scenarios such as urban transportation and long-distance travel.

[0003] However, mobile charging station technology faces challenges in energy distribution and scheduling. Due to the uncertainty and suddenness of charging demand, how to reasonably allocate limited power resources in real time is a major problem. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide an intelligent charging scheduling system and resource allocation method to achieve optimal scheduling and allocation of mobile charging piles.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] An intelligent charging scheduling system, comprising:

[0007] The data collection module is used to collect user order information, target city weather information, and target road traffic information to obtain energy demand data, weather data, and road condition data;

[0008] An energy allocation module, configured to input the energy demand data, the weather data, and the road condition data within a dynamic time window into a pre-trained energy allocation model for prediction, thereby obtaining an energy scheduling plan;

[0009] The charging pile transportation module is used to transport the mobile charging pile to the target charging node using a prepared mobile charging vehicle according to the energy scheduling plan, and place the mobile charging pile at the charging pile position of the target charging node;

[0010] A real-time statistics module, configured to calculate the energy supply-demand ratio of each charging node based on the energy demand data;

[0011] a rapid adjustment module, configured to, when the energy supply-demand ratio is lower than a preset normal power supply threshold, determine a charging pile movement plan based on a preset charging pile screening model, and use the mobile charging vehicle to move the mobile charging pile corresponding to the charging pile movement plan to a charging node where the energy supply-demand ratio is lower than the preset normal power supply threshold;

[0012] The charging pile recovery module is used to use the mobile charging vehicle to transport the mobile charging piles whose power levels in the charging nodes within the target range are lower than the normal power threshold back to the charging station for power replenishment when the mobile charging vehicle returns to the charging station.

[0013] Preferably, a smart charging resource allocation method includes:

[0014] The user's order information, the target city's weather information, and the target road section's traffic information are collected to obtain the energy demand data, the weather data, and the road condition data;

[0015] Inputting the energy demand data, the weather data, and the road condition data within a dynamic time window into the pre-trained energy allocation model for prediction to obtain the energy scheduling plan;

[0016] Transporting the mobile charging pile to the target charging node using the prepared mobile charging vehicle according to the energy scheduling plan, and placing the mobile charging pile at the charging pile position of the target charging node;

[0017] Calculating the energy supply-demand ratio of each charging node based on the energy demand data;

[0018] When the energy supply-demand ratio is lower than a preset normal power supply threshold, a charging pile movement plan is determined according to the preset charging pile screening model, and the mobile charging vehicle is used to move the mobile charging pile corresponding to the charging pile movement plan to a charging node where the energy supply-demand ratio is lower than the preset normal power supply threshold;

[0019] When the mobile charging vehicle returns to the charging station, the mobile charging piles whose power levels are lower than a normal power threshold among the charging nodes within the target range are transported by the mobile charging vehicle back to the charging station for power replenishment.

[0020] Preferably, the training process of the energy distribution model includes:

[0021] Collecting operating data of the study area and dividing the operating data into historical data and dynamic window data according to chronological order; the operating data includes: the energy demand data, the weather data, and the road condition data;

[0022] Determining energy redundancy data of the study area according to the dynamic window data, and determining an optimal energy deployment plan for the study area using the energy redundancy data;

[0023] Determine the locations of charging nodes in the study area to obtain static features;

[0024] Build a TFT network including encoder and decoder;

[0025] Constructing the loss function of the TFT network based on energy gap and energy redundancy;

[0026] The historical contemporaneous data, the dynamic window data, and the static features are used as network inputs, the optimal energy deployment plan is used as the network target output, and the TFT network is trained using the loss function according to sliding window sampling and early stopping strategy to obtain the trained energy allocation model.

[0027] Preferably, according to the energy scheduling plan, the mobile charging pile is transported to the target charging node by using a prepared mobile charging vehicle, and the mobile charging pile is placed at the charging pile position of the target charging node, including:

[0028] Using an image recognition algorithm and a laser sensor on the mobile charging vehicle, the mobile charging vehicle is transported back to the charging station, and the mobile charging pile is transferred to a preset guide rail module using a camera and a robotic arm device. The modular battery on the mobile charging pile is transferred to the guide rail module using the robotic arm device;

[0029] The guide rail module is used to transport the spare battery with a power ratio of 100% to the target location, the robotic arm device is used to install the spare battery on the mobile charging pile, and the robotic arm device is used to place the mobile charging pile on an empty position on the mobile charging vehicle according to the laser sensor;

[0030] When the number of mobile charging piles brought back to the charging station by the mobile charging vehicle is less than the full load, after the batteries of all the brought-back mobile charging piles have been replaced, the mobile charging piles remaining in the charging station are placed in the vacant positions on the mobile charging vehicle using the robotic arm device;

[0031] Determining a target node according to the energy scheduling plan;

[0032] Laser navigation technology is used to control the mobile charging vehicle to reach the target node, and the position sensor on the target node and the robotic arm device are used to fix the mobile charging pile on the mobile charging vehicle to the charging pile placement area of the target node.

[0033] Preferably, the energy supply-demand ratio is calculated as follows: ;

[0034] in, is the calculated value of the energy supply-demand ratio; is the available power of the i-th mobile charging pile; is the charging efficiency of the i-th mobile charging pile; is the time attenuation coefficient; The transportation time for the replacement charging pile to reach the i-th mobile charging pile; is the available time attenuation coefficient; Predicting the remaining battery life of the i-th mobile charging station; is the original predicted power demand; is the weather weight; is the weather influencing factor; is the traffic weight; is the traffic congestion factor; is the vehicle quantity weight; It is the predicted value of the number of charging vehicles in the future time window.

[0035] Preferably, the calculation formula of the electric pile screening model is: ;

[0036] in, ; Score the screening; They are time weight, busy weight, and power weight respectively; is the round trip time; is the node busyness; The remaining power of the mobile charging pile; It is a fixed value.

[0037] Preferably, the training process of the energy distribution model further includes:

[0038] freezing the decoder of the energy distribution model during a preset low-peak charging period;

[0039] The encoder and attention weight of the energy allocation model are adjusted using the newly collected dynamic window data to obtain the energy allocation model after online update.

[0040] Preferably, the historical contemporaneous data, the dynamic window data, and the static features are used as network inputs, the optimal energy deployment plan is used as the network target output, and the TFT network is trained using the loss function according to the sliding window sampling and early stopping strategy to obtain the trained energy allocation model, including:

[0041] Standardizing and concatenating the historical contemporaneous data, the dynamic window data, and the static features to obtain an input sequence;

[0042] Extracting temporal features of the input sequence using a gated recurrent unit of the encoder to obtain an encoded latent state;

[0043] Performing multi-head self-attention mechanism calculation, residual connection and normalization processing on the encoded hidden state to obtain an encoded output;

[0044] Using the self-attention mechanism and cross-attention mechanism of the decoder to locally model and interact with the encoded output to obtain a decoded hidden state;

[0045] Mapping the decoded hidden state using the fully connected layer of the TFT network to obtain a network prediction matrix;

[0046] The loss function is calculated according to the network prediction matrix to obtain a loss value, and the network parameters of the TFT network are optimized according to the loss value, and the process returns to the step of "using the gated recurrent unit of the encoder to extract the temporal features of the input sequence to obtain the encoded latent state" to obtain the trained energy allocation model.

[0047] Preferably, the loss function is expressed as: ;

[0048] in, ; is the calculated value of the loss function; are the first penalty weight and the second penalty weight respectively; for Regularization coefficient; is the energy gap at time step t; is the energy redundancy at time step t; is the indicator function; is the time window length; are the model parameters of the TFT network.

[0049] Preferably, the detection implementation process of the laser sensor includes:

[0050] Fixing the laser sensor in the cargo box of the mobile charging vehicle according to a preset placement space layout;

[0051] Utilizing the laser sensor to emit detection laser;

[0052] When the laser detection device corresponding to the laser sensor detects the detection laser, determining the placement state as not placed;

[0053] When the laser detection device does not detect the detection laser, the placement state is determined to be placed.

[0054] The present invention discloses the following technical effects:

[0055] The present invention provides an intelligent charging scheduling system and resource allocation method, which generates an energy scheduling plan through energy demand data, weather data, road condition data and an energy distribution model, solving the problem of poor energy scheduling and distribution in the existing technology, and realizing the design of energy distribution based on historical data and real-time data; through the calculation of energy supply and demand ratio and the determination of charging pile movement plan, it solves the problem of poor ability of the existing technology to deal with emergencies, and realizes energy transfer in emergency situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 A schematic diagram of an intelligent charging scheduling system provided by an embodiment of the present invention;

[0058] Figure 2 A schematic diagram of the intelligent charging resource allocation process provided by an embodiment of the present invention;

[0059] Figure 3 A schematic diagram of the energy allocation model training process provided by an embodiment of the present invention;

[0060] Figure 4 A schematic diagram of the transportation process of an electric charging pile provided in an embodiment of the present invention;

[0061] Figure 5 A schematic diagram of the TFT network iteration process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0063] The purpose of the present invention is to provide an intelligent charging scheduling system and resource allocation method to achieve optimal scheduling and allocation of mobile charging piles.

[0064] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0065] Figure 1 A schematic diagram of an intelligent charging scheduling system provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides an intelligent charging scheduling system, comprising:

[0066] The data collection module is used to collect user order information, target city weather information, and target road traffic information to obtain energy demand data, weather data, and road condition data;

[0067] The energy allocation module is used to input energy demand data, weather data, and road condition data within a dynamic time window into a pre-trained energy allocation model for prediction and obtain an energy scheduling plan;

[0068] The charging pile transportation module is used to transport the mobile charging pile to the target charging node using the prepared mobile charging vehicle according to the energy scheduling plan, and place the mobile charging pile at the charging pile position of the target charging node;

[0069] A real-time statistics module is used to calculate the energy supply-demand ratio of each charging node based on energy demand data;

[0070] A rapid adjustment module is used to determine a charging pile movement plan based on a preset charging pile screening model when the energy supply-demand ratio is lower than a preset normal power supply threshold, and use a mobile charging vehicle to move the mobile charging pile corresponding to the charging pile movement plan to a charging node where the energy supply-demand ratio is lower than the preset normal power supply threshold;

[0071] The charging pile recovery module is used to transport the mobile charging piles whose power levels are lower than the normal threshold in the charging nodes within the target range back to the charging station for power replenishment when the mobile charging vehicle returns to the charging station.

[0072] refer to Figure 2 , a smart charging resource allocation method, comprising:

[0073] Step 100: Collect the user's order information, the target city's weather information, and the target road section's traffic information to obtain energy demand data, weather data, and road condition data;

[0074] Step 200: Inputting energy demand data, weather data, and road condition data within a dynamic time window into a pre-trained energy allocation model for prediction to obtain an energy scheduling plan;

[0075] Step 300: transporting the mobile charging pile to the target charging node using a prepared mobile charging vehicle according to the energy scheduling plan, and placing the mobile charging pile at the charging pile position of the target charging node;

[0076] Step 400: Calculate the energy supply-demand ratio of each charging node based on the energy demand data;

[0077] Step 500: When the energy supply-demand ratio is lower than a preset power supply normal threshold, a charging pile relocation plan is determined according to a preset charging pile screening model, and a mobile charging vehicle is used to relocate the mobile charging pile corresponding to the charging pile relocation plan to a charging node where the energy supply-demand ratio is lower than the preset power supply normal threshold;

[0078] Step 600: When the mobile charging vehicle returns to the charging station, the mobile charging piles whose power levels are lower than a normal power threshold among the charging nodes within the target range are transported back to the charging station for power replenishment.

[0079] refer to Figure 3 ,The training process of the energy allocation model includes:

[0080] Step 201: Collecting operating data of the study area and dividing the operating data into historical data and dynamic window data according to chronological order; the operating data includes: energy demand data, weather data, and road condition data;

[0081] Step 202: determining energy redundancy data of the study area based on the dynamic window data, and using the energy redundancy data to determine an optimal energy deployment plan for the study area;

[0082] Step 203: Determine the locations of charging nodes in the study area and obtain static features;

[0083] Step 204: Construct a TFT network including an encoder and a decoder;

[0084] Step 205: constructing a loss function of the TFT network based on the energy gap and energy redundancy;

[0085] Step 206: Using historical data from the same period, dynamic window data, and static features as network inputs and the optimal energy deployment plan as the network target output, the TFT network is trained using a loss function based on sliding window sampling and early stopping strategies to obtain a trained energy allocation model.

[0086] refer to Figure 4 , according to the energy scheduling plan, use the prepared mobile charging vehicle to transport the mobile charging pile to the target charging node, and place the mobile charging pile at the charging pile position of the target charging node, including:

[0087] Step 301: Using an image recognition algorithm and a laser sensor on the mobile charging vehicle, the camera and robotic arm device are used to transport the mobile charging vehicle back to the mobile charging pile of the charging station and transfer it to a preset guide rail module. The robotic arm device is also used to transfer the modular battery on the mobile charging pile to the guide rail module.

[0088] Step 302: Use the guide rail module to transport the spare battery with a 100% charge ratio to the target location, use the robotic arm device to install the spare battery on the mobile charging pile, and use the robotic arm device to place the mobile charging pile on an empty position on the mobile charging vehicle based on the laser sensor;

[0089] Step 303: When the number of mobile charging piles returned to the charging station by the mobile charging vehicle is less than the full load, after the batteries of all the returned mobile charging piles have been replaced, the mobile charging piles remaining in the charging station are placed in the vacant positions on the mobile charging vehicle using a robotic arm device;

[0090] Step 304: Determine the target node according to the energy scheduling plan;

[0091] Step 305: Use laser navigation technology to control the mobile charging vehicle to reach the target node, and use the position sensor and mechanical arm device on the target node to fix the mobile charging pile on the mobile charging vehicle to the charging pile placement area of the target node.

[0092] Furthermore, the energy supply-demand ratio is calculated as: ;

[0093] in, is the calculated value of the energy supply-demand ratio; is the available power of the i-th mobile charging pile; is the charging efficiency of the i-th mobile charging pile; is the time attenuation coefficient; The transportation time for the replacement charging pile to reach the i-th mobile charging pile; is the available time attenuation coefficient; Predicting the remaining battery life of the i-th mobile charging station; is the original predicted power demand; is the weather weight; is the weather influencing factor; is the traffic weight; is the traffic congestion factor; is the vehicle quantity weight; It is the predicted value of the number of charging vehicles in the future time window.

[0094] Specifically, the calculation formula of the electric pile screening model is: ;

[0095] in, ; Score the screening; They are time weight, busy weight, and power weight respectively; is the round trip time; is the node busyness; The remaining power of the mobile charging pile; It is a fixed value.

[0096] Preferably, the training process of the energy distribution model further includes:

[0097] Freeze the decoder of the energy distribution model during the preset low-peak charging period;

[0098] The encoder and attention weights of the energy allocation model are adjusted using the newly collected dynamic window data to obtain an online updated energy allocation model.

[0099] refer to Figure 5 , using historical data, dynamic window data, and static features as network inputs, and the optimal energy deployment plan as the network target output, the TFT network is trained using a loss function based on sliding window sampling and early stopping strategies to obtain a trained energy allocation model, including:

[0100] Step 20601: Standardize and concatenate historical contemporaneous data, dynamic window data, and static features to obtain an input sequence;

[0101] Step 20602: Use the gated recurrent unit of the encoder to extract temporal features of the input sequence to obtain the encoded latent state;

[0102] Step 20603: Perform multi-head self-attention mechanism calculation, residual connection, and normalization on the encoded hidden state to obtain the encoded output;

[0103] Step 20604: Use the decoder's self-attention mechanism and cross-attention mechanism to locally model and interact with the encoded output to obtain the decoded hidden state;

[0104] Step 20605: Map the decoded hidden state using the fully connected layer of the TFT network to obtain a network prediction matrix;

[0105] Step 20606: Calculate the loss function based on the network prediction matrix to obtain the loss value, and optimize the network parameters of the TFT network based on the loss value, and return to the step "Use the encoder's gated recurrent unit to extract the temporal features of the input sequence to obtain the encoded hidden state" to obtain the trained energy allocation model.

[0106] Specifically, the loss function is expressed as: ;

[0107] in, ; is the calculated value of the loss function; are the first penalty weight and the second penalty weight respectively; for Regularization coefficient; is the energy gap at time step t; is the energy redundancy at time step t; is the indicator function; is the time window length; are the model parameters of the TFT network.

[0108] Preferably, the detection implementation process of the laser sensor includes:

[0109] Fix the laser sensor in the cargo box of the mobile charging vehicle according to the preset placement space layout;

[0110] Utilizing a laser sensor to emit a detection laser;

[0111] When the laser detection device corresponding to the laser sensor detects the detection laser, the placement state is determined to be not placed;

[0112] When the laser detection device does not detect the detection laser, the placement state is determined to be placed.

[0113] Specifically, historical data for the same period: data from the same time period in the past (such as the same period of the previous month, the time interval is less than 1 year to ensure the stability of the model) is selected to capture periodic patterns; dynamic window data: real-time data from the recent period (1 hour or 15 minutes) is extracted to reflect current dynamic changes.

[0114] Furthermore, energy gap: demand > supply; energy redundancy: supply > demand.

[0115] Preferably, the loss function includes three terms: an energy gap term, an energy redundancy term, and a regularization term. Only one of the energy gap and energy redundancy terms is retained, determined by an indicator function whose value is 1 if the conditions are met and 0 otherwise. This embodiment considers that energy shortages have a more severe impact on the user experience than energy redundancy. When designing the loss function, a greater weight is assigned to the energy gap term than to the energy redundancy term, prioritizing the reduction of power shortages. Furthermore, the regularization term is used to prevent parameter overfitting during training.

[0116] Specifically, the charging pile recycling and battery replacement process is as follows:

[0117] 1) Image recognition and laser detection:

[0118] After returning to the charging station, the mobile charging vehicle uses its onboard camera to capture images of the cargo box's interior and uses image recognition algorithms (such as the deep learning-based YOLO model) to identify the location of the mobile charging station requiring battery replacement. A laser sensor installed in the cargo box then emits a detection laser beam. If the laser receiving end is not obstructed (the laser is detected), the charging station is determined to be properly positioned. Conversely, if the laser is obstructed, the charging station is determined to be secure.

[0119] 2) Robotic arm operation:

[0120] Based on the recognition results, the robotic arm grabs the unsecured charging pile and transfers it to the pre-set guide rail module. The guide rail module uses a slide rail design that supports horizontal and vertical movement to ensure the precise positioning of the charging pile.

[0121] 3) Battery replacement process:

[0122] Old battery unloading: The robotic arm removes the modular battery from the charging station and transports it to the charging station's battery recycling area via the guide rail module. New battery installation: The guide rail module transports the spare battery with 100% charge to the target location. The robotic arm grabs the new battery and installs it on the charging station. Laser sensors verify the installation status in real time to ensure that the battery contacts are fully connected.

[0123] Furthermore, charging truck loading is optimized. If the mobile charging truck returns underloaded, a robotic arm will refill the empty space in the cargo box with a fully charged charger reserved at the charging station. The cargo box features a honeycomb-like layout, and laser sensors monitor empty spaces in real time. The robotic arm then completes loading based on the optimal path planning.

[0124] Target node navigation and charging station deployment are optimized. Path Planning and Navigation: Laser Navigation Technology: Mobile charging vehicles use onboard LiDAR (laser radar) to scan the surrounding environment and generate a navigation path in real time based on high-precision maps. Dynamic Obstacle Avoidance: If an obstacle is detected, the path planning algorithm automatically adjusts the route to ensure timely arrival at the target charging node.

[0125] Furthermore, the charging piles are precisely placed. Position sensor calibration: RFID tags or infrared beacons are pre-embedded in the charging pile placement area of the target charging node, and the mobile charging vehicle identifies the tag location through the on-board sensor. Robotic arm fixed operation: After the robotic arm grabs the charging pile from the cargo box, it adjusts the angle through the six-axis freedom to align the charging pile base with the card slot in the node placement area, and uses the pressure sensor to detect the embedded status of the card slot to ensure that the charging pile is firmly installed. If the charging pile is a movable device, then after it is placed at the target location, the moving wheels of the charging pile are controlled to retract to increase the contact area between the charging pile and the ground.

[0126] Optionally, if a mobile charger is delayed due to traffic congestion or mechanical failure, the system automatically triggers a priority adjustment, prioritizing charging pile resources to nodes with urgent needs. If battery installation fails (e.g., poor contact), the robotic arm automatically retries three times. If failure persists, an alarm is triggered, requiring manual intervention.

[0127] The beneficial effects of the present invention are as follows:

[0128] The present invention generates an energy scheduling plan through energy demand data, weather data, road condition data and energy distribution models, realizes the design of energy distribution based on historical data and real-time data, and improves the rationality and accuracy of distribution; through the calculation of energy supply and demand ratio and the determination of charging pile movement plan, energy transfer in emergency situations is realized, and the fault tolerance of the system is improved.

[0129] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0130] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A smart charging resource allocation method, characterized in that: The system is applied to an intelligent charging scheduling system, comprising: a data acquisition module for respectively collecting user order information, target city weather information, and target road section traffic information to obtain energy demand data, weather data, and road condition data; an energy allocation module for inputting the energy demand data, weather data, and road condition data within a dynamic time window into a pre-trained energy allocation model for prediction to obtain an energy scheduling plan; and a charging pile transportation module for transporting mobile charging piles to target charging nodes using prepared mobile charging vehicles according to the energy scheduling plan, and placing the mobile charging piles at the charging pile locations of the target charging nodes. A real-time statistics module is used to calculate the energy supply-demand ratio of each charging node based on the energy demand data; a rapid adjustment module is used to determine a charging pile movement plan based on a preset charging pile screening model when the energy supply-demand ratio is lower than a preset normal power supply threshold, and use the mobile charging vehicle to transfer the mobile charging pile corresponding to the charging pile movement plan to a charging node with an energy supply-demand ratio lower than the preset normal power supply threshold; a charging pile recovery module is used to use the mobile charging vehicle to transport the mobile charging piles with power levels lower than the normal power threshold in the charging nodes within the target range back to the charging station for power replenishment when the mobile charging vehicle returns to the charging station; The method comprises: The user's order information, the target city's weather information, and the target road section's traffic information are collected to obtain the energy demand data, the weather data, and the road condition data; Inputting the energy demand data, the weather data, and the road condition data within a dynamic time window into the pre-trained energy allocation model for prediction to obtain the energy scheduling plan; Transporting the mobile charging pile to the target charging node using the prepared mobile charging vehicle according to the energy scheduling plan, and placing the mobile charging pile at the charging pile position of the target charging node; Calculating the energy supply-demand ratio of each charging node based on the energy demand data; When the energy supply-demand ratio is lower than a preset normal power supply threshold, a charging pile movement plan is determined according to the preset charging pile screening model, and the mobile charging vehicle is used to move the mobile charging pile corresponding to the charging pile movement plan to a charging node where the energy supply-demand ratio is lower than the preset normal power supply threshold; When the mobile charging vehicle returns to the charging station, the mobile charging piles whose power levels are lower than a normal power threshold among the charging nodes within the target range are transported by the mobile charging vehicle back to the charging station for power replenishment; The training process of the energy allocation model includes: Collecting operating data of the study area and dividing the operating data into historical data and dynamic window data according to chronological order; the operating data includes: the energy demand data, the weather data, and the road condition data; Determining energy redundancy data of the study area according to the dynamic window data, and determining an optimal energy deployment plan for the study area using the energy redundancy data; Determine the locations of charging nodes in the study area to obtain static features; Build a TFT network including encoder and decoder; Constructing the loss function of the TFT network based on energy gap and energy redundancy; The historical contemporaneous data, the dynamic window data, and the static features are used as network inputs, the optimal energy deployment plan is used as the network target output, and the TFT network is trained using the loss function according to sliding window sampling and early stopping strategy to obtain the trained energy allocation model; The expression of the loss function is: ； in, ; is the calculated value of the loss function; 、 are the first penalty weight and the second penalty weight respectively; is the L2 regularization coefficient; is the energy gap at time step t; is the energy redundancy at time step t; is the characteristic function; is the time window length; are the model parameters of the TFT network.

2. The intelligent charging resource allocation method according to claim 1, characterized in that: The method includes: transporting a mobile charging pile to a target charging node using a prepared mobile charging vehicle according to the energy scheduling plan, and placing the mobile charging pile at the charging pile position of the target charging node, including: Using an image recognition algorithm and a laser sensor on the mobile charging vehicle, the mobile charging vehicle is transported back to the charging station, and the mobile charging pile is transferred to a preset guide rail module using a camera and a robotic arm device. The modular battery on the mobile charging pile is transferred to the guide rail module using the robotic arm device; The guide rail module is used to transport the spare battery with a power ratio of 100% to the target location, the robotic arm device is used to install the spare battery on the mobile charging pile, and the robotic arm device is used to place the mobile charging pile on an empty position on the mobile charging vehicle according to the laser sensor; When the number of mobile charging piles brought back to the charging station by the mobile charging vehicle is less than the full load, after the batteries of all the brought-back mobile charging piles have been replaced, the mobile charging piles remaining in the charging station are placed in the vacant positions on the mobile charging vehicle using the robotic arm device; Determining a target node according to the energy scheduling plan; Laser navigation technology is used to control the mobile charging vehicle to reach the target node, and the position sensor on the target node and the robotic arm device are used to fix the mobile charging pile on the mobile charging vehicle to the charging pile placement area of the target node.

3. The intelligent charging resource allocation method according to claim 1, characterized in that: The energy supply-demand ratio is calculated as follows: ； in, is the calculated value of the energy supply-demand ratio; is the available power of the i-th mobile charging pile; is the charging efficiency of the i-th mobile charging pile; is the time attenuation coefficient; The transportation time for the replacement charging pile to reach the i-th mobile charging pile; is the available time attenuation coefficient; Predicting the remaining battery life of the i-th mobile charging station; is the original predicted power demand; is the weather weight; is the weather influencing factor; is the traffic weight; is the traffic congestion factor; is the vehicle quantity weight; It is the predicted value of the number of charging vehicles in the future time window.

4. The intelligent charging resource allocation method according to claim 1, characterized in that: The calculation formula of the electric pile screening model is: ； in, ; Score the screening; 、 、 They are time weight, busy weight, and power weight respectively; is the round trip time; is the node busyness; The remaining power of the mobile charging pile; It is a fixed value.

5. The intelligent charging resource allocation method according to claim 1, characterized in that: The training process of the energy allocation model further includes: freezing the decoder of the energy distribution model during a preset low-peak charging period; The encoder and attention weight of the energy allocation model are adjusted using the newly collected dynamic window data to obtain the energy allocation model after online update.

6. The intelligent charging resource allocation method according to claim 1, characterized in that: The historical contemporaneous data, the dynamic window data, and the static features are used as network inputs, the optimal energy deployment plan is used as the network target output, and the TFT network is trained using the loss function according to sliding window sampling and early stopping strategy to obtain the trained energy allocation model, including: Standardizing and concatenating the historical contemporaneous data, the dynamic window data, and the static features to obtain an input sequence; Extracting temporal features of the input sequence using a gated recurrent unit of the encoder to obtain an encoded latent state; Performing multi-head self-attention mechanism calculation, residual connection and normalization processing on the encoded hidden state to obtain an encoded output; Using the self-attention mechanism and cross-attention mechanism of the decoder to locally model and interact with the encoded output to obtain a decoded hidden state; Mapping the decoded hidden state using the fully connected layer of the TFT network to obtain a network prediction matrix; The loss function is calculated according to the network prediction matrix to obtain a loss value, and the network parameters of the TFT network are optimized according to the loss value. Then, the process returns to the step of "using the gated recurrent unit of the encoder to extract the temporal features of the input sequence to obtain the encoded hidden state" to obtain the trained energy allocation model.

7. The intelligent charging resource allocation method according to claim 2, characterized in that: The detection implementation process of the laser sensor includes: Fixing the laser sensor in the cargo box of the mobile charging vehicle according to a preset placement space layout; Utilizing the laser sensor to emit detection laser; When the laser detection device corresponding to the laser sensor detects the detection laser, determining the placement state as not placed; When the laser detection device does not detect the detection laser, the placement state is determined to be placed.

Images