Charging parameter optimization method and system based on artificial intelligence

Abstract

The invention discloses a charging parameter optimization method and system based on artificial intelligence, and relates to the field of unmanned aerial vehicle charging, and the method comprises the steps: obtaining the battery type, current electric quantity and remaining flight tasks of an unmanned aerial vehicle; determining a charging mode and an initial charging parameter of the unmanned aerial vehicle based on the battery type, the current electric quantity and the remaining flight tasks; in response to the charging signal of the unmanned aerial vehicle, monitoring in real time to obtain a battery charging curve of the unmanned aerial vehicle; based on the battery charging curve of the current time period, an artificial intelligence model is adopted, charging parameters of the unmanned aerial vehicle battery in the next preset time period are dynamically adjusted at the end moment of the current time period, and the charging parameters comprise charging current and charging voltage; and charging the battery of the unmanned aerial vehicle according to the charging parameters at the starting moment of the next preset time period. The charging time can be shortened to the greatest extent on the premise of ensuring the safety and the service life of the battery of the unmanned aerial vehicle.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of UAV charging, and in particular to a method and system for optimizing charging parameters based on artificial intelligence. Background Art

[0002] With the rapid development of UAV technology, UAVs have been widely used in many fields such as surveying and mapping, agriculture, logistics, emergency rescue, etc. However, the battery life has always been the key bottleneck restricting the long-term operation of UAVs. At present, the flight time of most commercial UAVs is only 20 - 30 minutes, while the charging time often takes 1 - 2 hours. This contradiction of "short flight and long charging" seriously affects the work efficiency and mission continuity of UAVs. Especially in scenarios with high timeliness requirements such as emergency rescue and large-scale surveying and mapping, too long charging time will lead to mission interruption and missed key opportunities. Therefore, how to shorten the UAV charging time and improve the work efficiency of UAVs while ensuring the safety and life of the battery has become an urgent problem in the industry.

[0003] Traditional UAV charging methods mainly include three modes: constant current charging, constant voltage charging, and constant current-constant voltage charging.

[0004] Constant current charging refers to a charging method that keeps the charging current constant during the charging process. This method has a relatively fast charging speed, but it is easy to cause the battery to overheat and overcharge, affecting the battery life. Constant voltage charging is to keep the charging voltage constant during the charging process. The charging current is relatively large at the beginning of charging, and as the battery charge increases, the charging current gradually decreases. This method is relatively safe, but the charging speed is slow, especially the charging efficiency significantly decreases in the later stage.

[0005] At present, most UAVs adopt the constant current-constant voltage charging method, that is, first charge with a constant current, and when the battery voltage reaches a certain value, switch to constant voltage charging. This method is a combination of the first two methods, which balances the charging speed and safety to a certain extent. However, different types and different degrees of aging of the battery have different charging characteristics. The traditional charging method adopts the above fixed charging strategy to ensure safety, but sacrifices the charging speed to a certain extent, resulting in low charging efficiency or unnecessary loss of battery life.

[0006] Therefore, there is an urgent need for a method that can intelligently adjust the charging strategy according to the real-time state of the battery, so as to minimize the charging time while ensuring the safety and life of the battery, and improve the work efficiency and mission continuity of UAVs. Summary of the Invention

[0007] The embodiments of the present application provide a method and system for optimizing charging parameters based on artificial intelligence, which are used to minimize the charging time while ensuring the safety and life of the UAV battery.

[0008] To achieve the above object, the embodiments of the present application adopt the following technical solutions:

[0009] In a first aspect, a method for optimizing charging parameters based on artificial intelligence is provided, and the method includes:

[0010] Obtain the battery type, current battery level, and remaining flight mission of the drone;

[0011] Based on the battery type, current battery level, and remaining flight mission, determine the charging mode and initial charging parameters of the drone;

[0012] In response to the charging signal of the drone, obtain the battery charging curve of the drone in real time, where the battery charging curve is constructed based on the real-time charging voltage, real-time charging current, and real-time battery temperature;

[0013] Based on the battery charging curve in the current time period, use an artificial intelligence model to dynamically adjust the charging parameters of the drone battery in the next preset time period at the end of the current time period, and the charging parameters include charging current and charging voltage;

[0014] Charge the drone battery according to the charging parameters at the start time of the next preset time period.

[0015] In a possible implementation manner of the first aspect, based on the battery type, current battery level, and remaining flight mission, determining the charging mode and initial charging parameters of the drone includes:

[0016] According to the battery type of the drone, obtain the corresponding battery safety parameter threshold from the preset battery parameter database;

[0017] According to the current battery level, determine the charging stage of the battery;

[0018] According to the remaining flight mission, determine the flight urgency, where the flight urgency includes high, medium, and low;

[0019] Based on the battery safety parameter threshold, charging stage, and flight urgency, determine that the charging mode is any one of a constant current charging mode, a constant voltage charging mode, or a constant current and constant voltage combined charging mode, and determine the initial charging current value or charging voltage value according to the charging mode.

[0020] In another possible implementation manner of the first aspect, the remaining flight mission includes mission type and estimated flight time, and according to the remaining flight mission, determining the flight urgency includes:

[0021] When the mission type is the preset first mission type, determine that the flight urgency is high;

[0022] When the task type is not the first task type, determine that the flight emergency level is medium or low based on the estimated flight time;

[0023] Among them, when the estimated flight time is less than the preset time threshold, determine that the flight emergency level is medium; when the estimated flight time is not less than the preset time threshold, determine that the flight emergency level is low.

[0024] In another possible implementation of the first aspect, the battery safety parameter threshold includes a safety current threshold and a safety voltage threshold. Based on the battery safety parameter threshold, the charging stage, and the flight emergency level, determine that the charging mode is any one of a constant current charging mode, a constant voltage charging mode, or a constant current-constant voltage combined charging mode, and determine the initial charging current value or the charging voltage value according to the charging mode, including:

[0025] When the charging stage is the first charging stage and the flight emergency level is high, determine that the charging mode is a constant current-constant voltage combined charging mode, and set the initial charging current value to the first preset percentage of the safety current threshold;

[0026] When the charging stage is the first charging stage and the flight emergency level is medium or low, determine that the charging mode is a constant current charging mode, and set the initial charging current value to the second preset percentage of the safety current threshold, where the second preset percentage is less than the first preset percentage;

[0027] When the charging stage is the second charging stage, determine that the charging mode is a constant voltage charging mode, and set the initial charging voltage value to the third preset percentage of the safety voltage threshold.

[0028] In another possible implementation of the first aspect, based on the battery charging curve of the current time period, use an artificial intelligence model to dynamically adjust the charging parameters of the UAV battery in the next preset time period at the end of the current time period, including:

[0029] Analyze the change trend of the battery charging curve through the artificial intelligence model to obtain an analysis result, and the analysis result includes the battery health and the optimal charging parameters;

[0030] Based on the battery health and the optimal charging parameters, dynamically adjust the charging parameters of the UAV battery in the next preset time period at the end of the current time period.

[0031] In another possible implementation of the first aspect, the artificial intelligence model is a deep reinforcement learning model, including a convolutional neural network layer for extracting battery charging curve features, a long short-term memory network layer for learning the temporal change law of the battery charging state, a policy network for predicting the optimal charging parameters, and a value network.

[0032] In another possible implementation of the first aspect, the change trend of the battery charging curve is analyzed by an artificial intelligence model to obtain an analysis result, and the analysis result includes the battery health and the optimal charging parameters, including:

[0033] Collect the battery charging curve data within the current time period, and the battery charging curve data is time series data including charging voltage, charging current, and battery temperature;

[0034] Preprocess the time series data;

[0035] Input the preprocessed time series data into a convolutional neural network layer to extract the local features of the battery charging curve;

[0036] Input the local features into a long short-term memory network layer to learn the time series change law of the battery charging state;

[0037] Based on the deviation between the battery charging curve and the standard charging curve, calculate the battery internal resistance change rate, capacity attenuation rate, and temperature sensitivity index;

[0038] According to the internal resistance change rate, capacity attenuation rate, and temperature sensitivity index, use a preset battery health function to calculate the battery health;

[0039] Input the local features and the time series change law into a policy network and a value network, predict the charging efficiency and battery life under different charging parameters, and obtain the optimal charging parameters.

[0040] In another possible implementation of the first aspect, based on the battery health and the optimal charging parameters, the charging parameters of the UAV battery in the next preset time period are dynamically adjusted at the end of the current time period, including:

[0041] When the battery health is lower than the first preset threshold, at the end of the current time period, reduce the charging current in the next preset time period to the first preset ratio of the optimal charging parameter, and reduce the charging voltage to the second preset ratio of the optimal charging parameter;

[0042] When the battery health score is not lower than the first preset threshold and not higher than the second preset threshold, at the end of the current time period, set the charging parameters in the next preset time period to the optimal charging parameters;

[0043] When the battery health score is higher than the second preset threshold and the flight emergency level is high, at the end of the current time period, increase the charging current in the next preset time period to the third preset ratio of the optimal charging parameter, and not exceed the battery safety parameter threshold.

[0044] In a second aspect, the present application provides an electronic device, including:

[0045] A memory configured to store instructions; and

[0046] A processor configured to call the instructions from the memory and capable of implementing the above-mentioned artificial intelligence-based charging parameter optimization method when executing the instructions.

[0047] In a third aspect, the present application provides an artificial intelligence-based charging parameter optimization system, including:

[0048] An electronic device;

[0049] A drone connected to the electronic device.

[0050] Through the above technical solution, the battery type, current power, and remaining flight tasks of the drone are obtained to determine a suitable charging mode and initial parameters; by real-time monitoring of the battery charging curve and using a deep reinforcement learning model to dynamically adjust the charging parameters, intelligent management of the drone battery charging process is achieved. The charging strategy can be adaptively adjusted according to the real-time state, health, and task urgency of the battery, maximizing the charging efficiency while ensuring the safety and lifespan of the battery. Compared with traditional fixed charging strategies, this method can effectively shorten the charging time, extend the battery lifespan, significantly improve the working efficiency and task continuity of the drone, and is particularly suitable for scenarios with high timeliness requirements such as emergency rescue and large-scale mapping. In addition, this method also has good adaptability and scalability and can adapt to different types and aging degrees of batteries.

[0051] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. Brief Description of the Drawings

[0052] Figure 1 It is a flowchart of an artificial intelligence-based charging parameter optimization method provided by an embodiment of the present application;

[0053] Figure 2 It is a structural diagram of an artificial intelligence model provided by an embodiment of the present application;

[0054] Figure 3 It is a structural diagram of an artificial intelligence-based charging parameter optimization system provided by an embodiment of the present application. Detailed Description of the Embodiments

[0055] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. It should be understood that the specific embodiments described herein are only for explaining and illustrating the embodiments of this application, and are not used to limit the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts fall within the scope of protection of this application.

[0056] It should be noted that if there are directional indications (such as up, down, left, right, front, back,...) involved in the embodiments of this application, then such directional indications are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If this specific posture changes, then the directional indications will also change accordingly.

[0057] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of this application, then such descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0058] Figure 1 A flowchart of a method for optimizing charging parameters based on artificial intelligence according to an embodiment of this application is schematically shown. As Figure 1 shown, the embodiments of this application provide a method for optimizing charging parameters based on artificial intelligence, and the method may include the following steps.

[0059] S110. Obtain the battery type, current battery level, and remaining flight mission of the drone;

[0060] S120. Determine the charging mode and initial charging parameters of the drone based on the battery type, current battery level, and remaining flight mission;

[0061] S130. In response to the charging signal of the drone, monitor in real time to obtain the battery charging curve of the drone, where the battery charging curve is constructed based on the real-time charging voltage, real-time charging current, and real-time battery temperature;

[0062] S140. Based on the battery charging curve of the current time period, use an artificial intelligence model to dynamically adjust the charging parameters of the drone battery in the next preset time period at the end of the current time period. The charging parameters include charging current and charging voltage;

[0063] S150. Charge the drone battery according to the charging parameters at the start time of the next preset time period.

[0064] Obtain the battery type, current battery level, and remaining flight mission of the drone. Specifically, the battery type information can be obtained through the battery management system (BMS) on the drone, including but not limited to lithium polymer batteries (LiPo), lithium-ion batteries (Li-ion), lithium iron phosphate batteries (LiFePO4), etc. The battery type information usually contains parameters such as the chemical composition of the battery, nominal voltage, capacity, and maximum discharge rate.

[0065] The current battery level can be obtained through the battery level calculation module in the BMS. This module calculates the actual remaining battery percentage by measuring parameters such as the battery terminal voltage, internal resistance, and temperature, and combining the battery discharge characteristic curve.

[0066] The remaining flight mission information can be obtained from the drone's mission planning system, including mission type (such as emergency rescue, regular mapping, agricultural spraying, etc.), estimated flight time, flight path complexity, etc. In practical applications, this information can be transmitted through the wireless communication interface between the drone and the charging station, such as Wi-Fi, Bluetooth, ZigBee, or a dedicated wireless communication protocol.

[0067] Based on the battery type, current battery level, and remaining flight mission, determine the charging mode and initial charging parameters of the drone. Specifically, first, according to the battery type of the drone, obtain the corresponding battery safety parameter thresholds from the preset battery parameter database, including the maximum charging current, maximum charging voltage, highest safety temperature, etc. Then, based on the current battery level, determine the charging stage of the battery. For example, when the battery level is below 30%, it can be determined as the first charging stage (deep discharge state); when the battery level is between 30% and 80%, it can be determined as the second charging stage (normal charging state); when the battery level is above 80%, it can be determined as the third charging stage (near saturation state).

[0068] Determine the flight urgency based on the remaining flight mission. When the mission type is a preset first mission type (such as emergency rescue), determine the flight urgency as high; when the mission type is not the first mission type, determine the flight urgency as medium or low based on the estimated flight time. Among them, when the estimated flight time is less than the preset time threshold, determine the flight urgency as medium, and when the estimated flight time is not less than the preset time threshold, determine the flight urgency as low. Determine the charging mode and initial charging parameters according to the charging stage of the battery and the flight urgency. For example, when the charging stage is the first charging stage and the flight urgency is high, determine the charging mode as the constant current and constant voltage combined charging mode, and set the initial charging current value to the first preset percentage (such as 90%) of the safe current threshold; when the charging stage is the first charging stage and the flight urgency is medium or low, determine the charging mode as the constant current charging mode, and set the initial charging current value to the second preset percentage (such as 70%) of the safe current threshold, where the second preset percentage is less than the first preset percentage; when the charging stage is the second charging stage, determine the charging mode as the constant voltage charging mode, and set the initial charging voltage value to the third preset percentage (such as 95%) of the safe voltage threshold. In this way, the most suitable charging mode and initial parameters can be selected according to the actual state of the battery and the mission requirements, which can not only ensure the charging efficiency but also take into account the battery safety and life.

[0069] In response to the charging signal of the unmanned aerial vehicle (UAV), the battery charging curve of the UAV is monitored in real time, where the battery charging curve is constructed based on the real-time charging voltage, real-time charging current, and real-time battery temperature. Specifically, when the UAV establishes a physical connection with the charging station and sends a charging request signal, the charging control system of the UAV starts to charge the battery and simultaneously activates the real-time monitoring module. This module collects the real-time charging voltage, real-time charging current, and real-time battery temperature data of the battery at a preset sampling frequency (such as 10 times per second) through high-precision voltage sensors, current sensors, and temperature sensors.

[0070] The battery charging curve is a multi-dimensional time-series data set, including the voltage-time curve, current-time curve, and temperature-time curve. These curves can comprehensively reflect the state changes and response characteristics of the battery during the charging process. For example, the voltage-time curve can reflect the charging acceptance ability and internal resistance change of the battery; the current-time curve can reflect the charging control strategy and the charging efficiency of the battery; the temperature-time curve can reflect the thermal characteristics and safety state of the battery.

[0071] Based on the battery charging curve in the current time period, an artificial intelligence model is used to dynamically adjust the charging parameters of the drone battery in the next preset time period at the end of the current time period. The charging parameters include charging current and charging voltage. Specifically, first, collect the battery charging curve data in the current time period. The battery charging curve data is time-series data including charging voltage, charging current, and battery temperature. Then, preprocess the time-series data, including data standardization, outlier detection and processing, missing value filling, etc., to improve the data quality.

[0072] Input the preprocessed time-series data into the artificial intelligence model for analysis and prediction. The artificial intelligence model is a deep reinforcement learning model, including a convolutional neural network layer for extracting battery charging curve features, a long short-term memory network layer for learning the temporal variation law of the battery charging state, a policy network and a value network for predicting the optimal charging parameters. Specifically, input the preprocessed time-series data into the convolutional neural network layer to extract the local features of the battery charging curve; input the local features into the long short-term memory network layer to learn the temporal variation law of the battery charging state; calculate the battery internal resistance change rate, capacity attenuation rate, and temperature sensitivity index based on the deviation between the battery charging curve and the standard charging curve; calculate the battery health based on the internal resistance change rate, capacity attenuation rate, and temperature sensitivity index using a preset battery health function; input the local features and the temporal variation law into the policy network and the value network to predict the charging efficiency and battery life under different charging parameters, and obtain the optimal charging parameters.

[0073] Dynamically adjust the charging parameters of the next preset time period according to the battery health and flight urgency. When the battery health is lower than the first preset threshold (such as 60%), at the end of the current time period, reduce the charging current of the next preset time period to the first preset proportion (such as 80%) of the optimal charging parameters, and reduce the charging voltage to the second preset proportion (such as 90%) of the optimal charging parameters to protect the battery; when the battery health score is not lower than the first preset threshold and not higher than the second preset threshold (such as 80%), at the end of the current time period, set the charging parameters of the next preset time period to the optimal charging parameters; when the battery health score is higher than the second preset threshold and the flight urgency is high, at the end of the current time period, increase the charging current of the next preset time period to the third preset proportion (such as 110%) of the optimal charging parameters, and not exceed the battery safety parameter threshold to accelerate the charging speed. In this way, the charging parameters can be dynamically adjusted according to the real-time state and health of the battery, which can not only ensure the charging efficiency but also take into account the battery safety and life.

[0074] Charge the drone battery according to the charging parameters at the start of the next preset time period. Specifically, after the end of the current time period, the charging control system will immediately adjust the output current and voltage of the charging circuit according to the charging parameters determined in the previous step to charge the drone battery.

[0075] To ensure a smooth transition of the charging parameters and avoid shocks to the battery caused by sudden changes, a gradual adjustment strategy can be adopted, that is, gradually adjust the charging parameters from the current value to the target value within a short period (such as 100 milliseconds). During the charging process, continuously monitor the real-time status of the battery, including parameters such as voltage, current, and temperature. If abnormal situations occur (such as too high temperature, abnormal voltage fluctuations, etc.), the safety protection mechanism will be immediately triggered, such as reducing the charging power, pausing the charging, or issuing an alarm. At the same time, record the charging data for each time period, including charging parameters, battery response, charging efficiency, etc. On the one hand, it is used for real-time monitoring and adjustment, and on the other hand, it is stored for subsequent data analysis and model optimization. In addition, information such as the estimated charging completion time, the current charging speed, and the charged battery capacity can be calculated and displayed to facilitate the user to understand the charging progress and plan subsequent tasks. In this way, safe and efficient charging of the drone battery can be achieved, which can not only meet the fast charging requirements of emergency tasks but also protect the battery life under normal circumstances, improving the overall usage efficiency and economy of the drone.

[0076] In this embodiment, by obtaining the battery type, current battery level, and remaining flight tasks of the drone, the appropriate charging mode and initial parameters are determined; by real-time monitoring the battery charging curve and using a deep reinforcement learning model to dynamically adjust the charging parameters, intelligent management of the drone battery charging process is achieved. It can adaptively adjust the charging strategy according to the real-time status, health, and task urgency of the battery, and maximize the charging efficiency on the premise of ensuring battery safety and life. Compared with the traditional fixed charging strategy, this method can effectively shorten the charging time, extend the battery life, significantly improve the working efficiency and task continuity of the drone, and is especially suitable for scenarios with high timeliness requirements such as emergency rescue and large-scale mapping. In addition, this method also has good adaptability and scalability, and can adapt to different types and different aging degrees of batteries.

[0077] In one implementation manner of this embodiment, based on the battery type, current battery level, and remaining flight tasks, determine the charging mode and initial charging parameters of the drone, including the following steps:

[0078] S210. According to the battery type of the drone, obtain the corresponding battery safety parameter threshold from the preset battery parameter database;

[0079] S220. According to the current battery level, judge the charging stage of the battery;

[0080] S230. Determine the flight urgency according to the remaining flight mission, where the flight urgency includes high, medium, and low.

[0081] S240. Based on the battery safety parameter threshold, the charging stage, and the flight urgency, determine that the charging mode is any one of the constant current charging mode, the constant voltage charging mode, or the constant current-constant voltage combined charging mode, and determine the initial charging current value or the charging voltage value according to the charging mode.

[0082] Obtain the corresponding battery safety parameter threshold from the preset battery parameter database according to the battery type of the UAV. Specifically, the battery parameter database contains the UAV battery type and its corresponding safety parameter threshold. The battery types mainly include lithium polymer battery (LiPo), lithium-ion battery (Li-ion), lithium iron phosphate battery (LiFePO4), nickel-metal hydride battery (NiMH), etc. For each battery type, the database stores its nominal voltage, maximum charging current rate (C-rate), maximum charging voltage, highest safety temperature, lowest safety temperature, maximum discharge rate and other safety parameter thresholds. For example, for a typical 3S lithium polymer battery, its maximum charging current rate is usually 1C - 2C, the maximum charging voltage is 12.6V (4.2V per cell), the highest safety temperature is 45°C, and the lowest safety temperature is 0°C. When the UAV establishes a connection with the charging station, obtain the battery model identifier or directly obtain the battery type information through the communication interface of the battery management system. For example, a certain UAV uses a 4S lithium polymer battery with a capacity of 5000mAh and a model of "LP4S5000". After the charging system receives this information, it searches for the corresponding record in the battery parameter database to obtain the safety parameter threshold of this model battery: the maximum charging current is 5A (1C), the maximum charging voltage is 16.8V (4.2V per cell), the highest safety temperature is 45°C, and the lowest safety temperature is 5°C. The database can also dynamically adjust the safety parameter threshold according to the production date and usage times of the battery. For example, for a lithium polymer battery that has been used more than 300 times, the maximum charging current rate can be reduced by 20% to extend the battery life.

[0083] Judge the charging stage of the battery according to the current battery level. Specifically, first obtain the current battery level percentage of the battery through the battery management system. The current battery level of the battery is usually measured by the coulomb counting method or the open circuit voltage method. The coulomb counting method calculates the change in battery level by integrating the charge and discharge current of the battery, while the open circuit voltage method estimates the battery level by measuring the open circuit voltage of the battery and combining the voltage-capacity curve of the battery. In practical applications, these two methods are often combined and fused through algorithms such as Kalman filtering to improve the accuracy of battery level estimation.

[0084] After obtaining the current battery level, the charging process is divided into different stages according to a preset battery level threshold. Generally, the charging process of a lithium battery can be divided into three main stages: pre-charging stage, constant current charging stage, and constant voltage charging stage. In this embodiment, according to the current battery level percentage of the battery, the charging process is divided into the following three stages: The first charging stage (deep discharge state): When the battery level is lower than 30%, the battery is in a deep discharge state. At this time, the active substances inside the battery are in a highly unstable state, and a milder charging strategy needs to be adopted to avoid battery damage caused by a large current impact. The second charging stage (conventional charging state): When the battery level is between 30% and 80%, the battery is in a conventional charging state. At this time, the battery has a good charging acceptance ability, and a larger charging current can be adopted to improve the charging efficiency. The third charging stage (near saturation state): When the battery level is higher than 80%, the battery is near saturation. At this time, the charging acceptance ability of the battery begins to decline, and continuing to charge with a large current may cause the battery to overheat and shorten its lifespan. Therefore, the charging current needs to be gradually reduced. For example, the current battery level of a certain drone is 25%. According to the above division criteria, it is determined that the battery is in the first charging stage (deep discharge state). In different charging stages, the internal resistance, temperature sensitivity, and charging efficiency of the battery are different. Therefore, different charging strategies need to be adopted. By accurately judging the charging stage of the battery, it can provide an important basis for the subsequent selection of charging modes and parameter optimization, which can not only ensure charging safety but also maximize charging efficiency and extend the battery lifespan.

[0085] According to the remaining flight tasks, determine the flight urgency level, where the flight urgency level includes high, medium, and low. Specifically, first obtain the detailed information of the remaining flight tasks from the mission planning system of the drone, including the task type, estimated flight time, flight path complexity, task priority, etc. The task type refers to the specific nature of the task executed by the drone, such as emergency rescue, disaster monitoring, conventional mapping, agricultural spraying, recreational aerial photography, etc.

[0086] The estimated flight time refers to the estimated time required to complete the remaining tasks, usually calculated by the mission planning system based on factors such as the length of the flight path, flight speed, and hovering time. The flight path complexity refers to the complexity of the flight path, including factors such as the number of turns, frequency of altitude changes, and obstacle density. The task priority refers to the importance and urgency of the task, usually specified by the user during mission planning. According to the obtained task information, use a preset rule or machine learning model to determine the flight urgency level.

[0087] In this embodiment, the steps to determine the flight emergency level are as follows: First, judge the basic emergency level according to the mission type. The mission types are divided into three categories: The first category of missions includes high-urgency missions such as emergency rescue and disaster monitoring, and the basic emergency level is high; the second category of missions includes medium-urgency missions such as conventional mapping and inspection, and the basic emergency level is medium; the third category of missions includes low-urgency missions such as agricultural spraying and recreational aerial photography, and the basic emergency level is low. Then, adjust the emergency level according to the estimated flight time. If the estimated flight time is less than the preset time threshold T1 (such as 30 minutes), the emergency level is raised by one level (if it is already high level, it remains unchanged); if the estimated flight time is greater than the preset time threshold T2 (such as 2 hours), the emergency level is lowered by one level (if it is already low level, it remains unchanged).

[0088] Further adjust the emergency level according to the mission priority. If the mission priority is the highest level, the emergency level is set to high; if the mission priority is the lowest level, the emergency level is set to low. For example, the remaining flight mission of a certain drone is to conduct search and rescue in the earthquake-stricken area. The mission type is emergency rescue (the first category of missions), the estimated flight time is 45 minutes, and the mission priority is the highest level. According to the above rules, the basic emergency level is high, the estimated flight time is between 30 minutes and 2 hours, the emergency level is not adjusted, the mission priority is the highest level, and the emergency level is set to high. Therefore, the flight emergency level of this drone is finally determined to be high.

[0089] Based on the battery safety parameter threshold, the charging stage, and the flight emergency level, determine the charging mode to be any one of the constant current charging mode, the constant voltage charging mode, or the constant current-constant voltage combined charging mode, and determine the initial charging current value or the charging voltage value according to the charging mode. Specifically, first determine the safety upper limits of the charging current and the charging voltage according to the battery safety parameter threshold. For example, for a 4S lithium polymer battery with a capacity of 5000 mAh, its maximum safe charging current is 5A (1C), and its maximum safe charging voltage is 16.8V. Then, according to the charging stage the battery is in and the flight emergency level, use a preset decision matrix to determine the charging mode and the initial charging parameters.

[0090] The decision matrix is as follows: For the first charging stage (deep discharge state, battery level < 30%): If the flight urgency is high, select the constant current and constant voltage combined charging mode, set the initial charging current to 90% of the maximum safe charging current (i.e., 4.5 A), and set the initial charging voltage to 95% of the maximum safe charging voltage (i.e., 16.0 V); If the flight urgency is medium, select the constant current charging mode, set the initial charging current to 70% of the maximum safe charging current (i.e., 3.5 A); If the flight urgency is low, select the constant current charging mode, set the initial charging current to 50% of the maximum safe charging current (i.e., 2.5 A). For the second charging stage (normal charging state, 30% ≤ battery level ≤ 80%): If the flight urgency is high, select the constant current charging mode, set the initial charging current to 100% of the maximum safe charging current (i.e., 5 A); If the flight urgency is medium, select the constant current charging mode, set the initial charging current to 80% of the maximum safe charging current (i.e., 4 A); If the flight urgency is low, select the constant current charging mode, set the initial charging current to 60% of the maximum safe charging current (i.e., 3 A). For the third charging stage (near saturation state, battery level > 80%): If the flight urgency is high, select the constant voltage charging mode, set the initial charging voltage to 100% of the maximum safe charging voltage (i.e., 16.8 V); If the flight urgency is medium or low, select the constant voltage charging mode, set the initial charging voltage to 98% of the maximum safe charging voltage (i.e., 16.5 V).

[0091] For example, a certain UAV uses a 4S lithium polymer battery with a capacity of 5000 mAh, and the current battery level is 25% (the first charging stage), and the flight urgency is high. According to the above decision matrix, select the constant current and constant voltage combined charging mode, set the initial charging current to 4.5 A (90% of the maximum safe charging current of 5 A), and set the initial charging voltage to 16.0 V (95% of the maximum safe charging voltage of 16.8 V). In practical applications, the temperature factor of the battery also needs to be considered. If the battery temperature is close to the maximum safe temperature (such as 40 °C), the charging current should be appropriately reduced to prevent the battery from overheating; If the battery temperature is close to the minimum safe temperature (such as 5 °C), the charging current should also be reduced because the charging acceptance ability of the battery decreases at low temperatures. In this way, the most suitable charging mode and initial parameters can be selected according to the actual state of the battery and the mission requirements, which can not only ensure the charging efficiency but also take into account the battery safety and life.

[0092] The charging parameter optimization method based on artificial intelligence provided by this embodiment realizes the intelligent management of the charging process of the UAV battery by obtaining the battery safety parameter thresholds from the preset battery parameter database, judging the charging stage of the battery according to the current battery level, determining the flight urgency according to the remaining flight tasks, and determining the charging mode and initial charging parameters based on this information. This method can adaptively select the most suitable charging strategy according to the actual state of the battery and the task requirements, and maximize the charging efficiency on the premise of ensuring the safety and life of the battery. Compared with the traditional fixed charging strategy, this method can better balance the relationship between charging speed, battery safety and battery life, and is particularly suitable for scenarios with high timeliness requirements such as emergency rescue and large-scale mapping.

[0093] In one implementation manner of this embodiment, the remaining flight tasks include the task type and the estimated flight time. According to the remaining flight tasks, determining the flight urgency includes the following steps:

[0094] S310. When the task type is the preset first task type, determine that the flight urgency is high;

[0095] S320. When the task type is not the first task type, determine the flight urgency as medium or low based on the estimated flight time;

[0096] Among them, when the estimated flight time is less than the preset time threshold, determine that the flight urgency is medium; when the estimated flight time is not less than the preset time threshold, determine that the flight urgency is low.

[0097] When the task type is the preset first task type, determine that the flight urgency is high. Specifically, the preset first task type refers to tasks with high timeliness and high importance. Such tasks often involve personal safety or major property safety and require the UAV to complete them in the shortest time. Typical first task types include: emergency rescue tasks, such as search and rescue operations after natural disasters such as earthquakes, floods, and fires, which are directly related to the lives of people in the disaster area; safety monitoring tasks, such as forest fire prevention monitoring, border patrol, and important facility safety monitoring; medical assistance tasks, such as the transportation of emergency drugs, blood, and medical equipment, which usually have strict time window requirements; emergency incident response tasks, such as traffic accident scene investigation, hazardous chemical leakage monitoring, etc., which require rapid response to prevent the situation from expanding.

[0098] When the task management system of the drone receives a task instruction, it first parses the task type identifier. Once it is confirmed that the current task belongs to the preset first task type, the flight emergency level is set to high. Setting the flight emergency level to high means that in the subsequent charging strategy formulation, the charging speed will be given priority to ensure that the drone can complete charging and execute the task in the shortest time. This mechanism ensures that in the face of emergencies, the drone can respond quickly and maximize its role in emergency scenarios.

[0099] In the case where the task type is not the first task type, the flight emergency level is determined to be medium or low based on the estimated flight time. Specifically, when the task of the drone does not belong to the preset first task type, other characteristics of the task are further analyzed, and the flight emergency level can be determined through the estimated flight time. The estimated flight time refers to the estimated time required to complete the remaining tasks, which is usually calculated based on factors such as flight path length, flight speed, hovering time, and task operation time.

[0100] The calculation formula for the estimated flight time can be expressed as: Estimated flight time = Path flight time + Hovering operation time + Task execution time + Safety margin time. Among them, the path flight time refers to the time required for the drone to fly along the planned path, and the calculation method is the path length divided by the average flight speed; the hovering operation time refers to the time for the drone to hover and execute operations at a specific position; the task execution time refers to the time required to complete specific task operations (such as taking pictures, measuring, dropping items, etc.); the safety margin time is the time reserved to cope with uncertain factors such as possible wind speed changes and path adjustments. For example, a drone needs to perform a farmland mapping task. The planned path length is 5 kilometers, the average flight speed is 10 meters per second, it needs to hover and take pictures at 10 position points, the hovering time for each time is 30 seconds, the picture-taking operation time is 5 seconds, and the safety margin time is set to 10% of the total time. Then the estimated flight time is calculated as follows: Path flight time = 5000 meters ÷ 10 meters per second = 500 seconds; Hovering operation time = 10 times × 30 seconds per time = 300 seconds; Task execution time = 10 times × 5 seconds per time = 50 seconds; Subtotal time = 500 seconds + 300 seconds + 50 seconds = 850 seconds; Safety margin time = 850 seconds × 10% = 85 seconds; Estimated flight time = 850 seconds + 85 seconds = 935 seconds ≈ 15.6 minutes.

[0101] The non-first task type usually includes tasks such as regular mapping, agricultural monitoring, environmental monitoring, and recreational aerial photography, which usually do not have extremely high timeliness requirements.

[0102] In the case where the estimated flight time is less than the preset time threshold, the flight urgency is determined to be medium. In the case where the estimated flight time is not less than the preset time threshold, the flight urgency is determined to be low. Usually, the preset time threshold can be set to 30% - 50% of the average endurance time of the drone in a fully charged state. For example, if the average endurance time of a certain model of drone is 60 minutes, the preset time threshold can be set to 20 - 30 minutes. When the estimated flight time is less than this threshold, it means that the task is relatively short and can be completed in a relatively short time, but still requires a certain amount of time guarantee. Therefore, the flight urgency is set to medium. For example, a drone needs to perform an urban building inspection task, and the calculated estimated flight time is 15 minutes, while the preset time threshold is 20 minutes. Since 15 minutes is less than 20 minutes, the flight urgency is set to medium. A flight urgency of medium means that in the charging strategy, both the charging speed and the protection of the battery life need to be considered. Usually, a medium charging current is adopted to balance between the charging speed and the battery protection.

[0103] When the estimated flight time is not less than the preset time threshold, it means that the task is relatively long and there is sufficient time for preparation, and there is no need for particularly urgent charging. For example, a drone needs to perform a large - area farmland monitoring task, and the calculated estimated flight time is 35 minutes, while the preset time threshold is 20 minutes. Since 35 minutes is greater than 20 minutes, the flight urgency is set to low. A flight urgency of low means that in the charging strategy, the protection of the battery life will be given priority. Usually, a lower charging current is adopted to extend the charging time to maximize the protection of the battery and extend its service life.

[0104] In addition, the preset time threshold is not a fixed parameter and can be dynamically adjusted according to factors such as the actual application scenario, seasonal changes, and battery aging degree. For example, in cold weather, since low temperature will affect the battery performance, the preset time threshold can be appropriately reduced; for a battery with a longer usage time, due to capacity attenuation, the preset time threshold can also be adjusted accordingly. Through this dynamic adjustment mechanism, it can better adapt to various complex environments and task requirements.

[0105] The method for determining flight urgency based on task characteristics provided in this embodiment realizes the intelligent judgment of flight urgency by analyzing the task type and estimated flight time of the drone. This method first identifies whether the task belongs to the preset first task type with high urgency; for non - first task types, the urgency is further determined to be medium or low based on the comparison between the estimated flight time and the preset threshold. Compared with the traditional fixed charging strategy, this method can adaptively adjust the charging parameters according to the actual needs of the task, and on the premise of ensuring battery safety and life, maximize the timeliness requirements for task execution. It is especially suitable for the drone application environment with multiple tasks and multiple scenarios.

[0106] In one implementation of this embodiment, the battery safety parameter threshold includes a safety current threshold and a safety voltage threshold. Based on the battery safety parameter threshold, the charging stage, and the flight emergency level, a charging mode is determined to be any one of a constant current charging mode, a constant voltage charging mode, or a constant current-constant voltage combined charging mode, and an initial charging current value or a charging voltage value is determined according to the charging mode, including the following steps:

[0107] S410. When the charging stage is the first charging stage and the flight emergency level is high, determine the charging mode as the constant current-constant voltage combined charging mode, and set the initial charging current value to the first preset percentage of the safety current threshold;

[0108] S420. When the charging stage is the first charging stage and the flight emergency level is medium or low, determine the charging mode as the constant current charging mode, and set the initial charging current value to the second preset percentage of the safety current threshold, where the second preset percentage is less than the first preset percentage;

[0109] S430. When the charging stage is the second charging stage, determine the charging mode as the constant voltage charging mode, and set the initial charging voltage value to the third preset percentage of the safety voltage threshold.

[0110] When the charging stage is the first charging stage and the flight emergency level is high, determine the charging mode as the constant current-constant voltage combined charging mode, and set the initial charging current value to the first preset percentage of the safety current threshold. Specifically, the first charging stage usually refers to the initial stage of battery charging. At this time, the state of charge (SOC) of the battery is relatively low, the internal electrochemical reactions of the battery are active, and it can accept a relatively large charging current. During the charging process of a lithium-ion battery, the first charging stage generally corresponds to the stage where the battery SOC ranges from 0% to about 70% - 80%. If the current is in the first charging stage and the flight emergency level is high, it means that the battery needs to obtain sufficient power in a short time to perform an emergency task. In this case, the constant current-constant voltage combined charging mode is adopted.

[0111] The constant current-constant voltage combined charging mode is a charging method that first uses a constant current and then a constant voltage. In the initial stage of charging, a constant and relatively large current is used for charging. When the battery voltage reaches a specific value, it switches to constant voltage charging and the current gradually decreases. This charging mode can achieve a relatively fast charging speed while ensuring battery safety. The safety current threshold refers to the maximum charging current that the battery can safely accept. Exceeding this threshold may cause safety problems such as battery overheating, increased internal impedance, and electrolyte decomposition. The safety current threshold is usually provided by the battery manufacturer and is related to factors such as the chemical composition, structural design, and capacity of the battery.

[0112] The first preset percentage refers to the ratio of the initial charging current value to the safe current threshold in an emergency. Considering the timeliness requirements of emergency tasks, the first preset percentage is usually set at a relatively high level, such as 80% - 95%. This can not only ensure a relatively fast charging speed but also leave a certain margin for battery safety. Taking the above 5000mAh battery as an example, if the safe current threshold is 10000mA and the first preset percentage is set at 90%, then the initial charging current value is 10000mA × 90% = 9000mA. In practical applications, the setting of the initial charging current value also needs to consider the influence of environmental temperature. For example, in a low-temperature environment (such as below 0°C), the charging acceptance ability of lithium-ion batteries will be significantly reduced. At this time, even if the emergency level is high, it is necessary to appropriately reduce the first preset percentage to prevent safety problems such as lithium metal precipitation. Similarly, in a high-temperature environment (such as above 40°C), it is also necessary to reduce the charging current to prevent the battery from overheating. Therefore, this embodiment also introduces a temperature compensation mechanism to dynamically adjust the first preset percentage according to the environmental temperature. For example, in a 25°C environment, the first preset percentage is 90%; in a 0°C environment, it is reduced to 70%; in a 40°C environment, it is reduced to 80%. In this way, it is possible to ensure the safe charging of the battery under different environmental conditions while maximizing the timeliness requirements of emergency tasks.

[0113] When the charging stage is the first charging stage and the flight emergency level is medium or low, determine the charging mode as the constant current charging mode, and set the initial charging current value to the second preset percentage of the safe current threshold, where the second preset percentage is less than the first preset percentage. Specifically, if the current is in the first charging stage and the flight emergency level is medium or low, it means that the battery does not need to be charged particularly quickly, and a more gentle charging strategy can be adopted to extend the battery life. In this case, the constant current charging mode is adopted. The constant current charging mode means that a constant current is maintained during the charging process to charge the battery until the battery voltage reaches a specific value. Compared with the constant current and constant voltage combined charging mode, the pure constant current charging mode pays more attention to battery life protection in the first charging stage, and the charging speed is relatively slow. The second preset percentage refers to the ratio of the initial charging current value to the safe current threshold in a non-emergency situation. Considering the need for battery life protection, the second preset percentage is usually set at a relatively low level, such as 50% - 70%, to significantly reduce the heat generation during charging, reduce the internal stress of the battery, and thus extend the cycle life of the battery.

[0114] In practical applications, the specific value of the second preset percentage can also be further divided according to different levels of flight urgency. For example, when the flight urgency is medium, the second preset percentage can be set to 65% - 70%; when the flight urgency is low, the second preset percentage can be set to 50% - 60%, so as to more precisely balance the relationship between the charging speed and the battery life. In addition, the setting of the second preset percentage also needs to consider the state of health (SOH) of the battery. As the number of times the battery is used increases, its internal impedance will gradually increase, and its charging acceptance ability will gradually decrease. Therefore, this embodiment also introduces a battery state of health assessment mechanism to dynamically adjust the second preset percentage according to the actual state of health of the battery. For example, for a new battery (SOH > 90%), the standard second preset percentage can be used; for an aging battery (SOH < 70%), the second preset percentage can be reduced by 10% - 20% to relieve the charging pressure on the battery. In this way, the charging strategy can be adaptively adjusted according to the actual condition of the battery and the task requirements, and the battery life can be maximally extended on the premise of ensuring battery safety. At the same time, since a lower charging current is adopted, the heat generated during the charging process will also be reduced, which is particularly important for a charging system such as a drone that has high requirements for thermal management.

[0115] When the charging stage is the second charging stage, determine that the charging mode is the constant voltage charging mode, and set the initial charging voltage value to the third preset percentage of the safety voltage threshold. Specifically, the second charging stage usually refers to the later stage of battery charging, at this time the state of charge (SOC) of the battery is already relatively high, generally above 70% - 80%. During the charging process of a lithium-ion battery, when the battery voltage reaches a specific value, continuing to use constant current charging may cause the battery to overvoltage, increasing the safety risk. Therefore, adopting the constant voltage charging mode in the second charging stage is a necessary measure to protect the battery. The constant voltage charging mode means that a constant voltage is maintained during the charging process to charge the battery. As the battery charge increases, the charging current will gradually decrease until it reaches the preset cut-off current value. The safety voltage threshold refers to the maximum charging voltage that the battery can safely withstand. Exceeding this threshold may cause safety problems such as battery overcharging, electrolyte decomposition, and gas generation. The safety voltage threshold is usually provided by the battery manufacturer and is closely related to the chemical composition of the battery. For example, for a standard lithium-ion battery (graphite negative electrode, lithium cobalt oxide positive electrode), the safety voltage threshold of a single cell is usually 4.2V; for a lithium iron phosphate battery, the safety voltage threshold is usually 3.65V.

[0116] The third preset percentage refers to the ratio of the initial charging voltage value to the safety voltage threshold. Considering the balance between battery safety and life, the third preset percentage is usually set between 95% - 100%.

[0117] In practical applications, the setting of the third preset percentage needs to consider various factors. First is the battery life factor. Reducing the charging cut-off voltage can significantly extend the cycle life of lithium-ion batteries. For example, reducing the charging cut-off voltage from 4.2V to 4.1V may increase the cycle life of the battery by 30% - 50%, but at the same time, about 10% of the capacity will be lost. Therefore, in non-emergency situations, the third preset percentage can be reduced to extend the battery life. Second is the temperature factor. In a high-temperature environment, lithium-ion batteries are more likely to undergo side reactions, increasing the safety risk. Therefore, in a high-temperature environment (such as above 35°C), the third preset percentage should be reduced, for example, from 98% to 95%. In addition, the degree of battery aging also needs to be considered. As the battery is used, its internal structure will gradually change, increasing its sensitivity to high voltages. Therefore, for batteries with a longer usage time, the third preset percentage should also be appropriately reduced. By precisely controlling the voltage value in the constant-voltage charging stage, it is possible to balance the charging speed and battery life according to actual needs while ensuring battery safety.

[0118] In another embodiment, when the real-time battery temperature exceeds the preset temperature threshold, the charging current value is reduced to a safe level. Among them, the safe level of the current value is determined based on the current state of the battery. Specifically, when the real-time battery temperature exceeds the preset temperature threshold, the safety protection mechanism is immediately triggered to reduce the charging current value to a safe level to prevent safety hazards and performance damage caused by battery overheating.

[0119] During specific implementation, the battery temperature monitoring system uses high-precision thermistors or thermocouple sensors, which are arranged at multiple key positions of the battery pack, such as near the positive and negative terminal posts, the center area of the battery cells, and the hot spot area, etc. The temperature measurement accuracy of these sensors can reach ±0.5°C, and the sampling frequency is set to 1 - 2 times per second to ensure that small temperature changes can be captured in a timely manner. The collected temperature data is filtered and then compared with the preset temperature threshold.

[0120] For lithium polymer batteries, three levels of temperature thresholds can be set: the first-level warning threshold is 40°C, the second-level warning threshold is 43°C, and the third-level warning threshold is 45°C. When the battery temperature exceeds the first-level warning threshold but is lower than the second-level warning threshold, a warning is issued but the charging current does not need to be immediately reduced; when the temperature exceeds the second-level warning threshold but is lower than the third-level warning threshold, the charging current is moderately reduced; when the temperature exceeds the third-level warning threshold, the charging current is significantly reduced or charging is paused.

[0121] In practical applications, the temperature threshold can be dynamically adjusted according to the ambient temperature. For example, in a high-temperature environment (such as in summer above 35°C), each level of temperature threshold can be appropriately reduced by 2 - 3°C; in a low-temperature environment (such as in winter below 0°C), each level of temperature threshold can be appropriately increased by 1 - 2°C.

[0122] When the real-time battery temperature exceeds the preset temperature threshold, the system needs to reduce the charging current to a safe level. The safe level is not fixed but is dynamically determined based on the current state of the battery to achieve the best balance between safety and efficiency. The current state of the battery includes, but is not limited to, key parameters such as temperature, state of charge (SOC), state of health (SOH), internal resistance, and voltage.

[0123] The determination of the safe charging current adopts a multi-parameter comprehensive evaluation model, which takes into account factors such as the thermal characteristics, electrochemical characteristics, and safety boundaries of the battery.

[0124] After determining the safe charging current, the output of the charger is adjusted in real time, and the battery response is continuously monitored through a closed-loop feedback control mechanism to ensure the safety and stability of the charging process.

[0125] During the adjustment process, the key parameters of the battery are continuously monitored, especially the temperature change trend. If it is found that the temperature continues to rise or the rising rate does not slow down, the charging current is further reduced; if the temperature begins to drop or remains stable, the current value is maintained. This fine-tuning mechanism based on real-time feedback can keep the charging current always at the best balance point between safety and efficiency. In actual implementation, it can adapt to batteries of different models, different aging degrees, as well as different environmental conditions and usage scenarios. It can continuously learn and optimize the charging strategy to make the charging process more and more in line with the actual needs of specific batteries.

[0126] In this embodiment, different charging strategies are adopted according to different charging stages and the urgency of tasks: in the first charging stage and when the urgency is high, a constant current and constant voltage combined charging mode is adopted and a higher charging current is used to quickly increase the battery power; in the first charging stage and when the urgency is medium or low, a constant current charging mode is adopted and a lower charging current is used to balance the charging speed and battery life; in the second charging stage, regardless of the urgency, a constant voltage charging mode is adopted and the charging voltage is set to the third preset percentage of the safe voltage threshold to ensure battery safety. Thus, the charging parameters can be dynamically adjusted according to the actual situation, which not only meets the timeliness requirements of emergency tasks but also maximally protects battery safety and extends battery life. Compared with traditional fixed charging strategies, it can effectively shorten the charging time in case of emergencies and effectively extend the battery cycle life in non-emergency situations, significantly improving the overall performance and reliability of the UAV system.

[0127] In one implementation manner of this embodiment, based on the battery charging curve of the current time period, an artificial intelligence model is adopted to dynamically adjust the charging parameters of the UAV battery in the next preset time period at the end of the current time period, including the following steps:

[0128] S510. Analyze the changing trend of the battery charging curve through an artificial intelligence model to obtain an analysis result, where the analysis result includes the battery health and the optimal charging parameters;

[0129] S520. Based on the battery health and the optimal charging parameters, dynamically adjust the charging parameters of the drone battery in the next preset time period at the end of the current time period.

[0130] Analyze the changing trend of the battery charging curve through an artificial intelligence model to obtain an analysis result, where the analysis result includes the battery health and the optimal charging parameters. Specifically, the battery charging curve refers to a graph showing the variation of parameters such as battery voltage, current, and temperature over time during the charging process. By analyzing the curve, the health status of the battery and the most suitable charging strategy can be deeply understood.

[0131] In practical applications, it is first necessary to collect real-time data during the battery charging process, including but not limited to key parameters such as battery voltage, charging current, battery surface temperature, ambient temperature, and charging duration. After the data collection is completed, these time-series data are input into a pre-trained artificial intelligence model for analysis. The artificial intelligence model can be implemented using various algorithms, such as the long short-term memory network (LSTM), which is suitable for processing time-series data and capturing long-term dependencies therein.

[0132] Taking the LSTM model as an example, its network structure includes an input layer, multiple LSTM layers, and an output layer. The input layer receives the normalized time-series data of the battery parameters; the LSTM layers are responsible for extracting time-series features and learning the internal laws of the battery parameter changes; the output layer generates the battery health assessment and the optimal charging parameter suggestions. The training process of the model requires a large amount of historical charging data and the corresponding battery health status labels, and the model parameters are optimized through a supervised learning method. For example, for a lithium polymer battery with a capacity of 5000 mAh, the charge and discharge data during its process from a new battery to aging can be collected, and its actual capacity measured through experiments can be used as the health status label.

[0133] After the model training is completed, in actual applications, when the battery is being charged, the model will analyze the charging curve data in real time within the current time period (such as 15 minutes or 30 minutes). By analyzing features such as the voltage rise rate, current decline curve, and temperature change trend, the model can evaluate the current health status of the battery. For example, in the constant current charging stage of a healthy new battery, the voltage rise rate is relatively gentle and stable; while for an aging battery, the voltage rise rate is usually faster and unstable. At the same time, the model will also analyze the change trend of the battery's internal resistance. The internal resistance of a healthy battery is low and stable, while the internal resistance of an aging battery is high and may fluctuate significantly during the charging process. Based on these analyses, the model outputs the battery health assessment result, usually expressed as a percentage. For example, 90% indicates that the battery is in good condition, and 60% indicates that the battery has aged significantly. In addition to the battery health, the model will also predict the optimal charging parameters according to the current charging curve features, including the charging current for the next time period, the upper limit of the charging voltage, the charging cut-off condition, etc.

[0134] Based on the battery health and the optimal charging parameters, the charging parameters of the drone battery for the next preset time period are dynamically adjusted at the end of the current time period. Specifically, after obtaining the battery health and the optimal charging parameters analyzed by the artificial intelligence model, the charging strategy for the next preset time period can be dynamically adjusted according to the analysis results at the end of the current time period. This dynamic adjustment mechanism can make the charging process more adaptable to the real-time state changes of the battery.

[0135] According to the battery health assessment result, adjust the overall conservatism of the charging strategy. For batteries with a health assessment of over 90%, a relatively aggressive charging strategy can be adopted to improve the charging efficiency; while for aging batteries with a health below 70%, a more conservative charging strategy is required to extend the remaining service life. The specific adjustment methods include: when the battery health is in the range of 90% - 100%, the charging current can be maintained at a relatively high level, such as 0.8C - 1.0C (C is a multiple of the battery's rated capacity. For example, for a 5000mAh battery, the 1C charging current is 5000mA); when the battery health is in the range of 70% - 90%, the charging current is appropriately reduced to 0.5C - 0.8C; when the battery health is below 70%, the charging current is further reduced to 0.3C - 0.5C, and the charging cut-off voltage may be reduced, such as from 4.2V to 4.1V or 4.0V.

[0136] Based on the best charging parameter suggestions given by the artificial intelligence model, the specific charging parameters for the next preset time period are finely adjusted. These parameters include but are not limited to: the magnitude of the charging current, the upper limit of the charging voltage, the charging cut-off conditions (such as the cut-off current), the charging mode (constant current, constant voltage, or a combination of constant current and constant voltage), etc. For example, if it is analyzed that the current battery temperature is rising too fast, it is recommended to reduce the charging current in the next time period to prevent the battery from overheating; if the internal resistance of the battery increases abnormally, the constant current charging stage is ended in advance and the constant voltage charging stage is entered.

[0137] In practical applications, this kind of dynamic adjustment is usually carried out in units of preset time periods, and the length of the time period can be flexibly set according to the application scenario and computing resources. Common settings are 5 minutes, 15 minutes, or 30 minutes. Taking 15 minutes as an example, at the end of each 15-minute time period, based on the charging data collected during this time period, the battery health and the best charging parameters are analyzed through the artificial intelligence model, and then the charging strategy for the next 15-minute time period is adjusted accordingly. This segmented dynamic adjustment strategy can not only respond to the changes in the battery state in a timely manner but also avoid system instability caused by overly frequent parameter adjustments. Through the dynamic adjustment mechanism based on artificial intelligence analysis, it is possible to provide personalized charging strategies according to the real-time state and characteristics of the battery, protecting the battery to the greatest extent while ensuring the charging efficiency and prolonging the battery life.

[0138] This embodiment uses an artificial intelligence model to deeply analyze the battery charging curve, extract battery health information and the best charging parameter suggestions from it; then based on the analysis results, the charging strategy for the next time period is dynamically adjusted at the end of each preset time period to achieve real-time optimization of the charging process. This data-driven adaptive charging method has significant advantages compared with traditional fixed charging strategies: it can provide personalized charging solutions according to the actual state of the battery, ensuring both the charging efficiency and protecting the battery to the greatest extent; it can timely detect abnormal battery states and make corresponding adjustments, improving charging safety; by optimizing the charging parameters, it effectively prolongs the service life of the battery and reduces the maintenance cost of the drone. By adopting this method, the cycle life of the battery is effectively improved, the charging efficiency is increased, and at the same time, the risks of battery overheating and overcharging are significantly reduced, providing a strong guarantee for the long-term and highly reliable operation of the drone system.

[0139] In one implementation of this embodiment, the artificial intelligence model is a deep reinforcement learning model, including a convolutional neural network layer for extracting battery charging curve features, a long short-term memory network layer for learning the temporal change law of the battery charging state, a policy network and a value network for predicting the optimal charging parameters.

[0140] Refer to Figure 2, the artificial intelligence model is a deep reinforcement learning model, in which a convolutional neural network layer is first used to extract the characteristics of the battery charging curve. Specifically, the convolutional neural network (CNN) layer, as the front-end processing unit of the deep reinforcement learning model, is used to extract valuable feature information from the original battery charging curve data. The battery charging curve usually contains multi-dimensional time-series data such as voltage, current, and temperature, which are recorded in a high-frequency sampling manner to form a large number of data points. Directly inputting these raw data into the neural network will result in too high computational complexity and is prone to introducing noise interference. Therefore, feature extraction and dimensionality reduction processing need to be carried out through the CNN layer. In actual implementation, the input of the CNN layer is multi-channel time-series data within a fixed time window. For example, voltage, current, and temperature data sampled at 1-second intervals within a 10-minute window form a two-dimensional data matrix with a shape of [600, 3]. The CNN layer usually consists of multiple convolutional blocks, and each convolutional block contains a convolutional operation, batch normalization, and an activation function. Taking a typical CNN structure as an example, the first layer of convolution uses 32 convolutional kernels with a size of 5×1 and a stride of 1 to perform a convolutional operation on the input data to capture the parameter change characteristics within a local time period; then, a ReLU activation function is used to introduce a non-linear transformation; then, downsampling is performed through a max-pooling layer (pooling window size of 2×1 and stride of 2) to reduce the data dimension and retain significant features. The second layer of convolution uses 64 convolutional kernels with a size of 3×1 to further extract higher-level features; the ReLU activation and max-pooling operations are also performed. Finally, the third layer of convolution uses 128 convolutional kernels with a size of 3×1 to extract the highest-level abstract features. This multi-layer convolutional structure can effectively capture different-scale features in the battery charging curve, such as short-term fluctuations, medium-term trends, and long-term change patterns. For example, for the constant-current charging stage during the lithium battery charging process, the CNN can identify the slope change characteristics of the voltage rise curve; for the constant-voltage charging stage, the CNN can capture the exponential decay characteristics of the current drop curve. In addition, the CNN can also extract key features such as the temperature rise rate and temperature fluctuation from the temperature curve to evaluate the battery health status and safety. In this way, the CNN layer converts the original high-dimensional time-series data into a low-dimensional but information-rich feature vector, providing a more refined and effective input for the subsequent long short-term memory network layer, greatly improving the learning efficiency and prediction accuracy of the model.

[0141] The long short-term memory network layer is used to learn the temporal variation law of the battery charging state. Specifically, the long short-term memory (LSTM) layer receives the sequence of feature vectors from the convolutional neural network layer and is used to process the temporal dependence relationship during the battery charging process.

[0142] Battery charging is a typical sequential process. The battery state at the current moment is not only related to the current charging parameters but also closely related to the historical charging process. With its special gating mechanism (including input gate, forget gate, and output gate), the LSTM network can effectively learn long-term dependencies and capture the inherent laws of the battery state evolution over time. In actual implementation, the LSTM layer usually consists of multiple layers of bidirectional LSTM units to enhance the ability to capture sequential information. Taking a two-layer bidirectional LSTM structure as an example, the first layer of LSTM contains 128 hidden units and receives the feature sequence output by the CNN layer. Each LSTM unit internally contains three gating units: the input gate determines how much new information is stored in the cell state at the current moment; the forget gate determines how much of the cell state from the previous moment is retained; the output gate controls how much of the current cell state is output. This gating mechanism enables the LSTM to selectively remember or forget information, making it particularly suitable for processing long time series data such as battery charging. The design of the bidirectional LSTM allows the network to consider both past and future information simultaneously, further improving the ability to understand sequential patterns. The second layer of LSTM contains 64 hidden units to further extract higher-level sequential features. The stacking of the two layers of LSTM enables the network to learn more complex sequential patterns, such as the state transition laws of the battery at different charging stages (pre-charging, constant current charging, constant voltage charging). During the battery charging process, the LSTM layer can learn various important sequential laws. For example, for a healthy new battery, during the constant current charging stage, the voltage rise rate usually remains relatively stable; while for an aged battery, the voltage rise rate may gradually increase with the charging time. The LSTM can capture this trend that changes over time and use it as an important basis for evaluating the battery health state. Similarly, the LSTM can also learn the relationship between temperature changes and the charging state, such as the correlation between the temperature rise rate and the battery internal resistance during high-current charging. By learning these sequential change laws, the LSTM layer can accurately evaluate the current state of the battery and provide rich state representations for the subsequent policy network and value network, thus supporting more accurate prediction of charging parameters.

[0143] The policy network is used to predict the optimal charging parameters. Specifically, the policy network receives the state representation from the LSTM layer and outputs the optimal charging parameter distribution for the current battery state. In the reinforcement learning framework, the policy network is responsible for making charging decisions based on the observed battery state. The policy network adopts a probabilistic policy, that is, it outputs a probability distribution for each possible combination of charging parameters, rather than directly outputting deterministic parameter values. In actual implementation, the policy network is usually composed of multiple layers of fully connected neural networks. Taking a typical policy network structure as an example, the first fully connected layer contains 256 neurons, receives the state representation output by the LSTM layer, and introduces a non-linear transformation through the ReLU activation function; the second fully connected layer contains 128 neurons, also using the ReLU activation function; the number of neurons in the last fully connected layer depends on the dimension of the action space. For example, if two parameters, the charging current and the upper limit of the charging voltage, need to be controlled simultaneously, and each parameter is discretized into 10 possible values, then the output layer contains 20 neurons (10 + 10). The output layer uses the Softmax activation function to convert the original output of the neural network into a probability distribution. During the training process, the policy network continuously optimizes its parameters through the policy gradient algorithm to maximize the expected cumulative reward. The design of the reward function is the core of reinforcement learning. For the battery charging task, the reward function usually comprehensively considers multiple objectives: charging speed (the shorter the charging time, the higher the reward), charging efficiency (the higher the energy conversion efficiency, the higher the reward), and battery health protection (temperature control within the safe range, avoiding overcharging and over-discharging, etc.).

[0144] In practical applications, the policy network can dynamically adjust the charging parameters according to the real-time state of the battery. For example, for a 5000mAh lithium polymer battery, when it is detected that the battery temperature rises too fast, the policy network outputs a decision to reduce the charging current, such as from 1.0C (5000mA) to 0.5C (2500mA); when it is detected that the battery enters the constant voltage charging stage, the policy network will gradually reduce the charging current and precisely control the cut-off current to achieve the best charging effect. In this way, the policy network can achieve intelligent control of the charging process and find the best balance among charging speed, efficiency, and battery protection.

[0145] The value network is used to evaluate the long-term value of the current battery state and charging strategy. Specifically, the value network receives the same state representation as the policy network and outputs an estimate of the value of the current state, that is, the expected cumulative reward that can be obtained by acting according to the current policy starting from the current state. The value network provides a more accurate gradient estimate for the policy network, reduces the variance of policy updates, and improves learning efficiency and stability.

[0146] In practical implementation, the value network usually shares the first few layers of network structure with the policy network to improve computational efficiency. The independent part of the value network is also composed of multiple layers of fully connected neural networks. Taking a typical value network structure as an example, the first fully connected layer contains 128 neurons, receives the output of the shared layer, and passes through the ReLU activation function; the second fully connected layer contains 64 neurons and also uses the ReLU activation function; the last fully connected layer has only 1 neuron, directly outputs the state value estimation, and does not use an activation function. During the training process, the value network continuously optimizes its parameters through the temporal difference (TD) learning method to make the predicted state value approach the true expected return.

[0147] Specifically, the loss function of the value network is usually defined as the mean squared error between the predicted value and the target value. The introduction of the value network enables the deep reinforcement learning model to be applicable to solving complex control problems with continuous states and continuous actions such as battery charging. In practical applications, the value network can evaluate the long-term effects of different charging strategies and help the policy network make more informed decisions. For example, for two possible charging strategies: Strategy A uses a higher charging current (1.0C) for fast charging, and Strategy B uses a lower charging current (0.5C) for slow charging. On the surface, Strategy A can increase the battery charge faster and obtain a higher immediate reward; but by evaluating the long-term cumulative reward, the value network may find that although Strategy B has a slower charging speed, it can better protect the battery health, reduce capacity decay, and obtain a higher overall return in the long run. In this way, the value network provides a more comprehensive and long-term decision-making basis for the policy network, enabling the charging control strategy to consider both short-term benefits and long-term value.

[0148] This embodiment realizes the intelligent control of the UAV battery charging process through the organic combination of a convolutional neural network layer, a long short-term memory network layer, a policy network, and a value network. The model first uses the convolutional neural network layer to extract key features from multi-dimensional charging curve data, then learns the temporal variation law of the battery state through the long short-term memory network layer, and finally, based on the policy network and the value network, realizes the dynamic optimization decision of the charging parameters. Compared with traditional charging control methods, this deep reinforcement learning model can automatically learn complex battery charging patterns without manual design of control rules; can provide personalized charging strategies according to the individual differences and aging degree of the battery; can find the best balance point among charging speed, efficiency, and battery life, effectively improving the cycle life of the battery, shortening the charging time, and at the same time maintaining a low temperature rise and capacity decay rate, providing a solid guarantee for the long-term and highly reliable operation of the UAV in various complex environments.

[0149] In one implementation of this embodiment, the change trend of the battery charging curve is analyzed by an artificial intelligence model to obtain an analysis result, where the analysis result includes the battery health and the optimal charging parameters. The steps are as follows:

[0150] S710. Collect the battery charging curve data within the current time period. The battery charging curve data is time-series data including the charging voltage, charging current, and battery temperature.

[0151] S720. Preprocess the time-series data.

[0152] S730. Input the preprocessed time-series data into a convolutional neural network layer to extract the local features of the battery charging curve.

[0153] S740. Input the local features into a long short-term memory network layer to learn the time-series change law of the battery charging state.

[0154] S750. Calculate the battery internal resistance change rate, capacity attenuation rate, and temperature sensitivity index based on the deviation between the battery charging curve and the standard charging curve.

[0155] S760. Calculate the battery health according to the internal resistance change rate, capacity attenuation rate, and temperature sensitivity index by using a preset battery health function.

[0156] S770. Input the local features and the time-series change law into a policy network and a value network, and predict the charging efficiency and battery life under different charging parameters to obtain the optimal charging parameters.

[0157] In this embodiment, to ensure the integrity and continuity of the data, a sliding window mechanism is used to collect the battery charging curve data within the current time period. The battery charging curve data is time-series data including the charging voltage, charging current, and battery temperature. The window size is usually set to 10 minutes, and the step size is 1 minute, that is, the data window is updated every minute, and the time-series data of the most recent 10 minutes is retained for subsequent analysis.

[0158] The preprocessing of time series data includes four main steps: data cleaning, data normalization, data augmentation, and time alignment. Data cleaning is used to handle outliers and noise in the original data. The sliding window median filtering method is adopted to remove short-term noise. The data normalization step is used to convert data with different dimensions to the same scale so that the neural network can learn more effectively. For the three different physical quantities of voltage, current, and temperature, the Z-score normalization method is used respectively. For example, for voltage data, assuming the mean is 3.7V and the standard deviation is 0.3V, the original voltage value of 4.0V becomes (4.0 - 3.7) / 0.3 = 1.0 after normalization. This normalization process makes the data of different physical quantities concentrated in a distribution with a mean of 0 and a standard deviation of 1, which is beneficial to the training convergence of the neural network.

[0159] Data augmentation is used to expand the training samples by synthesizing data artificially and improve the generalization ability of the model. It can be achieved by adding Gaussian noise, that is, superimposing random noise with a mean of 0 and a standard deviation of 10% of the standard deviation of the original data on the original data. Time alignment is used to solve the problem of inconsistent sampling frequencies of different sensors. Since the sampling frequencies of voltage and current are 10Hz, while the sampling frequency of temperature is 1Hz, these data need to be aligned to a unified time axis. The linear interpolation method is used to interpolate the temperature data from 1Hz to 10Hz to make the three types of data completely aligned in the time dimension. After the preprocessing of the above four steps, the original irregular and noisy time series data is transformed into a standardized, aligned, and low-noise dataset, providing high-quality input for subsequent feature extraction and model training.

[0160] The preprocessed time series data is input into a convolutional neural network layer to extract local features of the battery charging curve. Specifically, the convolutional neural network (CNN) layer is used to extract local feature patterns from the standardized battery charging curve, such as the voltage rise rate, current decline characteristics, temperature fluctuations, etc. CNN can effectively capture local patterns in time series data regardless of position. In actual implementation, the input to the CNN layer is a two-dimensional tensor with a shape of [T, 3], where T is the time step (usually 6000, corresponding to the number of data points sampled at 10 Hz within a 10-minute window), and 3 represents the three channels of voltage, current, and temperature. The CNN layer consists of three consecutive convolutional blocks, and each convolutional block contains a convolutional layer, a batch normalization layer, an activation function layer, and a pooling layer. The first convolutional block uses 32 convolutional kernels with a size of 15×1 and a stride of 1 to perform a convolution operation on the input data. The convolutional kernel size of 15 means that a data window of 1.5 seconds (15 / 10 Hz) is considered each time, which can capture the characteristics of parameter changes in a short period. After the convolution operation, a batch normalization layer is connected to stabilize the training process; then a ReLU activation function is introduced to introduce a non-linear transformation; finally, a max pooling layer with a pooling window size of 2 and a stride of 2 is used to halve the data dimension while retaining significant features. The second convolutional block uses 64 convolutional kernels with a size of 10×1, and other settings are similar to those of the first convolutional block, but the pooling window is increased to 3 to further reduce the data dimension. The third convolutional block uses 128 convolutional kernels with a size of 5×1, and the pooling window size is 4. Through these three consecutive convolutional blocks, the original [6000, 3] input data is converted into a feature map with a shape of approximately [125, 128], greatly reducing the data dimension while retaining key information. In the analysis of the battery charging curve, the CNN layer can extract various valuable local features. In this way, the CNN layer converts the original high-dimensional time series data into a feature representation containing rich semantic information, providing high-quality input for the subsequent long short-term memory network layer and greatly improving the model's ability to understand the battery charging state.

[0161] The local features are input into a long short-term memory network layer to learn the temporal variation law of the battery charging state. Specifically, the long short-term memory network (LSTM) layer receives the local feature representation from the CNN layer, and its main task is to capture the long-term dependencies and variation laws of the battery charging state evolving over time.

[0162] In actual implementation, the input of the LSTM layer is the feature sequence output by the CNN layer, with a shape of approximately [125, 128], representing 125 time steps, and each step has 128-dimensional features. The LSTM layer adopts a bidirectional structure, including forward and backward LSTM sub-networks, which can consider past and future context information simultaneously, improving the ability to understand temporal patterns. The LSTM layer consists of two stacked bidirectional LSTM cells. The first layer contains 64 hidden units (32 in each direction), and the second layer contains 32 hidden units (16 in each direction). Each LSTM cell contains three gated units and a memory unit. Taking a cell in the first layer of LSTM as an example, the input gate is controlled by the sigmoid function, determining how much of the current input information is stored in the memory unit; the forget gate is also controlled by the sigmoid function, determining how much of the memory unit state at the previous moment is retained; the output gate controls how much of the current memory unit state is output as the hidden state. This refined gating mechanism enables LSTM to selectively remember or forget information, which is particularly suitable for capturing long-term change trends during the battery charging process. In battery charging analysis, the LSTM layer can learn various important temporal change rules. For example, for a healthy new battery, during the constant current charging stage, the voltage rise curve usually shows a specific non-linear pattern, and LSTM can learn this pattern and identify abnormal situations that deviate from the normal pattern; for an aging battery, LSTM can capture the relationship change between the temperature rise rate and charging time during the charging process, and this change usually reflects the increase in the internal impedance of the battery. In addition, LSTM can also learn the transition characteristics between different charging stages, such as the transition point characteristics from pre-charging to constant current charging and from constant current charging to constant voltage charging, and these characteristics are crucial for evaluating the battery health state. Through the two-layer stacked bidirectional LSTM structure, the model can extract high-level temporal patterns from local features, forming a comprehensive understanding of the battery charging state. The output of the LSTM layer is a feature sequence with a shape of [125, 32], containing the temporal change rules of the battery charging state, providing a solid foundation for subsequent battery health assessment and charging parameter optimization.

[0163] Based on the deviation between the battery charging curve and the standard charging curve, the battery internal resistance change rate, capacity attenuation rate, and temperature sensitivity index are calculated. Specifically, by comparing the deviation between the actual battery charging curve and the standard charging curve (i.e., the charging curve of a brand-new battery under ideal conditions), key indicators reflecting the battery health state can be extracted. Taking a 5000mAh lithium polymer battery as an example, its standard charging curve library includes the standard voltage, current, and temperature curves at three charging rates of 0.2C, 0.5C, and 1.0C.

[0164] After obtaining the actual battery charging curve, the matching standard curve is automatically selected for comparison according to the charging conditions. The calculation of the battery internal resistance change rate is based on Ohm's law and the battery equivalent circuit model. The specific calculation method is: in the constant current charging stage, select multiple time points (usually 10 evenly distributed points), and calculate the instantaneous internal resistance R = ΔV / I at each point, where ΔV is the voltage increment relative to the open circuit voltage, and I is the charging current; then calculate the average internal resistance of these points, and compare it with the standard internal resistance to obtain the internal resistance change rate ΔR% = (Ractual-Rstandard) / Rstandard×100%. For example, if the standard internal resistance is 20mΩ, and the actual measured average internal resistance is 24mΩ, the internal resistance change rate is (24-20) / 20×100%=20%, indicating that the battery internal resistance has increased by 20%.

[0165] The capacity decay rate is the second key indicator, which reflects the degree of decline in the battery's ability to store energy. Its calculation is based on the comparison of the integral area of the charging curve. The specific method is: calculate the current-time integral of the constant current charging stage to obtain the actual amount of electricity charged Qactual; compare this amount of electricity with the electricity Qstandard under the same conditions under the standard charging curve to obtain the capacity decay rate ΔQ% = (Qstandard-Qactual) / Qstandard×100%. For example, if the standard amount of electricity is 4800mAh, and the actual measured amount of electricity is 4320mAh, the capacity decay rate is (4800-4320) / 4800×100%=10%, indicating that the battery capacity has decreased by 10%.

[0166] The temperature sensitivity index is the third key indicator, which reflects the change in the heating characteristics of the battery during the charging process. Its calculation is based on the comparison of the slope of the temperature rise curve. The specific method is: calculate the average slope of the temperature rise curve during the charging process k_actual = ΔT / Δt; compare this slope with the standard temperature rise slope k_standard, and obtain the temperature sensitivity index ΔT% = (kactual-kstandard) / kstandard×100%. For example, if the standard temperature rise slope is 0.05℃ / minute, and the actual measured slope is 0.08℃ / minute, the temperature sensitivity index is (0.08-0.05) / 0.05×100%=60%, indicating that the temperature sensitivity of the battery has increased by 60%, that is, it is easier to heat up under the same charging conditions. By calculating these three key indicators, the system can comprehensively evaluate the health of the battery and provide a quantitative basis for the subsequent battery health calculation.

[0167] Based on the internal resistance change rate, capacity attenuation rate, and temperature sensitivity index, the battery health is calculated using a preset battery health function. Specifically, the battery health is a percentage index that comprehensively reflects the overall health state of the battery. Usually, it is based on the state of a brand-new battery (100%), and gradually decreases as the battery ages. The calculation of the battery health uses a weighted comprehensive scoring function to convert the three key indicators (internal resistance change rate, capacity attenuation rate, and temperature sensitivity index) calculated in the previous step into health scores through non-linear mapping. The preset battery health function takes the following form:

[0168] SOH = 100% - [w1×f1(ΔR%) + w2×f2(ΔQ%) + w3×f3(ΔT%)], where w1, w2, and w3 are weight coefficients, representing the influence degrees of the three indicators on the battery health respectively; f1, f2, and f3 are non-linear mapping functions used to convert each indicator into a health loss. Usually, w1 (internal resistance weight) is 0.3, w2 (capacity weight) is 0.5, and w3 (temperature weight) is 0.2, indicating that the capacity attenuation has the greatest impact on the battery health, followed by the internal resistance change, and finally the temperature sensitivity. The non-linear mapping function adopts a sigmoid-type function, which can handle the saturation effect of the indicator values, that is, when the indicator exceeds a certain threshold, the impact on the health tends to be stable.

[0169] Taking the mapping function of the internal resistance change rate as an example: f1(ΔR%) = 100%×[1 - exp(-ΔR% / α1)] / [1 + exp(-ΔR% / α1)], where α1 is an adjustment parameter, usually taking a value of 30%, indicating that when the internal resistance change rate reaches 30%, the impact on the health reaches a medium level. The mapping function forms of the capacity attenuation rate and temperature sensitivity index are similar, but the adjustment parameters are different, which are α2 = 20% and α3 = 50% respectively.

[0170] The calculation result of the battery health not only intuitively reflects the overall health state of the battery but also can be used to predict the remaining service life of the battery. Generally, it is considered that when the battery health drops below 80%, the battery enters the accelerated aging stage; when it drops below 60%, it is recommended to replace the battery. In addition, the battery health is also an important basis for formulating the charging strategy. Batteries with lower health usually require more conservative charging parameters to slow down the aging speed.

[0171] The local features and the temporal variation rules are input into the policy network and the value network, and the charging efficiency and battery life under different charging parameters are predicted to obtain the optimal charging parameters. Specifically, the policy network and the value network are used to predict the optimal charging parameters according to the current state of the battery. These two networks receive the local features from the CNN layer and the temporal variation rules from the LSTM layer, and output the optimal charging strategy for the current battery state. The policy network is responsible for generating actions (charging parameters); the value network is responsible for evaluating the value of the actions. Since the policy network and the value network have been specifically described above, they will not be elaborated in this application.

[0172] In this embodiment, the high-precision sensor network is used to collect the voltage, current, and temperature time-series data during the battery charging process, and then through data preprocessing, including data cleaning, standardization, enhancement, and time alignment. The preprocessed data is input into the convolutional neural network layer to extract local features, and then the long short-term memory network layer is used to learn the temporal variation rules. Based on the extracted features, the system calculates the battery internal resistance change rate, capacity attenuation rate, and temperature sensitivity index, and comprehensively evaluates the battery health state through a preset health function. Finally, the policy network and the value network predict and output the optimal charging parameters based on the current state and health of the battery, so as to be able to monitor the battery health state in real time, give early warnings of potential risks, provide personalized charging strategies according to the individual differences and aging degrees of the batteries, and find the best balance among the charging speed, efficiency, and battery life.

[0173] In one implementation of this embodiment, based on the battery health and the optimal charging parameters, the charging parameters of the UAV battery in the next preset time period are dynamically adjusted at the end of the current time period, including the following steps:

[0174] S810. When the battery health is lower than the first preset threshold, at the end of the current time period, reduce the charging current in the next preset time period to the first preset proportion of the optimal charging parameters, and reduce the charging voltage to the second preset proportion of the optimal charging parameters;

[0175] S820. When the battery health score is not lower than the first preset threshold and not higher than the second preset threshold, at the end of the current time period, set the charging parameters in the next preset time period as the optimal charging parameters;

[0176] S830. When the battery health score is higher than the second preset threshold and the flight emergency level is high, at the end of the current time period, increase the charging current in the next preset time period to the third preset proportion of the optimal charging parameters, and not exceed the battery safety parameter threshold.

[0177] At the end of the current time period, the charging parameters of the UAV battery for the next preset time period are dynamically adjusted according to the battery health and the optimal charging parameters. This method balances the relationship between battery life protection and charging efficiency by real-time monitoring of the battery health and combining with the urgency of the flight mission, and adopting corresponding charging strategies.

[0178] When the battery health score is lower than the first preset threshold, it indicates that the battery is in a poor health state and a protective charging strategy needs to be adopted. At the end of the current time period, the charging current for the next preset time period is automatically reduced to the first preset proportion of the optimal charging parameters, and at the same time the charging voltage is reduced to the second preset proportion of the optimal charging parameters. This way of reducing the charging parameters can effectively relieve the pressure on the battery and delay the attenuation rate of the battery performance.

[0179] Specifically, the first preset threshold can be set to 80, and this threshold can be appropriately adjusted according to the UAV battery type and the usage environment. The first preset proportion can be set to 0.7, that is, the charging current is reduced to 70% of the optimal charging current; the second preset proportion can be set to 0.9, that is, the charging voltage is reduced to 90% of the optimal charging voltage. For example, if the optimal charging current of a certain type of lithium battery is 2A and the optimal charging voltage is 4.2V, when its health score is 75, the system will automatically adjust the charging current for the next time period to 1.4A (2A × 0.7) and the charging voltage to 3.78V (4.2V × 0.9) after the end of the current charging time period.

[0180] This way of reducing the charging parameters can significantly reduce the heat accumulation and electrochemical stress inside the battery, and reduce the risk of electrolyte decomposition and electrode material structure damage. In practical applications, adopting this charging strategy can avoid further damage to the battery caused by fast charging for batteries with poor health conditions, and provide a buffer period for battery recovery.

[0181] When the battery health score is not lower than the first preset threshold and not higher than the second preset threshold, it indicates that the battery is in a normal health state and a standard charging strategy can be adopted. At the end of the current time period, the charging system will set the charging parameters for the next preset time period to the optimal charging parameters, which can ensure the charging efficiency and will not cause excessive damage to the battery.

[0182] In practical applications, the second preset threshold can be set to 90. When the battery health score is between 80 and 90, the optimal charging parameters are used for charging. The optimal charging parameters refer to the parameter combination that can achieve higher charging efficiency on the premise of ensuring battery safety and life.

[0183] Charging with the optimal charging parameters can achieve a relatively ideal charging speed while ensuring battery safety. Under this charging strategy, the battery can usually complete 80% of its capacity charging within 1 - 1.5 hours, meeting the needs of most regular usage scenarios. At the same time, since the parameters are set within the safe range, there will be no obvious additional loss to the battery, and the battery can maintain its normal cycle life. This balanced charging strategy is suitable for daily usage scenarios and can balance charging efficiency and battery life protection.

[0184] When the battery health score is higher than the second preset threshold and the flight mission urgency is high, it indicates that the battery is in good condition and there is a need for fast charging. At the end of the current time period, the charging current for the next preset time period will be increased to the third preset ratio of the optimal charging parameters to accelerate the charging speed, while ensuring that it does not exceed the battery safety parameter threshold to prevent safety hazards caused by overcharging.

[0185] In the specific implementation process, the third preset ratio can be set to 1.3, that is, the charging current is increased to 130% of the optimal charging current. The battery safety parameter threshold includes key indicators such as the maximum allowable charging current and the highest charging temperature. For example, if the optimal charging current of a certain model of battery is 3A and the maximum allowable charging current is 4.5A, when its health score is 95 points and the flight mission urgency is high, after the end of the current charging time period, the charging current for the next time period will be adjusted to 3.9A (3A × 1.3). Since this value does not exceed the maximum allowable charging current of 4.5A, it can be safely implemented.

[0186] It should be noted that although increasing the charging current can accelerate the charging speed, it will also increase the battery heat generation. Therefore, during the implementation process, the charging system will continuously monitor the battery temperature. Once it approaches the safety threshold (usually 45°C), the charging current will be automatically reduced to ensure battery safety. Since this strategy is only applicable to batteries in good health and is limited by the safety parameter threshold, the impact on battery life is relatively controllable.

[0187] In another embodiment, according to the battery temperature change trend, predict the battery temperature for the next preset time period. In the case where the predicted temperature is close to the safety threshold, pre - reduce the charging current to prevent the battery from overheating; through an adaptive control algorithm, fine - tune the charging parameters in real - time to make the actual charging curve gradually approach the theoretical optimal charging curve.

[0188] Specifically, according to the battery temperature change trend, the battery temperature for the next preset time period can be predicted, and in the case where the predicted temperature is close to the safety threshold, pre - reduce the charging current, thereby preventing the battery from overheating. Compared with traditional passive response - type temperature control, it can more effectively protect battery safety and extend battery life.

[0189] During specific implementation, the temperature change data within the most recent 30 minutes is recorded, and the temperature change rate and trend are analyzed through mathematical models such as linear regression or exponential smoothing. For example, if the current battery temperature is 38°C and the temperature has been rising at an average rate of 0.2°C per minute within the most recent 15 minutes, it can be predicted that if the current charging current is maintained, the battery temperature will reach 44°C at the end of the next 30-minute preset time period. If the battery safety temperature threshold is set at 45°C, then the predicted temperature is approaching the safety threshold, and preventive current reduction measures will be triggered.

[0190] The amplitude of preventive current reduction can be dynamically determined based on the proximity of the predicted temperature to the safety threshold. For example, a three-level current reduction strategy can be set: when the predicted temperature reaches 85% of the safety threshold, the charging current is reduced to 80% of the current value; when the predicted temperature reaches 90% of the safety threshold, the charging current is reduced to 60% of the current value; when the predicted temperature reaches 95% of the safety threshold, the charging current is reduced to 40% of the current value.

[0191] This preventive current reduction strategy based on temperature prediction can effectively avoid the risk of the battery temperature hitting the safety threshold and reduce the temperature fluctuations during the charging process. Although it may slightly extend the charging time in some cases (usually by 10% - 15%), it significantly improves the charging safety, reduces the damage to the battery materials caused by high temperature, and can extend the actual service life of the battery in the long run. In addition, a more stable temperature curve also helps to improve the consistency of the battery charging efficiency and charging depth, thereby enhancing the endurance and reliability of the drone.

[0192] Through an adaptive control algorithm, the charging parameters can be fine-tuned in real time to make the actual charging curve gradually approach the theoretical optimal charging curve, so as to maximize the charging efficiency and battery life while ensuring battery safety.

[0193] The adaptive control algorithm is based on a closed-loop feedback mechanism and mainly includes three core links: state observation, deviation calculation, and parameter adjustment. First, the key state parameters of the battery are monitored in real time, including terminal voltage, charging current, temperature, estimated internal resistance, etc. Second, these actual parameters are compared with the pre-established theoretical optimal charging model to calculate the deviation value. Finally, according to the deviation value and its change trend, algorithms such as PID (Proportional-Integral-Derivative) controller or fuzzy logic controller are used to calculate the increment of the charging parameters that need to be adjusted and apply it to the next control cycle.

[0194] The theoretical optimal charging curve is usually determined comprehensively based on the battery chemical characteristics and usage scenarios.

[0195] In practical applications, the adjustment period of the adaptive control algorithm can be set to once every 30 seconds, and the adjustment amplitude each time does not exceed 5% of the current parameter to ensure system stability. For example, if the current charging current is 2A and the actual charging curve is lower than the theoretical curve, the current is adjusted to 2.1A; if the battery temperature rise rate exceeds the expectation, the current is reduced to 1.9A. This fine-tuning process will continue until the deviation between the actual charging curve and the theoretical curve is controlled within an acceptable range (such as ±3%).

[0196] The charging parameter adjustment method based on the battery temperature change trend prediction and the adaptive control algorithm realizes the intelligent and refined management of the UAV battery charging process. This method effectively avoids the risk of battery overheating; at the same time, by means of the adaptive control algorithm, the charging parameters are fine-tuned in real time, making the actual charging process gradually approach the theoretical optimal state. The combined application of the two technologies forms a closed-loop optimized charging management system, which can not only ensure battery safety, but also maximize the charging efficiency and battery life.

[0197] For UAV operations, this means shorter ground preparation time, longer flight duration, and lower battery replacement frequency and maintenance costs. Especially in harsh environmental conditions or high-intensity mission scenarios, this method can significantly improve the reliability and mission execution ability of the UAV system.

[0198] In this embodiment, the charging strategy is divided into three modes: protective charging, standard charging, and fast charging according to the battery health status, and a comprehensive decision is made in combination with the urgency of the flight mission. Through real-time monitoring and dynamic adjustment, it is possible to maximize the balance between charging efficiency and battery life on the premise of ensuring battery safety, significantly enhancing the adaptability and reliability of the UAV in various complex mission scenarios, and providing a strong guarantee for the long-term, high-efficiency, safe and reliable flight mission of the UAV.

[0199] The embodiment of the present application also provides an electronic device, including:

[0200] A memory configured to store instructions; and

[0201] A processor configured to call the instructions from the memory and be able to implement the above-mentioned charging parameter optimization method based on artificial intelligence when executing the instructions.

[0202] Referring to Figure 3 , the embodiment of the present application also provides a charging parameter optimization system based on artificial intelligence, including:

[0203] An electronic device;

[0204] A UAV connected to the electronic device.

[0205] In this embodiment, the electronic device may be a device with a processor such as a tablet computer, a desktop computer, a laptop computer, a handheld computer, a wearable device, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, etc. Of course, the electronic device may also be a server. The embodiments of the present application do not impose special restrictions on the specific form of the electronic device.

[0206] The specific working process is as follows: The drone sensor system collects battery status data; the data is transmitted to the electronic device through the on-board communication unit; the data processing unit of the electronic device analyzes the data; the AI engine generates optimized charging parameters; the optimized parameters are sent to the drone through the communication module; the battery management system of the drone executes the optimized parameters for charging.

[0207] Those skilled in the art should understand that the embodiments of the present application may be provided as a method, a system, or a computer program product. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0208] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0209] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.

[0210] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 steps of the functions specified in one block or multiple blocks.

[0211] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0212] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0213] Computer-readable media includes permanent and non-permanent, removable and non-removable media and can be implemented by any method or technology for information storage. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0214] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, commodity or device comprising the element.

[0215] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. An artificial intelligence-based charging parameter optimization method, characterized in that, Including: Obtain the battery type, current battery level, and remaining flight mission of the drone; Based on the battery type, current battery level, and remaining flight mission, determine the charging mode and initial charging parameters of the drone; In response to the charging signal of the drone, obtain the battery charging curve of the drone in real time, where the battery charging curve is constructed based on the real-time charging voltage, real-time charging current, and real-time battery temperature; Based on the battery charging curve of the current time period, use an artificial intelligence model to dynamically adjust the charging parameters of the drone battery in the next preset time period at the end of the current time period, and the charging parameters include charging current and charging voltage; Charge the drone battery according to the charging parameters at the start of the next preset time period.

2. The method according to claim 1, wherein Based on the battery type, current battery level, and remaining flight mission, determining the charging mode and initial charging parameters of the drone includes: According to the battery type of the drone, obtain the corresponding battery safety parameter threshold from the preset battery parameter database; Based on the current battery level, determine the charging stage of the battery; Based on the remaining flight mission, determine the flight urgency level, where the flight urgency level includes high, medium, and low; Based on the battery safety parameter threshold, charging stage, and flight urgency level, determine that the charging mode is any one of a constant current charging mode, a constant voltage charging mode, or a constant current-constant voltage combined charging mode, and determine the initial charging current value or charging voltage value according to the charging mode.

3. The method according to claim 2, wherein The remaining flight mission includes the mission type and the estimated flight time. Based on the remaining flight mission, determining the flight urgency level includes: When the mission type is the preset first mission type, determine that the flight urgency level is high; When the mission type is not the first mission type, determine that the flight urgency level is medium or low based on the estimated flight time; Among them, when the estimated flight time is less than the preset time threshold, determine that the flight urgency level is medium, and when the estimated flight time is not less than the preset time threshold, determine that the flight urgency level is low.

4. The method according to claim 2, wherein The battery safety parameter threshold includes a safety current threshold and a safety voltage threshold. Based on the battery safety parameter threshold, charging stage, and flight urgency level, determining that the charging mode is any one of a constant current charging mode, a constant voltage charging mode, or a constant current-constant voltage combined charging mode, and determining the initial charging current value or charging voltage value according to the charging mode includes: When the charging stage is the first charging stage and the flight urgency level is high, determine that the charging mode is a constant current-constant voltage combined charging mode, and set the initial charging current value to the first preset percentage of the safety current threshold; When the charging stage is the first charging stage and the flight urgency level is medium or low, determine that the charging mode is a constant current charging mode, and set the initial charging current value to the second preset percentage of the safety current threshold, where the second preset percentage is less than the first preset percentage; When the charging stage is the second charging stage, determine that the charging mode is a constant voltage charging mode, and set the initial charging voltage value to the third preset percentage of the safety voltage threshold.

5. The method according to claim 2, wherein Based on the battery charging curve of the current time period, an artificial intelligence model is used to dynamically adjust the charging parameters of the UAV battery in the next preset time period at the end of the current time period, including: Analyze the change trend of the battery charging curve through the artificial intelligence model to obtain an analysis result, and the analysis result includes battery health and optimal charging parameters; Based on the battery health and optimal charging parameters, dynamically adjust the charging parameters of the UAV battery in the next preset time period at the end of the current time period.

6. The method according to claim 5, characterized in that, The artificial intelligence model is a deep reinforcement learning model, including a convolutional neural network layer for extracting battery charging curve features, a long short-term memory network layer for learning the temporal change law of the battery charging state, and a policy network and a value network for predicting optimal charging parameters.

7. The method according to claim 6, wherein Analyze the change trend of the battery charging curve through the artificial intelligence model to obtain an analysis result, and the analysis result includes battery health and optimal charging parameters, including: Collect the battery charging curve data in the current time period, and the battery charging curve data is time series data including charging voltage, charging current and battery temperature; Preprocess the time series data; Input the preprocessed time series data into the convolutional neural network layer to extract the local features of the battery charging curve; Input the local features into the long short-term memory network layer to learn the temporal change law of the battery charging state; Based on the deviation between the battery charging curve and the standard charging curve, calculate the battery internal resistance change rate, capacity attenuation rate and temperature sensitivity index; According to the internal resistance change rate, capacity attenuation rate and temperature sensitivity index, use a preset battery health function to calculate the battery health; Input the local features and the temporal change law into the policy network and the value network, predict the charging efficiency and battery life under different charging parameters, and obtain the optimal charging parameters.

8. The method according to claim 5, wherein Based on the battery health and optimal charging parameters, dynamically adjust the charging parameters of the UAV battery in the next preset time period at the end of the current time period, including: In the case where the battery health is lower than the first preset threshold, at the end of the current time period, reduce the charging current in the next preset time period to the first preset ratio of the optimal charging parameter, and reduce the charging voltage to the second preset ratio of the optimal charging parameter; In the case where the battery health score is not lower than the first preset threshold and not higher than the second preset threshold, at the end of the current time period, set the charging parameters in the next preset time period to the optimal charging parameters; In the case where the battery health score is higher than the second preset threshold and the flight emergency level is high, at the end of the current time period, increase the charging current in the next preset time period to the third preset ratio of the optimal charging parameter, and not exceed the battery safety parameter threshold.

9. An electronic device, characterized in that, Including: A memory configured to store instructions; And A processor configured to call the instructions from the memory and be able to implement the artificial intelligence-based charging parameter optimization method according to any one of claims 1 to 8 when executing the instructions.

10. An artificial intelligence-based charging parameter optimization system, characterized in that, Including: An electronic device according to claim 9; A UAV connected to the electronic device.

Images