Dynamic charging pile configuration method and system for intelligent parking lot

Abstract

The application discloses a dynamic charging pile configuration method and system of an intelligent parking lot, and relates to the technical field of charging pile configuration. The method comprises the following steps: collecting multi-source data of the parking lot in real time; predicting the charging demand space-time distribution of each area of the parking lot in a future period of time through a demand prediction model; taking the minimization of the total waiting time of users, the load fluctuation of the power grid and the cost of operators as the optimization target, performing dynamic configuration optimization solution of the charging pile, and outputting the scheduling path of the mobile charging unit and the power allocation scheme of the fixed charging pile; and executing the scheduling path and the power allocation scheme, and dynamically adjusting the matching relationship between the charging resources and the parking resources. The technical problems of the prior art, such as the fixed charging pile configuration mode, the inability to dynamically adjust the charging pile configuration mode according to the real-time parking and charging demand changes, the low resource utilization rate and the long user waiting time, are solved, and the technical effects of realizing the dynamic intelligent configuration of the charging pile, reducing the user waiting time, improving the operation efficiency and the resource utilization rate are achieved.

Description

Technical Field

[0001] This invention relates to the field of charging pile configuration technology, specifically to a dynamic charging pile configuration method and system for intelligent parking lots. Background Technology

[0002] With the rapid popularization of new energy vehicles, parking lots have gradually become important locations for centralized parking and charging of electric vehicles in cities. However, the current configuration of charging piles in parking lots is mainly based on a fixed number and fixed power, lacking the ability to comprehensively analyze multi-dimensional data such as parking space occupancy status, differences in vehicle battery charge, expected dwell time, and historical charging behavior. In actual operation, due to the randomness of vehicle arrival times and the obvious spatiotemporal imbalance of charging demand, some areas have idle charging resources while others experience queuing and congestion. At the same time, concentrated charging during peak hours may also cause fluctuations in local power grid load, increasing the operational risks and costs of the power distribution system. Summary of the Invention

[0003] This application provides a method and system for configuring dynamic charging piles in intelligent parking lots, which solves the technical problems in the prior art where the configuration of charging piles is fixed and cannot be dynamically adjusted according to changes in real-time parking and charging demand, resulting in low resource utilization and long user waiting time.

[0004] The first aspect of this application provides a method for configuring dynamic charging piles in a smart parking lot, the method comprising:

[0005] Real-time collection of multi-source data from parking lots, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns; based on this multi-source data, a demand forecasting model predicts the spatiotemporal distribution of charging demand in various areas of the parking lot during future periods; according to the spatiotemporal distribution of charging demand and parking lot resource constraints, with the optimization objectives of minimizing total user waiting time, grid load fluctuations, and operator costs, dynamic configuration optimization of charging piles is performed, outputting the scheduling path of mobile charging units and the power allocation scheme of fixed charging piles; the scheduling path and power allocation scheme are executed to dynamically adjust the matching relationship between charging resources and parking resources.

[0006] A second aspect of this application provides a dynamic charging pile configuration system for intelligent parking lots, the system comprising:

[0007] Data Acquisition Module: Collects multi-source data from the parking lot in real time, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns; Demand Forecasting Module: Based on the multi-source data, predicts the spatiotemporal distribution of charging demand in various areas of the parking lot during future periods using a demand forecasting model; Solution Module: Based on the spatiotemporal distribution of charging demand and parking lot resource constraints, optimizes the dynamic configuration of charging piles with the goal of minimizing total user waiting time, grid load fluctuations, and operator costs, and outputs the scheduling path for mobile charging units and the power allocation scheme for fixed charging piles; Adjustment Module: Executes the scheduling path and power allocation scheme to dynamically adjust the matching relationship between charging resources and parking resources.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, multi-source data from the parking lot is collected in real time, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns. Next, based on this multi-source data, a demand forecasting model predicts the spatiotemporal distribution of charging demand in different areas of the parking lot during future periods. Then, based on the spatiotemporal distribution of charging demand and parking lot resource constraints, with the optimization objectives of minimizing total user waiting time, grid load fluctuations, and operator costs, a dynamic configuration optimization solution for charging piles is performed, outputting the scheduling path for mobile charging units and the power allocation scheme for fixed charging piles. Finally, the scheduling path and power allocation scheme are executed to dynamically adjust the matching relationship between charging resources and parking resources. This solves the technical problems of existing technologies where the fixed configuration of charging piles cannot be dynamically adjusted according to real-time changes in parking and charging demand, resulting in low resource utilization and long user waiting times. It achieves the technical effect of realizing dynamic intelligent configuration of charging piles, reducing user waiting time, and improving operational efficiency and resource utilization. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of the dynamic charging pile configuration method for intelligent parking lots provided in this application embodiment;

[0012] Figure 2 This is a schematic diagram of the dynamic charging pile configuration system for an intelligent parking lot provided in an embodiment of this application.

[0013] Figure labeling: Data acquisition module 11, demand forecasting module 12, solution module 13, adjustment module 14. Detailed Implementation

[0014] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0015] Example 1, as Figure 1 As shown, this application provides a method for configuring dynamic charging piles in a smart parking lot, wherein the method includes:

[0016] Real-time collection of multi-source data from parking lots, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns.

[0017] In this embodiment, a parking space status detection module, a vehicle information interaction module, and a data aggregation server are deployed in the parking lot. The parking space status detection module includes a geomagnetic sensor, an ultrasonic sensor, or a video recognition camera, used to collect the occupancy status information of each parking space at a preset sampling period (preferably 1-10 seconds) and generate data records corresponding to parking space numbers and occupancy markers. The vehicle information interaction module obtains the vehicle's current remaining battery power (SOC value), battery capacity parameters, and the estimated stay time entered by the user on the mobile terminal or inferred by the system, through a charging interface communication protocol (such as GB / T27930) or a license plate recognition and mobile application binding method. The historical charging demand pattern is obtained by retrieving data from the parking lot's historical charging record database, which includes historical charging start time, end time, charging amount, area, and environmental information for the corresponding time period. Subsequently, the multi-source data is uniformly timestamped and formatted through edge computing nodes. The parking space occupancy status data, vehicle battery power data, estimated dwell time data, and historical charging demand pattern data are constructed into a unified data vector set and uploaded to the central dispatch server in real time for subsequent demand prediction model calls.

[0018] Based on the multi-source data, the spatiotemporal distribution of charging demand in each area of ​​the parking lot is predicted in the future using a demand forecasting model.

[0019] First, parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns are processed in sliding segments according to preset time windows (preferably 15 or 30 minutes). The parking lot is then divided into several spatial grid areas or parking space cluster areas based on parking lot space division rules. Next, feature construction is performed on the multi-source data corresponding to each area within each time window, forming a feature vector that includes the area's parking space utilization rate, average remaining battery power, probability of potential charging demand, and the average and standard deviation of historical charging intensity for the same time period. This feature vector is then input into a demand prediction model in chronological order, the demand prediction model including a spatial feature extraction module. The model comprises a block and a time-series dependency modeling module. The spatial feature extraction module uses a convolutional neural network to extract the demand correlation between different spatial grids, while the time-series dependency modeling module uses an attention mechanism to dynamically allocate the influence weights of different historical time slices on the prediction period. The model outputs the charging demand intensity value of each spatial region within multiple consecutive prediction time slices in the future. The charging demand intensity value is the expected number of charging vehicles or the expected charging power demand per unit time. Finally, the charging demand intensity values ​​of each prediction time slice and the corresponding spatial region are combined to generate a two-dimensional time-space matrix form of the charging demand spatiotemporal distribution map, which is used for subsequent resource allocation optimization calculations.

[0020] Furthermore, based on the aforementioned multi-source data, a demand forecasting model is used to predict the spatiotemporal distribution of charging demand in various areas of the parking lot during future periods, including:

[0021] The multi-source data is spatiotemporally aligned, missing values ​​are imputed, and outliers are cleaned to construct a feature sequence for model input. The feature sequence is then input into a deep learning model that integrates convolutional neural networks and attention mechanisms to output the predicted charging demand intensity values ​​of parking space groups in each area of ​​the parking lot within multiple consecutive time slices in the future, thus generating the spatiotemporal distribution of charging demand.

[0022] First, the multi-source data is aligned to a unified time reference. Data with different sampling frequencies are resampled and timestamped according to a preset time window Δt (preferably 15 minutes) to construct a unified time axis. For missing data points, imputation is performed using interpolation based on the mean of adjacent time windows or K-nearest neighbor interpolation algorithms. For outliers, anomaly identification and removal are performed by setting a 3σ discrimination rule based on historical statistical distribution or an isolated forest algorithm to complete data cleaning. After completing spatiotemporal alignment and data cleaning, the parking lot is divided into N parking space clusters according to spatial region division rules. For each region within each time window, feature indicators such as parking space occupancy rate, average remaining power, potential charging probability per unit time, historical average demand, and demand fluctuation coefficient are extracted to construct a time series feature matrix of length T, which serves as the input feature sequence for the model. Subsequently, the feature sequence is input into a deep learning model that integrates a convolutional neural network and an attention mechanism. The convolutional neural network is used to extract convolutional features from the spatial grid adjacency relationships, capturing the demand coupling correlation between different regions. The attention mechanism is used to weight the feature vectors of multiple historical time slices, assigning weights to the influence of different time slices on future prediction results. The model outputs the predicted charging demand intensity value of parking space groups in each region within H consecutive future time slices. The predicted charging demand intensity value is represented by the expected number of vehicles charging or the expected charging amount per unit time. Finally, the predicted demand intensity values ​​of each time slice are combined according to the spatial region index to form a data set with a time-space two-dimensional matrix structure, generating the spatiotemporal distribution of charging demand.

[0023] Furthermore, the convolutional neural network is used to extract demand correlation features between parking space grids, and the attention mechanism is used to capture the temporal dependency weights of the influence of different historical periods on the prediction period.

[0024] Preferably, a spatial adjacency matrix is ​​constructed based on the actual layout of the parking lot, and a grid topology is established for the N parking space groups divided into the parking lot according to physical proximity or traffic connectivity. The feature vectors of each region in the same time slice are used as input feature maps, and the charging demand change trend between adjacent regions is extracted through multi-layer two-dimensional convolution operation. The convolution kernel size is preferably 3×3 or set according to the adjacency radius. The ReLU activation function is used in the convolution process and combined with batch normalization to enhance the stability of the model. The spatial feature tensor, which includes the demand coupling strength between regions, spatial diffusion trend and hotspot clustering characteristics, is obtained by superimposing multiple convolutions to characterize the demand correlation characteristics between different regions of the parking lot.

[0025] The attention mechanism is used to capture the time-dependent weights of the influence of different historical time periods on the prediction time period. It inputs historical time series features of length T into the time encoding module to generate a time feature representation for each historical time slice. Then, it constructs a query vector Q (corresponding to the target prediction time period), a key vector K, and a value vector V (corresponding to the features of each historical time slice). By calculating the dot product similarity between Q and K and performing Softmax normalization, the attention weight coefficients corresponding to each historical time slice are obtained. V is then weighted and summed according to these attention weight coefficients to obtain a time-weighted feature vector, thereby achieving dynamic allocation of the influence degree of different historical time periods. The time-weighted feature vector is fused with the spatial feature tensor and used as input data for the final demand prediction output layer.

[0026] Furthermore, after generating the spatiotemporal distribution of charging demand, the process also includes:

[0027] The demand forecast for open-air parking spaces is revised based on collected weather forecast data, and the spatiotemporal distribution of charging demand is revised for the first time based on the revision results; the event scale of known large-scale events on site is collected, and the spatiotemporal distribution of charging demand is revised for the second time.

[0028] First, weather forecast data for the future prediction period is collected, including rainfall probability, rainfall intensity, ambient temperature, and extreme weather warning information. For the open-air parking area within the parking lot, a weather impact correction function is established, using rainfall intensity and ambient temperature as influencing variables. A weather correction coefficient α1 is constructed, obtained through historical data regression analysis, and expressed as α1 = 1 + k1·R + k2·(T_ref - T), where R represents the normalized rainfall intensity, T represents the current predicted temperature, T_ref is the comfort temperature baseline, and k1 and k2 are empirical weighting coefficients. The original predicted demand intensity value for the open-air area is multiplied by the weather correction coefficient α1 to obtain the first corrected charging demand intensity value, thus completing the first correction of the spatiotemporal distribution of charging demand.

[0029] Subsequently, known schedule information for large-scale events within the venue is collected, including the start time, duration, expected number of participants, and event area. Based on the event scale and the ratio of charging demand growth during similar historical events, an event impact correction coefficient α2 is constructed. This coefficient is determined by the ratio of the event scale S to the average daily passenger flow S_ref, expressed as α2 = 1 + k3·(S / S_ref), where k3 is the amplification factor obtained from historical fitting. For the spatial grid within the event impact coverage area, the first corrected demand intensity value is multiplied by the event impact correction coefficient α2 to obtain the second corrected charging demand intensity value. Finally, a target spatiotemporal distribution of charging demand is formed, comprehensively considering both weather and event scale factors.

[0030] Based on the spatiotemporal distribution of charging demand and parking lot resource constraints, with the optimization objectives of minimizing total user waiting time, grid load fluctuations, and operator costs, a dynamic configuration optimization solution for charging piles is performed, outputting the scheduling path of mobile charging units and the power allocation scheme of fixed charging piles.

[0031] Furthermore, based on the spatiotemporal distribution of charging demand and parking lot resource constraints, and with the optimization objectives of minimizing total user waiting time, grid load fluctuations, and operator costs, a dynamic configuration optimization solution for charging piles is performed, outputting the scheduling path of mobile charging units and the power allocation scheme for fixed charging piles, including:

[0032] A higher-level planning model is established, with decision variables including the deployment location and standby status of mobile charging units, and the start / stop and basic power levels of each fixed charging pile. The objective function is to minimize the total resource allocation cost and the expected operating penalty cost. Constraints include the number of mobile units, the total number of fixed piles, and the upper limit of total power consumption. A lower-level planning model is established, based on the uncertainty scenarios of resource allocation and demand forecasting given by the higher-level planning model. The decision variables are the start time, allocated resource units, and dynamic power curve for each specific charging task. The objective function is to minimize the average user waiting time and grid load variance under all scenarios. Constraints include the task time window, battery charging characteristic curve, and real-time safe power limit. The higher-level and lower-level planning models are solved to output the scheduling path of the mobile charging units and the power allocation scheme of the fixed charging piles.

[0033] First, an upper-level planning model is established for macro-level allocation decisions of charging resources. The decision variables of this upper-level planning model include: the deployment location variable xᵢ and standby status variable yᵢ of mobile charging units in each spatial area, and the start / stop status variable zⱼ and basic power level variable pⱼ of each fixed charging pile within the prediction period; where xᵢ is a 0-1 integer variable indicating whether a mobile unit is deployed in a certain area, zⱼ is a 0-1 variable indicating whether a fixed charging pile is activated, and pⱼ is a value from the discrete power level set. The upper-level objective function is constructed as: Min F1 = C_deploy + C_oper + C_penalty, where C_deploy represents the sum of the mobile unit deployment cost and the fixed charging pile start / stop cost, C_oper represents the electricity cost generated by the basic power configuration, and C_penalty represents the operational penalty cost caused by insufficient capacity or overflowing demand. The constraints include: the total number of mobile units not exceeding a preset upper limit, the number of fixed charging piles not exceeding the total number of piles in the system, and the total power of all activated charging equipment not exceeding the upper limit of the transformer capacity.

[0034] Subsequently, a lower-level planning model is established to schedule specific charging tasks given the determined upper-level resource allocation results. This lower-level planning model is based on the set of uncertain demand forecast scenarios Ω, and performs scheduling calculations for each scenario ω∈Ω. Its decision variables include: the initial charging time variable t for each charging task k. k 1. Resource allocation unit variable a k (Corresponding to the fixed pile or moving unit number) and the dynamic power curve variable P k (t). The lower-level objective function is constructed as: Min F2=E_ω[W_total(ω)]+λ·Var(P_grid(ω)), where W_total represents the sum of waiting times for all users in the scenario, Var(P_grid) represents the variance of the total charging load curve, and λ is the weighting coefficient. Constraints include: the charging task must be completed within the expected vehicle dwell time window; the charging power curve must meet the phased constraint of the battery charging characteristic curve; and the total charging power at any given time must not exceed the real-time safe power limit constraint. Finally, the upper-level planning model and the lower-level planning model are nested and iteratively solved. A bi-level programming decomposition algorithm or a solution strategy based on a combination of genetic algorithm and mixed integer programming is preferred: first, the upper-level model is solved to obtain candidate resource allocation schemes, then the lower-level model is solved under each candidate scheme and the objective function value is fed back, iteratively updated until the preset convergence threshold is met; finally, the optimal scheduling path sequence of the mobile charging unit and the time-sharing power allocation scheme of the fixed charging pile within the prediction period are output.

[0035] Furthermore, the mobile charging unit is an automated guided vehicle or an autonomous mobile robot, equipped with an energy storage battery, a bidirectional charger, and a robotic arm interface, which can autonomously drive to the target parking space and establish a physical charging connection with the vehicle according to the scheduling instructions.

[0036] The mobile charging unit is an automated guided vehicle (AGV) or autonomous mobile robot, comprising a chassis module, an energy storage battery module, a bidirectional charger module, a robotic arm interface module, and a control and communication module. The chassis module includes a drive motor, a steering actuator, and lidar or visual navigation sensors for path planning and autonomous driving. The energy storage battery module stores electrical energy and outputs it to the vehicle or performs reverse charging during low-load periods via the bidirectional charger module. The bidirectional charger module supports DC fast charging or AC charging protocols and has built-in voltage, current, and temperature monitoring units for real-time monitoring of charging safety. The robotic arm interface module includes a retractable robotic arm and an automatic charging gun insertion / removal mechanism for automatically docking with the target vehicle's charging port according to scheduling instructions.

[0037] When executing scheduling instructions, the control and communication module receives target parking space coordinates and path information from the central scheduling system or edge computing nodes, and generates a driving trajectory using a preset path planning algorithm (preferably the A* algorithm or a real-time path planning algorithm based on laser SLAM). Upon reaching the target parking space, the module confirms the target vehicle's identity through visual recognition or RFID identification and controls the robotic arm interface module to complete the charging gun docking action. After establishing a physical charging connection, the bidirectional charger module executes charging control according to the issued power curve and feeds back the charging status data to the scheduling system in real time. After charging is completed or a new scheduling instruction is received, the mobile charging unit automatically disconnects and moves to the next target location.

[0038] The scheduling path and power allocation scheme are executed to dynamically adjust the matching relationship between charging resources and parking resources.

[0039] When executing the scheduling path and power allocation scheme, the central scheduling system first sends the path sequence of the mobile charging unit and the time-sharing power curve of each fixed charging pile to the edge computing node at the parking lot end. The edge computing node parses the received scheduling path data and converts the path coordinate sequence into motion control commands that the mobile charging unit can recognize. At the same time, it converts the power allocation scheme into the target output power value and start-stop control commands of the fixed charging pile.

[0040] During execution, edge computing nodes collect real-time data on changes in parking space occupancy and charging task completion. When a vehicle leaves early, extends its stay, or generates new charging demand, a local rescheduling mechanism is triggered. This local rescheduling mechanism quickly re-optimizes the remaining resources in the current time period, adjusts the path order of unexecuted mobile charging units, or dynamically modifies the output power ratio of fixed charging piles, thereby ensuring optimal matching between charging resources and real-time parking status.

[0041] Meanwhile, the edge computing nodes continuously monitor the total charging load curve. When the instantaneous power exceeds the safety threshold or the load fluctuation exceeds the preset variance threshold, the power peak shaving strategy is automatically executed, including reducing the power output of some non-urgent charging tasks or delaying the start time of some deferred tasks. Finally, through a closed-loop control mechanism of real-time data acquisition, execution control, status feedback and local re-optimization, the dynamic adjustment of the matching relationship between charging resources and parking resources is realized.

[0042] Furthermore, executing the aforementioned scheduling path and power allocation scheme to dynamically adjust the matching relationship between charging resources and parking resources includes:

[0043] The scheduling path and power allocation scheme are sent to the field edge computing nodes; the edge nodes directly control the power output of the fixed charging piles and the path execution of the mobile charging units.

[0044] Preferably, the scheduling path data and power allocation scheme are sent to the field edge computing node via an internal private network or industrial Ethernet communication method. The scheduling path data includes the target parking space coordinate sequence, time node and path priority identifier of the mobile charging unit; the power allocation scheme includes the time-sharing power value, start-stop status and power adjustment time point of each fixed charging pile in the prediction period.

[0045] After receiving the data, the field edge computing node first performs instruction parsing and validity verification, converting the path coordinates into motion control instructions for the mobile charging unit, including speed setpoints, steering angles, and driving path sequences; at the same time, it converts the power allocation values ​​into power setting instructions recognizable by the charging pile control protocol, and sends real-time power adjustment commands to each fixed charging pile through the charging pile control interface to achieve precise control of the charging pile power output.

[0046] During execution, edge computing nodes collect real-time location data and operating status data of mobile charging units, as well as voltage, current, and load data of fixed charging piles, and compare and verify them with the target instructions. When a path deviation exceeds a preset threshold or a power output deviation exceeds the allowable error range, the edge computing nodes automatically execute corrective control or report abnormal information to the central dispatch system. Through the direct control and real-time feedback mechanism of edge nodes, the localized execution and rapid response of the scheduling path and power allocation scheme are realized, thereby dynamically adjusting the matching relationship between charging resources and parking resources.

[0047] In summary, the embodiments of this application have at least the following technical effects:

[0048] First, multi-source data from the parking lot is collected in real time, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns. Next, based on this multi-source data, a demand forecasting model predicts the spatiotemporal distribution of charging demand in different areas of the parking lot during future periods. Then, based on the spatiotemporal distribution of charging demand and parking lot resource constraints, with the optimization objectives of minimizing total user waiting time, grid load fluctuations, and operator costs, a dynamic configuration optimization solution for charging piles is performed, outputting the scheduling path for mobile charging units and the power allocation scheme for fixed charging piles. Finally, the scheduling path and power allocation scheme are executed to dynamically adjust the matching relationship between charging resources and parking resources. This solves the technical problems of existing technologies where the fixed configuration of charging piles cannot be dynamically adjusted according to real-time changes in parking and charging demand, resulting in low resource utilization and long user waiting times. It achieves the technical effect of realizing dynamic intelligent configuration of charging piles, reducing user waiting time, and improving operational efficiency and resource utilization.

[0049] Example 2, based on the same inventive concept as the dynamic charging pile configuration method for intelligent parking lots in the foregoing examples, such as... Figure 2 As shown, this application provides a dynamic charging pile configuration system for intelligent parking lots, wherein the system includes:

[0050] Data acquisition module 11: Collects multi-source data from the parking lot in real time, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns; Demand prediction module 12: Based on the multi-source data, predicts the spatiotemporal distribution of charging demand in various areas of the parking lot during future periods using a demand prediction model; Solution module 13: Based on the spatiotemporal distribution of charging demand and parking lot resource constraints, optimizes the dynamic configuration of charging piles with the goal of minimizing total user waiting time, grid load fluctuations, and operator costs, and outputs the scheduling path of mobile charging units and the power allocation scheme of fixed charging piles; Adjustment module 14: Executes the scheduling path and power allocation scheme to dynamically adjust the matching relationship between charging resources and parking resources.

[0051] Furthermore, the demand forecasting module 12 is used to perform the following method:

[0052] The multi-source data is spatiotemporally aligned, missing values ​​are imputed, and outliers are cleaned to construct a feature sequence for model input. The feature sequence is then input into a deep learning model that integrates convolutional neural networks and attention mechanisms to output the predicted charging demand intensity values ​​of parking space groups in each area of ​​the parking lot within multiple consecutive time slices in the future, thus generating the spatiotemporal distribution of charging demand.

[0053] Furthermore, the demand forecasting module 12 is used to perform the following method:

[0054] The convolutional neural network is used to extract demand correlation features between parking space grids, and the attention mechanism is used to capture the time dependence weights of the influence of different historical periods on the prediction period.

[0055] Furthermore, the demand forecasting module 12 is used to perform the following method:

[0056] The demand forecast for open-air parking spaces is revised based on collected weather forecast data, and the spatiotemporal distribution of charging demand is revised for the first time based on the revision results; the event scale of known large-scale events on site is collected, and the spatiotemporal distribution of charging demand is revised for the second time.

[0057] Furthermore, the solver module 13 is used to perform the following method:

[0058] A higher-level planning model is established, with decision variables including the deployment location and standby status of mobile charging units, and the start / stop and basic power levels of each fixed charging pile. The objective function is to minimize the total resource allocation cost and the expected operating penalty cost. Constraints include the number of mobile units, the total number of fixed piles, and the upper limit of total power consumption. A lower-level planning model is established, based on the uncertainty scenarios of resource allocation and demand forecasting given by the higher-level planning model. The decision variables are the start time, allocated resource units, and dynamic power curve for each specific charging task. The objective function is to minimize the average user waiting time and grid load variance under all scenarios. Constraints include the task time window, battery charging characteristic curve, and real-time safe power limit. The higher-level and lower-level planning models are solved to output the scheduling path of the mobile charging units and the power allocation scheme of the fixed charging piles.

[0059] Furthermore, the solver module 13 is used to perform the following method:

[0060] The mobile charging unit is an automated guided vehicle or an autonomous mobile robot, equipped with an energy storage battery, a bidirectional charger and a robotic arm interface, and can autonomously drive to the target parking space and establish a physical charging connection with the vehicle according to the scheduling instructions.

[0061] Furthermore, the adjustment module 14 is used to perform the following method:

[0062] The scheduling path and power allocation scheme are sent to the field edge computing nodes; the edge nodes directly control the power output of the fixed charging piles and the path execution of the mobile charging units.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for configuring dynamic charging piles in an intelligent parking lot, characterized in that, The method includes: Real-time collection of multi-source data from parking lots, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns; Based on the multi-source data, the spatiotemporal distribution of charging demand in each area of ​​the parking lot during future periods is predicted using a demand forecasting model. Based on the spatiotemporal distribution of charging demand and parking lot resource constraints, with the optimization objectives of minimizing total user waiting time, grid load fluctuations and operator costs, a dynamic configuration optimization solution for charging piles is performed, and the scheduling path of mobile charging units and the power allocation scheme of fixed charging piles are output. The scheduling path and power allocation scheme are executed to dynamically adjust the matching relationship between charging resources and parking resources.

2. The dynamic charging pile configuration method for intelligent parking lots as described in claim 1, characterized in that, Based on the aforementioned multi-source data, a demand forecasting model is used to predict the spatiotemporal distribution of charging demand in various areas of the parking lot during future periods, including: The multi-source data is spatiotemporally aligned, missing values ​​are imputed, and outliers are cleaned to construct a feature sequence for model input. The feature sequence is input into a deep learning model that integrates convolutional neural networks and attention mechanisms, and the predicted charging demand intensity values ​​of parking space groups in each area of ​​the parking lot are output in multiple consecutive time slices in the future, thus generating the spatiotemporal distribution of charging demand.

3. The dynamic charging pile configuration method for intelligent parking lots as described in claim 2, characterized in that, The convolutional neural network is used to extract demand correlation features between parking space grids, and the attention mechanism is used to capture the time dependence weights of the influence of different historical periods on the prediction period.

4. The dynamic charging pile configuration method for intelligent parking lots as described in claim 2, characterized in that, After generating the spatiotemporal distribution of the charging demand, the method further includes: The demand forecast for open-air parking spaces is corrected by collecting weather forecast data, and the spatiotemporal distribution of charging demand is corrected for the first time based on the correction results. The event scale of known large-scale events within the venue is collected to make a second correction to the spatiotemporal distribution of charging demand.

5. The method for configuring dynamic charging piles in an intelligent parking lot as described in claim 1, characterized in that, Based on the spatiotemporal distribution of charging demand and parking resource constraints, and with the optimization objectives of minimizing total user waiting time, grid load fluctuations, and operator costs, a dynamic configuration optimization solution for charging piles is performed. This outputs the scheduling path for mobile charging units and the power allocation scheme for fixed charging piles, including: Establish an upper-level planning model. The decision variables are the deployment location and standby status of the mobile charging units, as well as the start-up and shutdown and basic power level of each fixed charging pile. The objective function is to minimize the total resource allocation cost and the expected operating penalty cost. The constraints include the number of mobile units, the total number of fixed piles, and the upper limit of total power consumption. A lower-level planning model is established based on the uncertainty scenarios of resource allocation and demand forecast given by the upper-level planning model. The decision variables are the start time, allocated resource units, and dynamic power curve for each specific charging task. The objective function is to minimize the average user waiting time and grid load variance under all scenarios. The constraints include task time window, battery charging characteristic curve, and real-time safety power limit. Solve the upper-level planning model and the lower-level planning model to output the scheduling path of the mobile charging unit and the power allocation scheme of the fixed charging pile.

6. The dynamic charging pile configuration method for intelligent parking lots as described in claim 1, characterized in that, The mobile charging unit is an automated guided vehicle or an autonomous mobile robot, equipped with an energy storage battery, a bidirectional charger and a robotic arm interface, and can autonomously drive to the target parking space and establish a physical charging connection with the vehicle according to the scheduling instructions.

7. The dynamic charging pile configuration method for intelligent parking lots as described in claim 1, characterized in that, Execute the aforementioned scheduling path and power allocation scheme to dynamically adjust the matching relationship between charging resources and parking resources, including: The scheduling path and power allocation scheme are sent to the field edge computing nodes; The power output of fixed charging piles and the path execution of mobile charging units are directly controlled by edge nodes.

8. A dynamic charging pile configuration system for intelligent parking lots, characterized in that, The system for implementing the dynamic charging pile configuration method for intelligent parking lots according to any one of claims 1-7, the system comprising: Data acquisition module: Collects multi-source data from the parking lot in real time, including parking space occupancy status, vehicle battery level, estimated dwell time, and historical charging demand patterns; Demand forecasting module: Based on the multi-source data, predicts the spatiotemporal distribution of charging demand in each area of ​​the parking lot during future periods using a demand forecasting model; Solution module: Based on the spatiotemporal distribution of charging demand and parking lot resource constraints, with the optimization objectives of minimizing total user waiting time, grid load fluctuations and operator costs, it performs dynamic configuration optimization of charging piles and outputs the scheduling path of mobile charging units and the power allocation scheme of fixed charging piles. Adjustment module: Executes the scheduling path and power allocation scheme to dynamically adjust the matching relationship between charging resources and parking resources.

Images