Unmanned aerial vehicle charging safety early warning method, device and equipment

Abstract

This invention provides a method, apparatus, and device for early warning of drone charging safety, relating to the field of drone charging technology. The method includes: acquiring the impedance change rate sequence of a charging point and the linear acceleration sequence of the drone; establishing a wireless inductive charging connection between the charging point and the drone; identifying target data points in the linear acceleration sequence that meet preset vibration wake-up conditions; determining an analysis time window based on the target data points; calculating the spatiotemporal coupling degree of the impedance change rate sequence and the linear acceleration sequence within the analysis time window; determining whether physical charging misalignment has occurred based on the spatiotemporal coupling degree; if physical charging misalignment occurs, predicting a charging temperature curve based on the current charging power, impedance, and ambient temperature; and issuing a charging safety warning based on the charging temperature curve. This invention can timely and accurately monitor anomalies during wireless charging of drones and issue early warnings, ensuring charging safety.

Description

Technical Field

[0001] This invention relates to the field of drone charging technology, and in particular to a method, device and equipment for drone charging safety early warning. Background Technology

[0002] With the large-scale application of drone technology in fields such as power line inspection, border security, geographic surveying, and emergency rescue, the endurance of drones has become a core bottleneck restricting their operational radius and level of automation. Wireless inductive charging technology, with its advantages of being contactless, automated, and highly adaptable to various environments, has become the mainstream solution for automatic drone charging. However, wireless charging stations are mostly deployed in unattended outdoor scenarios, and risks such as coil misalignment, foreign object intrusion, battery thermal runaway, and malicious network attacks during the charging process can easily lead to safety accidents such as equipment burnout and explosions.

[0003] In related technologies, safety monitoring solutions for wireless charging of drones mostly rely on low-frequency macroscopic electrical quantities and temperature sensor sampling. These solutions can only capture serious faults that have already occurred and cannot identify abnormal transient microscopic electromagnetic characteristics such as slight coil misalignment, intrusion of tiny foreign objects, or precursors to internal short circuits in advance. At the same time, temperature sensors have serious thermal inertial hysteresis, which greatly delays the warning timing. Furthermore, in harsh outdoor environments, sensors are prone to aging drift, wire breakage, and short circuit faults, leading to false alarms or missed alarms in the system. Summary of the Invention

[0004] This invention provides a method, device, and equipment for early warning of drone charging safety, which can monitor abnormalities during drone wireless charging in a timely and accurate manner and issue early warnings to ensure charging safety.

[0005] In a first aspect, embodiments of the present invention provide a method for early warning of drone charging safety, comprising: Obtain the impedance change rate sequence of the charging point and the linear acceleration sequence of the drone; wherein, a wireless inductive charging connection is established between the charging point and the drone; Identify target data points in the linear acceleration sequence that meet the preset vibration awakening conditions, determine the analysis time window based on the target data points, and calculate the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window; Based on the spatiotemporal coupling degree, determine whether a physical misalignment during charging has occurred; If a physical misalignment occurs during charging, the charging temperature curve is predicted based on the current charging power, impedance, and ambient temperature. Based on the charging temperature curve, a charging safety warning is issued.

[0006] In one possible implementation, target data points in the linear acceleration sequence that satisfy preset vibration awakening conditions are identified. An analysis time window is determined based on these target data points, and the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window is calculated, including: If a preset number of consecutive data points in a linear acceleration sequence are all greater than a preset vibration wake-up threshold, then the first data point among the preset number of data points will be determined as the target data point that meets the preset vibration wake-up condition. The analysis time window is determined based on the preset time before and after the target data point; The spatiotemporal coupling degree of the impedance rate of change sequence and the linear acceleration sequence within the time window is calculated and analyzed using a dynamic time warping algorithm.

[0007] In one possible implementation, the determination of whether a physical charging misalignment has occurred is based on the spatiotemporal coupling degree, including: if the spatiotemporal coupling degree is greater than a preset spatiotemporal coupling degree threshold, then it is determined that a physical charging misalignment has occurred.

[0008] In one possible implementation, the charging temperature profile is predicted based on the current charging power, impedance, and measured ambient temperature, including: Calculate the absolute difference between the impedance and the reference impedance. Based on the absolute difference and the charging power, look up the preset three-dimensional mapping table to obtain the physical misalignment distance. The physical misalignment distance, charging power, and measured ambient temperature are input into a pre-trained temperature prediction model to obtain the charging temperature curve output by the temperature prediction model.

[0009] In one possible implementation, before predicting the charging temperature profile based on the current charging power, impedance, and measured ambient temperature, the following is also included: Based on a pre-built thermodynamic lumped parameter model of the charging process, the virtual ambient temperature is obtained in real time through extended Kalman filtering algorithm; Calculate the residual between the measured ambient temperature and the virtual temperature, and obtain the cumulative deviation based on the residual using the cumulative sum control chart algorithm; When the cumulative deviation exceeds the preset confidence collapse threshold, the temperature sensor is identified as faulty, and the measured ambient temperature is replaced with the virtual ambient temperature.

[0010] In one possible implementation, when the accumulated deviation exceeds a preset confidence collapse threshold, the following is also included: The limiting power is determined based on the difference between the cumulative deviation and the confidence collapse threshold; wherein the limiting power is negatively correlated with the difference and is less than the rated power; Adjust the charging power output of the charging point to the limited power.

[0011] In one possible implementation, the method further includes: When establishing a wireless inductive charging connection between the charging point and the drone, a preset low-power detection pulse is injected into the transmitting coil of the charging point, and the transient impedance decay oscillation curve of the transmitting coil is obtained as the physical unclonable fingerprint of the drone. Based on the drone's MAC address, obtain the drone's reference physical unclonable fingerprint stored in the preset feature database; The similarity between a physically unclonable fingerprint and a reference physically unclonable fingerprint is calculated. If the similarity is lower than a preset similarity threshold, it is determined that a malicious power injection attack has occurred, and the charging point is prohibited from outputting charging power, as well as a malicious attack warning is issued.

[0012] In one possible implementation, charging safety warnings are based on the charging temperature profile, including: If the charging temperature curve indicates a risk of thermal runaway, an edge data rescue strategy will be implemented after a charging safety warning is issued. The edge data rescue strategies include: A non-maskable interrupt is applied to the operating system kernel, forcibly suspending non-core threads to exclusively allocate computing resources to the data rescue thread; Through the data rescue thread, high-frequency transient electrical characteristics are losslessly compressed to generate death declaration data packets; Through an in-band escape mechanism, the death declaration data packet is frequency shift keyed and superimposed on the residual AC carrier, and then transmitted to the fireproof and explosion-proof solid-state memory of the drone.

[0013] Secondly, embodiments of the present invention provide a drone charging safety early warning device, comprising: The acquisition module is used to acquire the impedance change rate sequence of the charging point and the linear acceleration sequence of the drone; wherein, the charging point establishes a wireless inductive charging connection with the drone; The calculation module is used to identify target data points in the linear acceleration sequence that meet the preset vibration awakening conditions, determine the analysis time window based on the target data points, and calculate the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window. The judgment module is used to determine whether a physical misalignment during charging has occurred based on the spatiotemporal coupling degree. The prediction module is used to predict the charging temperature curve based on the current charging power, impedance, and ambient temperature if a charging physical misalignment occurs. The early warning module is used to provide charging safety warnings based on the charging temperature curve.

[0014] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0015] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention, through simultaneous acquisition of the impedance change rate sequence at the charging point and the linear acceleration sequence of the drone, identifies vibrations based on vibration wake-up conditions. By calculating the spatiotemporal coupling degree of the two sets of sequences within the corresponding analysis time window, it determines the physical misalignment of the charging coil. This allows for early identification of minor misalignment faults before traditional temperature sensors show signs of temperature rise or thermal inertia. Furthermore, by combining multi-dimensional parameters such as charging power, impedance, and ambient temperature, it predicts temperature change curves and provides safety warnings, enabling early fault prediction and handling. This embodiment significantly improves the timeliness and reliability of fault identification, effectively preventing safety accidents such as overheating and equipment damage caused by coil misalignment, and ensuring the stable and safe operation of the drone's wireless inductive charging process. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of the drone charging safety early warning method provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the drone charging safety early warning device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0017] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0018] To illustrate the technical solution described in this invention, specific embodiments are described below.

[0019] In related technologies, wireless charging safety often relies on visual camera positioning or temperature sensor monitoring. However, visual solutions are easily affected by environmental factors such as light, rain, snow, and obstructions, while temperature sensors suffer from significant thermal inertia hysteresis. This invention transforms the problem of minute spatial displacement of the charging coil into a multi-dimensional time-series morphological comparison problem. It eliminates the need for additional physical positioning hardware and, based on pure algorithm analysis, can identify potential safety hazards tens of seconds or even minutes before a significant temperature rise.

[0020] See Figure 1 The flowchart illustrating the implementation of the drone charging safety early warning method provided in this embodiment of the invention is described in detail below: Step S101: Obtain the impedance change rate sequence of the charging point and the linear acceleration sequence of the drone; wherein, the charging point establishes a wireless inductive charging connection with the drone.

[0021] This embodiment achieves reliable detection through the high-frequency transient electrical characteristics of the base (charging point) and the micro kinematic characteristics of the drone fuselage.

[0022] To accurately capture transient changes caused by minute anomalies such as slight misalignment of the wireless charging coil, intrusion of tiny foreign objects, or precursors to internal short circuits, the data acquisition capability of the base needs to overcome the performance limitations of conventional low-frequency sampling. The specific execution process is as follows: The base-side microprocessor uses a high-precision analog-to-digital converter to perform high-frequency sampling of the wireless charging transmitter coil input. The sampling channels cover the transmitter's DC bus voltage, DC bus current, inverter output high-frequency AC voltage, and high-frequency AC current. To satisfy the Nyquist sampling theorem and fully capture high-frequency harmonic signals, the AC channel sampling rate is set to more than 10 times the wireless charging operating frequency (typically 85kHz or 100kHz), i.e., a sampling rate of no less than 1MHz. The original analog signal is first filtered by a hardware anti-aliasing filter (such as a fourth-order Butterworth low-pass filter with a cutoff frequency of 500kHz) and then converted into a discrete digital time series. Due to the large amount of original high-frequency time series data, it cannot be directly applied to long-period security early warning; therefore, feature reduction is performed using a Fast Fourier Transform algorithm. Specifically, a sliding data window of length N (1024 in an example) is opened in memory. When the window is full, a Fast Fourier Transform is performed on the AC voltage and current data within the window to extract the fundamental frequency, fundamental amplitude, and amplitudes of the first five harmonics. By calculating the cross power spectrum of the fundamental voltage and fundamental current, the phase difference between the two is obtained. Combined with Ohm's law in complex form, the real part (equivalent resistance) and imaginary part (equivalent inductive reactance) of the current equivalent input impedance of the transmitting coil are solved in real time. The above frequency domain features and complex domain features are packaged at a frequency of 100Hz (outputting a set of feature vectors every 10ms) and stored in the ring buffer at the base end to obtain the impedance change rate sequence of the charging point.

[0023] On the drone side, the onboard computer reads the underlying operating data of the battery management system in real time via a CAN bus or UART serial port at a sampling frequency of 10Hz. The collected data includes conventional parameters such as total battery voltage, total current, and remaining charge, as well as cell maximum temperature, cell minimum temperature, maximum voltage difference between individual cells, estimated internal resistance of individual cells, and battery health. During drone docking and charging, the onboard computer continuously calls the inertial measurement unit (IMU), which integrates a three-axis accelerometer and a three-axis gyroscope, with a sampling frequency set to 200Hz. The raw IMU data is processed by the onboard computer's built-in Kalman filter to perform attitude calculations, removing the gravitational acceleration component and calculating only the linear acceleration reflecting fuselage vibration and wind disturbance, and the angular velocity characterizing the fuselage's tilt state. This kinematic characteristic data is ultimately stored in the drone's local memory, thus obtaining the drone's linear acceleration sequence.

[0024] In addition, the physical temperature of the transmitter can be collected by multiple thermistors or infrared temperature probes deployed at the bottom of the transmitting coil, and ambient temperature, humidity, and wind speed can be collected by meteorological sensors outside the base. The low-frequency data mentioned above are smoothed by a moving average filter with a window size of 10 and then synchronously stored in a circular buffer.

[0025] Considering that the base and the drone have separate independent clock crystals, crystal errors and temperature drift can easily cause clock asynchrony, thus affecting fault causal analysis. This embodiment relies on the electromagnetic coupling channel of the charging coil to achieve cross-domain clock synchronization, avoiding the problem of wireless communication delay. Specifically: the base periodically superimposes a synchronization beacon carrying an absolute timestamp onto a high-frequency AC carrier; the drone demodulates the beacon and records the local timestamp. After sending back an acknowledgment message, the system calculates the transmission delay and clock deviation according to the protocol algorithm. The drone then calibrates its local clock accordingly, achieving sub-millisecond time synchronization. After synchronization is complete, the drone uploads the timestamped data to the edge computing gateway. The edge computing gateway uses linear interpolation or a zero-order hold to uniformly resample electrical, kinematic, and temperature data from different sampling frequencies to 100Hz, generating a time-aligned global charging state feature matrix to provide data support for subsequent early warning algorithms.

[0026] Step S102: Identify target data points in the linear acceleration sequence that meet the preset vibration awakening conditions, determine the analysis time window based on the target data points, and calculate the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window.

[0027] In this embodiment, the impedance change rate and linear acceleration differ significantly in dimensions and magnitudes, making direct comparison impossible. Therefore, preprocessing of both data streams is necessary. A dynamic window of length W (exemplarily 1000, corresponding to 10 seconds of 100Hz sampling data) can be created in memory. The mean and standard deviation of the data within the window are calculated in real time, and Z-score standardization is performed on the impedance change rate and acceleration integral values ​​based on the mean and standard deviation. After standardization, both electrical abrupt changes and mechanical vibration data are converted into dimensionless relative fluctuation sequences with a mean of 0 and a variance of 1, providing a mathematical basis for subsequent time-series morphological comparison.

[0028] In one possible implementation, if a preset number of consecutive data points in the linear acceleration sequence are all greater than a preset vibration wake-up threshold, then the first data point among the preset number of data points is determined as the target data point that meets the preset vibration wake-up condition; the analysis time window is determined based on the preset time before and after the target data point; and the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window is calculated using a dynamic time warping algorithm.

[0029] For example: When the absolute value of the standardized acceleration exceeds the threshold for three consecutive sampling periods (for example, the threshold is 2.5, which means that the data deviates from the mean by 2.5 standard deviations, mostly caused by gusts and aftershocks), the analysis is triggered. The analysis time window is 5 seconds (500 data points) before and after the trigger point, and the standardized impedance change rate sequence and linear acceleration sequence within the window are extracted.

[0030] In real-world physical scenarios, the vibration of the drone fuselage and the relative displacement of the coils cause a nonlinear time delay of tens to hundreds of milliseconds in the impedance response. The traditional Pearson correlation coefficient cannot handle this distortion and misalignment on the time axis. Therefore, a dynamic time warping algorithm is introduced: the DSP unit of the edge computing gateway initializes a two-dimensional matrix in memory. Each element in the matrix , representing the sequence number The nth point and the sequence number Euclidean distance between points; synchronously construct a cumulative cost matrix. The boundary conditions of the matrix are initialized to infinity, starting from... Then, the dynamic programming state transition equation is executed through a double loop: This equation finds a continuous path from the top left to the bottom right corner of the matrix that minimizes the sum of local costs at all points along the path. This path allows for stretching and compression of the time axis, perfectly aligning with the nonlinear delay characteristics between mechanical vibration and electromagnetic response. After calculation, the endpoint value of the cumulative cost matrix is ​​extracted. Then, it is normalized by dividing it by the length of the regular path to obtain the spatiotemporal coupling degree.

[0031] Step S103: Determine whether a physical misalignment of the charging process has occurred based on the spatiotemporal coupling degree.

[0032] In this embodiment, if the spatiotemporal coupling degree is greater than a preset spatiotemporal coupling degree threshold, a charging physical misalignment is determined to have occurred. For example: if the spatiotemporal coupling degree... (High coupling) suggests that the drone may have been physically misaligned due to gusts of wind or external forces, resulting in a sharp increase in magnetic leakage.

[0033] In step S104, if a physical misalignment occurs during charging, the charging temperature curve is predicted based on the current charging power, impedance, and ambient temperature.

[0034] In this embodiment, the absolute difference between the impedance and the reference impedance can be calculated. Based on the absolute difference and the charging power, a preset three-dimensional mapping table is consulted to obtain the physical misalignment distance. In other words, the absolute difference and the charging power are used as indexes to look up the corresponding physical misalignment distance in the table.

[0035] Then, the physical misalignment distance, charging power, and measured ambient temperature are input into a pre-trained temperature prediction model to obtain the charging temperature curve output by the model. Here, an LSTM model can be used, utilizing its internal cell state memory of historical thermal inertia to perform forward propagation calculations. The model's output layer is a fully connected layer, outputting an array of 60 elements representing the predicted temperature curve for the next 1 to 60 seconds. .

[0036] Step S105: Based on the charging temperature curve, a charging safety warning is issued.

[0037] For example, the system traverses the predicted temperature curves. If any value is found in the array If the battery's critical safety temperature is reached, it is determined that without intervention, the system will experience thermal runaway after a certain period of time. At this point, the actual readings of the physical sensors may only indicate a safe level. Therefore, this embodiment gained valuable time for rescue by anticipating charging safety risks.

[0038] As can be seen, this embodiment of the invention simultaneously collects the impedance change rate sequence of the charging point and the linear acceleration sequence of the UAV, identifies vibration based on vibration wake-up conditions, and determines the physical misalignment of the charging coil by calculating and analyzing the spatiotemporal coupling degree of the two sets of sequences within a time window. This allows for the early identification of minor misalignment faults before traditional temperature sensors show signs of temperature rise and thermal inertia. Furthermore, by combining multi-dimensional parameters such as charging power, impedance, and ambient temperature, it predicts temperature change curves and provides safety warnings, enabling early fault prediction and handling. This embodiment significantly improves the timeliness and reliability of fault identification, effectively prevents safety accidents such as overheating and equipment damage caused by coil misalignment, and ensures the stable and safe operation of the UAV's wireless inductive charging process.

[0039] In some embodiments, considering that physical temperature sensors are highly susceptible to aging, drift, detachment, or short-circuit failures in harsh outdoor operating environments, relying entirely on data from damaged sensors in the early warning system would lead to catastrophic missed alarms or frequent false alarms. This embodiment constructs a virtual sensor using pure software algorithms, achieving high fault tolerance for physical hardware failures.

[0040] Specifically, before step S104 predicts the charging temperature curve based on the current charging power, impedance, and measured ambient temperature, the following steps are also included: Based on a pre-built thermodynamic lumped parameter model of the charging process, the virtual ambient temperature is obtained in real time through extended Kalman filtering algorithm; Calculate the residual between the measured ambient temperature and the virtual temperature, and obtain the cumulative deviation based on the residual using the cumulative sum control chart algorithm; When the cumulative deviation exceeds the preset confidence collapse threshold, the temperature sensor is identified as faulty, and the measured ambient temperature is replaced with the virtual ambient temperature.

[0041] In this embodiment, the wireless charging transmitting coil, receiving coil, and drone battery are abstracted as three coupled thermodynamic nodes. A thermal balance relationship is established based on the law of conservation of energy: the change in node heat is equal to the input heat minus the dissipated heat. Taking the receiving coil as an example, the heat generation mainly comes from the copper loss of the coil and the iron loss of the shielding material, while the heat dissipation is determined by the node temperature, the ambient temperature, and the overall thermal resistance. A continuous-time state-space equation can be constructed based on this.

[0042] Considering that wind speed causes dynamic changes in thermal resistance and power loss introduces measurement noise, an extended Kalman filter is used to solve for the optimal virtual temperature. The extended Kalman filter algorithm runs at a 10Hz cycle, performing three steps sequentially: First, combining historical temperature, real-time electrical and environmental data, a priori virtual temperature is obtained through discrete state equations, and the prediction error covariance matrix is ​​updated; second, based on the error covariance matrix and the measurement noise covariance matrix, the Kalman gain matrix is ​​calculated; third, the state is updated: when the sensor is functioning normally, the measured temperature is used as the observed value to correct the priori value, obtaining the posterior optimal virtual temperature; when the sensor is malfunctioning, the observation weight is reduced, and only the model's derivation result is used as the virtual temperature.

[0043] Finally, verification is performed using virtual temperature as a benchmark: the residual between the measured temperature and the virtual temperature is calculated, and then the cumulative deviation is statistically analyzed using a cumulative sum control chart algorithm. When the cumulative deviation exceeds the confidence collapse threshold, the sensor is confirmed to be faulty. Afterwards, the data input source for all subsequent processing logic can be redirected from the physical sensor register address to the virtual sensor output variable address by modifying the data routing pointer in memory. This switching process is completed in microseconds, ensuring uninterrupted operation of the early warning system.

[0044] In this embodiment, when the cumulative deviation is greater than a preset confidence collapse threshold, the method further includes: determining the limiting power based on the difference between the cumulative deviation and the confidence collapse threshold; and adjusting the charging power output by the charging point to the limiting power.

[0045] Since virtual sensors rely entirely on mathematical models and lack absolute calibration based on physical truth, they pose a certain risk of long-term cumulative errors. Therefore, this embodiment generates a degraded operation command, which forcibly sets the upper limit of wireless charging transmission power to a limited power, such as 50% of the rated power (e.g., reducing from 200W to 100W). This conservative charging mode ensures that even with minor deviations in the virtual model, the low heat generation will absolutely not trigger thermal runaway. Simultaneously with degraded operation, external auxiliary devices (such as infrared thermal imaging cameras) can capture images of the charging area and extract the highest surface temperature using image recognition algorithms for recalibrating the virtual sensor's model parameters.

[0046] In some embodiments, considering that when a drone lands and establishes a Wi-Fi or Bluetooth connection, a malicious attacker might intercept and tamper with communication messages, for example, by disguising the MAC address and request messages of a small drone that only supports 50W charging as a large industrial drone that supports 500W fast charging. If the base blindly trusts the digital messages and injects 500W of power instantly, it will directly cause the small drone's battery to explode. To defend against this digital deception attack, this embodiment introduces physical feature cross-verification logic. Before officially starting high-power charging, the base forcibly executes a physical flaw detection subroutine: When establishing a wireless inductive charging connection between the charging point and the drone, a preset low-power detection pulse is injected into the transmitting coil of the charging point, and the transient impedance decay oscillation curve of the transmitting coil is obtained as the physical unclonable fingerprint of the drone. Based on the drone's MAC address, obtain the drone's reference physical unclonable fingerprint stored in the preset feature database; The similarity between a physically unclonable fingerprint and a reference physically unclonable fingerprint is calculated. If the similarity is lower than a preset similarity threshold, it is determined that a malicious power injection attack has occurred, and the charging point is prohibited from outputting charging power, as well as a malicious attack warning is issued.

[0047] In this embodiment, a high-frequency probe pulse with a duration of only 10 milliseconds and a power of only 5W is injected into the transmitting coil via a base inverter. Within this 10-millisecond probe window, the base captures the transient voltage and current waveforms of the transmitting coil at a sampling rate of 2MHz. Because drone batteries of different models, capacities, and aging levels have completely different internal equivalent series resistances, equivalent series inductances, and parasitic capacitances of the receiving coil, these microscopic physical differences, under the excitation of the 5W probe pulse, will produce a unique transient impedance decay oscillation curve. This curve is the physical unclonable fingerprint of the drone battery. By extracting the envelope features and decay time constant of this curve, and performing cosine similarity calculations with reference physical unclonable fingerprints in a preset feature library, if a digital message requests 500W charging power, but the similarity calculation results indicate that the physical entity currently resting on the base is a small-capacity battery (similarity <0.6), the system determines that a malicious power injection attack has occurred. At this point, the inverter's hardware enable pin can be pulled low to completely lock the physical power output, prohibit the charging point from outputting charging power, and at the same time, the attacker's MAC address and radio frequency characteristics can be written into the local blacklist, and a malicious attack warning can be issued.

[0048] In some embodiments, if a risk of thermal runaway is determined based on the charging temperature profile, an edge data recovery strategy is also executed after a charging safety warning is issued: A non-maskable interrupt is applied to the operating system kernel, forcibly suspending non-core threads to exclusively allocate computing resources to the data rescue thread; Through the data rescue thread, high-frequency transient electrical characteristics are losslessly compressed to generate death declaration data packets; Through an in-band escape mechanism, the death declaration data packet is frequency shift keyed and superimposed on the residual AC carrier, and then transmitted to the fireproof and explosion-proof solid-state memory of the drone.

[0049] In this embodiment, when the LSTM model predicts that thermal runaway cannot be contained or the virtual sensor residuals become drastically abnormal, the system state machine switches from a warning state to a critical collapse state, initiating an emergency procedure for extreme conditions. In this situation, only 2-5 seconds remain for handling the situation, and conventional log upload methods cannot retain data from the moment of the fault. This embodiment can perform a non-maskable interrupt on the operating system kernel, forcibly suspending non-core threads and allocating all computing power and channel resources to the data rescue thread; simultaneously, the sampling rate is maximized to the highest level (5MHz in the example) to rapidly acquire voltage, current, and impedance waveform data before the fault. For massive high-frequency sampled data, a two-stage compression scheme of differential coding + Huffman coding is adopted: first, the sampling point difference is recorded through Delta differential coding, and then a Huffman coding tree resident in memory is called to complete the secondary compression, which can compress the data volume to within 15% of the original size. After compression, high-frequency features, predicted temperature, and physical unclonable fingerprints are integrated to generate a death declaration data packet. To ensure that at least one copy of the data packet is successfully transmitted, the edge computing gateway does not wait for any handshake confirmation and directly pushes the data packet to all available physical network interfaces (Wi-Fi interface, 5G cellular interface, Ethernet interface). If the external radio frequency antenna has been melted by high temperature at this time, the system will activate the last physical escape channel: the death declaration data packet is frequency shift keyed and superimposed on the residual AC carrier, and transmitted to the drone's fireproof and explosion-proof solid-state storage.

[0050] In some embodiments, considering that traditional solutions upload raw data from each charging station to the cloud for centralized model training, which not only consumes a large amount of bandwidth but also easily leads to the leakage of sensitive information such as location and device status, and that the operating environments of charging stations in different regions vary greatly, a unified static LSTM model is prone to false positives and false negatives. Therefore, in this embodiment, the LSTM model is updated through a cross-regional group security model evolution mechanism based on federated learning. Specifically, this includes: the edge computing gateway uses local historical time series with security labels to incrementally train the local long short-term memory network and extract the gradient matrix of the network layer parameters; the gradient matrix is ​​encrypted and uploaded to the central cloud server; the central cloud server uses a federated averaging algorithm to perform a weighted average of the gradient matrices uploaded by multiple edge computing gateways to generate global security warning model parameters; the edge computing gateway receives the global security warning model parameters as new initial weights and fine-tunes them using local data to adapt the local model to the unique local climate and the aging characteristics of specific models.

[0051] In some embodiments, when the current charging state is determined to deviate from the safety benchmark, the traditional control logic is to cut off the power when the temperature is high, which can easily lead to large-scale delays in drone missions. In this embodiment, a lightweight reinforcement learning agent is deployed in the edge computing gateway to achieve flexible intelligent control.

[0052] The state space of the intelligent agent includes: the current predicted temperature, the remaining battery power, the urgency of the drone's next mission (issued by the cloud scheduling system), and the current ambient temperature; The action space of the intelligent agent includes: adjusting the PWM duty cycle of the transmitter inverter (i.e., continuously adjusting the charging power, ranging from 10W to 500W), adjusting the cooling power of the semiconductor heat sink at the bottom of the base, and adjusting the speed of the cooling fan; The reward function of the agent is designed to maximize the charging speed while minimizing the energy consumption of the heat dissipation system, provided that the predicted temperature does not exceed the safety threshold.

[0053] Through reinforcement learning, the system can automatically find the dynamic balance point between the optimal charging power and heat dissipation power in high-temperature environments, achieving flexible derating charging instead of simply and brutally shutting down the system by blowing up the fuse.

[0054] Based on the degree of danger, the system generates four levels of early warning instructions to achieve tiered response: Level 1 (Hint - Slight Virtual Sensor Residual Drift): Only logged locally, does not change charging status, cloud-marked base sensor may require maintenance; Level 2 (Warning - Predicted temperature approaching threshold or slight displacement): Triggers the reinforcement learning agent to actively reduce charging power, activate maximum power cooling, and send attitude fine-tuning commands to the drone. Level 3 (Severe - Predicted thermal runaway is inevitable): Immediately pull down the inverter hardware enable pin, cut off wireless power output, and push a severe alert to the cloud containing a snapshot of high-frequency electrical characteristics. Level 4 (Meltdown attack or physical structure disintegration): Triggers the data rescue mechanism, forcefully pushes the death declaration data packet through multi-link and in-band communication, locks the base, and requires manual on-site reset.

[0055] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0056] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0057] Figure 2 A schematic diagram of the structure of the drone charging safety early warning device provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below: like Figure 2 As shown, the drone charging safety warning device 20 includes: The acquisition module 21 is used to acquire the impedance change rate sequence of the charging point and the linear acceleration sequence of the drone; wherein, the charging point establishes a wireless inductive charging connection with the drone. The calculation module 22 is used to identify target data points in the linear acceleration sequence that meet the preset vibration awakening conditions, determine the analysis time window based on the target data points, and calculate the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window. The judgment module 23 is used to determine whether a physical misalignment during charging has occurred based on the spatiotemporal coupling degree. The prediction module 24 is used to predict the charging temperature curve based on the current charging power, impedance and ambient temperature if a charging physical misalignment occurs. The early warning module 25 is used to provide charging safety warnings based on the charging temperature curve.

[0058] In one possible implementation, the computing module 22 is used for: If a preset number of consecutive data points in a linear acceleration sequence are all greater than a preset vibration wake-up threshold, then the first data point among the preset number of data points will be determined as the target data point that meets the preset vibration wake-up condition. The analysis time window is determined based on the preset time before and after the target data point; The spatiotemporal coupling degree of the impedance rate of change sequence and the linear acceleration sequence within the time window is calculated and analyzed using a dynamic time warping algorithm.

[0059] In one possible implementation, the judgment module 23 is used to: determine that a charging physical misalignment has occurred if the spatiotemporal coupling degree is greater than a preset spatiotemporal coupling degree threshold.

[0060] In one possible implementation, the prediction module 24 is used for: Calculate the absolute difference between the impedance and the reference impedance. Based on the absolute difference and the charging power, look up the preset three-dimensional mapping table to obtain the physical misalignment distance. The physical misalignment distance, charging power, and measured ambient temperature are input into a pre-trained temperature prediction model to obtain the charging temperature curve output by the temperature prediction model.

[0061] In one possible implementation, before predicting the charging temperature profile based on the current charging power, impedance, and measured ambient temperature, the prediction module 24 is further configured to: Based on a pre-built thermodynamic lumped parameter model of the charging process, the virtual ambient temperature is obtained in real time through extended Kalman filtering algorithm; Calculate the residual between the measured ambient temperature and the virtual temperature, and obtain the cumulative deviation based on the residual using the cumulative sum control chart algorithm; When the cumulative deviation exceeds the preset confidence collapse threshold, the temperature sensor is identified as faulty, and the measured ambient temperature is replaced with the virtual ambient temperature.

[0062] In one possible implementation, when the accumulated deviation exceeds a preset confidence collapse threshold, the prediction module 24 is further configured to: The limiting power is determined based on the difference between the cumulative deviation and the confidence collapse threshold; wherein the limiting power is negatively correlated with the difference and is less than the rated power; Adjust the charging power output of the charging point to the limited power.

[0063] In one possible implementation, module 21 is used for: When establishing a wireless inductive charging connection between the charging point and the drone, a preset low-power detection pulse is injected into the transmitting coil of the charging point, and the transient impedance decay oscillation curve of the transmitting coil is obtained as the physical unclonable fingerprint of the drone. Based on the drone's MAC address, obtain the drone's reference physical unclonable fingerprint stored in the preset feature database; The similarity between a physically unclonable fingerprint and a reference physically unclonable fingerprint is calculated. If the similarity is lower than a preset similarity threshold, it is determined that a malicious power injection attack has occurred, and the charging point is prohibited from outputting charging power, as well as a malicious attack warning is issued.

[0064] In one possible implementation, the early warning module 25 is used for: If the charging temperature curve indicates a risk of thermal runaway, an edge data rescue strategy will be implemented after a charging safety warning is issued. The edge data rescue strategies include: A non-maskable interrupt is applied to the operating system kernel, forcibly suspending non-core threads to exclusively allocate computing resources to the data rescue thread; Through the data rescue thread, high-frequency transient electrical characteristics are losslessly compressed to generate death declaration data packets; Through an in-band escape mechanism, the death declaration data packet is frequency shift keyed and superimposed on the residual AC carrier, and then transmitted to the fireproof and explosion-proof solid-state memory of the drone.

[0065] This invention, through simultaneous acquisition of the impedance change rate sequence at the charging point and the linear acceleration sequence of the drone, identifies vibrations based on vibration wake-up conditions. By calculating and analyzing the spatiotemporal coupling degree of the two sets of sequences within a time window, it determines the physical misalignment of the charging coil. This allows for the early identification of minor misalignment faults before traditional temperature sensors show signs of temperature rise or thermal inertia. Furthermore, by combining multi-dimensional parameters such as charging power, impedance, and ambient temperature, it predicts temperature change curves and provides safety warnings, enabling early fault prediction and handling. This embodiment significantly improves the timeliness and reliability of fault identification, effectively preventing safety accidents such as overheating and equipment damage caused by coil misalignment, and ensuring the stable and safe operation of the drone's wireless inductive charging process.

[0066] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 3 in this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module in the various device embodiments described above.

[0067] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.

[0068] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.

[0069] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0070] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0071] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for early warning of drone charging safety, characterized in that, include: Obtain the impedance change rate sequence of the charging point and the linear acceleration sequence of the drone; wherein, the charging point establishes a wireless inductive charging connection with the drone; Identify target data points in the linear acceleration sequence that meet preset vibration awakening conditions, determine the analysis time window based on the target data points, and calculate the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window; Based on the spatiotemporal coupling degree, determine whether a physical misalignment during charging has occurred; If a physical misalignment occurs during charging, the charging temperature curve is predicted based on the current charging power, impedance, and ambient temperature. Based on the charging temperature curve, a charging safety warning is issued.

2. The method for early warning of drone charging safety according to claim 1, characterized in that, The process of identifying target data points in the linear acceleration sequence that meet preset vibration awakening conditions, determining an analysis time window based on the target data points, and calculating the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window includes: If a preset number of consecutive data points in the linear acceleration sequence are all greater than a preset vibration wake-up threshold, then the first data point among the preset number of data points is determined as the target data point that satisfies the preset vibration wake-up condition. The analysis time window is determined based on the preset time before and after the target data point; The spatiotemporal coupling degree of the impedance rate of change sequence and the linear acceleration sequence within the analysis time window is calculated using a dynamic time warping algorithm.

3. The drone charging safety early warning method according to claim 1, characterized in that, The step of determining whether a charging physical misalignment has occurred based on the spatiotemporal coupling degree includes: if the spatiotemporal coupling degree is greater than a preset spatiotemporal coupling degree threshold, then it is determined that a charging physical misalignment has occurred.

4. The drone charging safety early warning method according to claim 1, characterized in that, The method for predicting the charging temperature curve based on the current charging power, impedance, and measured ambient temperature includes: Calculate the absolute difference between the impedance and the reference impedance. Based on the absolute difference and the charging power, look up a preset three-dimensional mapping table to obtain the physical misalignment distance. The physical misalignment distance, the charging power, and the measured ambient temperature are input into a pre-trained temperature prediction model to obtain the charging temperature curve output by the temperature prediction model.

5. The method for early warning of drone charging safety according to any one of claims 1 to 4, characterized in that, Before predicting the charging temperature profile based on the current charging power, impedance, and measured ambient temperature, the following steps are also included: Based on a pre-built thermodynamic lumped parameter model of the charging process, the virtual ambient temperature is obtained in real time through extended Kalman filtering algorithm; Calculate the residual between the measured ambient temperature and the virtual temperature, and obtain the cumulative deviation based on the residual using a cumulative sum control chart algorithm; When the cumulative deviation exceeds the preset confidence collapse threshold, the temperature sensor is determined to be faulty, and the measured ambient temperature is replaced with the virtual ambient temperature.

6. The drone charging safety early warning method according to claim 5, characterized in that, When the accumulated deviation exceeds a preset confidence collapse threshold, the method further includes: The power limit is determined based on the difference between the cumulative deviation and the confidence collapse threshold; wherein the power limit is negatively correlated with the difference and is less than the rated power. The charging power output by the charging point is adjusted to the limit power.

7. The method for early warning of drone charging safety according to any one of claims 1 to 4, characterized in that, Also includes: When a wireless inductive charging connection is established between the charging point and the drone, a preset low-power detection pulse is injected into the transmitting coil of the charging point, and the transient impedance decay oscillation curve of the transmitting coil is obtained as the physical unclonable fingerprint of the drone. Based on the MAC address of the drone, obtain the reference physical unclonable fingerprint of the drone stored in the preset feature database; Calculate the similarity between the physical unclonable fingerprint and the reference physical unclonable fingerprint. If the similarity is lower than a preset similarity threshold, it is determined that a malicious power injection attack has occurred, and the charging point is prohibited from outputting charging power, and a malicious attack warning is issued.

8. The method for early warning of drone charging safety according to any one of claims 1 to 4, characterized in that, The charging safety warning based on the charging temperature curve includes: If, based on the charging temperature curve, it is determined that there is a risk of thermal runaway, then after issuing a charging safety warning, an edge data rescue strategy will also be executed. The edge data rescue strategy includes: A non-maskable interrupt is applied to the operating system kernel, forcibly suspending non-core threads to exclusively allocate computing resources to the data rescue thread; The data rescue thread performs lossless compression on the high-frequency transient electrical characteristics to generate a death declaration data packet. Through an in-band escape mechanism, the death declaration data packet is frequency shift keyed and superimposed on the residual AC carrier, and then transmitted to the fireproof and explosion-proof solid-state memory of the UAV.

9. A drone charging safety early warning device, characterized in that, include: The acquisition module is used to acquire the impedance change rate sequence of the charging point and the linear acceleration sequence of the drone; wherein the charging point establishes a wireless inductive charging connection with the drone. The calculation module is used to identify target data points in the linear acceleration sequence that meet the preset vibration awakening conditions, determine the analysis time window based on the target data points, and calculate the spatiotemporal coupling degree between the impedance change rate sequence and the linear acceleration sequence within the analysis time window. The judgment module is used to determine whether a physical misalignment during charging has occurred based on the spatiotemporal coupling degree. The prediction module is used to predict the charging temperature curve based on the current charging power, impedance, and ambient temperature if a charging physical misalignment occurs. The early warning module is used to provide charging safety warnings based on the charging temperature curve.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.

Images