Mobile charging scheduling method and system based on time series data analysis

Abstract

The present application relates to the technical field of mobile charging scheduling, and particularly relates to a mobile charging scheduling method and system based on time series data analysis, comprising: collecting site information and charging demand time series data through a multi-source sensor, analyzing site availability, eliminating unusable space based on the analysis results, and reconstructing available charging areas; filtering serviceable charging demands in the areas as constraints, determining to-be-scheduled mobile charging equipment based on the charging demands; constructing a multi-dimensional scheduling evaluation model for the filtered charging demands, obtaining scheduling evaluation values of each charging demand through calculation, and generating scheduling priorities according to the scheduling evaluation values; based on the scheduling priorities, adopting a double-layer asynchronous architecture of space constraints and priority decision to make scheduling decisions, excluding invalid paths through distributed collaboration and redundant calculation, and outputting final scheduling instructions. The present application constructs a dynamic reconstruction reverse scheduling architecture of available charging space, thereby improving the reliability of scheduling.

Description

Technical Field

[0001] This invention relates to the field of mobile charging scheduling technology, and in particular to a mobile charging scheduling method and system based on time-series data analysis. Background Technology

[0002] With the increasing number of electric vehicles, the shortage of parking and charging resources has become a prominent issue. Mobile charging devices, due to their flexibility in providing services without requiring fixed parking spaces, have become a key means of alleviating the charging problem, and the optimization of their scheduling technology is a research hotspot in this field. Currently, existing mobile charging scheduling technologies can achieve basic spatial identification and demand-equipment matching. Some solutions introduce path planning and basic efficiency indicators, which can meet the scheduling needs of simple scenarios.

[0003] However, existing technologies have many shortcomings. First, problems such as fuel vehicles occupying spaces, lane congestion, and vehicles blocking lanes may occur during mobile dispatching, requiring the replanning of dispatching schemes and increasing user waiting time. Second, dispatching focuses on mechanical efficiency and does not take into account user experience and time window pressure, resulting in high order cancellation rates and low user satisfaction.

[0004] Therefore, there is an urgent need for a solution to address the above problems, overcome existing shortcomings, and promote the development of mobile charging technology. Summary of the Invention

[0005] This invention constructs a dynamic reconfiguration reverse scheduling architecture for available charging space, integrating global perception, probabilistic prediction, and safety redundancy constraints. It first divides the real reachable available domain and then dispatches orders in reverse, and configures a multi-pile collaborative avoidance and dynamic self-repair mechanism to solve the technical problem of space unreachability caused by dispatching orders first and then avoiding obstacles in the existing technology, thereby improving the reliability of scheduling.

[0006] The technical solution proposed in this invention is: a mobile charging scheduling method based on time-series data analysis, the method comprising: By collecting site information and charging demand time-series data from multiple source sensors, the site availability is analyzed, and unusable spaces are eliminated based on the analysis results to reconstruct available charging areas. Using available charging areas as constraints, filter the available charging demand within the area, and determine the mobile charging equipment to be dispatched based on the charging demand. For the selected charging demands, a multi-dimensional scheduling evaluation model is constructed. The scheduling evaluation value of each charging demand is calculated, and the scheduling priority is generated based on the scheduling evaluation value. Based on scheduling priority, a two-layer asynchronous architecture of space constraints and priority decision is adopted for scheduling decisions. Invalid paths are eliminated through distributed collaboration and redundant computation, and the final scheduling instruction is output.

[0007] Preferably, the process for collecting the site information and the charging demand time-series data is as follows: Multi-source sensors are activated to simultaneously collect site status data and charging demand time-series data. Site status data includes parking space location, passageway status and time-series change data, while charging demand time-series data includes charging request time, power demand and time-series change trend. A multi-source data fusion and denoising algorithm is used to deduplicatize and denoise the collected raw data, and remove invalid interference data and abnormal fluctuation data. Add collection timestamps to the processed valid data, organize them into a complete time-series data set according to the time sequence, and establish the time-series correlation of the data.

[0008] Preferably, the specific process for reconstructing the available charging area is as follows: Based on time series data sets, a time series feature extraction algorithm is used to extract the time series change features and stability features of site conditions; Set a threshold for site availability and a standard for continuous availability. Combine this with the characteristics of temporal changes to delineate the boundaries between available and unavailable spaces, mark unavailable spaces, and remove them. By integrating the location information, temporal variation characteristics, and continuous availability of available space, a spatial reconstruction algorithm is used to reconstruct a dynamically adaptable available charging area that adapts to temporal changes.

[0009] Preferably, the specific process for determining the mobile charging device to be scheduled is as follows: Obtain the reconstructed available charging area boundary information and time-series availability data, and filter out available charging demands that are within the available area and during the available time period; Collect data on the current location, remaining power, load capacity, and moving speed of each mobile charging device to establish a device status database; A matching algorithm is used to combine the location of available charging needs, power demand and device status data to perform precise matching, select suitable devices as mobile charging devices to be dispatched, and form a list of devices to be dispatched.

[0010] Preferably, the specific process for constructing the multi-dimensional scheduling evaluation model is as follows: Based on available charging demand and mobile charging devices to be dispatched, the core evaluation dimensions are determined. Define quantitative standards for each evaluation dimension, transform qualitative indicators into calculable quantitative parameters, and integrate parameters from each dimension to construct a multi-dimensional scheduling evaluation model. The weight allocation rules for each evaluation dimension are set using the analytic hierarchy process (AHP), and the model parameters are calibrated in conjunction with the actual scheduling scenario to ensure that the model adapts to the dynamic changes in available charging areas.

[0011] Preferably, the process for obtaining the scheduling priority is as follows: The quantitative parameters of available charging demand and the adaptation parameters of the mobile charging equipment to be scheduled are substituted into the multi-dimensional scheduling evaluation model. The comprehensive evaluation value of each available charging demand is obtained through model calculation, and the matching degree between each demand and the equipment to be scheduled is calculated simultaneously. The comprehensive evaluation values ​​are sorted in descending order. For demands with the same evaluation value, a second sort is performed based on the matching degree. The scheduling priority of each charging demand is generated according to the sorting results.

[0012] Preferably, the specific process of the scheduling decision is as follows: A two-layer asynchronous architecture of spatial constraints and priority decision-making is adopted, which divides the decision-making modules into two independent layers to decouple spatial constraints and priority decision-making. The spatial constraint layer performs preliminary planning of the equipment operation path based on the reconstructed available charging area and time-series change data, and delineates the spatial boundary of equipment operation. The priority decision-making layer optimizes the equipment scheduling order based on the scheduling priority and the list of equipment to be scheduled, and completes the scheduling decision by using a collaborative decision-making algorithm to avoid equipment operation conflicts.

[0013] Preferably, the process of obtaining the final scheduling instruction is as follows: Based on the scheduling decision results, distributed collaborative control is carried out on each mobile charging device to be scheduled, and a real-time interaction mechanism between devices is established. A redundancy calculation algorithm is used to perform multi-path verification on the initially planned equipment operation path, eliminate invalid paths that are outside the available charging area or have congestion or conflict risks, and select the optimal operation path. Integrate the optimal running path, scheduling priority, and device task allocation information to generate specific scheduling instructions for each mobile charging device, and output the final scheduling instructions after completing instruction verification.

[0014] The present invention also provides a mobile charging scheduling system based on time-series data analysis, the system being used to execute the mobile charging scheduling method based on time-series data analysis.

[0015] The present invention also provides a computer-readable storage medium storing a computer program that is executed by a processor to implement the mobile charging scheduling method based on time-series data analysis.

[0016] The beneficial effects of this invention are: 1. This invention constructs a dynamic reconstructed reverse scheduling underlying architecture for available charging space, integrating global spatial perception, probabilistic prediction, and safety redundancy constraints. It first delineates the truly reachable available charging domain and then generates scheduling tasks in reverse. Combined with multi-pile collaborative avoidance and dynamic spatial self-repair mechanism, it fundamentally solves the defects of traditional methods such as dispatching orders first and then avoiding obstacles, spatial unreachability, scheduling failure due to prediction errors, and congestion of multiple device paths. It significantly improves scheduling reachability and enhances the effectiveness and efficiency of mobile scheduling.

[0017] 2. This invention constructs a multi-dimensional scheduling evaluation model strongly bound to the available charging domain. It combines five core dimensions: charging urgency, spatial accessibility, equipment matching, resource utilization, and marginal revenue. It employs a weight allocation and model calibration mechanism that integrates linear precise quantification algorithms, analytic hierarchy process (AHP), and fully connected neural networks. Simultaneously, it incorporates dynamic pricing factors, grid health adaptation adjustments, and user credit management to achieve a refined, dynamic, and economically balanced scheduling evaluation. By dynamically adjusting the marginal revenue achievement threshold, quantification standards, and dimensional weights, it adapts to different electricity price periods, supply and demand states, and grid load conditions, effectively reducing order cancellation rates and suppressing malicious order placement. It balances user experience, system efficiency, and operational economy, and differs from conventional scheduling evaluation models, demonstrating outstanding creativity and providing accurate and reliable evaluation basis for scheduling priority generation and scheduling decisions.

[0018] 3. This invention adopts a two-layer asynchronous decoupled architecture of spatial constraints and priority decision-making, combined with A-Star path planning, KSP redundant path verification, distributed collaborative control, and timeout interruption deadlock resolution mechanism, to achieve collaborative optimization of spatial constraints and scheduling priorities, effectively avoiding scheduling disorder caused by equipment path conflicts, deadlocks or livelocks, and invalid paths. Through multi-level defense mechanisms such as grid health monitoring, power degradation scheduling, equipment power safety red line withdrawal, communication keep-alive error correction, and graceful rejection of system back pressure, it comprehensively covers extreme operating conditions such as grid fluctuations, equipment failures, demand surges, and communication interference, significantly improving the all-weather stability, anti-interference capability, and grid risk prevention and control capability of the scheduling system. At the same time, through distributed collaboration and redundant computing, it ensures the accuracy, timeliness, and compliance of scheduling instructions, forming a full-dimensional closed-loop industrial-grade scheduling system. Attached Figure Description

[0019] Figure 1 This is a flowchart of a mobile charging scheduling method based on time-series data analysis. Figure 2 A flowchart illustrating the dynamic reconstruction of available charging space in a mobile charging scheduling method based on time-series data analysis; Figure 3 A flowchart for generating scheduling priorities in a mobile charging scheduling method based on time-series data analysis is provided. Figure 4A flowchart for generating the final scheduling instruction of the mobile charging scheduling method based on time-series data analysis. Detailed Implementation

[0020] 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.

[0021] 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.

[0022] like Figure 1 As shown, a mobile charging scheduling method based on time-series data analysis is characterized by the following steps: First, collecting site information and charging demand time-series data from multiple sensors; analyzing site availability; eliminating unusable spaces based on the analysis results; reconstructing available charging areas; using available charging areas as constraints; filtering serviceable charging demands within the area; determining mobile charging devices to be scheduled based on the charging demands; constructing a multi-dimensional scheduling evaluation model for the filtered charging demands; calculating the scheduling evaluation value for each charging demand; generating scheduling priorities based on the scheduling evaluation values; and making scheduling decisions based on the scheduling priorities using a two-layer asynchronous architecture of spatial constraints and priority decision-making; eliminating invalid paths through distributed collaboration and redundant computation; and outputting the final scheduling instruction.

[0023] Furthermore, the process for collecting site information and charging demand time-series data is as follows: Multi-source sensors are activated to simultaneously collect site status data and charging demand time-series data. Site status data includes parking space location, passageway status, and time-series changes. Charging demand time-series data includes charging request time, power demand, and time-series change trends. A multi-source data fusion and denoising algorithm is used to deduplicate and denoise the collected raw data, removing invalid interference data and abnormal fluctuation data. A collection timestamp is added to the processed valid data, and the data is organized according to the time sequence to form a complete time-series data set, establishing a data time-series correlation relationship.

[0024] The multi-source sensor system consists of position sensors (such as geomagnetic sensors), accessibility sensors (such as environmental sensing cameras), charging demand acquisition sensors (NB-IoT IoT sensors), and environmental sensors (such as temperature and humidity sensors). An NTP time server combined with a hardware trigger bus is used to align the clock sources of each sensor. Position sensors and environmental sensing cameras are synchronized via the hardware trigger bus, while charging request data (from the app) has its timestamp calibrated via the NTP time server, ensuring that the time deviation of all sensor data acquisition does not exceed 10 milliseconds. After activating all the aforementioned multi-source sensors, the system controls them to operate synchronously and collaboratively, ensuring the time synchronization of various data acquisitions and avoiding data deviations caused by time differences.

[0025] Subsequently, corresponding data are collected through various sensors: parking space location data is collected through the location sensor; passage status and time sequence change data are collected through the passage status sensor; charging demand data collection sensor collects charging request time, power demand and time sequence change trend; and temperature and humidity data are collected through the environmental sensor, thus completing the synchronous collection of site status data and charging demand time sequence data.

[0026] After data acquisition, a multi-source data fusion denoising algorithm is used to process all raw data. This algorithm first removes high-frequency interference signals from the raw data using wavelet threshold denoising, then corrects fluctuations caused by data transmission using Kalman filtering, and simultaneously deduplicates repeatedly acquired data. Finally, it removes invalid interference data such as outliers caused by sensor malfunctions and distorted data caused by external electromagnetic interference, as well as abnormal fluctuation data deviating from the normal acquisition range by ±30%. In the initial stage when the system is powered on (e.g., the first 5 minutes), a dynamic baseline range is quickly established using median filtering combined with moving average. The median and moving average of the data acquired in the previous 10 seconds are calculated every 10 seconds, and the intersection of the two is taken as a temporary baseline range for preliminary judgment of data validity. After the baseline period, the system seamlessly switches to ±30% anomaly detection logic, using historical baseline data as a reference, to prevent the accidental deletion of valid data generated by sudden environmental changes (such as a sudden heavy rain causing a surge in humidity) during cold starts, ensuring the accuracy of data cleaning.

[0027] After processing, each valid data point is individually timestamped to the millisecond level. All valid data are then sorted and organized in ascending order of collection time to form a complete time-series data set. Simultaneously, the site status data and charging demand data at the same moment are bound together through the time dimension to establish a time-series correlation between the data, ensuring the integrity and correlation of the time-series data set and providing data support for subsequent site availability analysis and reconstruction of available charging areas.

[0028] Establish a sensor confidence assessment system to monitor the operating status of each sensor in real time. When an environmental sensor malfunctions and continuously outputs incorrect temperature and humidity data, causing a large area to be mistakenly judged as unusable, the system automatically switches to a redundancy scheme, calling historical temperature and humidity data from the same period within the past 30 days as temporary replacement parameters. At the same time, the faulty sensor is marked and an alarm is issued. Once the sensor recovers, the system automatically switches back to real-time data acquisition to avoid the failure of a single sensor affecting the normal operation of the system.

[0029] Furthermore, the specific process of reconstructing the available charging area is as follows: Based on the time-series data set, a time-series feature extraction algorithm is used to extract the time-series change features and stability features of the site status; a site availability judgment threshold and a continuous availability duration standard are set, and combined with the time-series change features, the boundaries between available and unavailable spaces are divided, and unavailable spaces are marked and removed; the location information, time-series features, and continuous availability duration of available spaces are integrated, and a spatial reconstruction algorithm is used to reconstruct a dynamically adaptable available charging area that adapts to time-series changes.

[0030] like Figure 2 As shown, the temporal feature extraction algorithm used is the LSTM neural network algorithm. This algorithm can effectively capture the temporal correlation of temporal data and is suitable for extracting the temporal change features of site conditions. Based on the obtained temporal data set, the LSTM neural network algorithm is started, and the temporal data set is sampled in a 1-minute time window to extract the temporal change features and stability features of site conditions from the temporal data set. The temporal change features include the frequency and duration of the available state switching of each area of ​​the site, and the stability features include the maximum duration of continuous availability of each area and the available state fluctuation coefficient.

[0031] When the LSTM outputs its prediction results, it simultaneously outputs the availability confidence score (range 0-100%) for the corresponding area, constructing a spatial model containing probabilistic information. Through an online learning mechanism, an LSTM prediction accuracy backtracking log is established to record in real time the deviation between the LSTM's predictions for each high-risk area and the actual site conditions. The specific backtracking rules are as follows: for each high-risk area (confidence 50%-80%), the matching between the prediction results and the actual conditions is statistically analyzed for 10 consecutive times. If 10 consecutive predictions indicate high risk but the actual site is unobstructed, the system automatically lowers the soft penalty coefficient of the A-Star algorithm cost function for that area from an initial 3 times to 1.5 times, guiding equipment to prioritize paths to that area and improving traffic efficiency. If 10 consecutive predictions indicate high risk and the actual site has obstacles (100% false negative rate), the penalty coefficient is automatically increased to 4 times, further reducing the probability of equipment selecting that area and avoiding scheduling errors. This online learning mechanism enables the spatial probability model to have self-evolution capabilities, dynamically adapting to the actual site conditions and balancing path tolerance and traffic efficiency.

[0032] Environmental perception cameras capture real-time images of the site's passageways and, equipped with semantic segmentation algorithms (such as the U-Net algorithm), classify and identify obstacles in the images, distinguishing between semi-static obstacles (such as overturned trash cans, randomly parked shared bicycles, and construction sandbags) and temporary obstacles (such as pedestrians and temporarily parked vehicles). A mechanism for tracking the duration of obstacle presence is established, recording the location and dwell time of various obstacles in real time. When a camera detects an obstacle remaining in the same location for more than 10 minutes without significant movement (movement distance < 0.5 meters), the system automatically marks it as a permanent roadblock and immediately updates the obstacle list in the Voronoi diagram. Simultaneously, the boundaries of available charging areas and the passageway's access status are updated, completely blocking path planning in that area and preventing devices from repeatedly attempting to pass through it and consuming power through repeated start-stop cycles. If an obstacle's dwell time is less than 10 minutes, it is marked as a temporary obstacle, incurring only temporary penalties during path planning and not included in the permanent obstacle list, ensuring flexibility in path planning.

[0033] Subsequently, site availability assessment thresholds, continuous availability duration standards, and confidence thresholds were set. The site availability assessment thresholds include parking space occupancy thresholds, passageway access thresholds, and environmental adaptability thresholds. The parking space occupancy threshold is set at an occupancy rate ≤ 80%, the passageway access threshold is set at a traffic congestion level ≤ 50%, and the environmental adaptability threshold is set at a temperature of -10℃. At 45℃ and humidity ≤85%, the standard for continuous usability is set as ≥5 minutes of continuous usability in a single area, and the confidence threshold is set at 80%.

[0034] Based on the extracted temporal variation features, LSTM prediction results, and confidence scores, and in accordance with the established criteria, the site space is divided into usable, unusable, and high-risk areas according to the following boundary coordinate rules: areas predicted to be usable with a confidence level ≥ 80% are classified as usable spaces; areas predicted to be unusable or with extremely low confidence levels (< 50%) are classified as unusable spaces and marked for removal; areas predicted to be usable but with confidence levels between 50% and 80% are marked as "high-risk areas," not directly removed, and retained as alternative areas for path planning.

[0035] The channel access status is determined using a combination of binary segmentation of available and unavailable channels and path penalties. For channels that are physically passable but not wide enough for mobile charging vehicles to pass (width < width of mobile charging device + 0.5-meter safety distance), they are not included in unavailable space. Instead, a penalty is set during path planning to reduce the path priority of that channel. Areas with occupancy rates exceeding thresholds, congestion levels exceeding standards, environmental parameters not meeting requirements, or insufficient continuous availability are identified as unavailable spaces, marked, and completely removed.

[0036] Finally, the Voronoi diagram segmentation algorithm is used as the spatial reconstruction algorithm. Combining the location coordinates, temporal characteristics, and continuous availability of the available space, the available space is divided into several independent charging sub-regions, each corresponding to a set of available charging needs. After Voronoi diagram segmentation, the maximum circumcircle radius of the mobile charging device is measured (set according to the actual size of the device, such as 0.75 meters). All obstacles (such as fixed parking spaces, walls) and passage boundaries in the site are expanded outward according to this maximum circumcircle radius, and the physical occupancy of the device and safety redundancy space are pre-deducted from the space. At the same time, a dynamic update mechanism is established. Every 5 minutes, based on the latest collected temporal data, LSTM prediction results, and confidence scores, the boundaries and sub-region divisions of the available charging area and high-risk area are updated. Simultaneously, the penalty coefficient adjusted by the online learning mechanism and the update status of permanent obstacles are combined to ensure that the available charging area can dynamically adapt to the temporal changes, uncertainties, prediction biases, and long-tail obstacles of the site. When a time difference occurs between the available charging area (updated every 5 minutes), the scheduling evaluation model (fine-tuned every 10 minutes), and the equipment status database (updated in real time), or when the site status changes due to unforeseen events such as temporary roadblocks during the issuance of a scheduling instruction, the system immediately triggers an interruption mechanism to suspend the issuance of the current scheduling instruction, synchronously update the site status data and available charging area information, recalculate the scheduling evaluation value and scheduling decision, and issue the adjusted scheduling instruction only after the data is synchronized and consistent, thus avoiding scheduling errors caused by data inconsistency.

[0037] Furthermore, the specific process for determining the mobile charging devices to be scheduled is as follows: Obtain the reconstructed available charging area boundary information and time-series availability data, and filter out the available charging demands that are within the available area and during the available time period; collect the current location, remaining power, load capacity, and movement speed data of each mobile charging device to establish a device status database; use a matching algorithm to combine the location, power demand, and device status data of available charging demands for precise matching, and filter out suitable devices as mobile charging devices to be scheduled, forming a list of devices to be scheduled.

[0038] First, the reconstructed available charging area boundary information and time-series availability data are obtained. The available charging area boundary information includes boundary coordinates and sub-area division information, while the time-series availability data includes the continuous availability period of each sub-area, availability fluctuations, and confidence scores. By comparing the available charging area boundary coordinates, it is confirmed whether the location of each charging demand is within the available area. Simultaneously, combined with the time-series availability data, charging demands whose time matches the available period of the sub-area are filtered out, and invalid demands with mismatched locations or time periods are eliminated, resulting in the available charging demands. A system backpressure and graceful rejection mechanism is added at the demand entry point. When the length of the service queue exceeds the total number of devices multiplied by the maximum concurrency multiple (the maximum concurrency multiple is set to 3, meaning that a single device can handle a maximum of 3 service requests at the same time), the system immediately triggers the system backpressure mechanism and actively pushes a circuit breaker prompt to the upstream (user APP) with the message "Currently busy, estimated waiting time XX minutes" (the estimated waiting time is calculated based on the length of the service queue and the average service time per device). For new charging requests, if the site channel throughput limit (calculated based on channel width and device movement speed, such as a maximum of 10 devices per minute) has reached the threshold, the request is gracefully rejected directly, and the user is recommended to go to an available charging area nearby. This protects the core scheduling system from being overwhelmed by demand surges and avoids problems such as CPU overload, system crashes, and user orders being stuck in the matching process for a long time.

[0039] Subsequently, relevant status data of each mobile charging device is collected. Current location data is collected via the device's GPS positioning module, accurate to the meter level. Remaining power data includes the current remaining percentage and the amount of power that can be output for charging. Load capacity data includes maximum charging power and the maximum number of devices that can be simultaneously served. Movement speed data includes idle movement speed and loaded movement speed. A device social radius parameter (set based on device size and site pedestrian density, e.g., 1.5 meters) is added to the device status database. Simultaneously, the system integrates the pedestrian detection function of an environmental perception camera to identify human activity around the device in real time. When a device approaches a densely populated area (such as an intersection or parking area) or detects human activity within its social radius, the system automatically reduces the device's movement speed to a crawling speed (e.g., 0.5 m / s) and activates an audible and visual alarm (an audible warning to avoid the device and a flashing yellow warning light).

[0040] A power grid health monitoring interface is set up in the equipment status database. This interface connects to the power grid dispatching system in real time to collect real-time power grid operating parameters, including core indicators such as grid voltage, frequency, and load factor. The allowable voltage fluctuation range is ±10% of the rated voltage, and the allowable frequency fluctuation range is 50Hz ±0.5Hz. When a grid voltage fluctuation exceeding ±10%, frequency instability (exceeding the 50Hz ±0.5Hz range), or a grid load factor ≥90% (extremely high grid load) is detected, the system automatically triggers a power degradation dispatch mechanism. Instead of prioritizing high-power (e.g., ≥60kW) and high-capacity (e.g., ≥50 kWh) charging demands, it prioritizes low-power (<30kW) and low-capacity (<20 kWh) charging demands, reducing the operating power of individual devices and alleviating grid load pressure. If grid fluctuations persist for more than 5 minutes or the fluctuation amplitude exceeds ±15%, the system immediately suspends new task assignment. Simultaneously, for devices currently performing high-power charging tasks, the charging power is gradually reduced to a safe threshold (e.g., 30kW) to prevent errors in the device's battery management system (BMS), large-scale device offlineness, or failure due to grid fluctuations. Once the power grid health returns to normal (voltage fluctuation ≤ ±10%, frequency stable at 50Hz ± 0.5Hz, load factor < 80%), the system will automatically remove the power degradation scheduling and restore the normal scheduling strategy.

[0041] Mobile charging devices employ a combination of vehicle-mounted laser SLAM and UWB positioning for autonomous obstacle avoidance and navigation. Vehicle-mounted laser SLAM scans the environment in real time, identifies obstacles, and generates dynamic obstacle avoidance paths. UWB positioning accurately pinpoints the device's location, ensuring the path output by the spatial constraint layer is accurately executed. A power safety threshold withdrawal mechanism is also implemented. A power safety threshold (remaining power ≥ 20%) is set for each device. The system monitors the device's power consumption during scheduling tasks. When a device's remaining power drops below the safety threshold during long-distance scheduling or queuing, the system automatically triggers a withdrawal command, suspending the current scheduling task and returning the device to a designated charging point to replenish its power, preventing the device from failing to provide service due to depleted power. The collected status data of each mobile charging device is categorized and stored by device number, creating a device status database. This database is updated in real time, with the update frequency consistent with the time-series data acquisition frequency, and simultaneously updates the grid health data and power degradation scheduling status.

[0042] The matching algorithm used is the Hungarian algorithm, which can accurately match available charging demands with mobile charging devices, improving matching efficiency and accuracy. After the algorithm is activated, it uses proximity, power matching, load matching, and marginal revenue achievement as core criteria. Combined with grid health status and power degradation scheduling requirements, it compares the location, power demand, and capacity of available charging demands with device status data in the device status database to calculate the matching degree between each available charging demand and each mobile charging device. Devices with a matching degree ≥80%, achieving marginal revenue (cost not exceeding 1.5 times revenue, adjustable flexibly after dynamic pricing), and meeting power degradation scheduling requirements are identified as suitable devices. All suitable devices are selected as mobile charging devices to be scheduled, arranged in descending order of matching degree to form a list of devices to be scheduled. The list includes core information such as device number, current status, matched charging demand number, estimated arrival time, marginal revenue assessment result, dynamic pricing coefficient, and grid health matching status, facilitating subsequent scheduling priority matching and scheduling decisions.

[0043] Furthermore, the specific process of constructing a multi-dimensional scheduling evaluation model is as follows: Based on the available charging demand and the mobile charging devices to be scheduled, the core evaluation dimensions are determined; the quantitative standards of each evaluation dimension are defined, the qualitative indicators are transformed into calculable quantitative parameters, and the parameters of each dimension are integrated to construct a multi-dimensional scheduling evaluation model; the weight allocation rules of each evaluation dimension are set using the analytic hierarchy process, and the model parameters are calibrated in combination with the actual scheduling scenario to ensure that the model adapts to the dynamic changes of available charging areas.

[0044] Based on available charging demand and a list of equipment to be dispatched, the core evaluation dimensions of the multi-dimensional dispatch evaluation model are determined, including charging urgency, spatial accessibility, equipment matching, resource utilization, and marginal revenue. The system acquires real-time grid electricity price data (updated every 5 minutes) and regional supply and demand data in real time, dynamically adjusting the charging unit price or service premium coefficient. During off-peak hours (late nights) when electricity demand is low (real-time price ≤ 0.8 times the average price) and the regional supply-demand ratio is < 0.5 (sparse demand), the service premium coefficient is lowered to 0.8 times, reducing the marginal revenue threshold (allowing cost ≤ 1.8 times revenue) to avoid equipment idleness and ensure equipment utilization. During peak grid load periods such as hot afternoons (real-time price ≥ average price), the system adjusts the service premium coefficient accordingly. When the electricity price is 1.5 times higher and the regional supply-demand ratio is greater than 2 (demand surge), the service premium coefficient is increased to 1.2-1.5 times. This dynamic premium covers high energy consumption costs. Even with higher costs, as long as users are willing to pay the premium, marginal revenue can be achieved, and dispatch can proceed. During flat periods (within ±20% of the real-time electricity price and flat period price) and when supply and demand are balanced (supply-demand ratio 0.5-2), the benchmark pricing and the threshold of cost ≤ revenue 1.5 times are maintained to ensure economic viability during normal periods. Simultaneously, considering the grid health status, when power degradation dispatch is triggered, the quantitative standard for marginal revenue assessment is dynamically adjusted. The marginal revenue score for low-power, low-electricity demands is appropriately weighted (weighting coefficient 1.2 times), increasing their dispatch priority and guiding the system to prioritize serving such demands, thus reducing grid load pressure.

[0045] The system comprehensively covers the three major scheduling elements of demand, equipment, and space, including charging urgency, spatial accessibility, equipment matching, and resource utilization. The fifth dimension, marginal revenue, is used to evaluate the economics of a single scheduling operation. Marginal revenue assessment uses a precise calculable model with a score of 0-100. The core calculation logic is as follows: Estimated energy cost = Equipment round-trip distance × Power consumption per unit distance × Real-time electricity price; Service revenue = User charging kilowatt-hours × (Benchmark charging unit price × Dynamic pricing factor) + User satisfaction points converted to monetary value; Marginal revenue = Service revenue - Estimated energy cost; Dynamic pricing factor = (Real-time electricity price fluctuation coefficient × 0.6) + (Regional supply-demand ratio coefficient × 0.4), where the real-time electricity price fluctuation coefficient = Current real-time electricity price / Average electricity price (the average electricity price is a preset benchmark price, such as 0.8 yuan / kilowatt-hour); Regional supply-demand ratio coefficient = Number of pending service demands in the current region / Number of pending scheduling devices in the current region (if the number of pending scheduling devices is 0, the coefficient is calculated as 3.0); Dynamic threshold... The coefficient is calculated as follows: 1.5 + (dynamic pricing factor - 1.0) × 0.3 (within the range of 1.5-1.8 times). The scoring rule uses linear scoring. When marginal revenue ≥ 0 and estimated energy cost ≤ service revenue × dynamic threshold coefficient, the score is 80 + 20 × [(service revenue - estimated energy cost) / (service revenue × dynamic threshold coefficient)] (100 points are awarded directly if the threshold is exceeded). When marginal revenue ≥ 0 and estimated energy cost > service revenue × dynamic threshold coefficient, the score is 40 + 39 × [(service revenue × dynamic threshold coefficient - estimated energy cost) / (service revenue × dynamic threshold coefficient)]. When marginal revenue < 0, the score is 39 × [marginal revenue / (service revenue × dynamic threshold coefficient)] (positive values ​​are taken; scores below 0 are counted as 0).

[0046] The other four dimensions all use a linear scoring model of 0-100 points, as follows: Charging urgency is based on the user's vehicle's remaining battery power (S, %). When S≤20%, the score = 80 + 20 × [(20-S) / 20]; when 20%<S<50%, the score = 40 + 39 × [(50-S) / 30]; when S≥50%, the score = 39 × [(50-S) / 50] (the result is positive); Spatial accessibility is based on the distance between the device and the demand point (D, meters). When D≤50 meters, the score = 80 + 20 × [(50-D) / 50]; when 50 meters<D<100 meters, the score = 40 + 39 × [(100-D) / 50]; when D≥100 meters, the score = 39 × [(100-D) / 100] (the result is positive); Device matching degree Based on the matching degree (M, %), the score is 80 + 20 × [(M-90) / 10] when M ≥ 90%, 40 + 39 × [(M-70) / 20] when 70% ≤ M < 90%, and 39 × [M / 70] when M < 70%. Resource utilization is based on the current load rate of the device (L, %), load rate = current service demand / maximum service capacity of the device. The score is 80 + 20 × [(60-L) / 60] when L ≤ 60%, 40 + 39 × [(80-L) / 20] when 60% < L ≤ 80%, and 39 × [(80-L) / 80] when L > 80% (the result is positive). All dimensions are calculated and rounded to one decimal place. Scores outside the 0-100 range are rounded to 0 or 100 based on the nearest value.

[0047] Using quantitative parameters from each dimension as model input, a multi-dimensional scheduling evaluation model is constructed using a fully connected neural network to achieve fusion calculation of parameters from each dimension. The fully connected neural network has three hidden layers: the input layer has 5 neurons (corresponding to the five evaluation dimensions), the first hidden layer has 16 neurons, the second has 8, the third has 4, and the output layer has 1 (corresponding to the comprehensive evaluation value). The ReLU function is used as the activation function, and the mean squared error (MSE) function is used as the loss function to calculate the deviation between the comprehensive evaluation value output by the model and the actual optimal evaluation value. The analytic hierarchy process (AHP) is used to define the weight allocation rules for each evaluation dimension. A judgment matrix is ​​constructed using AHP to determine the weight allocation of the five core evaluation dimensions: charging urgency (35%), spatial accessibility (20%), equipment matching degree (20%), resource utilization rate (10%), and marginal revenue evaluation (15%), balancing demand urgency and operational economy.

[0048] The interface mechanism combining the Analytic Hierarchy Process (AHP) and fully connected neural networks is as follows: Fixed weights for each dimension obtained from the AHP are used as the initial weights of the neural network's input layer. During model training, the weights of the hidden and output layers are updated using backpropagation, incorporating actual scheduling data samples. The initial weights do not participate in backpropagation updates; they remain fixed as input layer parameters, ensuring the rationality of weight allocation and the model's fitting accuracy. In practical scheduling scenarios, the model is trained using actual scheduling data samples covering different site conditions, charging demand scenarios, electricity price periods, and grid health states. Model parameter calibration is completed, and the model's evaluation error after calibration is ≤5%. A dynamic calibration mechanism is also established, fine-tuning the model parameters every 10 minutes based on temporal changes in available charging areas, dynamic pricing factor adjustments, and grid health fluctuations. During the fine-tuning process every 10 minutes, 10%-15% of historical typical scenario data (covering different scenarios such as holidays, severe weather, weekday peak hours, weekend off-peak hours, different electricity price periods, and different grid health states) are extracted from the replay buffer and used in conjunction with the latest 10-minute time series data for model training. The replay buffer adopts a first-in-first-out (FIFO) strategy to update and supplement new typical scenario data in real time, ensuring that the model can adapt to changes in site time series, dynamic pricing adjustments, and fluctuations in grid health, while not forgetting the old long-term patterns, avoiding model drift during long-term operation, and ensuring the stability and accuracy of model evaluation.

[0049] Furthermore, the process of obtaining scheduling priority is as follows: The quantitative parameters of available charging demand and the adaptation parameters of the mobile charging equipment to be scheduled are substituted into the multi-dimensional scheduling evaluation model. The comprehensive evaluation value of each available charging demand is calculated by the model, and the matching degree between each demand and the equipment to be scheduled is calculated simultaneously. The comprehensive evaluation values ​​are sorted in descending order. For demands with the same evaluation value, a second sort is performed based on the matching degree. The scheduling priority of each charging demand is generated according to the sorting result.

[0050] like Figure 3 As shown, the quantitative parameters of available charging demand and the adaptation parameters of the mobile charging equipment to be dispatched are substituted into the constructed and calibrated multi-dimensional dispatch evaluation model. The multi-dimensional dispatch evaluation model is used to calculate the quantitative parameters of each available charging demand according to their weights, so as to obtain the comprehensive evaluation value of each available charging demand. The comprehensive evaluation value ranges from 0 to 100 points. The quantitative results of the marginal benefit evaluation dimension (including the impact of dynamic pricing factors) and the grid health adaptation directly affect the comprehensive evaluation value, ensuring that economic efficiency, dynamic adaptability and grid risk prevention and control are taken into account in the dispatch priority consideration.

[0051] Simultaneously, the matching degree between each available charging demand and the mobile charging equipment to be scheduled is calculated. The matching degree is calculated using cosine similarity normalization. The specific calculation process is as follows: the required electricity and charging power are taken as the demand vector X (x1, x2), and the equipment load capacity and output power are taken as the equipment vector Y (y1, y2). First, the dot product of vectors X and Y is calculated, then the magnitude of vectors X and Y are calculated respectively, and finally the cosine similarity is calculated. The cosine similarity result is normalized to 0-100%, which is the matching degree. The higher the matching degree, the better the adaptability between the equipment and the demand.

[0052] Subsequently, the comprehensive evaluation values ​​of each available charging demand are sorted in descending order, with higher comprehensive evaluation values ​​indicating higher scheduling priority and priority allocation to mobile charging devices. For available charging demands with the same comprehensive evaluation value, a secondary sort is performed based on matching compatibility, with higher matching compatibility indicating higher priority in the secondary sort. Based on the final sorting results, available charging demands are divided into scheduling priority levels 1-5, with level 1 being the highest priority and level 5 being the lowest priority. Each priority level corresponds to a group of charging demands, clearly defining the scheduling order of each group of demands and providing a clear basis for subsequent scheduling decisions.

[0053] When a mobile charging device arrives at the coordinate point according to the dispatch instructions, but the user has already unplugged the charging gun or the vehicle has been moved (ghost order), the device immediately reports the on-site status to the system through a real-time interaction mechanism. After receiving the feedback, the system immediately triggers an on-site replanning process, searches for unassigned serviceable charging demands within a 30-meter radius of the device, recalculates the matching degree, marginal revenue, and dispatch priority, and assigns a new service task to the device. If there are no available demands in the vicinity, the device is dispatched back to the nearest temporary docking point and placed in an idle standby state to avoid wasting resources and energy costs by running the device empty.

[0054] A user credit profile is established, with the following credit deduction rules: For each confirmed ghost order (where the user cancels the order without a valid reason or there is no corresponding vehicle / charging requirement after the equipment arrives), 5 points will be deducted from the user ID, with an initial credit score of 100. The credit score levels and corresponding handling measures are as follows: Credit score ≥ 90 is Excellent, with no restrictions; 70-89 is Good, orders are accepted normally, with no additional restrictions; 50-69 is Average, a small deposit (e.g., 20 yuan, fully refundable upon service completion) is required when placing an order; < 50 is Poor, and the system automatically triggers a graceful rejection mechanism, no longer accepting any scheduling requests from that user until the user completes credit repair (e.g., submitting an explanation and waiting 7 calendar days without default). Simultaneously, the system will push a default reminder to the defaulting user, informing them of the default behavior and the change in credit score, fundamentally curbing malicious ordering behavior, ensuring system scheduling stability, and maintaining system reputation.

[0055] Furthermore, the specific process of scheduling decision-making is as follows: A two-layer asynchronous architecture of spatial constraints and priority decision-making is adopted, dividing the decision-making modules into two independently operating layers to achieve decoupling of spatial constraints and priority decision-making. The spatial constraint layer performs preliminary planning of the equipment operation path based on the reconstructed available charging area and time sequence change data, and delineates the spatial boundary of equipment operation. The priority decision-making layer optimizes the equipment scheduling order based on the scheduling priority and the list of equipment to be scheduled, using a collaborative decision-making algorithm to avoid equipment operation conflicts and complete the scheduling decision.

[0056] A two-layer asynchronous architecture of spatial constraints and priority decision-making is adopted. This architecture uses a dual-thread independent operation mode, dividing the scheduling decision into two independently operating decision modules: the spatial constraint layer and the priority decision layer. The two modules achieve real-time data synchronization through a data interaction interface, without interfering with each other. This decoupling design can prevent the entire scheduling system from paralyzing due to the failure of a single decision module, while improving the response speed of scheduling decisions. When the priority decision layer assigns a very high-priority task, but the spatial constraint layer finds that the only passage to the area is suddenly closed due to congestion or has a permanent roadblock, the spatial constraint layer sends a conflict warning to the priority decision layer through the data interaction interface, along with parameters such as the duration of the closed passage, information on alternative passages, and the location of the permanent roadblock. Upon receiving the warning, the priority decision layer immediately suspends the scheduling assignment of the high-priority task, and, in conjunction with the alternative passage information provided by the spatial constraint layer, re-optimizes the equipment scheduling order, adjusts the service equipment and running path of the task, completes the renegotiation, synchronously updates the scheduling decision results, and sends them to the relevant equipment, thus resolving the conflict.

[0057] The spatial constraint layer, based on reconstructed available charging areas, high-risk areas, and time-series change data, obtains real-time boundary information, sub-region division information, and time-series availability data for available charging areas and high-risk areas. Combined with permanent roadblock marking results, the A-Star algorithm is employed. This algorithm optimizes the cost function by incorporating site access status, high-risk area confidence scores, and penalty coefficients adjusted through an online learning mechanism. Dynamic soft penalties (penalty coefficient dynamically adjusted from 1.5 to 4 times) are applied to each high-risk area, while an infinite penalty (completely blocking the path) is applied to areas containing permanent roadblocks. This guides devices to prioritize paths within available space. If no feasible path exists within available space, a path in a high-risk area is then selected. This ensures that even if devices encounter sudden obstacles (such as LSTM prediction bias or illegally parked bicycles), they can complete scheduling by taking detours, avoiding dead ends and preventing repeated attempts to access areas with permanent roadblocks. Combining site access status and permanent roadblock distribution, preliminary operating paths are planned for each mobile charging device to be scheduled. Simultaneously, the spatial boundaries for device operation are defined, ensuring that devices do not exceed the available charging area and access range, thus avoiding conflicts between device operation and site conditions.

[0058] Based on the generated scheduling priorities, the priority decision layer combines the list of devices to be scheduled, the grid health status, and power degradation scheduling requirements to optimize the allocation of device scheduling order using a distributed collaborative decision-making algorithm. The distributed collaborative decision-making algorithm is executed as follows: Each mobile charging device acts as an independent decision node, generating its own optimal scheduling order and preliminary target path based on its location, remaining battery power, movement speed, and service range, targeting the charging demand with the highest priority. Each decision node broadcasts its location, planned path, target demand point, and occupied time window to other devices within a certain radius, while simultaneously receiving status and path information from neighboring devices, forming a local collaborative information set. Based on this local collaborative information set, the preliminary paths and time windows of each device are compared to identify path intersections, road segment occupancy, channel congestion, and other issues. The system identifies conflict types such as resource contention and marks conflict locations, time periods, and equipment numbers. It adjusts the scheduling order of conflicting equipment according to negotiation rules prioritizing high scheduling priority, high demand urgency, grid constraints, and low path cost. For conflicting road sections, it implements time-sharing passage, route detours, or demand reallocation to ensure only one device is allowed to pass through the same channel at the same time. Combining grid health status and power degradation scheduling requirements, it calculates the total charging power demand for each region and time period, limits the concurrent number of high-power charging tasks, and evenly distributes high-power demand across different time slots to keep grid load within safe thresholds. Through multiple rounds of local iterative adjustments, it eliminates all path conflicts, time conflicts, and power overload issues, achieving a globally conflict-free and globally optimal state for the scheduling order and operating paths of all equipment.

[0059] Through the aforementioned distributed collaborative decision-making algorithm, the scheduling order of equipment is optimized, and conflicts such as equipment path intersection and resource preemption are predicted and avoided. At the same time, the grid load pressure is taken into account to avoid multiple devices performing high-power charging tasks at the same time, ensuring that each device operates without conflict and the grid load is within a safe range. This results in a scheduling decision that includes core information such as the scheduling order of each device, operating path, service request number, marginal revenue assessment result, dynamic pricing coefficient, grid health adaptation status, and power limit parameters. This decision is then sent synchronously to the spatial constraint layer and subsequent instruction generation module to complete the scheduling decision.

[0060] When the collaborative decision-making algorithm gets stuck in an infinite loop of conflict prediction, or when the calculation time exceeds a set threshold (e.g., 500ms) and still fails to arrive at a conflict-free solution, the system immediately triggers a timeout interrupt. All devices to be scheduled within the conflict area are handed over to the central processing unit (CPU) for unified management. The CPU then performs a lock-unlock operation based on a strategy combining static priority and first-come, first-served. Specifically, devices are first sorted by scheduling priority, and those with the same priority are sorted by the time the scheduling request was initiated. Device running paths are locked and conflict nodes are unlocked in sequence, instantly breaking the deadlock or livelock impasse. After the impasse is broken, the CPU returns control of the devices to the distributed collaborative decision-making algorithm and resumes the normal scheduling process.

[0061] Furthermore, the process of obtaining the final scheduling instruction is as follows: Based on the scheduling decision results, distributed collaborative control is performed on each mobile charging device to be scheduled, and a real-time interaction mechanism between devices is established; a redundancy calculation algorithm is used to perform multi-path verification on the initially planned device operation path, eliminating invalid paths that exceed the available charging area or have congestion or conflict risks, and selecting the optimal operation path; the optimal operation path, scheduling priority, and device task allocation information are integrated to generate specific scheduling instructions for each mobile charging device, and the final scheduling instruction is output after instruction verification is completed.

[0062] like Figure 4 As shown, based on the obtained scheduling decision results, distributed collaborative control is implemented for each mobile charging device to be scheduled. A distributed control protocol is adopted, specifically the DDS distributed control protocol, which features low latency and high reliability, and is suitable for real-time interaction between devices. A real-time interaction mechanism between devices is established, where each device independently executes its own control commands and shares its own operating status through this mechanism, including current location, operating speed, remaining battery power, human activity within the social radius, and charging power. To address the impact of multipath effects on communication in complex underground parking lots or dense building complexes, a communication keep-alive and error correction mechanism combining frequency hopping communication and CRC cyclic redundancy check is adopted. Frequency hopping communication is used to avoid signal interference, and CRC cyclic redundancy check is used to detect and correct errors in the data transmission process, ensuring that the communication latency of the real-time interaction mechanism is ≤100 milliseconds, and guaranteeing the real-time synchronization of device operating status.

[0063] A redundant computation algorithm is employed to perform multi-path verification on the initially planned equipment operation paths in the spatial constraint layer. This algorithm is the KSP (K Shortest Paths) algorithm, where redundancy refers to multiple alternative paths with non-overlapping geometric paths and no time window conflicts, used to ensure the verifiability of path verification. The algorithm uses a multi-path parallel computation approach, planning 3-5 alternative operation paths for each device. Combining available charging area and high-risk area time-series change data, channel traffic status data, confidence scores, penalty coefficients adjusted through online learning, and permanent roadblock marking results, the feasibility of each alternative path is verified. The verification focuses on the fault tolerance and traffic efficiency of paths in high-risk areas, and the path avoidance capabilities in areas with permanent roadblocks. Invalid paths include those that exceed the boundary of the available charging area, have a channel congestion level of ≥70%, conflict with other equipment paths (occupying the same channel segment within the same time window), have a high-risk area confidence level of <50%, or pass through areas with permanent roadblocks. After removing the above invalid paths, the optimal operating path is selected based on the shortest path, the highest traffic efficiency, no conflict, the best marginal benefit (including the impact of dynamic pricing), and compliance with the requirements of grid health and power degradation scheduling.

[0064] Subsequently, the system integrates core elements such as optimal operating path, scheduling priority, equipment task allocation information, marginal revenue assessment results, dynamic pricing coefficient, grid health adaptation status, power limitation parameters, audible and visual alarm activation conditions, and charging unit price after dynamic pricing to generate specific scheduling instructions for each mobile charging device. For scheduling tasks that fail to meet the marginal revenue assessment criteria (cost > revenue × dynamic threshold coefficient) or do not meet the power degradation scheduling requirements, a graceful rejection or waiting for a better match mechanism is triggered during the instruction generation stage. The scheduling of this task is suspended, and a new suitable charging demand is retrieved to ensure the economic efficiency of system operation and grid safety. A combination of logical verification and scenario adaptation verification is used to verify the specific scheduling instructions. This verifies the consistency between the instruction content and the scheduling decision results, the status of available charging areas, equipment status, privacy compliance requirements, dynamic pricing rules, user credit rating, grid health status, and the distribution of permanent roadblocks. After successful verification, the final scheduling instructions are accurately sent to each mobile charging device to be scheduled via the wireless communication module, ensuring the accuracy, compliance, economy, stability, and grid risk control capabilities of the instruction execution, thus completing the acquisition of the final scheduling instructions.

[0065] 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.

[0066] 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.

[0067] 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 mobile charging scheduling method based on time-series data analysis, characterized in that, The method includes: By collecting site information and charging demand time-series data from multiple source sensors, the site availability is analyzed, and unusable spaces are eliminated based on the analysis results to reconstruct available charging areas. Using available charging areas as constraints, filter the available charging demand within the area, and determine the mobile charging equipment to be dispatched based on the charging demand. For the selected charging demands, a multi-dimensional scheduling evaluation model is constructed. The scheduling evaluation value of each charging demand is calculated, and the scheduling priority is generated based on the scheduling evaluation value. Based on scheduling priority, a two-layer asynchronous architecture of space constraints and priority decision is adopted for scheduling decisions. Invalid paths are eliminated through distributed collaboration and redundant computation, and the final scheduling instruction is output.

2. The mobile charging scheduling method based on time-series data analysis according to claim 1, characterized in that, The process for collecting the site information and the charging demand time-series data is as follows: Multi-source sensors are activated to simultaneously collect site status data and charging demand time-series data. Site status data includes parking space location, passageway status and time-series change data, while charging demand time-series data includes charging request time, power demand and time-series change trend. A multi-source data fusion and denoising algorithm is used to deduplicatize and denoise the collected raw data, and remove invalid interference data and abnormal fluctuation data. Add collection timestamps to the processed valid data, organize them into a complete time-series data set according to the time sequence, and establish the time-series correlation of the data.

3. The mobile charging scheduling method based on time-series data analysis according to claim 2, characterized in that, The specific process for reconstructing the available charging area is as follows: Based on time series data sets, a time series feature extraction algorithm is used to extract the time series change features and stability features of site conditions; Set a threshold for site availability and a standard for continuous availability. Combine this with the characteristics of temporal changes to delineate the boundaries between available and unavailable spaces, mark unavailable spaces, and remove them. By integrating the location information, temporal variation characteristics, and continuous availability of available space, a spatial reconstruction algorithm is used to reconstruct a dynamically adaptable available charging area that adapts to temporal changes.

4. The mobile charging scheduling method based on time-series data analysis according to claim 3, characterized in that, The specific process for determining the mobile charging device to be scheduled is as follows: Obtain the reconstructed available charging area boundary information and time-series availability data, and filter out available charging demands that are within the available area and during the available time period; Collect data on the current location, remaining power, load capacity, and moving speed of each mobile charging device to establish a device status database; A matching algorithm is used to combine the location of available charging needs, power demand and device status data to perform precise matching, select suitable devices as mobile charging devices to be dispatched, and form a list of devices to be dispatched.

5. The mobile charging scheduling method based on time-series data analysis according to claim 4, characterized in that, The specific process of constructing the multi-dimensional scheduling evaluation model is as follows: Based on available charging demand and mobile charging devices to be dispatched, the core evaluation dimensions are determined. Define quantitative standards for each evaluation dimension, transform qualitative indicators into calculable quantitative parameters, and integrate parameters from each dimension to construct a multi-dimensional scheduling evaluation model. The weight allocation rules for each evaluation dimension are set using the analytic hierarchy process (AHP), and the model parameters are calibrated in conjunction with the actual scheduling scenario to ensure that the model adapts to the dynamic changes in available charging areas.

6. The mobile charging scheduling method based on time-series data analysis according to claim 5, characterized in that, The process of obtaining the scheduling priority is as follows: The quantitative parameters of available charging demand and the adaptation parameters of the mobile charging equipment to be scheduled are substituted into the multi-dimensional scheduling evaluation model. The comprehensive evaluation value of each available charging demand is obtained through model calculation, and the matching degree between each demand and the equipment to be scheduled is calculated simultaneously. The comprehensive evaluation values ​​are sorted in descending order. For demands with the same evaluation value, a second sort is performed based on the matching degree. The scheduling priority of each charging demand is generated according to the sorting results.

7. The mobile charging scheduling method based on time-series data analysis according to claim 6, characterized in that, The specific process of the scheduling decision is as follows: A two-layer asynchronous architecture of spatial constraints and priority decision-making is adopted, which divides the decision-making modules into two independent layers to decouple spatial constraints and priority decision-making. The spatial constraint layer performs preliminary planning of the equipment operation path based on the reconstructed available charging area and time-series change data, and delineates the spatial boundary of equipment operation. The priority decision-making layer optimizes the equipment scheduling order based on the scheduling priority and the list of equipment to be scheduled, and completes the scheduling decision by using a collaborative decision-making algorithm to avoid equipment operation conflicts.

8. The mobile charging scheduling method based on time-series data analysis according to claim 7, characterized in that, The process of obtaining the final scheduling instruction is as follows: Based on the scheduling decision results, distributed collaborative control is carried out on each mobile charging device to be scheduled, and a real-time interaction mechanism between devices is established. A redundancy calculation algorithm is used to perform multi-path verification on the initially planned equipment operation path, eliminate invalid paths that are outside the available charging area or have congestion or conflict risks, and select the optimal operation path. Integrate the optimal running path, scheduling priority, and device task allocation information to generate specific scheduling instructions for each mobile charging device, and output the final scheduling instructions after completing instruction verification.

9. A mobile charging scheduling system based on time-series data analysis, characterized in that, The system is used to execute the mobile charging scheduling method based on time-series data analysis as described in any one of claims 1-8.

Images