Unmanned aerial vehicle battery replacement scheduling system and method based on multi-state battery prediction

Abstract

The application discloses a UAV battery replacement scheduling system and method based on multi-state battery prediction, and particularly relates to the technical field of UAV battery management, and is used for solving the problem of energy shortage and mismatching of replacement decision of a patrol UAV; the application collects multi-source data such as voltage, current, temperature, internal resistance and environmental wind speed, extracts key states such as charge, health, power, energy and temperature, forms a state feature vector related to a task context, estimates task energy consumption and return energy based on a prediction model, generates a scheduling instruction in continuing a task, going to a charging station or a battery replacement station, maintenance charging and cooling fast charging, in a battery replacement scene, an improved ant colony algorithm is introduced by introducing a battery health factor to construct a path cost function, distance, energy consumption and time are considered, optimal battery replacement station selection and safe path planning are realized, and the energy management accuracy and task stability of the UAV patrol task are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) battery management technology, and more specifically, to a UAV battery swapping scheduling system and method based on multi-state battery prediction. Background Technology

[0002] With the widespread application of drones in tasks such as power line inspection, logistics delivery, and disaster relief, their operation time and mission reliability increasingly depend on battery performance. However, most existing battery management systems are based on equivalent circuit models, which can only estimate the remaining battery capacity and cannot comprehensively characterize key parameters such as battery health, power state, energy state, and temperature state. At the same time, the battery degradation process is nonlinear and uncertain, and traditional methods lack the ability to dynamically predict the evolution trend of battery state, resulting in inaccurate timing of battery swapping.

[0003] Existing drone battery swapping systems typically use fixed thresholds or single rules to trigger swapping, which can lead to premature battery replacement and reduced utilization, as well as flight interruptions or even crashes due to battery swapping delays. Furthermore, battery swapping scheduling often prioritizes the shortest path as the sole optimization objective, failing to comprehensively consider multi-dimensional constraints such as energy consumption, time, and battery health, and lacking a global optimization mechanism based on the coordinated resources of drone swarms and battery swapping stations. Summary of the Invention

[0004] To overcome the aforementioned deficiencies in the prior art, this invention provides a UAV battery swapping scheduling system and method based on multi-state battery prediction, in order to solve the problem of insufficient energy and mismatch between battery swapping decisions of inspection UAVs in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The UAV battery swapping scheduling method based on multi-state battery prediction includes the following steps:

[0007] Step 1: By using sensor modules installed on the drone, real-time data are collected on battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment, to obtain multi-source data on battery operating status and environmental status.

[0008] Step 2: Based on multi-source acquired data, calculate the battery's state of charge, state of health, state of power, state of energy, and temperature state to form a set of multi-state parameters for comprehensively characterizing the battery's operating state.

[0009] Step 3: Use a long short-term memory network model to model the historical state data containing multiple state parameters and environmental parameters, predict the battery state decay trend within a preset time window, and predict the energy required to complete the current task and the minimum energy required for the drone to return safely.

[0010] Step 4: Based on real-time multi-state parameters, prediction results and task priorities, generate scheduling instructions according to preset decision rules. The scheduling instructions include at least one of the following: continue the task, go to the charging station to charge, go to the battery swapping station to swap batteries, perform trickle charging, and perform cooling fast charging.

[0011] Step 5: When the scheduling instruction is to go to the battery swapping station, an improved ant colony algorithm is used to construct multiple feasible paths from the current position of the UAV to the candidate battery swapping station on the operation environment graph model. The battery health status factor is introduced, and the optimal path to the target battery swapping station is searched under the condition of simultaneously considering the path distance, energy consumption and flight time.

[0012] Step six: Control the drone to fly along the optimal path to the target battery swapping station, where the mechanical actuators of the swapping station will automatically complete the battery removal and installation, and select the corresponding charging mode for the replaced battery according to the scheduling instructions.

[0013] In a preferred embodiment, step one includes: acquiring battery voltage through a voltage sensor, acquiring battery current through a current sensor, acquiring battery temperature through a temperature sensor, obtaining battery internal resistance through an internal resistance detection unit, and acquiring wind speed and ambient temperature in the flight environment through an environmental sensor.

[0014] The data from each acquisition channel are time-aligned and interpolated according to a unified time base, and the data format is standardized and outlier is removed to form a time-continuous and structured multi-source acquisition data sequence.

[0015] In a preferred embodiment, step two, battery multi-state calculation and evaluation, includes:

[0016] The state of charge is updated by integrating the sampled current over the sampling time interval and comparing it with the battery's nominal capacity.

[0017] The health status is calculated by combining the ratio of current available capacity to nominal capacity, the change in current internal resistance relative to initial internal resistance, and the number of charge-discharge cycles already experienced as a decay index.

[0018] By estimating the current maximum allowable continuous discharge current and combining it with the terminal voltage, the maximum power that the battery can output is determined, and the power state is obtained.

[0019] The remaining usable energy can be obtained by the correspondence between the state of charge and the nominal energy of the battery, or by integrating the power change curve over time.

[0020] By comparing real-time temperature samples with preset safe temperature ranges, the temperature status is obtained, and it is determined whether there is a risk of overheating or overcooling.

[0021] In a preferred embodiment, step three, battery state and energy consumption prediction, includes:

[0022] A long short-term memory network model is constructed with multiple state parameters and environmental parameters within a continuous time window as input, and the predicted values ​​of battery state of charge, health state and temperature state at future time steps are output.

[0023] Based on the predicted values ​​and the planned flight distance, predetermined flight speed and typical power consumption curve of the current mission, a regression estimate is made of the energy required to complete the current mission.

[0024] Based on the distance the UAV needs to return from its current operating position to a safe landing position, its return flight speed, and conservative power consumption assumptions, the minimum energy required for the UAV to return safely is estimated, and the energy safety margin is calculated accordingly.

[0025] In a preferred embodiment, step four, the intelligent scheduling decision generation includes:

[0026] When the health status is below the first preset threshold, or the energy security margin is negative, a dispatch instruction is generated to go to the battery swapping station for battery swapping.

[0027] When the temperature exceeds the second preset threshold, a scheduling instruction is generated to proceed to the charging station to perform cooling fast charging.

[0028] When the task urgency parameter is higher than the third preset threshold and the prediction results show that the task can be completed and the safe return conditions can be met through fast charging, a scheduling instruction to go to the charging station to perform fast charging is generated.

[0029] When the health status is below the fourth preset threshold and the energy safety margin is positive, a scheduling instruction to execute trickle maintenance charging is generated.

[0030] If none of the conditions are met, a scheduling instruction to continue the task is generated.

[0031] In a preferred embodiment, step five includes:

[0032] The operating environment is abstracted as a weighted directed graph. The nodes of the graph include the current location node of the UAV, multiple candidate battery swapping station nodes, and relay nodes. The edges of the graph represent feasible flight paths between nodes, and each edge carries attribute parameters such as flight distance, expected energy consumption, and expected flight time.

[0033] With the optimization goal of simultaneously minimizing path distance, energy consumption, and arrival time, algorithm parameters such as ant colony size, maximum number of iterations, and pheromone volatilization coefficient are set.

[0034] In each generation of the search process, the virtual ant is controlled to start from the current location node of the drone and select the next node in turn on the weighted directed graph to construct a complete path from the current location node to a certain battery swapping station node.

[0035] In a preferred embodiment, the improved ant colony algorithm in step five specifically includes:

[0036] When the virtual ant selects the next node, the transition probability of each candidate edge is calculated based on the pheromone intensity on the candidate edge and the heuristic factor obtained by combining the path distance, energy consumption, flight time and current battery health status.

[0037] After completing each generation of path search, the path quality is calculated based on the total distance, total energy consumption and total flight time of the path, and the original pheromones on each side are attenuated according to the preset pheromone volatilization coefficient.

[0038] New pheromone increments are added to the edges of the traversed paths based on path quality and current battery health status;

[0039] When the improvement in path quality is less than a preset threshold or the maximum number of iterations is reached, the path with the highest pheromone concentration is selected as the optimal path to the target battery swapping station.

[0040] In a preferred embodiment, step six, the drone battery swapping process includes:

[0041] The drone is controlled to fly along the optimal path to the preset docking position of the target battery swapping station, and automatically docks with the mechanical actuators of the battery swapping station based on visual alignment and attitude adjustment mechanisms.

[0042] The mechanical actuator removes the old batteries from the drone and installs fully charged new batteries in a predetermined sequence.

[0043] According to the dispatch instructions, the old batteries to be replaced are sent to the fast charging station, trickle charging station, or cooling charging station for charging and maintenance, and the charging mode, charging time, and health status update results of each battery are recorded.

[0044] In a preferred embodiment, during the entire process of the UAV performing the mission, the multi-state parameters of the battery, the summary of multi-source collected data, the scheduling instructions, the path planning results and the battery swapping execution results are reported to the cloud management platform.

[0045] The cloud management platform uses data from drones to jointly train and update the parameters of the long short-term memory network model and scheduling decision rules using federated learning or distributed training methods, so as to obtain the updated model parameters and decision rules without uploading the original data in a centralized manner.

[0046] The updated model parameters and decision rules were distributed to each drone.

[0047] A UAV battery swapping scheduling system based on multi-state battery prediction is used to implement the aforementioned UAV battery swapping scheduling method based on multi-state battery prediction, including:

[0048] The state awareness layer is used to collect battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment in real time through a sensor array set on the drone. The collected data is time-aligned and format-unified to generate a structured multi-source data sequence.

[0049] The intelligent decision-making layer is used to receive multi-source collected data and has a built-in multi-state estimation module and a prediction decision-making module. The multi-state estimation module is used to calculate the battery's state of charge, state of health, state of power, state of energy, and temperature state based on the multi-source collected data. The prediction decision-making module is used to predict the battery state and task energy consumption using a long short-term memory network model, and generate scheduling instructions based on real-time multi-state parameters, prediction results, and task priorities, including at least one of the following: continue task, go to charging station for charging, go to battery swapping station for battery swapping, perform trickle charging, and perform cooling fast charging.

[0050] The execution control layer receives scheduling instructions. When the scheduling instruction is to go to the battery swapping station for battery swapping, an improved ant colony algorithm with the introduction of a battery health status factor is used to search for the optimal path to the target battery swapping station on the operation environment graph model, taking into account path distance, energy consumption, and flight time. The drone is then controlled to fly along the optimal path to the target battery swapping station, where the mechanical actuators of the battery swapping station automatically complete the battery removal and installation, and select the corresponding charging mode for the replaced battery according to the scheduling instructions.

[0051] The technical effects and advantages of this invention are as follows:

[0052] This invention constructs an autonomous battery swapping control mechanism for UAVs based on multi-state battery perception and task-priority scheduling, enabling dynamic perception and differentiated control of battery status and energy safety during inspection missions of multi-rotor UAVs. The mechanism collects and standardizes multi-source data such as battery voltage, current, temperature, internal resistance, and ambient wind speed and temperature through a state perception layer. It extracts multi-state indicators including state of charge, health status, power status, energy status, and temperature status, generating a battery state feature vector that integrates task context and environmental constraints. Based on a long short-term memory network, it predicts task energy consumption and safe return energy. Combining task urgency and safety strategies, it generates a scheduling instruction structure including options for continuing the task, proceeding to a charging station, proceeding to a battery swapping station, maintenance charging, and cooling fast charging. After determining the battery swapping requirement, an improved ant colony algorithm incorporating battery health status factors is used to construct a path scheduling cost function, taking into account path distance, energy consumption, and arrival time, achieving optimal selection of battery swapping stations, optimal path planning, and adaptive control with safety margins.

[0053] After the strategy is executed, the inspection interruption points, battery swapping paths and results are analyzed to evaluate the task completion rate, energy safety margin utilization rate and the evolution trend of battery temperature and health. Management features such as scheduling response delay, path cost offset and battery swapping station load distribution are extracted to construct data management coefficients to describe the matching degree between energy management and task scheduling, and generate model parameter update signals or decision threshold adjustment signals. This reduces the problems of task interruption and forced landing caused by inaccurate energy assessment, conservative path planning or battery swapping timing mismatch in UAV inspection tasks, and significantly improves the battery management accuracy, task execution stability and overall operational safety of UAV swarms in long-endurance inspection scenarios. Attached Figure Description

[0054] Figure 1 This is a flowchart of the UAV battery swapping scheduling method based on multi-state battery prediction according to the present invention.

[0055] Figure 2 This is a schematic diagram of the structure of the UAV battery swapping scheduling system based on multi-state battery prediction according to the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Example 1: As Figure 1 As shown, the UAV battery swapping scheduling method based on multi-state battery prediction includes the following steps:

[0058] Step 1: By using sensor modules installed on the drone, real-time data are collected on battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment, to obtain multi-source data on battery operating status and environmental status.

[0059] Step 2: Based on multi-source acquired data, calculate the battery's state of charge, state of health, state of power, state of energy, and temperature state to form a set of multi-state parameters for comprehensively characterizing the battery's operating state.

[0060] Step 3: Use a long short-term memory network model to model the historical state data containing multiple state parameters and environmental parameters, predict the battery state decay trend within a preset time window, and predict the energy required to complete the current task and the minimum energy required for the drone to return safely.

[0061] Step 4: Based on real-time multi-state parameters, prediction results and task priorities, generate scheduling instructions according to preset decision rules. The scheduling instructions include at least one of the following: continue the task, go to the charging station to charge, go to the battery swapping station to swap batteries, perform trickle charging, and perform cooling fast charging.

[0062] Step 5: When the scheduling instruction is to go to the battery swapping station, an improved ant colony algorithm is used to construct multiple feasible paths from the current position of the UAV to the candidate battery swapping station on the operation environment graph model. The battery health status factor is introduced, and the optimal path to the target battery swapping station is searched under the condition of simultaneously considering the path distance, energy consumption and flight time.

[0063] Step six: Control the drone to fly along the optimal path to the target battery swapping station, where the mechanical actuators of the swapping station will automatically complete the battery removal and installation, and select the corresponding charging mode for the replaced battery according to the scheduling instructions.

[0064] Step one involves using sensor modules mounted on the drone to collect real-time data on battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment. This yields multi-source data on battery operating status and environmental conditions. Specific steps include:

[0065] In this embodiment, a multi-rotor UAV performing a power line inspection mission is used as an example to illustrate the workflow and decision-making process of the system described in this invention. The UAV is assigned to complete a power line inspection mission with a total length of 8 kilometers. The mission type is routine inspection, and the flight path and work area have been pre-set through the mission planning module.

[0066] In this embodiment, the power battery carried by the UAV is a common six-cell lithium polymer battery with a nominal capacity of 6 amp-hours and a nominal voltage of 22.2 volts, corresponding to a nominal battery energy of approximately 133.2 watt-hours. This nominal energy is written into the battery parameter file during the system initialization phase, serving as one of the benchmark parameters for subsequent energy assessment, mission energy consumption estimation, and return-to-home safety margin judgment.

[0067] Once the mission begins, the state awareness layer activates, collecting multi-source state data on the battery and environment in real time. Specifically, the state awareness layer acquires data through a sensor array deployed on the drone, including:

[0068] The battery terminal voltage output by the battery voltage measurement unit is used to reflect the current voltage level and its correspondence with the state of charge.

[0069] The discharge current output by the current sensor is used to describe the current load size and charge consumption rate.

[0070] The battery temperature sensor outputs the battery casing or cell temperature to monitor thermal safety status;

[0071] The battery's equivalent internal resistance, estimated by the internal resistance detection unit, is used to characterize the battery's aging degree and transient discharge capability.

[0072] The external wind speed and ambient temperature output by the environmental sensors are used to characterize the impact of the flight environment on energy consumption and heat dissipation conditions.

[0073] The aforementioned sensors synchronously collect data according to the sampling period set by the system. In this embodiment, the sampling time interval can be set to 60 seconds. The state perception layer adds a unified timestamp to the data of each channel and performs time alignment, outlier removal, and format normalization operations. It encapsulates the battery voltage, current, temperature, internal resistance, ambient wind speed, and ambient temperature within the same sampling period into structured state data frames.

[0074] Step two involves calculating the battery's state of charge, state of health, state of power, state of energy, and temperature state based on multi-source acquired data, forming a multi-state parameter set for comprehensively characterizing the battery's operating state. Specific steps include:

[0075] After completing the task scenario setting and multi-source state data collection, the battery multi-state update calculation is first performed to obtain the remaining power and available energy assessment results that are closer to the current operating conditions, providing a basic input for subsequent energy consumption prediction and scheduling decisions.

[0076] Specifically, at the initial moment of this embodiment, the estimation results of the previous sampling period provide the initial values ​​of the battery's multiple states:

[0077] The current state of charge (SOC) is 60% (i.e. The battery's state of health (SOH) is 85%; its state of temperature (SOT) is 45°C; its rated capacity C_rated = 6 Ah (i.e., the aforementioned nominal capacity C_t); and its nominal voltage U_t = 22.2 V.

[0078] Within the current sampling period, the sensor readings uploaded by the state perception layer show that the instantaneous discharge current is I_t = 12 A; and the sampling time interval is Δt = 60 s.

[0079] The multi-state estimation module in the intelligent decision-making layer first uses the state of charge update formula adopted in this invention to correct the SOC. For ease of online calculation, this embodiment uses hours as the time unit, converting the sampling time interval to: Based on this, the SOC update formula is: ;

[0080] Substituting the values ​​from this embodiment: the amount of charge removed is: ;

[0081] The corresponding SOC correction is: Therefore, the state of charge at the end of the current sampling period is updated as follows: That is, the current remaining power is approximately 56.7%, which can be approximated as 57% in this embodiment.

[0082] After updating the remaining capacity, the multi-state estimation module further calculates the current available energy (SOE). In this embodiment, an energy estimation formula based on remaining capacity and nominal voltage is used: Substituting the updated SOC and battery parameters, we get: ;

[0083] Therefore, the key state quantities obtained at the end of this sampling period in this embodiment are: updated state of charge SOC_t ≈ 0.57; current available energy SOE_t ≈ 75.9 Wh; battery health state SOH remains at 85% (no structural change update of health state was triggered during this period); temperature state SOT remains at 45 ℃ and is marked as high temperature state for subsequent cooling or charging strategy evaluation.

[0084] The updated multi-state parameters are packaged into a new state vector and written into the state sequence cache, and simultaneously reported to the intelligent decision-making layer as the input basis for task energy consumption prediction and safe return energy assessment in subsequent steps.

[0085] Step 3: Utilize a Long Short-Term Memory (LSTM) network model to model historical state data containing multiple state parameters and environmental parameters, predict the battery state degradation trend within a preset time window, and predict the energy required to complete the current task and the minimum energy required for the drone's safe return. Specific steps include:

[0086] After completing the battery multi-state update described in step 2, the intelligent decision-making layer in this embodiment enters the battery state prediction and task energy consumption assessment stage, with the aim of determining whether the current power is sufficient to support this 8-kilometer inspection task and safe return.

[0087] In this embodiment, the intelligent decision-making layer integrates a Long Short-Term Memory (LSTM) prediction module. This module uses a multi-state sequence within a continuous historical time window as input to predict the battery state degradation trend within several future time windows. To take into account the battery's intrinsic state, current load, and environmental conditions, the input state sequence constructed in this embodiment can be represented as follows: in: , , , , These are the state of charge, health, power, energy, and temperature at the previous sampling time, respectively. This represents the discharge current at the current sampling moment; This is a quantity related to the current motor or propeller speed, used to characterize the load on the propulsion system; Ambient temperature; External wind speed or wind field intensity; This represents a historical sequence window formed by tracing back k time steps.

[0088] Based on the input sequence described above, the LSTM prediction module outputs predicted values ​​for the battery's state of charge, state of health, and temperature over a future period, used to assess the battery's continued availability under the current task strategy. In this embodiment, for simplicity, the focus is on calculating and comparing the energy requirements for this task in conjunction with the currently updated state of energy.

[0089] On one hand, this embodiment estimates the energy required to complete the current inspection task. The task parameters are: inspection task distance... Average cruising speed of the drone during the inspection phase The mission flight time is then: ;

[0090] Based on historical power consumption curves and flight logs, the typical average power consumption of this type of UAV in this type of inspection mission can be taken as: ;

[0091] This average power consumption value can be continuously corrected by the LSTM model, regression model, or online learning module over a long period of time, but in this embodiment, it serves as the power consumption benchmark at the current moment. Based on this, the energy required to complete this inspection task is estimated as follows: On the other hand, to ensure a safe return after the mission, this embodiment estimates the energy required for the return trip separately. Considering the return distance, possible hovering time, and energy consumption during landing, the energy required for a safe return trip is taken as follows in this embodiment: ;

[0092] Meanwhile, to address environmental disturbances, model errors, and unforeseen operating conditions, this embodiment further introduces a safety margin factor. This is used to reserve a certain proportion of redundant energy for the total energy demand. In this embodiment, it is taken as: Therefore, the total minimum energy requirement required to complete the mission and return safely can be obtained. Substituting the values, we get: ; ;

[0093] therefore:

[0094] Comparison with the current available energy obtained in step two: Current available energy: This is far less than the calculated minimum energy requirement. ;

[0095] Therefore, the intelligent decision-making layer draws a clear conclusion in this embodiment: based on the current remaining battery energy and the predicted task energy consumption, this 8-kilometer inspection mission cannot guarantee both the completion of the mission and a safe return without midway battery swapping or charging.

[0096] The results of this energy-level assessment will serve as an important constraint on scheduling decisions in subsequent steps. Even if the subsequent health and temperature conditions remain within acceptable limits to some extent, the scheduling strategy must prioritize energy security constraints to provide a basis for subsequent battery swapping priority determination and route planning.

[0097] Step four: Based on real-time multi-state parameters, prediction results, and task priorities, generate scheduling instructions according to preset decision rules. These instructions include at least one of the following: continue the task, proceed to a charging station for charging, proceed to a battery swapping station for battery swapping, perform trickle charging, and perform cooling fast charging. Specific steps include:

[0098] In this embodiment, upon entering step 4, the intelligent decision-making layer has obtained the following key state variables and evaluation results: Current available energy: The minimum energy required for the mission and return. ,in,

[0099] The calculations have already been performed in the aforementioned steps. Battery health status Battery temperature status

[0100] The system pre-defines multiple decision thresholds, such as: energy security judgment conditions: if If the current battery level is insufficient to complete the task and return safely simultaneously, then the health maintenance trigger threshold is considered to be present. When the SOH is below this threshold, maintenance or trickle charging should be prioritized; temperature safety threshold. ;

[0101] When the SOT is above this temperature, a cooling or low-stress charging strategy should be adopted.

[0102] The scheduling rule set and priority are designed as follows:

[0103] To make choices among multiple possible actions, the scheduling engine defines a hierarchical set of rules. The key rules can be summarized as: Energy safety priority rule: if... If the current mission is deemed to pose a risk of not being able to guarantee a return to base, then a strategy that can quickly restore power must be prioritized, with the first choice being to go to a battery swapping station.

[0104] Temperature safety rules: If If energy conditions permit (i.e., there is redundancy in energy to meet the needs of the mission and return), it is advisable to choose to go to a charging station to perform charging with a cooling process or to adopt a cooling fast charging mode.

[0105] Health maintenance rules: If If the current energy is sufficient to ensure a safe return, then trickle charging should be prioritized to slow down battery degradation.

[0106] Normal operating rules: When When the SOT is within a safe temperature range and the task is urgent, instructions can be generated to continue the task or to continue the task after a short fast charge.

[0107] In the aforementioned rule set, the energy security priority rule has the highest priority. When the energy security conditions are not met, other rules will no longer be triggered individually, but will be reflected as additional constraints in the battery swapping or return-to-base strategy.

[0108] The specific decision-making process in this embodiment

[0109] State variables in this embodiment: Currently available energy Minimum energy required Obviously, there are:

[0110] Therefore, the "energy safety priority rule" was triggered first. The system determined that without mid-journey refueling, the existing energy was insufficient to complete the 8-kilometer inspection mission and return safely. Continuing to execute the mission would face a serious risk of losing contact or forced landing.

[0111] Meanwhile, other state variables show that the battery health status SOH = 85% is higher than the maintenance trigger threshold α = 70%, indicating that the battery has not yet entered the serious aging range. There is no need to immediately exit the task due to health issues. It is only necessary to continue to monitor its long-term degradation trend in subsequent scheduling.

[0112] The battery temperature SOT = 45℃ is higher than the temperature threshold β = 40℃. From a temperature perspective, a cooling and gentle charging strategy should be implemented to avoid high-power charging and discharging at high temperatures.

[0113] However, due to the failure to meet energy safety requirements, the scheduling engine prioritized "ensuring mission and return safety" as its primary objective, and eliminated and screened all candidate actions according to priority rules:

[0114] Continue the task: due to The mission was rejected outright because it did not meet energy safety requirements.

[0115] Heading to a charging station for charging or performing a cooling-down fast charge: In this embodiment, the drone currently has limited remaining energy. Flying from its current location to a charging station for charging and then continuing to perform tasks after charging would introduce additional flight distance and time overhead. Furthermore, it is not guaranteed that the inspection and return to base can still be completed within the overall mission window after charging is completed. At the same time, considering that the current SOT is already high, going to a charging station for high-power fast charging when the energy is severely insufficient may actually exacerbate the equipment risk. Therefore, in this scenario, both "heading to a charging station for charging" and "performing a cooling-down fast charge" are covered by the higher priority of battery swapping needs.

[0116] Perform trickle charge maintenance: Trickle charge maintenance is mainly for scenarios where SOH drops significantly but energy is still sufficient, in order to delay battery aging; currently, SOH = 85% is higher than the threshold α, and the energy itself is seriously insufficient, so this strategy is not applicable to the current emergency scenario.

[0117] Based on the above analysis and according to preset priority rules, the scheduling engine ultimately determines the optimal scheduling action in this embodiment as: immediately generating a scheduling instruction to "go to the battery swapping station for battery swapping". This scheduling instruction contains at least the following information:

[0118] Command type: Battery swap command;

[0119] Execution mode: Interrupt the current inspection task and prioritize switching to the nearest battery swapping station with service capabilities;

[0120] Constraints: In subsequent route planning, priority should be given to selecting sites that are closer, have lower energy consumption, and can quickly complete battery swapping, and the routes and sites should be screened in combination with the current battery SOH and temperature status;

[0121] Additional note: Replaced batteries should be cooled down during the charging phase before entering regular charging or maintenance mode.

[0122] After generating the battery swapping instruction, the intelligent decision-making layer encapsulates the instruction into a standardized scheduling message and sends it to the execution control layer. The execution control layer then initiates the subsequent path planning and ant colony algorithm optimization process, thus entering the next step of path planning and execution.

[0123] Step 5: When the scheduling instruction is to go to a battery swapping station, an improved ant colony algorithm is used to construct multiple feasible paths from the UAV's current location to the candidate battery swapping station on the operational environment graph model. A battery health status factor is introduced, and the optimal path to the target battery swapping station is searched while simultaneously considering path distance, energy consumption, and flight time. Specific steps include:

[0124] An improved ant colony algorithm is invoked to comprehensively optimize the path from the drone's current location to each candidate battery swapping station. The goal is to prioritize the selection of battery swapping stations and their flight paths that are closer, have lower energy consumption, and can quickly obtain services, while meeting energy and safety constraints.

[0125] The goals for setting operational environment modeling and optimization include:

[0126] First, the current task environment is abstracted as a weighted directed graph. , where: node set Including the current location node of the drone Each candidate battery swapping station node , and relay nodes introduced when encountering complex terrain or obstacles that need to be bypassed;

[0127] edge set Represents the feasible flight paths between nodes, each edge Assigning ternary weights , respectively corresponding to the nodes To the node The flight distance, estimated energy consumption, and estimated flight time.

[0128] To unify multiple objectives into a single optimization framework, this embodiment sets the path optimization objective function as follows: ,in: This represents the total distance from the current location to the target battery swapping station, in meters. This represents the total energy consumption for the path, in watt-hours. Indicates the total flight time along the route, in seconds or hours; It is an adjustable weighting coefficient used to make trade-offs between distance, energy consumption, and time, and is calibrated in combination with actual application scenarios.

[0129] The steps for designing transition probabilities during path construction are as follows:

[0130] During the solution construction process, each virtual ant starts from its current position node. Departure, in the map Step by step, select the next node and construct from To a candidate battery swapping station node The complete path. To balance historical search experience with the cost of the current path, this embodiment adopts a node transfer mechanism based on pheromones and heuristic factors.

[0131] Specifically, the virtual ant at the current node Select the next node When the transition probability is defined as: ,in: For a moment Time Side The intensity of pheromones on that side reflects the historical search preference for that side; For the edge The heuristic factor is used to measure the merits of the current local cost. In this embodiment, the following can be taken: In other words, the smaller the combined cost of distance, energy consumption, and time, the better. The larger; and The weighting parameter is used to balance the influence of pheromones and heuristic factors on the transition decision. This is a set of candidate nodes that have not yet been visited, used to avoid creating invalid loops.

[0132] Through the above-mentioned transition probability design, the algorithm can prioritize exploring path segments with lower local costs based on existing pheromone guidance, thereby achieving a balance between exploration and utilization.

[0133] A pheromone update mechanism for battery state of health (SOH) is introduced, the specific contents of which include:

[0134] To enable the path planning results to adapt to the current battery health status, this embodiment introduces a battery health status factor into the pheromone update rule of the traditional ant colony algorithm, which makes it more inclined to choose a safe path with shorter distance and lower energy consumption when the battery health status is poor.

[0135] After each iteration, pheromones are updated according to the following rules: ,in: This is the pheromone evaporation coefficient, used to control the decay rate of old pheromones;

[0136] In this embodiment, the pheromone increment related to battery health is set as a function that increases as SOH decreases. That is, when SOH is low, under the premise of meeting flight safety constraints, the pheromone increment is increased for paths with shorter distances and lower expected energy consumption. Apply a larger pheromone increment.

[0137] With this design, when the battery is in good health, the algorithm can balance factors such as distance, time, and queue waiting over a wider range; while when the battery health declines and the system tends to adopt a more conservative strategy, pheromones will accelerate their concentration on shorter, more energy-efficient paths, causing the drone to prioritize battery swapping routes with higher safety margins.

[0138] After multiple iterations, the distribution of pheromones along multiple paths gradually converged. Finally, considering distance, time, energy consumption, and the current battery health, the objective function was selected. The path with the smallest pheromone concentration is selected as the optimal path from the current location to the target battery swapping station.

[0139] The specific process for selecting a battery swapping station is explained below. In the given scenario of this embodiment, the candidate battery swapping stations include: battery swapping station A, which is about 1.5 km away from the current location; and battery swapping station B, which is about 2.2 km away from the current location.

[0140] Based on the planned path and average flight speed The estimated flight time from the current location to station A is: ;

[0141] Average power consumption in this embodiment The estimated energy consumption for a flight to station A is... The currently available energy is the amount calculated in steps three and four: ;

[0142] Therefore, from an energy perspective, the energy consumption to fly to station A is significantly less than the currently available energy, and the drone can safely reach it. Similarly, the time and energy consumption to fly to station B can be estimated, where the flight time is approximately: The corresponding energy consumption is approximately: Similarly satisfied This indicates that, from an energy constraint perspective, both paths can reach the corresponding battery swapping station.

[0143] Based on this, the improved ant colony algorithm incorporates the distances of the two paths (1.5 km and 2.2 km), the corresponding energy consumption (approximately 20.8 Wh and 30.6 Wh), the estimated flight time, and the current queuing status of the battery swapping station into the cost and heuristic factor calculations. Combined with the current battery health status (SOH) and the principle of prioritizing energy safety, after multiple iterations, the pheromones are gradually concentrated on the path that is "closer, has lower energy consumption, and shorter service waiting time".

[0144] Ultimately, the algorithm converges to the path leading to battery swapping station A. Based on this, the execution control layer determines battery swapping station A as the target station for this battery swapping operation and sends the optimal path and target station information as control command parameters to the UAV flight control and battery swapping station coordination process, providing path and station information for autonomous flight and battery swapping execution in subsequent steps.

[0145] Step Six: Control the drone to fly along the optimal path to the target battery swapping station. The mechanical actuators at the swapping station will automatically remove and install the battery, and select the corresponding charging mode for the replaced battery according to the scheduling instructions. Specific steps include:

[0146] Autonomous flight and docking preparation along the optimal path includes:

[0147] The execution control layer first sends the optimal path obtained based on the improved ant colony algorithm to the UAV flight control system, using key waypoints, flight altitude, speed constraints, and safety buffer zones in the path as flight plan parameters. While maintaining current attitude stability, the UAV autonomously flies along the optimal path to the pre-defined docking airspace near the target battery swapping station A.

[0148] As the drone approaches battery swapping station A, the flight control system gradually reduces its altitude and speed, and activates a fine-tuning positioning and alignment mechanism. Visual markers, positioning lights, or other alignment indicators are pre-positioned in the station's work area. The drone identifies these indicators using onboard visual sensors or positioning devices, adjusting its attitude and lateral position in real time to gradually align its fuselage with the station's mechanical interface. The entire alignment process is monitored by the execution control layer. Only when positional deviation, attitude error, and speed all fall within preset safety thresholds is the final docking stage permitted.

[0149] The robotic arm battery swapping process begins with the drone hovering and locking its attitude in the docking area. The execution control layer then sends a "battery swapping allowed" handshake signal to the battery swapping station control unit. After the safety interlock logic confirms that the drone is under control and the propellers are at a safe speed, the robotic arm at the battery swapping station begins the battery swapping operation.

[0150] Specifically, in this embodiment, the robotic arm of the battery swapping station executes the following actions sequentially according to a pre-set mechanical interface standard and operating sequence:

[0151] Unlock the securing mechanism: Release the mechanical locks or clamps securing the drone battery to ensure that the battery can be safely removed;

[0152] Removing old batteries: The robotic arm uses a customized gripper to hold the battery module and smoothly pulls out the old battery along the guide trajectory. During the process, the torque and position changes are monitored to prevent jamming or impact.

[0153] Install a new battery: Take out a new battery module that has been charged and tested from the battery buffer compartment, align it with the drone battery compartment guide rail and positioning hole, push the battery in smoothly and lock it in;

[0154] Electrical connection confirmation: After the mechanical locking is completed, the system quickly checks the battery voltage, contact resistance and communication link. After confirming that the power supply and communication are normal, it returns a "battery swap complete" signal to the execution control layer.

[0155] Throughout the battery swapping process, the station control unit continuously collects the robotic arm's posture, force conditions, and interface connection status. Once abnormal resistance or positional deviation exceeding the threshold is detected, a protection action will be triggered to suspend the battery swapping and issue an alarm.

[0156] Temperature management and charging strategies for the replaced battery are implemented, including:

[0157] In this embodiment, the removed old battery had reached a high temperature during the mission, with a SOT of 45°C, exceeding the preset temperature threshold β of 40°C. When generating the battery swapping command, the intelligent decision-making layer sent the battery's temperature information and suggested charging strategy to the battery swapping station control unit to guide the subsequent charging and maintenance process.

[0158] After receiving an old battery, the battery swapping station first performs a cooling process based on its temperature condition. A typical process includes:

[0159] Move the old battery into a buffer area with heat dissipation conditions, and start the forced air cooling or cooling circulation system to accelerate the temperature of the battery surface and internal parts down to a safe range.

[0160] After the temperature drops below the safe limit, select an appropriate charging mode based on the battery health status (SOH) and the current task load. For example, first use a low current constant current stage to restore the charge to a moderate level, and then switch to a constant voltage or trickle stage to slow down battery capacity decay and internal resistance increase.

[0161] Throughout the charging process, the battery voltage, temperature, and internal resistance changes are continuously monitored. If an abnormal temperature rise or voltage abnormality is detected, the charging current is automatically reduced or charging is stopped, and the abnormality is recorded in the battery life cycle management file.

[0162] The above-mentioned cooling and flexible charging strategies not only ensure that the replaced battery is charged within a safe temperature range, but also take into account the long-term health maintenance and life management needs of the battery.

[0163] New battery status confirmed and task resumed.

[0164] After the new battery is installed, the execution control layer will perform a quick status check, which mainly includes:

[0165] Read the new battery's State of Charge (SOC) to confirm that it has reached the preset lower limit required by the current mission; read the new battery's State of Health (SOH) to confirm that it is in a healthy range suitable for performing this type of mission; read the new battery's State of Temperature (SOT) to confirm that its temperature is within the safe operating range; verify that the data communication between the battery and the drone is normal, including data channels such as voltage, current, temperature, and status indicators.

[0166] After the check is passed, the execution control layer reports a "battery swapping completion and status confirmation" event to the ground or cloud management platform. This event includes information such as: swapping time, target swapping station identifier, SOC, SOH, SOT of the old and new batteries, task number, and swapping time. The management platform writes this information to the task log and battery management database for subsequent statistical analysis and model training.

[0167] Subsequently, the drone receives updated scheduling instructions from the intelligent decision-making layer, typically "continue the inspection mission." Based on the current mission progress and planned route, the flight control system takes off again from the interrupted position or the nearest available resuming point to rejoin the inspection route and continue to complete the remaining power line inspection tasks.

[0168] Throughout the entire battery swapping execution and task recovery process, the system continuously records multi-source state data, scheduling decisions, path planning results, and battery swapping execution status: battery-related records include changes in SOC, SOH, and SOT before and after each task, charging mode, and charging duration; task-related records include the number of battery swaps triggered by energy and temperature, the distribution and utilization of battery swapping stations, and the task completion rate; and algorithm-related records include the ant colony algorithm convergence rounds, the cost of the selected path, and the battery swapping station selection results.

[0169] This recorded data will be centrally aggregated by the cloud management platform and used for analyzing LSTM prediction models, scheduling rule parameters, and path planning weight coefficients. Offline analysis and online adaptive adjustments are performed. Through repeated iterations, the accuracy of mission energy consumption prediction, the rationality of battery swapping decisions, and the overall efficiency of path planning are gradually improved, better adapting to different UAV models, mission types, and environmental conditions.

[0170] Among the parameters involved are, for example, the nominal battery capacity. Nominal voltage Average power consumption Return energy Safety margin coefficient All of these can be calibrated during the system deployment phase based on the specific drone platform, mission characteristics, and field experience, and dynamically corrected during operation based on historical data to achieve adaptive adaptation to different application scenarios.

[0171] It should be noted that the thresholds involved in the embodiments can be determined according to specific scenarios and needs.

[0172] This invention constructs an autonomous battery swapping control mechanism for UAVs based on multi-state battery perception and task-priority scheduling, enabling dynamic perception and differentiated control of battery status and energy safety during inspection missions of multi-rotor UAVs. The mechanism collects and standardizes multi-source data such as battery voltage, current, temperature, internal resistance, and ambient wind speed and temperature through a state perception layer. It extracts multi-state indicators including state of charge, health status, power status, energy status, and temperature status, generating a battery state feature vector that integrates task context and environmental constraints. Based on a long short-term memory network, it predicts task energy consumption and safe return energy. Combining task urgency and safety strategies, it generates a scheduling instruction structure including options for continuing the task, proceeding to a charging station, proceeding to a battery swapping station, maintenance charging, and cooling fast charging. After determining the battery swapping requirement, an improved ant colony algorithm incorporating battery health status factors is used to construct a path scheduling cost function, taking into account path distance, energy consumption, and arrival time, achieving optimal selection of battery swapping stations, optimal path planning, and adaptive control with safety margins.

[0173] After the strategy is executed, the inspection interruption points, battery swapping paths and results are analyzed to evaluate the task completion rate, energy safety margin utilization rate and the evolution trend of battery temperature and health. Management features such as scheduling response delay, path cost offset and battery swapping station load distribution are extracted to construct data management coefficients to describe the matching degree between energy management and task scheduling, and generate model parameter update signals or decision threshold adjustment signals. This reduces the problems of task interruption and forced landing caused by inaccurate energy assessment, conservative path planning or battery swapping timing mismatch in UAV inspection tasks, and significantly improves the battery management accuracy, task execution stability and overall operational safety of UAV swarms in long-endurance inspection scenarios.

[0174] Example 2: A UAV battery swapping scheduling system based on multi-state battery prediction, such as Figure 2 As shown, it specifically includes:

[0175] The state awareness layer is used to collect battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment in real time through a sensor array set on the drone. The collected data is time-aligned and format-unified to generate a structured multi-source data sequence.

[0176] The intelligent decision-making layer is used to receive multi-source collected data and has a built-in multi-state estimation module and a prediction decision-making module. The multi-state estimation module is used to calculate the battery's state of charge, state of health, state of power, state of energy, and temperature state based on the multi-source collected data. The prediction decision-making module is used to predict the battery state and task energy consumption using a long short-term memory network model, and generate scheduling instructions based on real-time multi-state parameters, prediction results, and task priorities, including at least one of the following: continue task, go to charging station for charging, go to battery swapping station for battery swapping, perform trickle charging, and perform cooling fast charging.

[0177] The execution control layer receives scheduling instructions. When the scheduling instruction is to go to the battery swapping station for battery swapping, an improved ant colony algorithm with the introduction of a battery health status factor is used to search for the optimal path to the target battery swapping station on the operation environment graph model, taking into account path distance, energy consumption, and flight time. The drone is then controlled to fly along the optimal path to the target battery swapping station, where the mechanical actuators of the battery swapping station automatically complete the battery removal and installation, and select the corresponding charging mode for the replaced battery according to the scheduling instructions.

[0178] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0179] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0180] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0181] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0182] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0183] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0184] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0185] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A UAV battery swapping scheduling method based on multi-state battery prediction, characterized in that, Includes the following steps: Step 1: By using sensor modules installed on the drone, real-time data are collected on battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment, to obtain multi-source data on battery operating status and environmental status. Step 2: Based on multi-source acquired data, calculate the battery's state of charge, state of health, state of power, state of energy, and temperature state to form a set of multi-state parameters for comprehensively characterizing the battery's operating state. Step 3: Use a long short-term memory network model to model the historical state data containing multiple state parameters and environmental parameters, predict the battery state decay trend within a preset time window, and predict the energy required to complete the current task and the minimum energy required for the drone to return safely. The difference between the predicted energy required to complete the current task and the minimum energy required for the drone to return safely is used as the energy safety margin. Step 4: Based on real-time multi-state parameters, prediction results and task priorities, generate scheduling instructions according to preset decision rules. The scheduling instructions include at least one of the following: continue the task, go to the charging station to charge, go to the battery swapping station to swap batteries, perform trickle charging, and perform cooling fast charging. Step 5: When the scheduling instruction is to go to the battery swapping station, an improved ant colony algorithm is used to construct multiple feasible paths from the current position of the UAV to the candidate battery swapping station on the operation environment graph model. The battery health status factor is introduced, and the optimal path to the target battery swapping station is searched under the condition of simultaneously considering the path distance, energy consumption and flight time. Step 6: Control the drone to fly along the optimal path to the target battery swapping station. The mechanical actuator of the battery swapping station will automatically complete the battery removal and installation, and select the corresponding charging mode for the replaced battery according to the scheduling instructions. Step four, the intelligent scheduling decision generation includes: When the health status is below the first preset threshold, or the energy security margin is negative, a dispatch instruction is generated to go to the battery swapping station for battery swapping. When the temperature exceeds the second preset threshold, a scheduling instruction is generated to proceed to the charging station to perform cooling fast charging. When the task urgency parameter is higher than the third preset threshold and the prediction results show that the task can be completed and the safe return conditions can be met through fast charging, a scheduling instruction to go to the charging station to perform fast charging is generated. When the health status is below the fourth preset threshold and the energy safety margin is positive, a scheduling instruction to execute trickle maintenance charging is generated. If none of the conditions are met, a scheduling instruction to continue the task is generated; The improved ant colony algorithm described in step five specifically includes: When the virtual ant selects the next node, the transition probability of each candidate edge is calculated based on the pheromone intensity on the candidate edge and the heuristic factor obtained by combining the path distance, energy consumption, flight time and current battery health status. After completing each generation of path search, the path quality is calculated based on the total distance, total energy consumption and total flight time of the path, and the original pheromones on each side are attenuated according to the preset pheromone volatilization coefficient. New pheromone increments are added to the edges of the traversed paths based on path quality and current battery health status; When the improvement in path quality is less than a preset threshold or the maximum number of iterations is reached, the path with the highest pheromone concentration is selected as the optimal path to the target battery swapping station.

2. The UAV battery swapping scheduling method based on multi-state battery prediction according to claim 1, characterized in that: Step one includes: acquiring battery voltage through a voltage sensor, acquiring battery current through a current sensor, acquiring battery temperature through a temperature sensor, obtaining battery internal resistance through an internal resistance detection unit, and acquiring wind speed and ambient temperature in the flight environment through an environmental sensor. The data from each acquisition channel are time-aligned and interpolated according to a unified time base, and the data format is standardized and outlier is removed to form a time-continuous and structured multi-source acquisition data sequence.

3. The UAV battery swapping scheduling method based on multi-state battery prediction according to claim 2, characterized in that: Step two, the multi-state calculation and evaluation of the battery includes: The state of charge is updated by integrating the sampled current over the sampling time interval and comparing it with the battery's nominal capacity. The health status is calculated by combining the ratio of current available capacity to nominal capacity, the change in current internal resistance relative to initial internal resistance, and the number of charge-discharge cycles already experienced as a decay index. By estimating the current maximum allowable continuous discharge current and combining it with the terminal voltage, the maximum power that the battery can output is determined, and the power state is obtained. The remaining usable energy can be obtained by the correspondence between the state of charge and the nominal energy of the battery, or by integrating the power change curve over time. By comparing real-time temperature samples with preset safe temperature ranges, the temperature status is obtained, and it is determined whether there is a risk of overheating or overcooling.

4. The UAV battery swapping scheduling method based on multi-state battery prediction according to claim 3, characterized in that: Step three, battery status and energy consumption prediction, includes: A long short-term memory network model is constructed with multiple state parameters and environmental parameters within a continuous time window as input, and the predicted values ​​of battery state of charge, health state and temperature state at future time steps are output. Based on the predicted values ​​and the planned flight distance, predetermined flight speed, and power consumption curves of the current mission, a regression estimate is made of the energy required to complete the current mission. Based on the distance the UAV needs to return from its current operating position to a safe landing position, its return flight speed, and conservative power consumption assumptions, the minimum energy required for the UAV to return safely is estimated, and the energy safety margin is calculated accordingly.

5. The UAV battery swapping scheduling method based on multi-state battery prediction according to claim 4, characterized in that: Step five includes: The operating environment is abstracted as a weighted directed graph. The nodes of the graph include the current location node of the UAV, multiple candidate battery swapping station nodes, and relay nodes. The edges of the graph represent feasible flight paths between nodes, and each edge carries attribute parameters such as flight distance, expected energy consumption, and expected flight time. With the optimization goal of simultaneously minimizing path distance, energy consumption, and arrival time, algorithm parameters such as ant colony size, maximum number of iterations, and pheromone evaporation coefficient are set. In each generation of the search process, the virtual ant is controlled to start from the current location node of the drone and select the next node in turn on the weighted directed graph to construct a complete path from the current location node to a certain battery swapping station node.

6. The UAV battery swapping scheduling method based on multi-state battery prediction according to claim 5, characterized in that: Step six, the drone battery swapping process includes: The drone is controlled to fly along the optimal path to the preset docking position of the target battery swapping station, and automatically docks with the mechanical actuators of the battery swapping station based on visual alignment and attitude adjustment mechanisms. The mechanical actuator removes the old batteries from the drone and installs fully charged new batteries in a predetermined sequence. According to the dispatch instructions, the old batteries to be replaced are sent to the fast charging station, trickle charging station, or cooling charging station for charging and maintenance, and the charging mode, charging time, and health status update results of each battery are recorded.

7. The UAV battery swapping scheduling method based on multi-state battery prediction according to claim 6, characterized in that: Throughout the entire process of the drone mission, the multi-state parameters of the battery, the summary of multi-source collected data, the scheduling instructions, the path planning results and the battery swapping results are reported to the cloud management platform. The cloud management platform uses data from drones to jointly train and update the parameters of the long short-term memory network model and scheduling decision rules using federated learning or distributed training methods, so as to obtain the updated model parameters and decision rules without uploading the original data in a centralized manner. The updated model parameters and decision rules were distributed to each drone.

8. A UAV battery swapping scheduling system based on multi-state battery prediction, used to implement the UAV battery swapping scheduling method based on multi-state battery prediction as described in any one of claims 1-7, characterized in that, include: The state awareness layer is used to collect battery voltage, battery current, battery temperature, battery internal resistance, as well as wind speed and ambient temperature in the flight environment in real time through a sensor array set on the drone. The collected data is time-aligned and format-unified to generate a structured multi-source data sequence. The intelligent decision-making layer is used to receive multi-source collected data and has a built-in multi-state estimation module and a prediction decision-making module. The multi-state estimation module is used to calculate the battery's state of charge, state of health, state of power, state of energy, and temperature state based on the multi-source collected data. The prediction decision-making module is used to predict the battery state and task energy consumption using a long short-term memory network model, and generate scheduling instructions based on real-time multi-state parameters, prediction results, and task priorities, including at least one of the following: continue task, go to charging station for charging, go to battery swapping station for battery swapping, perform trickle charging, and perform cooling fast charging. The execution control layer receives scheduling instructions. When the scheduling instruction is to go to the battery swapping station for battery swapping, an improved ant colony algorithm with the introduction of a battery health status factor is used to search for the optimal path to the target battery swapping station on the operation environment graph model, taking into account path distance, energy consumption, and flight time. The drone is then controlled to fly along the optimal path to the target battery swapping station, where the mechanical actuators of the battery swapping station automatically complete the battery removal and installation, and select the corresponding charging mode for the replaced battery according to the scheduling instructions.

Images