A mobile storage and charging system scheduling and path planning method based on multi-objective optimization

Abstract

The application relates to the technical field of mobile storage and charging collaborative scheduling, and particularly discloses a mobile storage and charging system scheduling and path planning method based on multi-target optimization, which comprises the following steps: in response to charging demand requests in a region and state data of renewable energy access points, multi-source data collection and collaborative preprocessing are performed to generate a system state set containing time and space characteristics; based on the system state set, a multi-target optimization system containing a user side, an operation side, a power grid side, an energy storage side and a renewable energy consumption side is constructed, and a conflict resolution factor is introduced to balance the conflict relationship among the targets. By constructing a multi-dimensional target system containing users, operations, power grids and renewable energy consumption and introducing a hierarchical collaborative optimization and conflict resolution mechanism, the comprehensive scheduling efficiency and robustness of the system in a complex dynamic environment are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of mobile energy storage and charging collaborative scheduling technology, specifically a scheduling and path planning method for mobile energy storage and charging systems based on multi-objective optimization. Background Technology

[0002] With the surge in the number of new energy vehicles and the large-scale grid connection of distributed renewable energy, traditional fixed charging infrastructure is proving inadequate in dealing with sudden charging demands due to uneven spatial and temporal distribution. Mobile charging and storage systems have emerged as a solution. In existing technologies, the scheduling of mobile charging and storage vehicles typically relies solely on single-objective path optimization algorithms, such as shortest path or minimum time, or treats them simply as mobile power sources, failing to deeply integrate them into the interactive system of source-grid-load-storage.

[0003] However, in actual operation, mobile energy storage and charging systems face complex challenges due to the coupling of multiple physical fields: on the one hand, the output of distributed renewable energy is highly intermittent and volatile, while users' charging needs are also highly random in time and space, making it difficult to match the two; on the other hand, traffic congestion in urban road networks directly affects vehicle energy consumption and arrival time, making scheduling strategies based solely on geographical distance often ineffective.

[0004] Therefore, the core technological conflict that urgently needs to be resolved is: how to strike a balance between pursuing the operational economy of mobile energy storage and charging systems and ensuring the mobile survivability of vehicles, and how to solve the spatiotemporal misalignment between logical resource release and physical power grid response in abnormal takeover scenarios.

[0005] Specifically, when performing high-yield discharge tasks, existing technologies often overlook the surge in energy consumption caused by sudden changes in road network traffic conditions, which can easily lead to vehicles being locked in congested sections due to depleted power. At the same time, when handling task handover for equipment failures, due to the lack of consideration for the physical inertia of the power grid, simple resource release commands can easily cause power contention or superposition between old and new tasks, thereby triggering the overload protection tripping of the distribution network.

[0006] Therefore, how to construct a scheduling method that can coordinate multi-dimensional objectives, has a traffic-energy deadlock prevention mechanism, and can achieve seamless atomic takeover is a key issue of concern for those skilled in the art. Summary of the Invention

[0007] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a scheduling and path planning method for mobile energy storage and charging systems based on multi-objective optimization, so as to achieve safe, efficient, and continuous operation of mobile energy storage and charging systems under multi-domain coupling.

[0008] To achieve the above objectives, a first aspect of the present invention proposes a scheduling and path planning method for a mobile energy storage and charging system based on multi-objective optimization, applied to the central control system of the mobile energy storage and charging system, comprising:

[0009] In response to charging demand requests within the region and status data from renewable energy access points, multi-source data acquisition and collaborative preprocessing are performed to generate a system status set containing spatiotemporal characteristics;

[0010] Based on the system state set, a multi-objective optimization system is constructed, which includes the user side, operation side, grid side, energy storage side and renewable energy consumption side, and a conflict resolution factor is introduced to balance the conflict relationship between the objectives.

[0011] The multi-objective optimization system is input into the hierarchical collaborative optimization model for solution, and the output is the scheduling decision and path planning scheme for mobile storage and charging vehicles;

[0012] The scheduling decision and path planning scheme are sent to the mobile energy storage vehicle to trigger the mobile energy storage vehicle to perform charging tasks, dynamic obstacle avoidance and renewable energy coordinated charging and discharging operations;

[0013] The solution process of the hierarchical collaborative optimization model includes:

[0014] Initiate global layer optimization, using the first preset time period as a benchmark, and based on the spatiotemporal prediction data of renewable energy, determine the regional allocation of mobile energy storage and charging vehicles and the charging priority of access points, and generate a global path framework.

[0015] Local layer optimization is initiated, and based on the second preset time period, the service sequence and obstacle avoidance path of a single mobile energy storage and charging vehicle are refined according to the real-time output data of renewable energy and the dynamic obstacle information of the road network; the second preset time period is shorter than the first preset time period.

[0016] In response to a sudden state signal during system operation, an emergency adjustment process is triggered to perform power adjustment within a first preset response time or task redistribution within a second preset response time.

[0017] To achieve the above objectives, a second aspect of the present invention proposes a mobile energy storage and charging system scheduling and path planning system based on multi-objective optimization, applied in the central control system of the mobile energy storage and charging system, comprising:

[0018] The data collaborative processing module is used to respond to charging demand requests and renewable energy access point status data within the region, perform multi-source data acquisition and collaborative preprocessing, and generate a system status set containing spatiotemporal characteristics.

[0019] A multi-objective construction module is used to construct a multi-objective optimization system based on the system state set, including the user side, operation side, grid side, energy storage side and renewable energy consumption side, and introduces a conflict resolution factor to balance the conflict relationship between the objectives.

[0020] The hierarchical optimization solution module is used to input the multi-objective optimization system into the hierarchical collaborative optimization model for solution, and output the scheduling decision and path planning scheme for mobile storage and charging vehicles; the hierarchical collaborative optimization model is configured to execute three-level optimization logic including global layer optimization, local layer optimization and emergency adjustment;

[0021] The scheduling and execution module is used to send the scheduling decision and path planning scheme to the mobile energy storage vehicle to trigger the mobile energy storage vehicle to perform charging tasks, dynamic obstacle avoidance and renewable energy collaborative charging and discharging operations.

[0022] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for scheduling and path planning of a mobile storage and charging system based on multi-objective optimization.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] This invention discloses a multi-objective optimization-based scheduling and path planning method for mobile energy storage and charging systems. By constructing a multi-dimensional objective system encompassing users, operations, the power grid, and renewable energy consumption, and introducing hierarchical collaborative optimization and conflict resolution mechanisms, it significantly improves the overall scheduling efficiency and robustness of the system in complex dynamic environments. Its beneficial effects are mainly reflected in:

[0025] First, by integrating the spatiotemporal output characteristics and fluctuation risk levels of renewable energy, deep synergy between mobile energy storage and charging vehicles and green energy has been achieved, effectively improving the renewable energy consumption rate in the region.

[0026] Secondly, by using Riemannian manifolds to map road network traffic characteristics into energy potential, and combining dynamic return energy consumption threshold verification, a mobility safety lock was constructed to effectively prevent vehicles from falling into a deadlock state due to deteriorating road conditions when performing discharge tasks.

[0027] Finally, in the anomaly handling stage, an innovative resource atomization handover mechanism based on shadow tokens was adopted to resolve the time delay conflict between logical switching and physical response, ensuring a smooth power transition of the distribution network during task hot switching, avoiding grid failures caused by load changes or resource preemption, and realizing the safe, efficient and continuous operation of the mobile energy storage and charging system under the multi-domain coupling of source-grid-load-storage-circuit. Attached Figure Description

[0028] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0029] Figure 1 This is a flowchart illustrating the scheduling and path planning method for mobile energy storage and charging systems based on multi-objective optimization provided by the present invention.

[0030] Figure 2 This is a comparison chart of the original values ​​and the effect of moving average smoothing processing of real-time renewable energy output data in the mobile energy storage and charging system scheduling and path planning method based on multi-objective optimization provided by this invention;

[0031] Figure 3 This is a simulation diagram of the three-dimensional Pareto Front distribution under the multi-objective optimization system in the scheduling and path planning method for mobile energy storage and charging systems based on multi-objective optimization provided by the present invention.

[0032] Figure 4 This is a comparison curve of the convergence speed of the optimization algorithms based on transfer learning and random initialization in the multi-objective optimization-based mobile storage and charging system scheduling and path planning method provided by this invention.

[0033] Figure 5 This is a diagram showing the nonlinear mapping relationship between the Riemann curvature tensor modulus and the road network friction coefficient in the multi-objective optimization-based mobile energy storage and charging system scheduling and path planning method provided by this invention.

[0034] Figure 6 This is a simulation diagram of the mobile energy storage and charging vehicle deadlock risk area determination based on dynamic return energy consumption threshold in the mobile energy storage and charging system scheduling and path planning method based on multi-objective optimization provided by the present invention.

[0035] Figure 7 This is a schematic diagram illustrating the implementation of the mobile energy storage and charging system scheduling and path planning system based on multi-objective optimization provided by the present invention.

[0036] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0037] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0038] The following description, with reference to the accompanying drawings, describes a method, system, and electronic device for scheduling and path planning of a mobile energy storage and charging system based on multi-objective optimization, according to embodiments of the present invention.

[0039] Example 1:

[0040] This embodiment describes a scheduling and path planning method for mobile energy storage and charging systems based on multi-objective optimization. This method is applied to the central control system of a mobile energy storage and charging system, achieving intelligent scheduling of mobile energy storage and charging resources through real-time communication with sensor networks distributed within the region, mobile energy storage and charging vehicle terminals, power distribution network monitoring equipment, and distributed energy monitoring systems. Figure 1 As shown, the method in this embodiment specifically includes the following steps:

[0041] Step S1: Multi-source data acquisition and collaborative preprocessing.

[0042] The central control system first responds to charging demand requests within the region and status data from renewable energy access points, performing multi-source data acquisition and collaborative preprocessing to generate a system state set containing spatiotemporal characteristics. This forms the perception foundation of the entire dispatching system.

[0043] For example, the system collects charging orders from users within the area in real time via 5G communication modules or fiber optic networks. These orders include information such as the location of the demand, the amount of electricity required, and the expected service time window. Simultaneously, the system connects to the SCADA system of the power distribution network to acquire road network traffic data (such as traffic flow and average vehicle speed), power distribution network operation data (such as node voltage and transformer load rate), and status data of mobile energy storage and charging vehicles (such as current location, state of charge (SOC), and state of battery health (SOH)). Furthermore, the system also needs to collect data related to distributed renewable energy sources to support subsequent source-load-storage coordination.

[0044] For data related to distributed renewable energy, the central control system specifically collects the following data items: real-time output data of renewable energy access points, spatiotemporal forecast data, access point information, and meteorological data. Access point information includes the geographical coordinates, rated capacity, and grid connection voltage level of the access point; meteorological data includes real-time meteorological parameters such as solar irradiance, wind speed, wind direction, temperature, and cloud cover in the area.

[0045] Optionally, to eliminate noise interference in the raw data, the system performs collaborative preprocessing on the collected data. For the real-time output data of renewable energy access points, the system uses a moving average method for smoothing to suppress data fluctuation noise. Specifically, the system sets a preset time window size, such as 5 minutes. The system maintains a data queue of length N, where N corresponds to the number of sampling points within 5 minutes. For the smoothed output value at the current time t... Its calculation logic is based on the original output values ​​of the past N sampling points. Calculate the arithmetic mean. This processing method can effectively filter out second-level power fluctuations caused by gusts or rapidly passing clouds, allowing subsequent optimization models to be calculated based on a more stable power trend.

[0046] To delve deeper into the features of the data, the system invokes a pre-built spatial feature extraction network model to extract the spatial features of renewable energy output. This spatial feature extraction network model adopts the EfficientNet architecture. EfficientNet uses a composite scaling method to balance the scaling of the network in the three dimensions of depth, width, and resolution, thereby reducing the computational load while maintaining the accuracy of feature extraction. The system maps the output status of each renewable energy access point in the region to a two-dimensional or three-dimensional mesh map and inputs it into the EfficientNet model. The feature vector output by the model represents the spatial distribution pattern of renewable energy output in the region at the current moment.

[0047] Next, the system simultaneously invokes a pre-built time-series feature extraction network model to extract the time-series features of renewable energy output. This time-series feature extraction network model employs a Long Short-Term Memory (LSTM) network. By incorporating input gates, forget gates, and output gates, the LSTM network can effectively capture long-term dependencies in time-series data. The system inputs historical output data sequences over a past period into the LSTM model, and the model's output hidden state vector represents the trend of output changes, such as being in a ramp-up phase or about to enter a trough.

[0048] Subsequently, the system combines the aforementioned meteorological data and extracted spatiotemporal features to generate a renewable energy output fluctuation risk level. This risk level quantifies the uncertainty of renewable energy output at the current moment and incorporates it into the system state set. The risk level is divided into at least three preset levels, such as low risk, medium risk, and high risk. When meteorological data is stable and LSTM prediction error is small, it is marked as low risk; when meteorological data shows signs of abrupt change but output is currently stable, it is marked as medium risk; when the weather is severe and output fluctuates drastically, it is marked as high risk. This risk label directly affects the degree of relaxation of safety constraints in subsequent optimization algorithms.

[0049] like Figure 2 This chart compares the raw values ​​of real-time renewable energy output data with the smoothing effect of moving average processing. The horizontal axis represents the time dimension in minutes, and the vertical axis represents the power output value of the renewable energy access point in kilowatts.

[0050] Figure 2The thin, volatile blue curve represents the raw power output data collected in real time from the distributed energy monitoring equipment. The curve exhibits a dense and large-amplitude sawtooth shape, which intuitively reflects the high-frequency power fluctuation noise at the second level caused by gusts of wind, rapid cloud movement and obstruction, or measurement errors of the equipment itself in the actual operating environment. If such violent random fluctuations are directly input into the subsequent optimization model, it will lead to non-convergence of the calculation results or oscillation of the scheduling instructions.

[0051] Figure 2 The thicker red curve with a stable trajectory represents the smoothed power output data after processing by the moving average method in the collaborative preprocessing step described in this invention. It can be seen that the red curve accurately matches the overall trend of the original blue data, that is, it retains the real energy flow characteristics such as the ramp-up or decline of renewable energy output over time, while effectively filtering out high-frequency random noise interference, making the power curve smooth and continuous.

[0052] The comparison shows that at a specific moment, such as around the 30th minute, the original data may fluctuate dramatically due to instantaneous interference, while the smoothed data remains on a stable trend line. This proves that the data preprocessing mechanism adopted in this invention can significantly improve the data quality of the system state set, laying a solid foundation for the subsequent extraction of accurate spatial and temporal features, thereby ensuring the stability of multi-objective optimization solutions and the reliability of scheduling decisions.

[0053] Step S2: Construction of a multi-objective system and conflict resolution.

[0054] After completing data acquisition and preprocessing, the central control system constructs a multi-objective optimization system based on the generated system state set, which includes the user side, operation side, grid side, energy storage side and renewable energy consumption side, and introduces conflict resolution factors to balance the conflict relationship between the various objectives.

[0055] For example, the system establishes a seven-dimensional multi-objective system, which is quantitatively characterized by a total objective function F. The mathematical expression of the total objective function F is as follows:

[0056] ;

[0057] In the formula, The objective function on the user side aims to maximize user satisfaction. Specifically, this function includes four sub-dimensions: minimizing waiting time, maximizing service coverage, maximizing charging response speed, and maximizing charging fairness. The system quantifies waiting time by calculating the time difference between a user's request and the vehicle's arrival, and quantifies service coverage by the proportion of users served to the total number of requests.

[0058] The objective function on the operation side focuses on the operator's economic benefits. This function includes operating costs, charging and discharging costs, penalty costs, and obstacle avoidance efficiency costs. Operating costs are directly proportional to the vehicle's total mileage; charging and discharging costs involve grid purchase costs and battery depreciation; penalty costs correspond to virtual fines for failing to meet user needs or violating grid dispatch instructions.

[0059] The objective function on the grid side aims to ensure the safe and stable operation of the distribution network. This function includes voltage qualification rate, transformer load rate, and peak-valley difference control. The system obtains the voltage deviation of each node through power flow calculation and strives to maintain the node voltage within the allowable range of the rated voltage. At the same time, the optimization objective aims to smooth the load fluctuations of the distribution network and achieve peak shaving and valley filling.

[0060] The objective function for the energy storage side is to minimize the number of energy storage cycles. To extend the lifespan of the onboard battery in mobile charging and storage vehicles, this objective function penalizes frequent shallow charge and discharge operations, guiding the vehicle to perform scientific charge and discharge cycles.

[0061] The objective function for dynamic obstacle avoidance is to maximize the success rate of dynamic obstacle avoidance. The smaller the value of this objective function, the lower the probability of encountering dynamic obstacles such as temporary construction or traffic accidents in the planned path, and the higher the passage efficiency of the path.

[0062] The objective function for abnormal losses is to minimize the combined losses of users, power grid, and equipment under abnormal scenarios. This serves as a defensive objective, guiding the scheduling strategy towards lower-risk areas when the system anticipates potential failure risks.

[0063] The objective function for renewable energy consumption aims to improve the utilization rate of clean energy. Its expression can be given as:

[0064] ;

[0065] In the formula, Represents the renewable energy absorption rate within the region; This represents the total charging power of all mobile charging vehicles at the current moment; This represents the total real-time power output of all renewable energy access points within the region; This represents the maximum allowable power curtailment in the distribution network. The design logic of this objective function is: the more power charged by mobile energy storage vehicles, and the closer it gets to the sum of the real-time output of renewable energy and the allowable power curtailment, the higher the absorption rate. The higher the value, the better the objective function value. The smaller the value, the better it aligns with the minimization optimization approach.

[0066] to The weights are dynamic and satisfy the following conditions: To adapt to different energy scenarios, the system dynamically adjusts the weights based on the real-time monitoring of renewable energy output status. When the system detects an oversupply of renewable energy, i.e., supply exceeds demand, it will... The value is set to 0.25, thereby significantly increasing the importance of the consumption target and guiding vehicles to the charging point; when the output is normal, The value is 0.15; when the output is insufficient, When the value is 0.1, the system focuses more on meeting user needs or ensuring grid stability.

[0067] Optionally, to resolve the inherent contradictions among multiple objectives, such as the possibility that minimizing user waiting time may increase operational costs, the system introduces a conflict resolution factor. This factor is used to dynamically correct the conflict between objectives, and its value range is limited to [0.8, 1.2]. During the optimization process, the system calculates the gradient direction of each sub-objective function. When the angle between the gradient directions of two objectives exceeds a preset threshold, it indicates a strong conflict, and the system adjusts accordingly. The value is nonlinearly scaled to the weights, thereby finding a compromise solution on the Pareto front and avoiding the solution from getting trapped in a local optimum of a single objective.

[0068] like Figure 3 This is a simulation diagram of the three-dimensional Pareto front distribution under a multi-objective optimization system. The diagram visually illustrates the solution set distribution of the present invention when handling multi-dimensional objective conflicts using a three-dimensional coordinate system. The X-axis represents user waiting time in minutes; a smaller value indicates a better user experience. The Y-axis represents operating costs in yuan; a smaller value indicates higher economic efficiency. The Z-axis represents the renewable energy absorption rate as a percentage; a larger value indicates more efficient utilization of green energy.

[0069] Figure 3 Each scatter point in the diagram represents a non-dominated solution, i.e., a Pareto optimal solution, obtained after calculation by the hierarchical collaborative optimization model. These scatter points together form a surface that bulges upward into space, vividly revealing the mutual constraints and trade-offs among the three core objectives. That is, in order to pursue an extremely high renewable energy consumption rate, it is often necessary to sacrifice a certain amount of user waiting time for spatiotemporal matching, or increase operating costs to perform reverse power transmission operations.

[0070] Figure 3The color gradient of the scatter points from blue to yellow corresponds to the numerical change of the conflict resolution factor. This factor dynamically takes values ​​within a preset range. The brighter the color, the larger the value of the conflict resolution factor, and the more the system decision weights will favor the renewable energy consumption target. Conversely, the darker the color, the more it will focus on traditional indicators on the user side or the operation side.

[0071] As can be seen from the simulation diagram, the multi-objective optimization system proposed in this invention can effectively search for a series of continuous and uniformly distributed optimal compromise solutions in the complex solution space. This proves that the introduction of the conflict resolution factor mechanism can flexibly find the best balance point among different stakeholders, avoiding the local extreme value trap caused by single-objective optimization, thereby ensuring that the mobile energy storage and charging system maximizes the utilization efficiency of clean energy while meeting user needs.

[0072] Step S3: Construction of a multi-constraint system.

[0073] To ensure the physical feasibility and security of the dispatching scheme, the central control system establishes an eight-dimensional constraint system, including vehicle resource constraints, road network traffic constraints, charging demand constraints, distribution network operation constraints, dynamic obstacle avoidance constraints, abnormal state constraints, optimization efficiency constraints, and renewable energy synergy constraints. This section focuses on elaborating the constraints related to renewable energy synergy and road network traffic.

[0074] Specifically, the system imposes renewable energy coordination constraints, including spatiotemporal output constraints, access point capacity constraints, and fluctuation adaptation constraints. Regarding spatiotemporal output constraints, the system limits mobile energy storage and charging vehicles to any renewable energy access point. Charging power during charging The following inequalities must be satisfied:

[0075] ;

[0076] In the formula, For this access point At any moment Real-time output; This is the local base load of the distribution transformer area where the access point is located. This constraint ensures that vehicles only use net capacity for charging and do not preempt the local load's power quota. At the same time, the system also limits the charging time of the vehicles. The output duration of this access point must be covered. This means maintaining connectivity as much as possible during the periods when renewable energy is available, in order to maximize its absorption.

[0077] Regarding access point capacity constraints, the system limits the total power of mobile charging vehicles that can be charged simultaneously at a single access point. The rated capacity of the access point must not be exceeded. The preset percentage threshold, such as setting it to 80%, is... This is to prevent transformer overload.

[0078] For fluctuation adaptation constraints, considering battery life and grid impact, the system limits the power change rate of mobile energy storage and charging vehicles during charging and discharging operations. Do not exceed the preset power ramp rate threshold, such as 5kW / minute.

[0079] Specifically, when constructing road network traffic constraints, the system employs an innovative dual-graph constraint mechanism combining a hierarchical topology map and a Riemannian manifold. First, the system constructs a hierarchical topology map, comprising a bottom layer, a middle layer, and a top layer. The bottom layer is constructed based on the geometric spatial segmentation of lattice and point clouds, describing the geometric boundaries of physical roads, lane lines, and static obstacles. The middle layer is constructed based on a semantic segmentation network and obstacle type labeling, marking areas with specific semantic attributes, such as school zones and construction zones. The top layer constructs a topology graph with passable areas as nodes and semantic risks as edge weights, used for macro-level path search. Based on this hierarchical topology map, the system establishes the first obstacle avoidance branch, which is planned primarily based on geometric distance and static semantic rules.

[0080] Secondly, the system constructs a Riemannian manifold graph. To handle highly dynamic traffic flow, the system maps the motion characteristics of dynamic obstacles, such as fast-moving vehicles and congested traffic, to curvature tensors in Riemannian geometry. Specifically, the system treats the road network as a Riemannian manifold, on which the metric tensor... Instead of simply using Euclidean distance, a weighted term related to traffic congestion and accident risk is introduced. The presence of dynamic obstacles causes the manifold to bend in corresponding local regions, the degree of which is determined by the Riemann curvature tensor. This is used to characterize the system. Based on the Riemannian manifold, a second obstacle avoidance branch is established. This branch uses geodesic equations for path search and can naturally bypass areas with high curvature, i.e., congested or high-risk areas.

[0081] When generating route planning schemes, the system monitors the Riemann curvature tensor values ​​of each node in the road network in real time. If the Riemann curvature tensor of a certain route segment exceeds a preset curvature threshold, the system will take action. If the system determines that the area has high dynamic risks, such as sudden severe congestion or chain accidents, it will trigger obstacle avoidance constraint logic, forcing the path planning algorithm to switch from the first obstacle avoidance branch based on the hierarchical topology map to the second obstacle avoidance branch based on the Riemann manifold for path replanning, ensuring that the vehicle can quickly leave the high-risk area.

[0082] Step S4: Solve the hierarchical collaborative optimization problem.

[0083] After constructing the objective function and constraints, the central control system inputs the multi-objective optimization system into a hierarchical collaborative optimization model for solution. This solution process employs a three-level optimization architecture: global coarse programming, local fine optimization, and emergency fast adjustment, combined with a transfer learning mechanism.

[0084] For example, the system first initiates global-level optimization. This optimization operates on a first preset time period, such as a 1-hour basis. Based on future spatiotemporal prediction data of renewable energy sources, such as the photovoltaic output curve for the next hour, the system uses an improved Non-Dominated Sorting Genetic Algorithm II (NSGA-II) to perform regional-level allocation of mobile energy storage and charging vehicles, and determines the charging priority of each vehicle at different renewable energy access points. The output is a comprehensive framework for vehicle scheduling across the entire region, clearly defining the approximate destination and task type of each vehicle.

[0085] The system then initiates local-level optimization, which operates on a second preset time period, such as 15 minutes, and this second preset time period is strictly shorter than the first preset time period. At this level, the system uses a lightweight near-end strategy optimization (PPO) model to fine-tune the specific behavior of individual mobile energy storage and charging vehicles based on real-time renewable energy output data (sampled at the second or minute level) and real-time dynamic obstacle information of the road network. The optimization includes specific service sequence ordering, precise setting of charging and discharging durations, and generation of local obstacle avoidance paths.

[0086] To improve solution efficiency, the hierarchical collaborative optimization model employs a transfer learning mechanism. The system pre-constructs a scenario transfer library, which stores feature parameters of various typical renewable energy scenarios and pre-trained model parameters for these scenarios. The specific characteristic parameters include, but are not limited to: the shape characteristics of the time-series curve of renewable energy output (such as peak and trough distribution), power fluctuation variance, output ramp-up rate, and spatial correlation index between different access points (used to characterize the direction of cloud movement), etc. Before solving the optimization problem at the current time t, the system also needs to calculate the real-time scene feature vector. Feature vectors of typical scenes in the scene transfer library The matching degree is calculated, for example, using cosine similarity. If the matching degree of a typical scene is found to be higher than a preset matching threshold (set to 70%), the system determines that the current scene is highly similar to historical experience. At this time, the system directly calls the pre-trained model parameters corresponding to that typical scene. Instead of starting with random values, the pre-trained model parameters are used to initialize the population or neural network weights of the current optimization algorithm. For the NSGA-II algorithm, which optimizes the global layer, the pre-trained model parameters refer to the preferred initial population distribution; while for the PPO model, which optimizes the local layer, the pre-trained model parameters refer to the weight matrices and bias terms of the policy network (Actor) and the value network (Critic).

[0087] In this way, the convergence speed of the optimization algorithm is greatly improved, and the number of iterations required (such as 20 times) can be reduced to less (such as 10 times), thereby meeting the timeliness requirements of real-time scheduling.

[0088] like Figure 4 This graph compares the convergence speed of the optimization algorithms based on transfer learning and random initialization. It visually demonstrates the significant improvement in convergence performance achieved by introducing the transfer learning mechanism in the hierarchical collaborative optimization process. The horizontal axis represents the number of iterations of the optimization algorithm, and the vertical axis represents the fitness value of the objective function; a lower fitness value indicates a better scheduling scheme.

[0089] Figure 4 The blue dashed line represents the convergence trajectory when using the traditional random initialization strategy. It can be seen that because the initial population is randomly generated, its initial fitness value is high and decreases slowly during the search process, only gradually stabilizing and converging to the optimal solution near the twentieth iteration (shown on the horizontal axis). In contrast, Figure 4 The solid red line represents the convergence trajectory after the introduction of the transfer learning mechanism in this invention. Because the system calculates the matching degree between real-time scene features and typical scene features in the scene transfer library, and directly calls the pre-trained model parameters corresponding to high-matching scenes to initialize the population, the initial fitness value of this curve is significantly lower than that of the blue curve, indicating a substantial improvement in the quality of the initial solution. Simultaneously, the red curve exhibits a very rapid downward trend, reaching a stable state and converging to the same optimal solution level as the blue curve near the tenth iteration (displayed on the horizontal axis).

[0090] The comparison of the two curves clearly demonstrates that the transfer learning mechanism adopted in this invention, by utilizing historical experience knowledge, successfully reduces the number of iterations required for the optimization algorithm from about twenty in the traditional way to about ten. This significantly shortens the computation time while ensuring the quality of the solution, and verifies that this invention has excellent timeliness and response speed when dealing with high-frequency real-time scheduling requirements.

[0091] Step S5: Scheduling and Dynamic Adaptation.

[0092] Once the optimization model outputs the optimal solution, the central control system will send the scheduling decision and path planning scheme to the mobile energy storage and charging vehicles to trigger the mobile energy storage and charging vehicles to perform charging tasks, dynamic obstacle avoidance, and renewable energy coordinated charging and discharging operations.

[0093] For example, during the execution phase, the system makes dynamic logical judgments based on electricity prices and energy status. The system determines in real time the current grid electricity price status (peak, flat, or valley) and the renewable energy output status (surplus, balance, or shortage). If the system determines that the current electricity price is in a valley phase and renewable energy output is in surplus, it will generate a deep charging command. This command instructs the mobile energy storage vehicle to extend its charging time at the access point, for example, by 30% beyond the original plan, and to charge the battery's state of charge (SOC) to a first preset SOC threshold (such as 80% or 90%) or higher. This operation utilizes low-priced electricity while maximizing the absorption of clean energy.

[0094] If the system determines that the current electricity price is at its peak and renewable energy output is insufficient, meaning that the grid supply is tight and clean energy cannot compensate, the system will generate a discharge service instruction. This instruction directs the mobile energy storage vehicle to stop drawing power from the grid and instead prioritize using the energy stored in its own battery to supply power to users (V2L mode) or to transmit power back to the grid (V2G mode). Simultaneously, the instruction will shorten the vehicle's time at the access point or directly stop charging operations to alleviate the grid burden and generate operating revenue by taking advantage of the high peak electricity price difference.

[0095] Optionally, before executing the discharge service command, the system will also perform a mobility safety verification process based on road network traffic potential energy. The system calculates the minimum energy consumption for a vehicle to travel from its current location to the nearest guaranteed refueling point based on a Riemannian manifold graph. And introduce the road network friction coefficient To adjust energy consumption estimates. Only when the vehicle's estimated remaining battery power... satisfy ,in Discharge is only permitted when the risk level is high; otherwise, a forced keep-alive mode is triggered, discharge is prohibited, and the vehicle is guided to evacuate.

[0096] Step S6: Exception handling and collaborative feedback.

[0097] The system continuously monitors its status during operation, and triggers a processing procedure immediately if any abnormality occurs.

[0098] For example, in response to sudden state signals during system operation, such as sudden changes in renewable energy output, connection point failures, or distribution network voltage exceeding limits, the system triggers an emergency adjustment procedure. This procedure performs power adjustment within a first preset response time (e.g., milliseconds) or task reallocation within a second preset response time (e.g., within 1 minute).

[0099] Specifically, the anomaly handling process includes anomaly identification, secondary confirmation, task takeover, and resource release. The central control system monitors the status of mobile charging vehicles and communication links in real time through redundant communication modules, switching to a backup channel if the primary channel fails. Then, when anomalies such as lost heartbeat packets or out-of-limit parameters are detected, a primary anomaly signal is generated. To prevent false alarms, the system responds to the primary anomaly signal by triggering a paging process for the associated mobile charging vehicles. The system sends a specific status query command to the vehicle to perform secondary status confirmation. If the vehicle still does not respond or the anomaly is not resolved after the paging process is completed, the system determines that the anomaly status has been confirmed.

[0100] Subsequently, the system initiates the task takeover process. Within a preset allocation period (e.g., 1 minute), the system searches for healthy mobile charging / storage vehicles that are in normal operation and whose location meets a preset scheduling threshold, i.e., are sufficiently close, within the currently idle or low-priority vehicle pool. The system then reassigns the charging tasks associated with the abnormal vehicles to the healthy vehicles.

[0101] Simultaneously, the system performs resource release operations. To avoid grid fluctuations, the system employs a shadow locking mechanism: instead of immediately releasing the power quota occupied by the abnormal vehicle at the physical layer, it marks it as a shadow state and generates a resource inheritance token, which is then issued to the takeover vehicle. When the takeover vehicle arrives and submits the token, the edge computing unit performs atomic identity replacement, achieving seamless takeover.

[0102] In summary, this embodiment achieves efficient and safe scheduling of mobile energy storage and charging systems in complex dynamic environments by constructing a multi-dimensional objective system, introducing conflict resolution factors, adopting a hierarchical optimization architecture and transfer learning mechanism, and combining Riemannian manifold diagrams with renewable energy collaborative strategies. This method not only solves the multi-objective conflict problem but also significantly improves the system's robustness and renewable energy absorption capacity through refined anomaly handling and constraint control.

[0103] Example 2:

[0104] This embodiment focuses on detailing the safety assurance mechanism of mobile energy storage and charging systems when performing high-risk discharge tasks. In actual mobile energy storage and charging scheduling scenarios, especially when the grid is experiencing peak electricity prices and renewable energy output is insufficient, the system tends to schedule mobile energy storage and charging vehicles to perform V2G or V2L discharge services to obtain economic benefits and alleviate grid pressure. However, if economic benefits are pursued solely while ignoring the dynamic changes in road network traffic conditions, it is highly likely that the remaining power of the vehicle after discharge will be insufficient to support its departure from congested areas, thereby causing a serious energy-traffic deadlock accident. To address this, this embodiment elaborates on a mobility safety verification process based on road network traffic potential energy. This process, as a prerequisite for generating discharge service instructions, is integrated into the safety decision module of the central control system.

[0105] For example, before the central control system issues a discharge service command to a mobile charging vehicle, the system first initiates a mobility safety verification process based on road network traffic potential energy. The core of this process lies in introducing the physical concept of traffic potential energy, quantifying the ease of passage through the road network as a potential barrier to energy consumption. The system no longer uses geographical distance or simply time as a metric, but instead utilizes the methodology of Riemannian geometry to reconstruct the energy properties of the road network. Based on a pre-constructed Riemannian manifold, the system calculates the dynamic return energy consumption threshold in real time. Here, the Riemannian manifold treats the road network as a curved space, where the degree of curvature reflects changes in traffic flow density and speed.

[0106] To accurately calculate the dynamic return trip energy consumption threshold, the system first needs to determine the minimum energy consumption required for a mobile charging and storage vehicle to travel from its current location to the nearest guaranteed recharge node. Here, the guaranteed recharge node is strictly defined as a charging station or battery swapping station within the road network that is closest to the vehicle's current location, is currently free from queuing or congestion, and possesses dedicated power supply resources. This node selection not only considers spatial proximity but also emphasizes the certainty and immediacy of recharge resources, ensuring that the vehicle can immediately recharge upon arrival without facing the uncertain risk of queuing.

[0107] Next, in calculating the aforementioned minimum energy consumption value, the system introduces the key parameter of the road network friction coefficient. The road network friction coefficient is a dimensionless physical quantity used to characterize the hindering effect of the road network environment on vehicle energy consumption. In traditional route planning, energy consumption is usually considered a linear function of distance, but in this embodiment, energy consumption is modeled as a nonlinear function of the road network state. The system sets a positive correlation between the road network friction coefficient and the Riemann curvature tensor in the Riemann manifold diagram. The Riemann curvature tensor geometrically describes the curvature of the manifold; in traffic semantics, it accurately maps the uneven distribution of traffic density and the local capacity collapse caused by traffic accidents or construction. When the Riemann curvature tensor value of a certain road segment increases, it means that traffic congestion in that area intensifies, vehicles frequently start and stop, leading to a significant increase in energy consumption per unit mileage, i.e., an increase in the frictional force of the road network on vehicle movement.

[0108] The system employs a specific mathematical model to quantify this process, defining the first... The length of the Riemann curvature tensor modulus of each segment is Then the road network friction coefficient corresponding to this road segment It can be calculated using the following exponential function model:

[0109] ;

[0110] In the formula, It is the basic damping coefficient, used to adjust the weight of the effect of curvature on the friction coefficient; This is a congestion sensitivity factor used to reflect the sensitivity of a specific vehicle type to changes in road conditions. When road conditions are unobstructed, the Riemann curvature tensor modulus is [value missing]. The friction coefficient of the road network approaches zero. A value close to 1 indicates that energy consumption is at a baseline level. As congestion worsens, Increase The value increases rapidly and nonlinearly, representing the nonlinear amplification effect of traffic congestion on driving energy consumption.

[0111] like Figure 5 This is a nonlinear mapping relationship between the Riemann curvature tensor magnitude and the road network friction coefficient. The horizontal axis represents the Riemann curvature tensor magnitude, whose value quantifies the degree of congestion and dynamic complexity of road network traffic flow. The vertical axis represents the road network friction coefficient, whose value quantifies the energy consumption inhibition factor of the road network environment on vehicle movement.

[0112] Figure 5 The main feature is an orange-red nonlinear correlation curve exhibiting exponential growth. This curve vividly and accurately depicts the nonlinear amplification effect of traffic congestion on vehicle energy consumption. Specifically, in the initial stage of the curve, when the Riemann curvature tensor modulus is low, the road network friction coefficient remains at a baseline level close to one. This physically corresponds to an ideal scenario of smooth traffic flow, meaning that at this time, vehicle energy consumption mainly depends on the physical travel distance and is less affected by road conditions.

[0113] However, as the value of the horizontal axis gradually increases, the slope of the curve becomes steeper rapidly, and the road network friction coefficient shows a sharp upward trend. This physically corresponds to the phenomenon of a surge in energy consumption per unit mileage caused by frequent braking, starting, and low-speed crawling of vehicles when traffic congestion intensifies.

[0114] The simulation results strongly demonstrate the scientific nature and necessity of the road network traffic potential energy modeling method proposed in this invention. By introducing this nonlinear mapping mechanism, the system can fully consider the additional energy consumption under extreme congestion conditions when calculating the dynamic return energy consumption threshold, thereby accurately assessing the real survival energy consumption required for the vehicle to leave the current area. This builds a solid energy security firewall for mobile storage and charging vehicles, effectively avoiding vehicle breakdowns or deadlocks caused by underestimating return energy consumption when performing discharge tasks.

[0115] Based on the aforementioned road network friction coefficient, the system calculates the dynamic return trip energy consumption threshold. Assume the planned path for the vehicle from its current location to the backup refueling node includes... There are 1 road segment, and the physical length of each road segment is 1. The benchmark energy consumption per unit distance for a vehicle under standard operating conditions is: The dynamic return trip energy consumption threshold The calculation formula is as follows:

[0116] ;

[0117] In the formula, To account for the power consumption of vehicle accessories, such as the additional redundant energy consumption of air conditioning and on-board terminals, a road network friction coefficient is introduced. This formula can dynamically convert the risk of traffic congestion into specific electricity demand, ensuring that the calculated threshold can truly reflect the energy consumption for survival under current road conditions.

[0118] After completing the threshold calculation, the system proceeds to the crucial step of discharging rejection. This step aims to determine whether to approve the current discharge request by rehearsing the future state of energy. The system monitors the current state of charge of the mobile charging vehicle in real time through the onboard battery management system (BMS). Simultaneously, the system combines the discharge power from the proposed discharge service command. and planned discharge duration Predict the estimated remaining state of charge of the vehicle after executing the instruction. The prediction calculation follows the principle of energy conservation, and its expression is:

[0119] ;

[0120] In the formula, This refers to the total capacity of the vehicle's power battery.

[0121] For example, the system will calculate the estimated remaining state of charge. The system is compared and verified against a dynamic return trip energy consumption threshold. To further improve the system's robustness and address potential deterioration in road conditions in the near future, a preset road network fluctuation risk coefficient is introduced. This coefficient is derived from variance statistics of historical road condition big data and is used to quantify the uncertainty of road network status. The trigger condition for the discharge rejection mechanism is defined by the following inequality:

[0122] ;

[0123] If the above inequality holds true, it means that if the vehicle executes the current discharge command, its remaining battery power will not be able to safely reach the nearest backup charging node, even considering the risk of road network fluctuations. In this case, the system determines that there is an extremely high risk of deadlock and immediately triggers the discharge rejection mechanism.

[0124] like Figure 6This is a simulation diagram for determining the deadlock risk area of ​​mobile energy storage and charging vehicles based on a dynamic return trip energy consumption threshold. The diagram visually illustrates the core determination logic of the mobility safety verification process in this invention using a two-dimensional plane. The horizontal axis represents the physical travel distance from the current geographical location of the mobile energy storage and charging vehicle to the nearest backup energy replenishment node, in kilometers, and the vertical axis represents the vehicle's current state of charge, in percentage.

[0125] Figure 6 The gray solid line extending from the lower left to the upper right represents the dynamic return trip energy consumption threshold curve. This curve is not a nonlinear distance energy consumption relationship, but a nonlinear growth curve that incorporates the road network friction coefficient. It vividly reflects the accelerating upward trend of the minimum survival energy consumption required for the vehicle's return trip as the driving distance increases and the potential traffic congestion risk accumulates.

[0126] The threshold curve divides the entire plane into two key functional regions. The green-filled area above the curve is defined as the safe discharge zone. When a vehicle's status point falls within this zone, it indicates that its current battery level, after deducting the return trip's needs and the reserve power to cope with road network fluctuations, still has surplus power. The system allows the generation of discharge service commands for peak shaving or auxiliary services. Conversely, the red-filled area below the curve is defined as the deadlock risk zone. When a vehicle's status point falls within this zone, it means that its remaining battery level is insufficient to support its safe arrival at the nearest charging point, considering the friction of severe road conditions, posing an extremely high risk of breakdown on the road.

[0127] Once a vehicle enters the red zone, the system will immediately trigger the discharge rejection mechanism and forcibly activate the keep-alive mode, prohibiting any operation that outputs electrical energy and guiding the vehicle to evacuate urgently. This visualized zone determination mechanism effectively prevents mobile energy storage vehicles from falling into a double deadlock of energy and transportation, ensuring the asset safety and operational continuity of the system.

[0128] Furthermore, once the discharge rejection mechanism is triggered, the system will not execute the original discharge service command, but will immediately activate the forced keep-alive mode. This is the highest priority system protection state, designed to ensure the survival and evacuation capability of mobile charging vehicles in extremely adverse environments. In response to the discharge rejection mechanism, the central control system forcibly intercepts and cancels all pending discharge service commands, and switches the mobile charging vehicle's operating status from operational service mode to mobility locked mode.

[0129] In mobility-locked mode, the system implements strict control strategies on the vehicle's electrical system. The central control system remotely locks the vehicle's bidirectional charging and discharging motor's external discharge function, prohibiting the mobile charging vehicle from outputting any electrical energy, including prohibiting V2G from feeding back into the grid and V2L from supplying power to user loads. This measure is to completely cut off unnecessary energy loss channels, reserving every bit of electricity for the vehicle's power drive.

[0130] Optionally, in conjunction with the mobility lockout status, the system simultaneously generates an emergency evacuation route plan. This plan differs from conventional economically optimal route planning, prioritizing survival. The system instructs mobile storage and charging vehicles to immediately proceed to the calculated backup energy replenishment node. When generating this emergency evacuation route, the system again utilizes the Riemannian manifold to find the path with the minimum Riemann curvature tensor integral—that is, to find the route with the least friction and highest certainty of passage within the road network, rather than the route with the shortest physical distance. This can be understood as, under energy constraints, vehicles moving along the valleys of traffic potential energy to escape the danger zone at the lowest possible cost.

[0131] To ensure the effectiveness of the backup power supply node, the system sends a resource reservation request to the target backup power supply node simultaneously with the evacuation command. Since the backup power supply node is defined as a site with dedicated power supply resources, upon receiving the request, the site will immediately lock a charging gun or battery swapping station and reject access requests from other vehicles in the electronic fence system, ensuring that the mobile energy storage and charging vehicle can be plugged in and charged immediately upon arrival.

[0132] In summary, the entire mobility safety verification process based on road network traffic potential energy is a dynamic and cyclical process. Even before a vehicle is scheduled to discharge, the system will silently run the verification logic in the background at certain intervals, such as every 5 minutes. If the system detects a sharp deterioration in road network congestion, causing a surge in the dynamic return trip energy consumption threshold, and thus posing a risk of the vehicle running out of power even without discharging, the system will proactively activate a forced keep-alive mode to guide the vehicle to avoid the risk in advance. This proactive defense mechanism fundamentally solves the energy depletion dilemma that mobile charging vehicles may encounter in complex urban road networks, demonstrating the absolute priority given to the survival objective in multi-objective optimization of this invention.

[0133] As can be seen from the detailed description of the above embodiments, the method of this embodiment is not merely a simple scheduling algorithm, but a deep integration of traffic attributes in the physical world and energy management in the information world. Through the mathematical mapping of the Riemann curvature tensor and the road network friction coefficient, this invention successfully establishes a cross-domain risk assessment system. The linkage mechanism between discharge rejection and forced keep-alive mode constitutes a robust firewall, effectively preventing the systemic risk of vehicle paralysis due to sudden changes in road conditions caused by the pursuit of short-term discharge gains, thereby significantly improving the operational reliability and safety of the mobile energy storage and charging system under all weather and road conditions.

[0134] Example 3:

[0135] This embodiment, building upon Embodiments 1 and 2, further elaborates on the resource scheduling mechanism of mobile energy storage and charging systems under abnormal takeover scenarios. Particularly in high-load distribution network environments, when a mobile energy storage and charging vehicle undergoing high-power charging or discharging suddenly experiences a communication interruption or control module failure, the key technical challenge addressed by this system is how to safely and smoothly transfer the charging task associated with that vehicle to another healthy mobile energy storage and charging vehicle, while simultaneously preventing the activation of grid relay protection devices due to drastic load fluctuations. This embodiment details a resource atomization handover process based on shadow tokens, which ensures seamless integration of logical resource release and physical resource takeover during task redistribution.

[0136] For example, when the central control system confirms that a mobile energy storage vehicle (hereinafter referred to as an abnormal vehicle) is in an abnormal state and cannot be restored, the system initiates a task takeover process. In conventional scheduling logic, the system usually releases the resources occupied by the abnormal vehicle first, and then a new vehicle re-applies. However, in a complex system involving the interaction of power generation, grid, load, and energy storage, this "break and rebuild" approach carries significant risks. If grid resources are released directly, the distribution network monitoring equipment will detect a sudden drop in load, and when the takeover vehicle (hereinafter referred to as a healthy vehicle) connects, the load will surge again. This pulse-like power surge can easily lead to voltage flicker or frequency oscillation. Therefore, this embodiment employs an interception and freezing strategy. When initiating the task takeover process, the central control system does not perform conventional resource release operations for the grid resources and access point resources occupied by the abnormal vehicle, but instead forcibly intercepts conventional resource release commands. The system marks the status of the grid resources (such as transformer capacity quotas) and access point resources (such as physical pile IDs) as shadow locked in the database and memory.

[0137] In shadow-locked mode, the access point resource exhibits a unique dual attribute. For queries from external public dispatch pools and other third-party vehicles, the resource remains occupied or invisible, preventing unauthorized vehicles from seizing it and avoiding resource contention. However, for specific takeover processes within the system, the resource is in a suspended state awaiting inheritance. Crucially, in shadow-locked mode, the system fully preserves the original power quota parameters of the abnormal vehicle. Let's assume the active power setpoint, negotiated with the power grid before the abnormal vehicle's failure, is... The reactive power setting value is The relevant set of Policy and Charge Control (PCC) rules is as follows: During shadow locking, these parameters , and The edge computing units are frozen and remain unchanged, continuing to operate according to... Control signals are output to the physical port. Even though the abnormal vehicle may no longer be able to respond at this time, the control logic on the power grid side remains consistent, preserving a physical channel for subsequent seamless switching.

[0138] For example, to ensure the security and uniqueness of the takeover process, the central control system generates a digital credential, namely a resource inheritance token. This token is the core medium for achieving task handover. The system encapsulates the aforementioned retained power quota parameters and key information used for authentication within the token. Specifically, the resource inheritance token... The data structure includes: the power quota originally occupied by the abnormal vehicle. Billing session identifier for abnormal vehicles and a unique verification signature signed by the private key of the central control system. The signature verification uses an asymmetric encryption algorithm to ensure that the token cannot be tampered with or forged during transmission. The system inherits the token from this resource. Along with the reassigned charging task instruction package, it is sent to the selected mobile charging vehicles (healthy vehicles) that are in normal operation via an encrypted communication link.

[0139] Next, upon receiving the task instruction and resource inheritance token, the mobile energy storage and charging vehicle immediately drives towards the target access point. During this process, the vehicle's internal energy management system (EMS) parses the token... The battery thermal management system is preheated to adapt to the upcoming power level, thus shortening the preparation time after connection. When a mobile charging vehicle in normal operation arrives at the access point and physically connects to the charging interface, it does not initiate a new charging handshake request as in the conventional process. Instead, it submits the resource inheritance token to the edge computing unit where the access point is located. In this step, the vehicle declares to the edge device that it is the rightful heir to the rights of the faulty vehicle.

[0140] After receiving the resource inheritance token, the edge computing unit where the access point is located first uses the pre-set public key to... Decryption verification is performed. Once verification is successful and the takeover request is confirmed as legitimate, the edge computing unit immediately executes an atomic identity substitution operation. Here, "atomic" means that the operation is completed within the system's minimum instruction cycle, is indivisible, and either succeeds completely or fails entirely, with no intermediate states. The core objective of this operation is to deceive the grid-side monitoring logic, preventing it from detecting the change in the load entity. During the atomic identity substitution operation, the edge computing unit maintains the output power setpoint of the access point unchanged, i.e., maintaining the duty cycle of the pulse width modulation (PWM) signal or the power limit in the BMS communication message constant. .

[0141] Optionally, within the same clock cycle maintaining constant physical layer power output, the edge computing unit performs identity switching at the logical layer. The system modifies the mapping in memory, switching the current billing and control object identifier from the ID of the abnormal vehicle to the ID of the mobile charging / storage vehicle in normal operation. This can be achieved using a state vector. To describe the state of the access point during handover. Before, At the switching time after, .

[0142] It is worth noting that throughout the process, the power component Maintain continuity and smoothness, that is This approach ensures that the current waveform monitored by the current transformers and relay protection devices on the distribution network side is continuous, preventing load surges or overcurrent protection alarms, thus achieving a true hot-switching handshake.

[0143] While the logical handover is complete, the physical connection status may still require processing. Typically, although an abnormal vehicle is logically offline, the physical charging port may still be plugged in, or a healthy vehicle may be connected to the secondary charging port of the same charging station—this applies to dual-gun DC charging station scenarios. To prevent premature resource release from causing system instability, this embodiment employs a strict delayed physical release mechanism. The system only releases resources within a preset safe time window after the hot-swap handshake is complete. Only after that is the shadow lock status allowed to be released. This time window The setting is typically from a few seconds to a minute, the purpose of which is to give the automatic voltage regulator (AVR) on the grid side and the BMS of the takeover vehicle sufficient response time to establish steady-state equilibrium.

[0144] Optionally, in addition to the time window mechanism, the system also performs dual confirmation via physical signals. When the edge computing unit detects a physical disconnection signal from an abnormal vehicle, such as detecting a change in the CP control lead voltage at the abnormal vehicle's connection port from a connected state to a disconnected state, or detecting a loop current... Once the system reaches zero, it determines that the abnormal vehicle has completely physically disconnected from the system. Subsequently, regardless of whether the safe time window has been reached, the system will trigger a release command to officially release the shadow lock status of the power grid resources and access point resources.

[0145] After the shadow lock is released, the access point resource returns to its normal occupied or idle state, depending on whether a healthy vehicle continues to be used. At this time, the edge computing unit allows new power adjustment requests. Healthy mobile charging vehicles can request adjustments to their power settings from the system based on their own battery status and the current grid policy, such as those inherited from... Adjusted to be better The adjustments made at this point are considered normal power regulation and are no longer subject to the freeze restrictions of the shadow state.

[0146] Optionally, to further quantify the contribution of atomized identity permutation operations to grid stability, a grid impact index can be defined. In the traditional disconnect-then-connect switching mode, the impact index is related to the power change rate and duration, and can be expressed as:

[0147] ;

[0148] This means the power grid experienced one full-power unloading and one full-power loading. However, in the atomized switching mode proposed in this embodiment, due to power... Keeping constant, the theoretical impact index is:

[0149] ;

[0150] This comparison shows that the resource atomization handover process based on shadow tokens greatly reduces the disturbance to the power grid during the task takeover process, making it possible to schedule large-scale mobile energy storage and charging vehicles in fragile power grid environments where the grid capacity is close to saturation or the output of renewable energy fluctuates drastically.

[0151] This procedure also applies to fault takeover of mobile energy storage and charging vehicles in V2G reverse discharge mode. When the malfunctioning vehicle is feeding power to the grid (i.e., the power is negative), If a fault occurs and the power grid is directly disconnected, the grid will face a sudden power shortage, which may lead to a local frequency drop. Through the above process, healthy vehicles can quickly take over and inherit the negative power output task via shadow tokens, that is, continue to discharge, support the grid frequency stability, until the dispatch center completes the new power flow calculation and issues new instructions. This seamless takeover capability across all modes and processes demonstrates the significant advantages of the method in this embodiment in improving the resilience of the power system.

[0152] Example 4:

[0153] like Figure 7 A schematic diagram of a mobile energy storage and charging system scheduling and path planning system based on multi-objective optimization is presented. This system is applied to the central control system of the mobile energy storage and charging system and serves as the hardware and software logic carrier for implementing the methods described in Embodiments 1 to 3. Addressing the multi-physics coupling challenges faced by mobile energy storage and charging systems mentioned in the background art, particularly the difficulty in matching the intermittency of distributed renewable energy output with the high randomness of user demand, and the vehicle energy consumption deadlock problem caused by urban road network traffic congestion, the system provided in this embodiment achieves multi-domain coordination of source, grid, load, storage, and road through modular functional design. The system mainly includes a data collaborative processing module, a multi-objective construction module, a hierarchical optimization solution module, and a scheduling execution module.

[0154] For example, the data collaborative processing module serves as the sensing front-end of the entire system. It responds to charging demand requests and renewable energy access point status data within the region, performing multi-source data acquisition and collaborative preprocessing to generate a system status set containing spatiotemporal characteristics. In actual operation, this module connects in real-time to the regional distributed energy monitoring system and the distribution network SCADA system via a high-frequency communication interface. To address the issues of noise interference and inconsistent spatiotemporal dimensions in the raw data, the data collaborative processing module integrates a deep learning preprocessing algorithm.

[0155] Specifically, this module utilizes a pre-built EfficientNet spatial feature extraction network and an LSTM temporal feature extraction network to transform the collected real-time renewable energy output data into high-dimensional feature vectors with spatiotemporal semantics. Simultaneously, the module smooths the power data using a moving average method and combines it with meteorological correlation data to generate fluctuation risk levels. This processing directly corresponds to step S1 in Example 1, ensuring that the data subsequently input into the optimization model possesses high accuracy and robustness, providing a solid data foundation for addressing the data source heterogeneity and volatility issues mentioned in the background technology.

[0156] For example, the multi-objective construction module is used to construct a multi-objective optimization system based on the system state set, including the user side, operation side, grid side, energy storage side, and renewable energy consumption side, and introduces a conflict resolution factor to balance the conflict relationships between the objectives. The core function of this module is to resolve the core technical conflict mentioned in the background technology, namely, how to achieve a balance between pursuing operational economics and ensuring vehicle mobility survivability and grid stability. The module internally stores a preset seven-dimensional objective function library, covering key indicators such as user waiting time, operating costs, grid voltage qualification rate, energy storage lifetime, and renewable energy consumption rate. During operation, the multi-objective construction module dynamically adjusts the weight allocation of each sub-objective according to the real-time calculated conflict resolution factor.

[0157] For example, when the data collaborative processing module detects an oversupply of renewable energy, it automatically increases the weight of the consumption-side target, guiding the system to prioritize vehicle charging for green charging. This dynamic balancing mechanism ensures that the system's decisions do not fall into local optima for a single objective, but rather seek a globally comprehensive optimal solution on the Pareto front.

[0158] For example, the hierarchical optimization solution module is the computational core of the system, used to input the multi-objective optimization system into the hierarchical collaborative optimization model for solution, and output scheduling decisions and path planning schemes for mobile storage and charging vehicles. To address the contradiction between computational complexity and real-time requirements, the hierarchical collaborative optimization model is configured to execute a three-level optimization logic including global layer optimization, local layer optimization, and emergency adjustment.

[0159] Specifically, this module first utilizes an improved NSGA-II algorithm at the global layer to determine the macro-regional allocation of vehicles based on long-term spatiotemporal prediction data. Then, at the local layer, it employs a lightweight PPO algorithm, combined with real-time road network obstacle information, to refine the service sequence of vehicles. Notably, this module also integrates a transfer learning mechanism, enabling it to quickly call pre-trained parameters based on the matching degree between the current scene and historical scene sets, significantly shortening the solution time.

[0160] Furthermore, at the route planning level, this module incorporates a Riemannian manifold construction engine, capable of calculating the Riemann curvature tensor of the road network, thereby avoiding high traffic potential energy areas during the planning phase. This is closely related to step S4 in Example 1 and the safety verification logic in Example 2, ensuring that the generated route scheme is not only theoretically optimal but also possesses extremely high physical reliability.

[0161] For example, the scheduling execution module is used to issue the scheduling decision and route planning scheme to the mobile charging vehicle to trigger the mobile charging vehicle to perform charging tasks, dynamic obstacle avoidance, and renewable energy collaborative charging and discharging operations. This module is not a simple command transmitter, but an execution center with logic verification and exception takeover capabilities. Before issuing the command, the scheduling execution module performs a mobility safety check based on the road network traffic potential energy and calculates the dynamic return trip energy consumption threshold. Once a deadlock risk is detected in the vehicle, the discharge command is immediately intercepted and a forced keep-alive mode is triggered, which directly corresponds to the core technical solution in Embodiment 2.

[0162] Meanwhile, addressing the resource contention and grid impact issues in abnormal takeover scenarios mentioned in the background technology, this module is configured to generate a resource inheritance token containing power quota parameters and a unique verification signature when an abnormal vehicle is detected. Through a shadow locking mechanism, the scheduling execution module can freeze the resources occupied by the abnormal vehicle at the logic layer and guide healthy vehicles to complete atomic identity replacement by submitting the token. This mechanism ensures that the physical load curve of the distribution network remains smooth during task handover, achieving seamless synchronization between logical resource release and physical response, corresponding to the technical solution in Embodiment 3.

[0163] In summary, the four core modules in this embodiment, through tight logical coupling and data interaction, collectively constitute a closed-loop intelligent scheduling system. The data collaborative processing module provides the system with accurate situational awareness, the multi-objective construction module establishes a scientific optimization direction, the hierarchical optimization solution module provides efficient computing power, and the scheduling execution module ensures the safe implementation of decisions and the ability to recover from anomalies. This system effectively overcomes the shortcomings of existing technologies, such as single scheduling strategies, lack of multi-domain collaboration, and insufficient ability to cope with emergencies. It significantly improves the overall scheduling efficiency of mobile energy storage and charging systems in complex dynamic environments, the level of renewable energy consumption, and the safety of grid operation.

[0164] Example 5:

[0165] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0166] like Figure 3The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0167] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0168] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0169] The memory 103 stores a computer program corresponding to the multi-objective optimization-based mobile storage and charging system scheduling and path planning method of the above embodiments of the present invention. This computer program is controlled and executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0170] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 3 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0171] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A scheduling and path planning method for a mobile energy storage and charging system based on multi-objective optimization, characterized in that, The central control system used in mobile energy storage and charging systems includes: In response to charging demand requests within the region and status data of renewable energy access points, multi-source data acquisition and collaborative preprocessing are performed to generate a system status set containing spatiotemporal characteristics. Specifically, this includes: collecting real-time output data, spatiotemporal prediction data, and meteorological correlation data from renewable energy access points; smoothing the real-time output data using a moving average method to suppress data fluctuation noise; extracting spatial features from the real-time output data using a pre-set spatial feature extraction network model, and extracting temporal features from the real-time output data using a pre-set temporal feature extraction network model; and combining the meteorological correlation data to generate a renewable energy output fluctuation risk level containing at least three preset levels, and incorporating it into the system status set. Based on the system state set, a multi-objective optimization system is constructed, encompassing the user side, operation side, grid side, energy storage side, and renewable energy consumption side. A conflict resolution factor is introduced, dynamically balancing the conflict relationships between objectives based on the gradient direction or weight ratio of each objective function. The construction of this multi-objective optimization system includes: establishing a total objective function, which is the weighted sum of the user-side objective function, operation-side objective function, grid-side objective function, energy storage-side objective function, dynamic obstacle avoidance objective function, abnormal loss objective function, and renewable energy consumption objective function, each multiplied by its corresponding dynamic weight; setting a conflict resolution factor for the total objective function, the value of which is within a preset numerical correction range, used to dynamically correct numerical conflicts between sub-objective functions during the optimization process; wherein, the calculation logic of the renewable energy consumption objective function is: based on the ratio of the total power charged by mobile energy storage vehicles to the sum of the real-time output power of renewable energy and the maximum allowable power curtailment of the distribution network, the regional renewable energy consumption rate is calculated and maximized. The multi-objective optimization system is input into the hierarchical collaborative optimization model for solution, and the output is the scheduling decision and path planning scheme for mobile storage and charging vehicles; The scheduling decision and path planning scheme are sent to the mobile energy storage vehicle to trigger the mobile energy storage vehicle to perform charging tasks, dynamic obstacle avoidance and renewable energy coordinated charging and discharging operations; The solution process of the hierarchical collaborative optimization model includes: Initiate global layer optimization, using the first preset time period as a benchmark, and based on the spatiotemporal prediction data of renewable energy, determine the regional allocation of mobile energy storage and charging vehicles and the charging priority of access points, and generate a global path framework. Local layer optimization is initiated, and based on the second preset time period, the service sequence and obstacle avoidance path of a single mobile energy storage and charging vehicle are refined according to the real-time output data of renewable energy and the dynamic obstacle information of the road network; the second preset time period is shorter than the first preset time period. In response to a sudden state signal during system operation, an emergency adjustment process is triggered to perform power adjustment within a first preset response time or task redistribution within a second preset response time.

2. The method according to claim 1, characterized in that, The step of inputting the multi-objective optimization system into the hierarchical collaborative optimization model for solution also includes applying renewable energy collaborative constraints, which include: The charging power of mobile energy storage and charging vehicles at renewable energy access points is limited to be less than or equal to the difference between the real-time output of the access point and the local load of the distribution network, and the charging time is limited to cover the output duration of the access point. The total power of mobile energy storage and charging vehicles that are simultaneously charging at a single renewable energy access point is limited to a preset percentage threshold of the rated capacity of that access point. The power change rate of mobile energy storage and charging vehicles during charging and discharging operations is limited to not exceeding a preset power ramp rate threshold.

3. The method according to claim 1, characterized in that, The step of sending the scheduling decision and path planning scheme to the mobile charging vehicle to trigger the mobile charging vehicle to perform charging tasks, dynamic obstacle avoidance, and renewable energy coordinated charging and discharging operations includes: Determine the current electricity price status and renewable energy output status; If the electricity price is in a low-price period and there is a surplus of renewable energy output, a deep charging command is generated, instructing the mobile energy storage vehicle to extend the charging time at the access point and charge the battery state of charge to above the first preset state of charge threshold. If the electricity price is during peak hours and renewable energy output is insufficient, a discharge service instruction is generated, instructing the mobile energy storage vehicle to prioritize using its own stored energy to supply power to the user, and to shorten the charging time at the access point or stop charging.

4. The method according to claim 1, characterized in that, The hierarchical collaborative optimization model includes road network traffic constraints for path planning, and the process of constructing the road network traffic constraints includes: Construct a hierarchical topology map, and establish a first obstacle avoidance path on the hierarchical topology map based on geometric space segmentation and semantic risk weight; Construct a Riemannian manifold graph, and map the motion characteristics of dynamic obstacles in the road network to a Riemannian curvature tensor. Then, establish a second obstacle avoidance branch based on curvature calculation on the Riemannian manifold graph. When generating a path planning scheme, the Riemann curvature tensor is monitored in real time. If the Riemann curvature tensor exceeds a preset curvature threshold, it is determined that there is a high dynamic risk, and the path is switched from the first obstacle avoidance branch to the second obstacle avoidance branch for path replanning.

5. The method according to claim 1, characterized in that, The emergency adjustment process triggered in response to sudden state signals during system operation includes: The status of mobile storage and charging vehicles and communication links is monitored in real time through redundant communication modules. When abnormal indicators are detected, a primary abnormal signal is generated. In response to the primary abnormal signal, a paging process for the associated mobile storage and charging vehicle is triggered, and a secondary status confirmation is performed. If the abnormal indicator is not resolved after the paging process is completed, it is determined to be an abnormal state and the task takeover process is initiated: the charging task associated with the abnormal vehicle is reallocated to a mobile storage and charging vehicle that is in normal operation and whose location meets the preset scheduling threshold within a preset allocation period, and the power grid resources and access point resources occupied by the abnormal vehicle are released simultaneously.

6. The method according to claim 1, characterized in that, The solution to the hierarchical collaborative optimization model employs a transfer learning mechanism, including: Construct a scenario transfer library, which pre-stores feature parameters and corresponding pre-trained model parameters for a variety of typical renewable energy scenarios. Before solving the optimization problem at the current moment, calculate the matching degree between the real-time scene features and the typical scene features in the scene migration library; If the matching degree is higher than the preset matching threshold, the corresponding pre-trained model parameters are directly called to initialize the population of the optimization algorithm and reduce the number of iterations of the optimization algorithm.

7. The method according to claim 4, characterized in that, The issuance of the scheduling decision and route planning scheme includes generating a discharge service instruction; before executing the discharge service instruction, a mobility safety verification process based on road network traffic potential energy must also be performed, the mobility safety verification process includes: Based on the Riemann manifold, the minimum energy consumption required for the mobile energy storage vehicle to travel from its current geographical location to the nearest backup energy replenishment node is calculated in real time, serving as a dynamic return trip energy consumption threshold. The calculation of the minimum energy consumption value introduces the road network friction coefficient, which is positively correlated with the Riemann curvature tensor and is used to characterize the nonlinear amplification effect of traffic congestion on driving energy consumption. The current state of charge of the mobile energy storage vehicle is monitored in real time, and the estimated remaining state of charge after the discharge service command is executed is predicted. If the estimated remaining state of charge is less than the product of the dynamic return trip energy consumption threshold and the preset road network fluctuation risk coefficient, the discharge rejection mechanism is triggered. In response to the discharge rejection mechanism, the discharge service command is forcibly intercepted, and the working state of the mobile charging vehicle is switched to a mobility locked state. In the mobility locked state, the mobile charging vehicle is prohibited from outputting electrical energy, and an emergency evacuation route planning scheme is generated, instructing the mobile charging vehicle to immediately drive to the backup power replenishment node. The backup power replenishment node is defined as the charging station or battery swapping station in the road network that is closest to the current location and has dedicated power supply resources.

8. The method according to claim 5, characterized in that, The specific steps for reallocating the charging tasks associated with abnormal vehicles to mobile storage and charging vehicles that are in normal operation and whose locations meet the preset scheduling thresholds within a preset allocation period, and simultaneously releasing the grid resources and access point resources occupied by the abnormal vehicles, are as follows: When the task takeover process is initiated, the normal resource release command is intercepted for the power grid resources and access point resources occupied by the abnormal vehicle, and the status of the power grid resources and access point resources is marked as shadow locked. In the shadow locked state, the access point resources remain invisible to the public scheduling pool, and the original power quota parameters of the abnormal vehicle are retained. Generate a resource inheritance token containing the power quota parameters and a unique verification signature, and send the resource inheritance token along with the reassigned charging task to the mobile charging vehicle in normal operation. In response to the arrival of the mobile charging vehicle in normal operation at the access point and submission of the resource inheritance token, the edge computing unit where the access point is located performs an atomic identity replacement operation: keeping the output power setting of the access point unchanged, and switching the billing and control object identifier from the abnormal vehicle to the mobile charging vehicle in normal operation within the same clock cycle, thereby realizing a hot handover handshake. The shadow lock state is released and a new power adjustment request is allowed only after a preset safe time window following the completion of the hot-swap handshake, or when a physical disconnection signal is detected from the abnormal vehicle.

Images