Unmanned aerial vehicle battery health monitoring method and system based on intelligent charging management cabinet

Abstract

The invention provides an unmanned aerial vehicle battery health monitoring method and system based on an intelligent charging management cabinet, and relates to the technical field of battery monitoring, and the method comprises the steps: arranging a magnetic flux sensing detection unit in a charging cabin of the intelligent charging management cabinet, and recognizing a current density distortion region in an unmanned aerial vehicle battery; monitoring the stress distribution state of the battery shell of the unmanned aerial vehicle in real time, and calculating the internal damage degree of the battery of the unmanned aerial vehicle according to a stress-current coupling model; when the internal damage degree exceeds a preset threshold value, an active balance repairing mode of the intelligent charging management cabinet is triggered, uniform deposition of active substances in the unmanned aerial vehicle battery is promoted by generating mechanical vibration in the charging bin, and current in the unmanned aerial vehicle battery is guided to be redistributed through an alternating magnetic field generated by the three-axis magnetic induction coil; and dynamic repair and balance of the battery performance of the unmanned aerial vehicle are realized.

Description

Technical Field

[0001] The present invention relates to battery monitoring technology, and in particular to a method and system for monitoring the health of unmanned aerial vehicle batteries based on an intelligent charging management cabinet. Background Art

[0002] As a key power component of the drone system, the performance and life of drone batteries directly affect the endurance and safety of drones. With the continuous expansion of drone application scenarios, the battery usage environment is becoming more complex, and problems such as uneven charging and discharging and internal stress accumulation are becoming increasingly prominent, leading to accelerated battery performance degradation and even safety hazards.

[0003] Existing battery repair technologies mainly focus on passive balancing and active balancing. Passive balancing uses bypass resistors to consume the energy of high-power cells to achieve balancing, which has the problem of large energy loss. Although active balancing can achieve energy transfer and reuse, due to the lack of accurate perception of the internal state of the battery, the balancing effect is often not ideal. In addition, most existing repair methods are limited to electrical adjustments and fail to fully consider the impact of mechanical stress on battery performance.

[0004] Therefore, it is urgent to develop an intelligent charging management system that can monitor the internal state of the battery in real time, accurately assess the health status, and perform repair and adjustment in a timely manner based on the monitoring results. The system should have electro-mechanical coupling monitoring capabilities, be able to perceive the degradation characteristics of the battery from multiple dimensions, and provide accurate decision-making basis for subsequent repair strategies. At the same time, the system also needs to have an active repair function to achieve dynamic optimization of battery performance by collaboratively controlling mechanical vibration and magnetic field effects. Summary of the invention

[0005] The embodiments of the present invention provide a method and system for monitoring the health of a drone battery based on an intelligent charging management cabinet, which can solve the problems in the prior art.

[0006] According to a first aspect of an embodiment of the present invention, a method for monitoring the health of a drone battery based on a smart charging management cabinet is provided, comprising: a magnetic flux sensing detection unit is arranged in a charging compartment of the smart charging management cabinet, the magnetic flux sensing detection unit comprises an orthogonally arranged three-axis magnetic induction coil, the leakage magnetic flux change generated by the drone battery during the charging and discharging process is detected by the three-axis magnetic induction coil, a current density distribution diagram inside the drone battery is constructed according to the spatial distribution characteristics of the leakage magnetic flux change, and the current density distortion area inside the drone battery is identified; at the same time, a piezoelectric film stress sensing array is arranged on the charging compartment wall of the smart charging management cabinet to monitor the stress distribution state of the drone battery shell in real time and obtain deformation characteristic data of the drone battery. ; Based on the current density distribution diagram and the deformation characteristic data, a self-organizing mapping neural network is used to establish a stress-current coupling model of the drone battery. The stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery. The internal damage degree of the drone battery is calculated according to the stress-current coupling model; when the internal damage degree exceeds a preset threshold, the active balancing repair mode of the intelligent charging management cabinet is triggered, and mechanical vibration is generated in the charging compartment to promote the uniform deposition of active substances inside the drone battery, and the alternating magnetic field generated by the three-axis magnetic induction coil is used to guide the current redistribution inside the drone battery, thereby realizing dynamic repair and balance of the drone battery performance.

[0007] In an optional embodiment, the leakage magnetic flux changes generated by the drone battery during the charging and discharging process are detected by the three-axis magnetic induction coil, and the current density distribution map inside the drone battery is constructed according to the spatial distribution characteristics of the leakage magnetic flux changes. The identification of the current density distortion area inside the drone battery includes: the three-axis magnetic induction coil array is connected to the data acquisition unit through a signal conditioning circuit, and the signal conditioning circuit includes an operational amplifier and a low-pass filter; the leakage magnetic flux data generated by the drone battery during the charging and discharging process is collected, and the three-axis components of the leakage magnetic flux are obtained after orthogonal modulation and demodulation, including an X-axis component, a Y-axis component, and a Z-axis component; based on the three-axis components of the leakage magnetic flux, the inside of the drone battery is constructed as a three-dimensional grid model composed of a plurality of voxel units of equal size, and each voxel unit is regarded as an independent current source; a pre-constructed magnetic dipole model is used to establish a relationship equation between the current source in the voxel unit and the leakage magnetic flux, and the relationship equation includes a current density vector, an observation point position vector, and a source point position vector; a transfer matrix, a weighting matrix, and a weighting matrix are constructed. The objective function of the data item and the regularization term containing the regularization operator is obtained; the conjugate gradient method is used to solve the objective function to generate an initial three-dimensional current density distribution map; the gradient value of the initial three-dimensional current density distribution map is calculated, and the current density gradient value of each voxel unit is compared with the preset gradient threshold, and the area exceeding the preset gradient threshold is determined as a high gradient area; the voxel units in the high gradient area are subdivided using an octree structure, and each voxel unit is divided into eight sub-voxel units; the curve integral constraint based on the Ampere loop theorem is introduced into the encrypted grid model composed of the sub-voxel units, and the three-axis components of the leakage magnetic flux are re-solved based on the encrypted grid model, the curve integral constraint and the three-axis components of the leakage magnetic flux to generate an optimized three-dimensional current density distribution map; the current density mean value and the current density deviation coefficient of the local area in the optimized three-dimensional current density distribution map are calculated, and the local area exceeding the preset deviation threshold is marked as a current density distortion area, and the position, area and deviation coefficient of the current density distortion area are marked to form a final current density distortion distribution map inside the UAV battery.

[0008] In an optional embodiment, the current density mean and current density deviation coefficient of the local area in the optimized three-dimensional current density distribution map are calculated, the local area exceeding the preset deviation threshold is marked as the current density distortion area, and the position, area and deviation coefficient of the current density distortion area are marked to form the final current density distortion distribution map inside the UAV battery, including: dividing the optimized three-dimensional current density distribution map into multiple local analysis areas, each of the local analysis areas includes multiple basic voxel units, and calculating the adaptive weight coefficient based on the normalized distance from the basic voxel unit to the center of the area; using the adaptive weight coefficient to perform weighted calculation on the current density value of the basic voxel unit in each local analysis area to obtain the weighted current density mean, and at the same time calculating the weighted standard deviation of the current density value of each basic voxel unit in the local analysis area, combining the weighted current density means of all local analysis areas to calculate the global current density mean, and calculating the weighted current density of each local analysis area. The current density deviation coefficient is obtained by the ratio of the current density mean to the global current density mean; the local analysis area where the current density deviation coefficient exceeds the preset deviation threshold is marked as a potential distortion area, and the potential distortion area is used as a seed point. A regional growth criterion is constructed based on the current density deviation coefficient difference between adjacent local analysis areas, the current density spatial gradient value and the Euclidean distance between the regional center points, and the potential distortion area is iteratively grown according to the regional growth criterion to obtain the current density distortion area; the regional volume, center of gravity position coordinates, average distortion index and current density gradient tensor of the current density distortion area are extracted to construct a feature descriptor, and the position, area and deviation coefficient of the current density distortion area are calculated based on the feature descriptor; the position, area and deviation coefficient are marked at the corresponding position of the current density distortion area, and gradient colors are used to identify areas with different deviation coefficients to form a current density distortion distribution map of the internal battery of the drone with a distortion degree distribution.

[0009] In an optional embodiment, based on the current density distribution map and the deformation feature data, a stress-current coupling model of the drone battery is established using a self-organizing map neural network, the stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery, and calculating the internal damage degree of the drone battery according to the stress-current coupling model includes: performing maximum and minimum value normalization processing on the current density distribution map and the deformation feature data to obtain a normalized current density value and a normalized deformation value, combining the normalized current density value and the normalized deformation value with the corresponding spatial coordinates to construct a multidimensional input vector; constructing a two-layer self-organizing map neural network, including a competition layer and an output layer, the competition layer includes a competition neuron matrix, the output layer includes an output neuron matrix, using a Gaussian neighborhood function to calculate the topological distance between each neuron and the winning neuron, substituting the topological distance into a weight update function to calculate an adaptive learning rate, and using the adaptive learning rate to update the neuron. The element weights are iteratively updated to obtain a stress-current characteristic mapping matrix; the gradient vector of the stress-current characteristic mapping matrix on the spatial coordinates is calculated, the modulus of the gradient vector is multiplied by the cosine value of the angle between the stress gradient and the current gradient vector to obtain a stress-current coupling coefficient matrix, the equivalent stress value is introduced, and the stress-induced current density distortion factor matrix is ​​calculated in combination with the stress-current coupling coefficient matrix and the normalized current density; a damage evolution equation is established, the independent variables of the damage evolution equation include the stress-induced current density distortion factor matrix and the damage variable matrix, the damage evolution equation is numerically integrated in the time dimension to obtain a damage distribution matrix that characterizes the degree of damage of each region inside the battery; the maximum damage value and the damage change rate matrix in the damage distribution matrix are extracted, the maximum damage value, the damage change rate matrix and the battery health state parameter are input into a preset nonlinear mapping function, the battery remaining life prediction value is calculated, and the battery internal damage degree is evaluated according to the battery remaining life prediction value.

[0010] In an optional embodiment, the maximum damage value and the damage change rate matrix in the damage distribution matrix are extracted, the maximum damage value, the damage change rate matrix and the battery health status parameter are input into a preset nonlinear mapping function, and the battery remaining life prediction value is calculated. The degree of damage inside the battery is evaluated according to the battery remaining life prediction value, including: using a local binary pattern operator to perform texture analysis on the damage distribution matrix, extracting the spatial distribution characteristics of the damage area, calculating the gradient histogram features of each direction of the damage distribution matrix to obtain the directional information of the damage evolution, performing optical flow estimation on the damage values ​​at adjacent moments in the damage distribution matrix to obtain the motion characteristics of the damage expansion, and obtaining the spatial distribution The basic feature set for damage characterization is constructed by combining features, directional information and motion features; the basic feature set is input into a transfer learning framework including a feature extraction network, a domain adaptation layer and a prediction network, the feature extraction network adopts a dense connection structure to extract features, the domain adaptation layer reduces the feature distribution differences between different battery groups through adversarial training, and the prediction network combines residual connections and gating mechanisms for nonlinear mapping; short-term damage evolution trends, medium-term charge and discharge curve features and long-term cumulative effects of operating conditions are extracted respectively, and the time attention mechanism is used to adaptively weighted fuse features of different time scales, and the fused features are input into the transfer learning framework to obtain the battery damage degree assessment result.

[0011] In an optional embodiment, mechanical vibration is generated in the charging compartment to promote uniform deposition of active materials inside the drone battery, and the alternating magnetic field generated by the three-axis magnetic induction coil is used to guide the current redistribution inside the drone battery to achieve dynamic repair and balance of the drone battery performance, including: calculating initial values ​​of vibration parameters and magnetic field parameters according to the internal state of the drone battery, wherein the vibration parameters include vibration frequency, vibration amplitude and vibration phase, and the magnetic field parameters include magnetic field intensity, magnetic field direction and magnetic field gradient; according to the initial values ​​of the vibration parameters, controlling the arrayed piezoelectric ceramic vibrator to generate mechanical vibration in the charging compartment, collecting real-time vibration data of the mechanical vibration through an acceleration sensor, calculating the vibration energy density distribution based on the real-time vibration data, and adjusting the vibration parameters of the arrayed piezoelectric ceramic vibrator according to the vibration energy density distribution to generate a directional vibration field; according to the initial values ​​of the magnetic field parameters, controlling the orthogonally arranged magnetic induction coil array to generate an alternating magnetic field in the charging compartment, and collecting the alternating magnetic field through a Hall sensor The real-time magnetic field data of the drone is collected, the magnetic field intensity distribution is calculated based on the real-time magnetic field data, and the magnetic field parameters of the magnetic induction coil array are adjusted according to the magnetic field intensity distribution to generate a controllable magnetic field; the terminal voltage data, temperature data and internal resistance data of the drone battery under the action of the directional vibration field and the controllable magnetic field are collected, and the data are input into the state observer to obtain the drone battery state estimation value, and the drone battery state space model is established based on the drone battery state estimation value; a predictive controller is constructed according to the drone battery state space model, and the optimal control quantity is calculated with the uniformity of the drone battery internal state as the optimization target, and the vibration parameters and magnetic field parameters are updated respectively according to the optimal control quantity, and the array piezoelectric ceramic vibrator and the magnetic induction coil array are controlled to generate updated directional vibration field and controllable magnetic field; the capacity recovery index and the internal resistance uniformity index are calculated based on the drone battery state estimation value, and when the capacity recovery index and the internal resistance uniformity index reach the preset target threshold, the directional vibration field and the controllable magnetic field are turned off to complete the drone battery repair process.

[0012] In an optional embodiment, a predictive controller is constructed according to the drone battery state space model, an optimal control quantity is calculated with the uniformity of the internal state of the drone battery as the optimization target, vibration parameters and magnetic field parameters are respectively updated according to the optimal control quantity, and the array piezoelectric ceramic vibrator and the magnetic induction coil array are controlled to generate an updated directional vibration field and a controllable magnetic field, including: constructing a predictive control state equation including a state vector and a control input vector, the state vector including an electrolyte concentration distribution and an active material distribution, the control input vector including vibration parameters and magnetic field parameters, wherein the vibration parameters include vibration frequency, vibration amplitude and vibration phase, and the magnetic field parameters include magnetic field intensity, magnetic field direction and magnetic field gradient; based on the predictive control state equation, an objective function with the uniformity of the internal state of the battery as the optimization target is established, the objective function including a state weight matrix and a control weight matrix, the state weight matrix is ​​used to balance the importance of each state variable, and the control weight matrix is ​​used to balance the importance of each state variable. The heavy matrix is ​​used to weigh the control effect and the control cost; the vibration parameter constraint conditions and the magnetic field parameter constraint conditions are set, the vibration parameter constraint conditions include the vibration frequency range, the vibration amplitude range and the vibration phase range, and the magnetic field parameter constraint conditions include the magnetic field intensity range, the magnetic field direction range and the magnetic field gradient range; the quadratic programming method is used to solve the objective function with the vibration parameter constraint conditions and the magnetic field parameter constraint conditions to obtain the optimal control sequence in the prediction time domain; the first control quantity in the optimal control sequence is executed in each control cycle, and the array piezoelectric ceramic vibrator is controlled to generate a vibration field according to the first control quantity, and the magnetic induction coil array is controlled to generate a magnetic field at the same time; the real-time state data of the drone battery under the action of the vibration field and the magnetic field is collected, the state estimation is updated based on the real-time state data, and the objective function is re-solved according to the updated state estimation to obtain a new optimal control sequence to achieve rolling optimization control of the vibration field and the magnetic field.

[0013] According to a second aspect of an embodiment of the present invention, a drone battery health monitoring system based on an intelligent charging management cabinet is provided, comprising: a first unit, for setting a magnetic flux sensing detection unit in a charging compartment of the intelligent charging management cabinet, wherein the magnetic flux sensing detection unit comprises an orthogonally arranged three-axis magnetic induction coil, and detecting the leakage magnetic flux change generated by the drone battery during the charging and discharging process through the three-axis magnetic induction coil, constructing a current density distribution map inside the drone battery according to the spatial distribution characteristics of the leakage magnetic flux change, and identifying the current density distortion area inside the drone battery; at the same time, a piezoelectric film stress sensing array is set on the charging compartment wall of the intelligent charging management cabinet to monitor the stress distribution state of the drone battery shell in real time and obtain the deformation characteristic data of the drone battery; a second unit The first unit is used to establish a stress-current coupling model of the drone battery by using a self-organizing mapping neural network based on the current density distribution diagram and the deformation characteristic data. The stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery, and calculate the internal damage degree of the drone battery according to the stress-current coupling model; the third unit is used to trigger the active balancing repair mode of the intelligent charging management cabinet when the internal damage degree exceeds a preset threshold, promote the uniform deposition of active materials inside the drone battery by generating mechanical vibration in the charging compartment, and guide the current redistribution inside the drone battery through the alternating magnetic field generated by the three-axis magnetic induction coil, so as to realize dynamic repair and balance of the drone battery performance.

[0014] According to a third aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0015] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0016] In this embodiment, by combining the internal current density distribution map of the battery obtained by the magnetic flux sensing detection unit and the battery shell deformation characteristic data obtained by the piezoelectric film stress sensing array, the internal state of the drone battery can be more comprehensively and accurately perceived, potential battery damage can be identified, and the error caused by relying on a single parameter for judgment can be avoided. Accurately identify the current density distortion area inside the battery and the stress distribution state of the battery shell to achieve refined monitoring of the battery health state. A stress-current coupling model of the drone battery is established using a self-organizing map neural network, and the stress distribution inside the battery is associated with the current distribution. The internal damage degree of the battery is calculated using this model, which realizes a quantitative assessment of the battery health state, overcomes the shortcomings of the traditional method that relies on experience judgment, and makes the battery health state assessment more objective and scientific. When the internal damage degree of the battery exceeds the preset threshold, the intelligent charging management cabinet can automatically start the active equalization repair mode, and through the combined action of mechanical micro-vibration and alternating magnetic field, promote the uniform deposition of active substances inside the battery and the redistribution of current, realize dynamic repair and balance of battery performance, extend battery life, and reduce battery maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method for monitoring the health of a drone battery based on an intelligent charging management cabinet according to an embodiment of the present invention; Figure 2 The figure is a schematic diagram of the structure of a drone battery health monitoring system based on an intelligent charging management cabinet according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0020] Figure 1 FIG. 1 is a flow chart of a method for monitoring the health of a drone battery based on an intelligent charging management cabinet according to an embodiment of the present invention. Figure 1As shown, the method includes: S101. A magnetic flux sensing detection unit is set in the charging compartment of the intelligent charging management cabinet, and the magnetic flux sensing detection unit includes an orthogonally arranged three-axis magnetic induction coil. The leakage magnetic flux changes generated by the drone battery during the charging and discharging process are detected by the three-axis magnetic induction coil. According to the spatial distribution characteristics of the leakage magnetic flux changes, a current density distribution map inside the drone battery is constructed to identify the current density distortion area inside the drone battery; at the same time, a piezoelectric film stress sensing array is set on the charging compartment wall of the intelligent charging management cabinet to monitor the stress distribution state of the drone battery shell in real time and obtain the deformation characteristic data of the drone battery; S102. Based on the current density distribution map and The deformation feature data uses a self-organizing mapping neural network to establish a stress-current coupling model of the drone battery. The stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery. The internal damage degree of the drone battery is calculated according to the stress-current coupling model; S103. When the internal damage degree exceeds a preset threshold, the active balancing repair mode of the intelligent charging management cabinet is triggered, and mechanical vibration is generated in the charging compartment to promote the uniform deposition of active substances inside the drone battery, and the alternating magnetic field generated by the three-axis magnetic induction coil is used to guide the current redistribution inside the drone battery, thereby realizing dynamic repair and balance of the drone battery performance.

[0021] Among them, the charging compartment of the intelligent charging management cabinet is equipped with a magnetic flux sensing detection unit and a piezoelectric film stress sensing array for real-time monitoring of the health status of the drone battery. The magnetic flux sensing detection unit is composed of orthogonally arranged three-axis magnetic induction coils, which can detect the changes in leakage magnetic flux generated by the battery during the charging and discharging process. By analyzing the spatial distribution characteristics of these changes, the current density distribution map inside the battery can be constructed, and the current density distortion area can be identified. For example, by analyzing the changes in the magnetic induction intensity Bx, By, and Bz in the three directions of X, Y, and Z, the current density vector J=(Jx, Jy, Jz) can be determined and mapped to a two-dimensional plane to form a current density distribution map. Assuming that during a certain charging process, the Jz value of the top area of ​​the battery is significantly higher than that of other areas, it can be determined that there is current density distortion in this area.

[0022] At the same time, the piezoelectric film stress sensor array on the charging compartment wall monitors the stress distribution state of the battery shell in real time and obtains the deformation characteristic data of the battery. For example, by measuring the voltage output of each sensor on the piezoelectric film array, the stress values ​​at different positions of the battery shell can be obtained and converted into deformation data. For example, if the stress value of the top area of ​​the battery is significantly higher than that of other areas, it can be determined that there is deformation in this area.

[0023] Using the collected current density distribution map and deformation feature data, a self-organizing map neural network is used to establish a stress-current coupling model of the battery. Among them, the self-organizing map (SOM) is an unsupervised learning neural network structure used for pattern recognition and data dimension reduction. In the construction of the battery stress-current coupling model, the SOM neural network maps the complex relationship between stress and current distribution into a two-dimensional space by learning the current density distribution map and deformation feature data to reveal the coupling relationship between stress and current distribution inside the battery. The network can adaptively organize and classify different stress and current distribution patterns to establish a reliable stress-current coupling model. This model is used to characterize the interaction between stress distribution and current distribution inside the battery. For example, by taking current density data and deformation data as inputs to the neural network, a model that can reflect the relationship between the two can be obtained by training the network. It is assumed that the higher the model output value, the greater the degree of damage inside the battery.

[0024] According to the established stress-current coupling model, the internal damage degree of the battery can be calculated. For example, the actual measured current density data and deformation data are input into the trained neural network model to obtain an output value, which represents the internal damage degree of the battery. When the internal damage degree of the battery exceeds the preset threshold, the intelligent charging management cabinet will automatically trigger the active equalization repair mode. Mechanical micro-vibrations of a specific frequency, such as 100Hz micro-vibrations, will be generated in the charging compartment to promote the uniform deposition of active substances inside the battery. At the same time, the three-axis magnetic induction coil will generate an alternating magnetic field, such as an alternating magnetic field with a frequency of 50Hz and an intensity of 10mT, to guide the redistribution of the current inside the battery and realize the dynamic repair and balance of the battery performance. For example, after a period of repair, the current density distribution and deformation data of the battery are measured again and input into the stress-current coupling model, and the output value is 0.6, which is lower than the preset threshold, indicating that the battery performance has been repaired.

[0025] In this embodiment, by real-time monitoring and active repair of battery internal damage, the battery life can be effectively extended. By balancing the deposition of active substances and current distribution inside the battery, the battery charging and discharging efficiency and capacity can be improved. By timely discovering and repairing potential problems inside the battery, safety hazards such as overcharging and over-discharging of the battery can be effectively prevented.

[0026] In an optional embodiment, the leakage magnetic flux change generated by the drone battery during the charging and discharging process is detected by the three-axis magnetic induction coil, and the current density distribution map inside the drone battery is constructed according to the spatial distribution characteristics of the leakage magnetic flux change. The identification of the current density distortion area inside the drone battery includes: the three-axis magnetic induction coil array is connected to the data acquisition unit through a signal conditioning circuit, and the signal conditioning circuit includes an operational amplifier and a low-pass filter; the leakage magnetic flux data generated by the drone battery during the charging and discharging process is collected, and the three-axis components of the leakage magnetic flux are obtained after orthogonal modulation and demodulation processing, including an X-axis component, a Y-axis component, and a Z-axis component; based on the three-axis components of the leakage magnetic flux, the inside of the drone battery is constructed as a three-dimensional grid model composed of a plurality of voxel units of equal size, and each voxel unit is regarded as an independent current source; a pre-constructed magnetic dipole model is used to establish a relationship equation between the current source in the voxel unit and the leakage magnetic flux, and the relationship equation includes a current density vector, an observation point position vector, and a source point position vector; a transfer matrix, a weighted matrix, and a weighted matrix are constructed. The objective function of the regularization term of the matrix and the regularization term containing the regularization operator is obtained; the conjugate gradient method is used to solve the objective function to generate an initial three-dimensional current density distribution map; the gradient value of the initial three-dimensional current density distribution map is calculated, the current density gradient value of each voxel unit is compared with the preset gradient threshold, and the area exceeding the preset gradient threshold is determined as a high gradient area; the voxel units in the high gradient area are subdivided using an octree structure, and each voxel unit is divided into eight sub-voxel units; the curve integral constraint condition based on the Ampere loop theorem is introduced into the encrypted grid model composed of the sub-voxel units, and the optimized three-dimensional current density distribution map is generated based on the encrypted grid model, the curve integral constraint condition and the three-axis component of the leakage magnetic flux. The average current density and the current density deviation coefficient of the local area in the optimized three-dimensional current density distribution map are calculated, and the local area exceeding the preset deviation threshold is marked as a current density distortion area, and the position, area and deviation coefficient of the current density distortion area are marked to form a final current density distortion distribution map inside the UAV battery.

[0027] For example, during the charging and discharging process of the drone battery, the uneven internal current distribution may cause local overheating and even lead to safety accidents. In order to effectively detect the current density distortion area inside the battery, this embodiment proposes a method for detecting the internal current density distortion of the drone battery based on a three-axis magnetic induction coil.

[0028] First, build a detection system. Arrange multiple three-axis magnetic induction coils in a certain space to form an array and place them around the drone battery. The three axes of each coil sense the magnetic field changes in the X, Y, and Z directions respectively. The output signal of the coil array is amplified and filtered by the signal conditioning circuit, and then connected to the data acquisition unit for data acquisition. In the signal conditioning circuit, the operational amplifier amplifies the weak magnetic field induction signal, and the low-pass filter removes high-frequency noise interference. For example, a low-pass filter with a cutoff frequency of 1kHz can be selected. The data acquisition unit converts the analog signal into a digital signal and stores it in the computer.

[0029] Next, data preprocessing is performed. The collected data is the leakage flux data generated during the charging and discharging process of the drone battery, which needs to be processed by orthogonal modulation and demodulation. For example, if amplitude modulation is used, the collected signal needs to be multiplied by the reference signal, and then the high-frequency component is filtered out through a low-pass filter to obtain the X, Y, and Z axial components of the leakage flux. Assuming that the collected data is 1000 sampling points and the sampling frequency is 10kHz, after orthogonal modulation and demodulation, the X, Y, and Z axial components of the leakage flux at 1000 time points can be obtained.

[0030] Then, a battery model and relationship equations are constructed. The internal space of the drone battery is discretized into a three-dimensional grid model consisting of multiple voxel units of equal size. The size of each voxel unit can be set to 1mmx1mmx1mm, for example. Each voxel unit is regarded as an independent current source. Using the pre-built magnetic dipole model, the relationship between the current source in the voxel unit and the leakage magnetic flux measured at the coil is established. This relationship equation describes the relationship between the current density vector in each voxel unit, the coil (observation point) position vector, and the voxel unit (source point) position vector.

[0031] Subsequently, the current density distribution is solved. An objective function is constructed, which consists of two parts: a data term and a regularization term. The data term reflects the difference between the calculated leakage flux and the actually measured leakage flux, and the regularization term is used to constrain the smoothness of the current density distribution. The conjugate gradient method is used to iteratively solve the objective function to obtain an initial three-dimensional current density distribution map. For example, after 100 iterations, an initial current density distribution map containing 10x10x10 voxel units is obtained.

[0032] Next, mesh encryption and optimization are performed. The current density gradient value of each voxel unit in the initial three-dimensional current density distribution map is calculated. The gradient value is compared with the preset gradient threshold (for example, 0.1A / m²). The area where the gradient value exceeds the threshold is considered to be a high-gradient area and mesh encryption is required. The voxel units in the high-gradient area are subdivided using an octree structure, and each voxel unit is divided into eight smaller sub-voxel units. In the encrypted mesh model, the curve integral constraint based on Ampere's circuit theorem is introduced, the objective function is constructed again and the conjugate gradient method is used to solve it to obtain the optimized three-dimensional current density distribution map.

[0033] Finally, identify the current density distortion area. Calculate the current density mean and current density deviation coefficient (e.g., standard deviation) of the local area (e.g., 3x3x3 voxel unit) in the optimized three-dimensional current density distribution map. Mark the local area whose deviation coefficient exceeds the preset deviation threshold (e.g., 0.05) as the current density distortion area, and annotate its position, area, and deviation coefficient, and finally form the current density distortion distribution map inside the drone battery. For example, in a local area, the current density mean is 1A / m², the deviation coefficient is 0.1, which exceeds the preset threshold 0.05, then the area is marked as the current density distortion area.

[0034] In this embodiment, by mesh encryption and the introduction of curve integral constraints, the current density distortion area can be identified more accurately, thereby improving the detection accuracy. The generated three-dimensional current density distortion distribution map can intuitively display the current density distribution inside the battery, making it convenient for users to quickly understand the battery status. By detecting the current density distortion area, the potential failure risk of the battery can be warned in advance to avoid accidents caused by battery failure.

[0035] In an optional embodiment, the current density mean and current density deviation coefficient of the local area in the optimized three-dimensional current density distribution map are calculated, the local area exceeding the preset deviation threshold is marked as the current density distortion area, and the position, area and deviation coefficient of the current density distortion area are marked to form the final current density distortion distribution map inside the UAV battery, including: dividing the optimized three-dimensional current density distribution map into multiple local analysis areas, each of the local analysis areas includes multiple basic voxel units, and calculating the adaptive weight coefficient based on the normalized distance from the basic voxel unit to the center of the area; using the adaptive weight coefficient to perform weighted calculation on the current density value of the basic voxel unit in each local analysis area to obtain the weighted current density mean, and at the same time calculating the weighted standard deviation of the current density value of each basic voxel unit in the local analysis area, combining the weighted current density means of all local analysis areas to calculate the global current density mean, and calculating the weighted current density of each local analysis area. The current density deviation coefficient is obtained by the ratio of the current density mean to the global current density mean; the local analysis area where the current density deviation coefficient exceeds the preset deviation threshold is marked as a potential distortion area, and the potential distortion area is used as a seed point. A regional growth criterion is constructed based on the current density deviation coefficient difference between adjacent local analysis areas, the current density spatial gradient value and the Euclidean distance between the regional center points, and the potential distortion area is iteratively grown according to the regional growth criterion to obtain the current density distortion area; the regional volume, center of gravity position coordinates, average distortion index and current density gradient tensor of the current density distortion area are extracted to construct a feature descriptor, and the position, area and deviation coefficient of the current density distortion area are calculated based on the feature descriptor; the position, area and deviation coefficient are marked at the corresponding position of the current density distortion area, and gradient colors are used to identify areas with different deviation coefficients to form a current density distortion distribution map of the internal battery of the drone with a distortion degree distribution.

[0036] For example, to obtain the current density distortion distribution map inside the drone battery, an optimized three-dimensional current density distribution map is first required as input. The distribution map is composed of many tiny voxel units, each of which stores the corresponding current density value.

[0037] Next, the entire three-dimensional current density distribution map is divided into multiple local analysis areas. Each local analysis area contains multiple voxel units, for example, a 10x10x10 cube area contains 1000 voxel units. In order to avoid sudden changes in current density at the boundaries of the area, an adaptive weight coefficient is used for smoothing. The distance from each voxel unit to the center of its local analysis area is calculated, and the distance is normalized, for example, the distance is divided by the maximum distance of the area to obtain a value between 0 and 1. Then, based on the normalized distance value, the adaptive weight coefficient of each voxel unit is calculated using methods such as Gaussian function or inverse distance weighting. The closer the voxel unit is to the center of the area, the greater its weight coefficient.

[0038] The calculated adaptive weight coefficients are used to weight the current density values ​​of all voxel units in each local analysis area to obtain the weighted current density mean of the area. At the same time, the weighted standard deviation of the current density values ​​of all voxel units in the area is calculated. The weighted current density means of all local analysis areas are averaged to obtain the global current density mean. The weighted current density mean of each local analysis area is divided by the global current density mean to obtain the current density deviation coefficient of the area.

[0039] Set a preset deviation threshold, such as 0.2. Mark the local analysis area where the current density deviation coefficient exceeds the threshold as a potential distortion area. Assuming that the current density deviation coefficient of a certain area is 0.25, which exceeds the preset threshold of 0.2, the area is marked as a potential distortion area. Next, these potential distortion areas are used as seed points for region growth. The region growth criterion is based on the difference in current density deviation coefficients between adjacent local analysis areas, the current density spatial gradient value, and the distance between the center points of the area. For example, if the difference in current density deviation coefficients between adjacent areas is less than 0.05, the current density spatial gradient value is less than 0.1, and the distance between the center points of the area is less than the length of 5 voxel units, the adjacent area is merged into the current distortion area. Repeat this process until there are no adjacent areas that meet the growth criterion, and finally obtain the current density distortion area.

[0040] For each current density distortion region, extract its regional volume (number of voxel units), center of gravity position coordinates (average value of all voxel unit coordinates), average distortion index (average value of current density deviation coefficients of all voxel units in the region) and current density gradient tensor. This information constitutes the feature descriptor of the distorted region. Based on the feature descriptor, calculate the position, area and deviation coefficient of the distorted region. For example, the area of ​​the region can be estimated by volume and shape. Finally, the calculated position, area and deviation coefficient of the distorted region are marked at the corresponding position on the three-dimensional current density distribution map. Use gradient colors to identify regions with different deviation coefficients. For example, the larger the deviation coefficient, the darker the color, thereby forming a current density distortion distribution map of the drone battery with a degree of distortion distribution. Assuming that the deviation coefficient of a distorted region is 0.3, it is marked as dark red; if the deviation coefficient of another distorted region is 0.22, it is marked as light red.

[0041] In this embodiment, the current density distortion area inside the battery can be accurately identified and located through adaptive weight coefficients and regional growing algorithms, avoiding the influence of noise interference and boundary effects on traditional methods. By calculating indicators such as current density deviation coefficient and gradient tensor, the severity of the distortion area can be quantitatively evaluated, providing a reliable basis for battery design and optimization. By marking the position, area and deviation coefficient of the distortion area on the three-dimensional current density distribution map, and using gradient colors to mark areas with different deviation coefficients, the distribution of current density distortion inside the battery can be intuitively displayed, which is convenient for users to understand and analyze.

[0042] In an optional embodiment, based on the current density distribution map and the deformation feature data, a stress-current coupling model of the drone battery is established using a self-organizing map neural network, and the stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery. Calculating the internal damage degree of the drone battery according to the stress-current coupling model includes: performing maximum and minimum value normalization processing on the current density distribution map and the deformation feature data to obtain a normalized current density value and a normalized deformation value, combining the normalized current density value and the normalized deformation value with the corresponding spatial coordinates to construct a multidimensional input vector; constructing a two-layer self-organizing map neural network, which includes a competition layer and an output layer, the competition layer includes a competition neuron matrix, and the output layer includes an output neuron matrix, using a Gaussian neighborhood function to calculate the topological distance between each neuron and the winning neuron, substituting the topological distance into a weight update function to calculate an adaptive learning rate, and using the adaptive learning rate to update the neuron. The element weights are iteratively updated to obtain a stress-current characteristic mapping matrix; the gradient vector of the stress-current characteristic mapping matrix on the spatial coordinates is calculated, the modulus of the gradient vector is multiplied by the cosine value of the angle between the stress gradient and the current gradient vector to obtain a stress-current coupling coefficient matrix, the equivalent stress value is introduced, and the stress-induced current density distortion factor matrix is ​​calculated in combination with the stress-current coupling coefficient matrix and the normalized current density; a damage evolution equation is established, the independent variables of the damage evolution equation include the stress-induced current density distortion factor matrix and the damage variable matrix, the damage evolution equation is numerically integrated in the time dimension to obtain a damage distribution matrix that characterizes the degree of damage of each region inside the battery; the maximum damage value and the damage change rate matrix in the damage distribution matrix are extracted, the maximum damage value, the damage change rate matrix and the battery health state parameter are input into a preset nonlinear mapping function, the battery remaining life prediction value is calculated, and the battery internal damage degree is evaluated according to the battery remaining life prediction value.

[0043] Exemplarily, first, the current density distribution diagram and deformation characteristic data of the drone battery under different charge and discharge states are collected. The current density distribution diagram can be measured by a current sensor array, and the deformation characteristic data can be measured by a fiber grating sensor or a strain gauge. For example, the current sensor array is arranged inside the battery, and a sensor is set every 1 mm to collect the current density data of the battery at different discharge rates. At the same time, the fiber grating sensor is pasted on the surface of the battery to collect the deformation data of the battery under different charge and discharge states.

[0044] Then, the collected current density distribution map and deformation feature data are normalized to the maximum and minimum values. For example, assuming that the current density value at a certain location is 10A / m², the maximum current density value is 20A / m², and the minimum current density value is 2A / m², then the normalized current density value at this location is (10-2) / (20-2)=0.44. Similarly, the deformation feature data is normalized to obtain a normalized deformation value. The normalized current density value, the normalized deformation value, and the corresponding spatial coordinates (e.g., x, y, z coordinates) are combined to construct a multidimensional input vector. For example, assuming that the normalized current density value at a certain location is 0.44, the normalized deformation value is 0.6, and the spatial coordinates are (1, 2, 3), then the multidimensional input vector at this location is [0.44, 0.6, 1, 2, 3].

[0045] Next, a two-layer self-organizing map neural network is constructed, which includes a competition layer and an output layer. The competition layer contains a competition neuron matrix, and the output layer contains an output neuron matrix. The initial parameters of the network are set, such as the number of neurons in the competition layer, the number of neurons in the output layer, the learning rate, the neighborhood radius, etc. The topological distance between each neuron and the winning neuron is calculated using the Gaussian neighborhood function. The topological distance is substituted into the weight update function to calculate the adaptive learning rate. The neuron weights are iteratively updated using the adaptive learning rate until the network converges. Finally, a stress-current feature mapping matrix is ​​obtained, which reflects the mapping relationship between stress distribution and current distribution.

[0046] Subsequently, the gradient vector of the stress-current characteristic mapping matrix in the spatial coordinates is calculated. The modulus of the gradient vector is multiplied by the cosine value of the angle between the stress gradient vector and the current gradient vector to obtain the stress-current coupling coefficient matrix. The equivalent stress value is introduced, and the stress-current coupling coefficient matrix and the normalized current density are combined to calculate the stress-induced current density distortion factor matrix. This matrix reflects the degree of influence of stress on the current density distribution.

[0047] A damage evolution equation is established, and the independent variables of the equation include the stress-induced current density distortion factor matrix and the damage variable matrix. The damage evolution equation is numerically integrated in the time dimension to obtain a damage distribution matrix that characterizes the degree of damage in each region inside the battery. For example, a 2x2 damage distribution matrix can be expressed as [[0.1, 0.2], [0.3, 0.4]], where each element represents the degree of damage in the region.

[0048] Finally, extract the maximum damage value and damage change rate matrix in the damage distribution matrix. Input the maximum damage value, damage change rate matrix and battery health status parameters (e.g., number of cycles, capacity attenuation rate, etc.) into a preset nonlinear mapping function (e.g., support vector machine, neural network, etc.) to calculate the remaining battery life prediction value. Evaluate the degree of damage inside the battery based on the remaining battery life prediction value.

[0049] In this embodiment, by establishing a stress-current coupling model, the damage evolution process inside the battery can be more accurately characterized, thereby improving the accuracy of battery life prediction. This method can monitor the stress distribution, current distribution and degree of damage inside the battery in real time, and provide data support for battery health management. Optimized battery design and use strategy: By analyzing the stress-current coupling model, the weak links in the battery design can be identified, and the battery use strategy can be optimized, thereby extending the battery life.

[0050] In an optional embodiment, the maximum damage value and the damage change rate matrix in the damage distribution matrix are extracted, the maximum damage value, the damage change rate matrix and the battery health status parameter are input into a preset nonlinear mapping function, and the battery remaining life prediction value is calculated. The degree of damage inside the battery is evaluated according to the battery remaining life prediction value, including: using a local binary pattern operator to perform texture analysis on the damage distribution matrix, extracting the spatial distribution characteristics of the damaged area, calculating the gradient histogram features of each direction of the damage distribution matrix to obtain the directional information of the damage evolution, performing optical flow estimation on the damage values ​​at adjacent moments in the damage distribution matrix to obtain the motion characteristics of the damage expansion, and obtaining the spatial distribution The basic feature set for damage characterization is constructed by combining features, directional information and motion features; the basic feature set is input into a transfer learning framework including a feature extraction network, a domain adaptation layer and a prediction network, the feature extraction network adopts a dense connection structure to extract features, the domain adaptation layer reduces the feature distribution differences between different battery groups through adversarial training, and the prediction network combines residual connections and gating mechanisms for nonlinear mapping; short-term damage evolution trends, medium-term charge and discharge curve features and long-term cumulative effects of operating conditions are extracted respectively, and the time attention mechanism is used to adaptively weighted fuse features of different time scales, and the fused features are input into the transfer learning framework to obtain the battery damage degree assessment result.

[0051] Exemplarily, first, obtain the damage distribution matrix of the battery. The matrix can be obtained by the electrochemical impedance spectroscopy (EIS) of the battery or other technologies that can reflect the internal state distribution of the battery. For example, by performing an EIS test on the battery and spatially resolving the test results, a matrix reflecting the impedance distribution of different regions inside the battery can be obtained, which can be used as a damage distribution matrix. Assume that the damage distribution matrix obtained is a 10x10 matrix, and each element in the matrix represents the damage value of the corresponding area inside the battery. Then, extract the maximum damage value and the damage change rate matrix in the damage distribution matrix. The maximum damage value is the maximum value of all elements in the matrix. For example, the maximum value in the example matrix is ​​0.28. The damage change rate matrix can be obtained by calculating the difference between the elements of the corresponding positions of the two damage distribution matrices at adjacent moments.

[0052] Then, the maximum damage value, the damage change rate matrix, and the battery health status parameters (such as capacity decay rate, internal resistance increase, etc.) are input into a preset nonlinear mapping function to calculate the remaining battery life prediction value. The nonlinear mapping function can be a model trained based on a machine learning algorithm, such as a neural network model.

[0053] In order to evaluate the degree of damage inside the battery, it is necessary to extract features from the damage distribution matrix. First, the local binary pattern (LBP) operator is used to perform texture analysis on the damage distribution matrix to extract the spatial distribution features of the damage area. For example, the size relationship between each element and its eight surrounding adjacent elements can be calculated, and the result can be encoded into a binary number to obtain the LBP feature of the element. Then, the directional gradient histogram (HOG) features of the damage distribution matrix are calculated to obtain the directional information of the damage evolution. For example, the gradient of each element in the matrix in the horizontal and vertical directions can be calculated, and the gradient values ​​can be counted in different directional intervals to form a gradient histogram. Next, the damage values ​​at adjacent moments in the damage distribution matrix are estimated by optical flow to obtain the motion features of the damage extension. Optical flow estimation can reflect the motion direction and speed of the damage area. Finally, the obtained spatial distribution features (LBP features), directional information (HOG features) and motion features (optical flow features) are combined to construct the basic feature set for damage characterization.

[0054] The extracted basic feature set is input into a transfer learning framework. The framework includes a feature extraction network, a domain adaptation layer, and a prediction network. The feature extraction network uses a densely connected structure to extract features. The domain adaptation layer uses adversarial training to reduce the difference in feature distribution between different battery packs. The prediction network combines residual connections and gating mechanisms for nonlinear mapping and outputs the battery damage assessment results.

[0055] In order to further improve the evaluation accuracy, the short-term damage evolution trend, the medium-term charge and discharge curve characteristics and the long-term use condition cumulative effect are extracted respectively. The short-term damage evolution trend can be obtained by analyzing the damage change rate in the recent period. The medium-term charge and discharge curve characteristics can reflect the performance changes of the battery in different charge and discharge stages. The long-term use condition cumulative effect can reflect the aging degree of the battery during long-term use. The temporal attention mechanism is used to adaptively weight the features of different time scales, and the fused features are input into the transfer learning framework to obtain the final battery damage degree evaluation result.

[0056] In this embodiment, the accuracy of the remaining battery life prediction is improved: by combining the damage distribution matrix, the damage change rate matrix and the battery health status parameters, and using a nonlinear mapping function for prediction, the battery aging state can be more accurately reflected, thereby improving the prediction accuracy. A refined assessment of the degree of damage inside the battery is achieved: by performing multi-dimensional feature extraction on the damage distribution matrix and combining it with a transfer learning framework, the damage state inside the battery can be more comprehensively characterized to achieve a refined assessment. The generalization ability of the model is improved: through the domain adaptation layer, the feature distribution differences between different battery packs can be reduced, thereby improving the generalization ability of the model and making it applicable to different types of batteries.

[0057] In an optional embodiment, mechanical vibration is generated in the charging compartment to promote the uniform deposition of active materials inside the drone battery, and the alternating magnetic field generated by the three-axis magnetic induction coil is used to guide the current redistribution inside the drone battery to achieve dynamic repair and balance of the drone battery performance, including: calculating initial values ​​of vibration parameters and magnetic field parameters according to the internal state of the drone battery, wherein the vibration parameters include vibration frequency, vibration amplitude and vibration phase, and the magnetic field parameters include magnetic field intensity, magnetic field direction and magnetic field gradient; according to the initial values ​​of the vibration parameters, controlling the arrayed piezoelectric ceramic vibrator to generate mechanical vibration in the charging compartment, collecting real-time vibration data of the mechanical vibration through an acceleration sensor, calculating the vibration energy density distribution based on the real-time vibration data, and adjusting the vibration parameters of the arrayed piezoelectric ceramic vibrator according to the vibration energy density distribution to generate a directional vibration field; according to the initial values ​​of the magnetic field parameters, controlling the orthogonally arranged magnetic induction coil array to generate an alternating magnetic field in the charging compartment, collecting the alternating magnetic field through a Hall sensor The invention relates to a method for controlling the battery state of a drone, and a method for controlling the battery state of a drone, wherein the battery state of the drone is controlled by the state observer, and the battery state of the drone is controlled by the state observer, and the battery state of the drone is controlled by the state observer. The method comprises the following steps: collecting the real-time magnetic field data of the field, calculating the magnetic field intensity distribution based on the real-time magnetic field data, adjusting the magnetic field parameters of the magnetic induction coil array according to the magnetic field intensity distribution, and generating a controllable magnetic field; collecting the terminal voltage data, temperature data and internal resistance data of the drone battery under the action of the directional vibration field and the controllable magnetic field, and inputting the data into the state observer to obtain the estimated value of the drone battery state, and establishing a drone battery state space model based on the estimated value of the drone battery state; constructing a predictive controller according to the drone battery state space model, calculating the optimal control quantity with the uniformity of the internal state of the drone battery as the optimization target, updating the vibration parameters and the magnetic field parameters respectively according to the optimal control quantity, and controlling the array piezoelectric ceramic vibrator and the magnetic induction coil array to generate the updated directional vibration field and controllable magnetic field; calculating the capacity recovery index and the internal resistance uniformity index based on the estimated value of the drone battery state, and when the capacity recovery index and the internal resistance uniformity index reach the preset target threshold, closing the directional vibration field and the controllable magnetic field to complete the drone battery repair process.

[0058] The drone battery repair device includes a charging compartment, which is equipped with an array of piezoelectric ceramic vibrators, an orthogonally arranged magnetic induction coil array, an acceleration sensor and a Hall sensor. The device is also connected to a control unit and a state observer.

[0059] First, the initial state information of the battery is obtained by performing preliminary tests on the drone battery, such as measuring its terminal voltage, internal resistance, and temperature. For example, the battery terminal voltage is measured to be 3.7V, the internal resistance is 20mΩ, and the temperature is 25°C. Based on this initial state information, the initial values ​​of the vibration parameters and magnetic field parameters are calculated. For example, the initial vibration frequency is set to 10kHz, the amplitude is 1μm, and the phase is 0°; the initial magnetic field strength is set to 10mT, the direction is the positive direction of the X-axis, and the gradient is 1mT / mm.

[0060] The control unit controls the array piezoelectric ceramic vibrator to work according to the calculated initial vibration parameters, generating mechanical micro-vibrations of a specific frequency in the charging compartment. The acceleration sensor collects vibration data in real time, for example, the vibration acceleration values ​​of each piezoelectric ceramic vibrator. The control unit calculates the vibration energy density distribution based on these real-time vibration data. For example, if it is found that the vibration energy density is lower in the edge area of ​​the battery, the control unit will adjust the vibration parameters of some piezoelectric ceramic vibrators, such as increasing the amplitude of the vibrators in the edge area, so that the vibration energy density distribution is more uniform, forming a directional vibration field.

[0061] At the same time, the control unit controls the orthogonally arranged magnetic induction coil array according to the initial magnetic field parameters to generate an alternating magnetic field in the charging compartment. The Hall sensor collects magnetic field data in real time, such as the magnetic field strength values ​​at each position. The control unit calculates the magnetic field strength distribution based on these real-time data. For example, if the magnetic field strength is weak in the center area of ​​the battery, the control unit will adjust the current of the corresponding coil to enhance the magnetic field strength in the center area, thereby forming a controllable magnetic field.

[0062] Under the action of directional vibration field and controllable magnetic field, the terminal voltage, temperature and internal resistance data of the drone battery are collected. For example, the collected voltage is 3.8V, the temperature is 26℃, and the internal resistance is 18mΩ. These data are input into the state observer to obtain the estimated value of the drone battery state. For example, it is estimated that the deposition uniformity of the active material inside the battery has increased by 5%. Based on the estimated value of the battery state, a battery state space model is established, which describes the change law of the battery state with time and control input.

[0063] According to the established battery state space model, a predictive controller is constructed. Taking the uniformity of the internal state of the battery as the optimization target, the optimal control quantity, that is, the updated vibration parameters and magnetic field parameters, is calculated. For example, the optimal vibration frequency is calculated to be 12kHz, the amplitude is 1.2μm, and the optimal magnetic field strength is 12mT. The control unit controls the piezoelectric ceramic vibrator and the magnetic induction coil array respectively according to these optimal control quantities to generate an updated directional vibration field and controllable magnetic field.

[0064] Based on the estimated battery status, the capacity recovery index and the internal resistance uniformity index are calculated. For example, the capacity recovery index is the ratio of the recovered capacity to the initial capacity, and the internal resistance uniformity index is the standard deviation of the internal resistance difference of each part of the battery. These two indicators are continuously monitored. For example, the capacity recovery index reaches 95% and the internal resistance uniformity index is reduced to 1mΩ. When these two indicators reach the preset target threshold, the control unit turns off the directional vibration field and the controllable magnetic field to complete the drone battery repair process.

[0065] In this embodiment, directional vibration and controllable magnetic field are used to promote uniform deposition of active materials inside the battery and redistribution of current, effectively restore battery capacity, reduce internal resistance, and improve overall battery performance. This repair method can repair the internal microstructure of the battery, slow down the aging of the battery, and thus extend the battery life. The repair process is automatically controlled by the control unit, without manual intervention, and is easy to operate and efficient.

[0066] In an optional embodiment, a predictive controller is constructed according to the drone battery state space model, an optimal control quantity is calculated with the uniformity of the internal state of the drone battery as the optimization target, vibration parameters and magnetic field parameters are respectively updated according to the optimal control quantity, and the array piezoelectric ceramic vibrator and the magnetic induction coil array are controlled to generate an updated directional vibration field and a controllable magnetic field, including: constructing a predictive control state equation including a state vector and a control input vector, the state vector including an electrolyte concentration distribution and an active material distribution, the control input vector including vibration parameters and magnetic field parameters, wherein the vibration parameters include vibration frequency, vibration amplitude and vibration phase, and the magnetic field parameters include magnetic field intensity, magnetic field direction and magnetic field gradient; based on the predictive control state equation, an objective function with the uniformity of the internal state of the battery as the optimization target is established, the objective function including a state weight matrix and a control weight matrix, the state weight matrix is ​​used to balance the importance of each state variable, the control The weight matrix is ​​used to weigh the control effect and the control cost; the vibration parameter constraints and the magnetic field parameter constraints are set, the vibration parameter constraints include the vibration frequency range, the vibration amplitude range and the vibration phase range, and the magnetic field parameter constraints include the magnetic field intensity range, the magnetic field direction range and the magnetic field gradient range; the quadratic programming method is used to solve the objective function with the vibration parameter constraints and the magnetic field parameter constraints to obtain the optimal control sequence in the prediction time domain; the first control quantity in the optimal control sequence is executed in each control cycle, and the array piezoelectric ceramic vibrator is controlled to generate a vibration field according to the first control quantity, and the magnetic induction coil array is controlled to generate a magnetic field at the same time; the real-time state data of the drone battery under the action of the vibration field and the magnetic field is collected, the state estimation is updated based on the real-time state data, and the objective function is re-solved according to the updated state estimation to obtain a new optimal control sequence to achieve rolling optimization control of the vibration field and the magnetic field.

[0067] Exemplarily, first, a predictive control state equation for the drone battery is established. This equation describes how the internal state of the battery, such as the electrolyte concentration distribution and the active material distribution, changes over time and control input. The control input includes vibration parameters (vibration frequency, vibration amplitude, and vibration phase) and magnetic field parameters (magnetic field intensity, magnetic field direction, and magnetic field gradient). For example, assume that the electrolyte concentration distribution can be represented by a two-dimensional array and the active material distribution can also be represented by a two-dimensional array. The control input vector contains six elements, corresponding to three vibration parameters and three magnetic field parameters.

[0068] Then, an objective function with the uniformity of the internal state of the battery as the optimization target is established. The purpose of the objective function is to minimize the inhomogeneity of the internal state of the battery, such as differences in electrolyte concentration and active material distribution. The objective function contains a state weight matrix and a control weight matrix. The state weight matrix is ​​used to adjust the contribution of different state variables (such as electrolyte concentration and active material distribution) to the objective function, and the control weight matrix is ​​used to balance the control effect and control energy consumption. For example, if you are more concerned about the uniformity of the electrolyte concentration, you can increase the value of the weight matrix corresponding to the electrolyte concentration. If you want to control the energy consumption to be smaller, you can increase the value of the control weight matrix.

[0069] Next, set the constraints of the vibration parameters and magnetic field parameters. These constraints limit the actual working range of the vibrator and magnetic field generator. For example, the vibration frequency range may be set to 100Hz to 1000Hz, the vibration amplitude range may be set to 0.1mm to 1mm, and the magnetic field strength range may be set to 0.1T to 1T.

[0070] Subsequently, the quadratic programming method is used to solve the objective function with constraints to obtain the optimal control sequence in the prediction time domain. The prediction time domain refers to a period of time in the future, such as the next 1 second. The optimal control sequence contains the optimal control input for each control cycle in the future period of time. For example, if the control cycle is 0.1 seconds and the prediction time domain is 1 second, the optimal control sequence contains 10 groups of control inputs, each group of control inputs contains six parameters, corresponding to three vibration parameters and three magnetic field parameters. Assume that the first group of optimal control inputs obtained by solving are: vibration frequency 500Hz, vibration amplitude 0.5mm, vibration phase 0 degrees, magnetic field intensity 0.5T, magnetic field direction perpendicular to the battery surface, and magnetic field gradient 0.1T / mm.

[0071] In each control cycle, the first control quantity in the optimal control sequence is executed. For example, in the first control cycle, the vibration frequency is set to 500 Hz, the vibration amplitude is set to 0.5 mm, the vibration phase is set to 0 degrees, the magnetic field intensity is set to 0.5 T, the magnetic field direction is set to be perpendicular to the battery surface, and the magnetic field gradient is set to 0.1 T / mm. The array piezoelectric ceramic vibrator and the magnetic induction coil array are controlled to generate corresponding vibration fields and magnetic fields.

[0072] Collect real-time status data of drone batteries under vibration and magnetic fields, such as electrolyte concentration distribution and active material distribution. Update state estimates based on real-time status data. Resolve the objective function based on the updated state estimate to obtain a new optimal control sequence, and achieve rolling optimization control of vibration and magnetic fields.

[0073] In this embodiment, by optimizing the uniformity of the internal state of the battery, the internal resistance of the battery is reduced, and the charging and discharging efficiency and capacity of the battery are improved. The uniform internal state can reduce the side reactions and losses inside the battery and extend the service life of the battery. Based on the feedback control of the real-time state of the battery, the control strategy can be dynamically adjusted according to the actual situation of the battery to achieve accurate control of the battery state.

[0074] Figure 2 FIG. 1 is a schematic diagram of the structure of a drone battery health monitoring system based on an intelligent charging management cabinet according to an embodiment of the present invention. Figure 2 As shown, the system includes: a first unit, which is used to set a magnetic flux sensing detection unit in the charging compartment of the intelligent charging management cabinet, and the magnetic flux sensing detection unit includes an orthogonally arranged three-axis magnetic induction coil, and the leakage magnetic flux changes generated by the drone battery during the charging and discharging process are detected by the three-axis magnetic induction coil, and a current density distribution map inside the drone battery is constructed according to the spatial distribution characteristics of the leakage magnetic flux changes, and the current density distortion area inside the drone battery is identified; at the same time, a piezoelectric film stress sensing array is set on the charging compartment wall of the intelligent charging management cabinet to monitor the stress distribution state of the drone battery shell in real time and obtain the deformation characteristic data of the drone battery; a second unit is used to detect the deformation characteristic data of the drone battery based on the current density distribution map and the deformation feature data, and a self-organizing mapping neural network is used to establish a stress-current coupling model of the drone battery, wherein the stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery, and the internal damage degree of the drone battery is calculated according to the stress-current coupling model; the third unit is used to trigger the active balancing repair mode of the intelligent charging management cabinet when the internal damage degree exceeds a preset threshold, promote the uniform deposition of active materials inside the drone battery by generating mechanical vibration in the charging compartment, and guide the current redistribution inside the drone battery through the alternating magnetic field generated by the three-axis magnetic induction coil, so as to realize dynamic repair and balance of the drone battery performance.

[0075] According to a third aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0076] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0077] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the health of drone batteries based on an intelligent charging management cabinet, characterized in that: include: A magnetic flux sensing detection unit is arranged in the charging compartment of the intelligent charging management cabinet. The magnetic flux sensing detection unit includes an orthogonally arranged three-axis magnetic induction coil. The three-axis magnetic induction coil is used to detect the leakage magnetic flux change generated by the drone battery during the charging and discharging process. According to the spatial distribution characteristics of the leakage magnetic flux change, a current density distribution map inside the drone battery is constructed to identify the current density distortion area inside the drone battery. At the same time, a piezoelectric film stress sensing array is arranged on the charging compartment wall of the intelligent charging management cabinet to monitor the stress distribution state of the drone battery shell in real time and obtain the deformation characteristic data of the drone battery. Based on the current density distribution diagram and the deformation characteristic data, a stress-current coupling model of the drone battery is established by using a self-organizing map neural network. The stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery. The internal damage degree of the drone battery is calculated according to the stress-current coupling model. When the degree of internal damage exceeds the preset threshold, the active balancing repair mode of the intelligent charging management cabinet is triggered. By generating mechanical vibration in the charging compartment, the uniform deposition of active substances inside the drone battery is promoted, and the alternating magnetic field generated by the three-axis magnetic induction coil is used to guide the current redistribution inside the drone battery, thereby achieving dynamic repair and balance of the drone battery performance.

2. The method according to claim 1, characterized in that The leakage magnetic flux change generated by the drone battery during the charging and discharging process is detected by the three-axis magnetic induction coil, and the current density distribution map inside the drone battery is constructed according to the spatial distribution characteristics of the leakage magnetic flux change. The current density distortion area inside the drone battery is identified, including: The three-axis magnetic induction coil array is connected to the data acquisition unit through a signal conditioning circuit, and the signal conditioning circuit includes an operational amplifier and a low-pass filter; the leakage magnetic flux data generated by the drone battery during the charging and discharging process is collected, and the three-axis components of the leakage magnetic flux are obtained after orthogonal modulation and demodulation processing, including an X-axis component, a Y-axis component, and a Z-axis component; Based on the three-axis components of the leakage magnetic flux, the interior of the drone battery is constructed as a three-dimensional grid model consisting of a plurality of voxel units of equal size, and each voxel unit is regarded as an independent current source; a pre-constructed magnetic dipole model is used to establish a relationship equation between the current source in the voxel unit and the leakage magnetic flux, and the relationship equation includes a current density vector, an observation point position vector, and a source point position vector; Constructing an objective function including data items of a transfer matrix and a weighting matrix and a regularization item including a regularization operator; solving the objective function using a conjugate gradient method to generate an initial three-dimensional current density distribution map; The gradient value of the initial three-dimensional current density distribution map is calculated, the current density gradient value of each voxel unit is compared with a preset gradient threshold, and the area exceeding the preset gradient threshold is determined as a high-gradient area; the voxel units in the high-gradient area are subdivided using an octree structure, and each voxel unit is divided into eight sub-voxel units; a curve integral constraint condition based on the Ampere loop theorem is introduced into the encrypted grid model composed of the sub-voxel units, and the three-axis components of the leakage magnetic flux are re-solved based on the encrypted grid model, the curve integral constraint condition and the three-axis components of the leakage magnetic flux to generate an optimized three-dimensional current density distribution map; The current density mean and current density deviation coefficient of the local area in the optimized three-dimensional current density distribution map are calculated, the local area exceeding the preset deviation threshold is marked as the current density distortion area, and the position, area and deviation coefficient of the current density distortion area are annotated to form the final current density distortion distribution map inside the drone battery.

3. The method according to claim 2, characterized in that Calculating the current density mean and current density deviation coefficient of the local area in the optimized three-dimensional current density distribution map, marking the local area exceeding the preset deviation threshold as the current density distortion area, marking the position, area and deviation coefficient of the current density distortion area, and forming the final current density distortion distribution map inside the drone battery includes: Dividing the optimized three-dimensional current density distribution map into a plurality of local analysis regions, each of the local analysis regions comprising a plurality of basic voxel units, and calculating an adaptive weight coefficient based on a normalized distance from the basic voxel unit to the region center; The current density values ​​of the basic voxel units in each local analysis area are weightedly calculated using the adaptive weight coefficient to obtain a weighted current density mean value, and the weighted standard deviation of the current density values ​​of each basic voxel unit in the local analysis area is calculated at the same time, and the global current density mean value is calculated by combining the weighted current density mean values ​​of all local analysis areas, and the current density deviation coefficient is obtained by calculating the ratio of the weighted current density mean value of each local analysis area to the global current density mean value; The local analysis area where the current density deviation coefficient exceeds the preset deviation threshold is marked as a potential distortion area, and the potential distortion area is used as a seed point. A regional growth criterion is constructed based on the current density deviation coefficient difference between adjacent local analysis areas, the current density spatial gradient value and the Euclidean distance between the regional center points, and the potential distortion area is iteratively grown according to the regional growth criterion to obtain the current density distortion area; Extracting the regional volume, center of gravity position coordinates, average distortion index and current density gradient tensor of the current density distortion region to construct a feature descriptor, and calculating the position, area and deviation coefficient of the current density distortion region based on the feature descriptor; The position, area and deviation coefficient are marked at the corresponding position of the current density distortion area, and gradient colors are used to identify areas with different deviation coefficients to form a current density distortion distribution map inside the drone battery with a distortion degree distribution.

4. The method according to claim 1, characterized in that: Based on the current density distribution diagram and the deformation characteristic data, a stress-current coupling model of the drone battery is established using a self-organizing map neural network. The stress-current coupling model is used to characterize the interaction between the stress distribution and the current distribution inside the drone battery. Calculating the internal damage degree of the drone battery according to the stress-current coupling model includes: Performing maximum and minimum value normalization processing on the current density distribution diagram and the deformation characteristic data to obtain a normalized current density value and a normalized deformation value, and combining the normalized current density value and the normalized deformation value with corresponding spatial coordinates to construct a multidimensional input vector; A two-layer self-organizing map neural network is constructed, which includes a competition layer and an output layer, wherein the competition layer includes a competition neuron matrix, and the output layer includes an output neuron matrix. A Gaussian neighborhood function is used to calculate the topological distance between each neuron and a winning neuron, and the topological distance is substituted into a weight update function to calculate an adaptive learning rate. The neuron weights are iteratively updated using the adaptive learning rate to obtain a stress-current feature mapping matrix. Calculate the gradient vector of the stress-current characteristic mapping matrix on the spatial coordinates, multiply the modulus of the gradient vector by the cosine value of the angle between the stress gradient and the current gradient vector to obtain a stress-current coupling coefficient matrix, introduce the equivalent stress value, and calculate the stress-induced current density distortion factor matrix in combination with the stress-current coupling coefficient matrix and the normalized current density; Establishing a damage evolution equation, wherein the independent variables of the damage evolution equation include the stress-induced current density distortion factor matrix and the damage variable matrix, numerically integrating the damage evolution equation in the time dimension, and obtaining a damage distribution matrix characterizing the degree of damage of each region inside the battery; The maximum damage value and the damage change rate matrix in the damage distribution matrix are extracted, the maximum damage value, the damage change rate matrix and the battery health status parameter are input into a preset nonlinear mapping function, the battery remaining life prediction value is calculated, and the degree of damage inside the battery is evaluated according to the battery remaining life prediction value.

5. The method according to claim 4, characterized in that Extracting the maximum damage value and the damage change rate matrix in the damage distribution matrix, inputting the maximum damage value, the damage change rate matrix and the battery health state parameter into a preset nonlinear mapping function, calculating the battery remaining life prediction value, and evaluating the damage degree inside the battery according to the battery remaining life prediction value includes: The local binary pattern operator is used to perform texture analysis on the damage distribution matrix to extract the spatial distribution characteristics of the damage area, and the directional gradient histogram characteristics of the damage distribution matrix are calculated to obtain the directional information of the damage evolution. The damage values ​​at adjacent moments in the damage distribution matrix are estimated by optical flow to obtain the motion characteristics of the damage extension. The obtained spatial distribution characteristics, directional information and motion characteristics are combined to construct a basic feature set for damage characterization. Inputting the basic feature set into a transfer learning framework including a feature extraction network, a domain adaptation layer and a prediction network, wherein the feature extraction network extracts features using a dense connection structure, the domain adaptation layer reduces the feature distribution differences between different battery groups through adversarial training, and the prediction network performs nonlinear mapping in combination with residual connections and a gating mechanism; The short-term damage evolution trend, mid-term charge and discharge curve characteristics and long-term cumulative effects of operating conditions are extracted respectively, and the temporal attention mechanism is used to perform adaptive weighted fusion of features at different time scales. The fused features are input into the transfer learning framework to obtain the battery damage degree assessment result.

6. The method according to claim 1, characterized in that By generating mechanical vibration in the charging compartment, the uniform deposition of active materials inside the drone battery is promoted, and the alternating magnetic field generated by the three-axis magnetic induction coil guides the redistribution of the current inside the drone battery, so as to achieve dynamic repair and balance of the drone battery performance, including: Calculating initial values ​​of vibration parameters and magnetic field parameters according to the internal state of the drone battery, wherein the vibration parameters include vibration frequency, vibration amplitude and vibration phase, and the magnetic field parameters include magnetic field intensity, magnetic field direction and magnetic field gradient; According to the initial value of the vibration parameter, the array-type piezoelectric ceramic vibrator is controlled to generate mechanical vibration in the charging compartment, real-time vibration data of the mechanical vibration is collected through an acceleration sensor, vibration energy density distribution is calculated based on the real-time vibration data, and vibration parameters of the array-type piezoelectric ceramic vibrator are adjusted according to the vibration energy density distribution to generate a directional vibration field; According to the initial values ​​of the magnetic field parameters, the orthogonally arranged magnetic induction coil array is controlled to generate an alternating magnetic field in the charging compartment, the real-time magnetic field data of the alternating magnetic field is collected by the Hall sensor, the magnetic field intensity distribution is calculated based on the real-time magnetic field data, and the magnetic field parameters of the magnetic induction coil array are adjusted according to the magnetic field intensity distribution to generate a controllable magnetic field; The terminal voltage data, temperature data and internal resistance data of the drone battery under the action of the directional vibration field and the controllable magnetic field are collected and input into a state observer to obtain an estimated value of the drone battery state, and a drone battery state space model is established based on the estimated value of the drone battery state; A predictive controller is constructed according to the drone battery state space model, an optimal control quantity is calculated with the uniformity of the internal state of the drone battery as the optimization target, vibration parameters and magnetic field parameters are updated respectively according to the optimal control quantity, and the array piezoelectric ceramic vibrator and the magnetic induction coil array are controlled to generate an updated directional vibration field and a controllable magnetic field; A capacity recovery index and an internal resistance uniformity index are calculated based on the estimated value of the drone battery state. When the capacity recovery index and the internal resistance uniformity index reach preset target thresholds, the directional vibration field and the controllable magnetic field are turned off to complete the drone battery repair process.

7. The method according to claim 6, characterized in that Constructing a predictive controller according to the drone battery state space model, calculating the optimal control quantity with the uniformity of the drone battery internal state as the optimization target, updating the vibration parameters and magnetic field parameters respectively according to the optimal control quantity, and controlling the array piezoelectric ceramic vibrator and the magnetic induction coil array to generate the updated directional vibration field and controllable magnetic field include: Constructing a predictive control state equation including a state vector and a control input vector, wherein the state vector includes electrolyte concentration distribution and active material distribution, and the control input vector includes vibration parameters and magnetic field parameters, wherein the vibration parameters include vibration frequency, vibration amplitude and vibration phase, and the magnetic field parameters include magnetic field intensity, magnetic field direction and magnetic field gradient; Based on the predictive control state equation, an objective function with the uniformity of the internal state of the battery as the optimization target is established, wherein the objective function includes a state weight matrix and a control weight matrix, wherein the state weight matrix is ​​used to balance the importance of each state variable, and the control weight matrix is ​​used to weigh the control effect and the control cost; Setting vibration parameter constraints and magnetic field parameter constraints, wherein the vibration parameter constraints include a vibration frequency range, a vibration amplitude range, and a vibration phase range, and the magnetic field parameter constraints include a magnetic field intensity range, a magnetic field direction range, and a magnetic field gradient range; A quadratic programming method is used to solve the objective function with the vibration parameter constraint condition and the magnetic field parameter constraint condition to obtain an optimal control sequence in the prediction time domain; Execute the first control quantity in the optimal control sequence in each control cycle, control the array piezoelectric ceramic vibrator to generate a vibration field according to the first control quantity, and control the magnetic induction coil array to generate a magnetic field; The real-time status data of the drone battery under the vibration field and the magnetic field are collected, the state estimation is updated based on the real-time status data, and the objective function is re-solved according to the updated state estimation to obtain a new optimal control sequence, thereby realizing rolling optimization control of the vibration field and the magnetic field.

8. A drone battery health monitoring system based on an intelligent charging management cabinet, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to set a magnetic flux sensing detection unit in the charging compartment of the intelligent charging management cabinet, and the magnetic flux sensing detection unit includes an orthogonally arranged three-axis magnetic induction coil, through which the leakage magnetic flux change generated by the drone battery during the charging and discharging process is detected, and the current density distribution map inside the drone battery is constructed according to the spatial distribution characteristics of the leakage magnetic flux change, and the current density distortion area inside the drone battery is identified; at the same time, a piezoelectric film stress sensing array is set on the charging compartment wall of the intelligent charging management cabinet to monitor the stress distribution state of the drone battery shell in real time and obtain the deformation characteristic data of the drone battery; The second unit is used to establish a stress-current coupling model of the drone battery using a self-organizing map neural network based on the current density distribution map and the deformation characteristic data, wherein the stress-current coupling model is used to characterize the interaction relationship between the stress distribution and the current distribution inside the drone battery, and calculate the internal damage degree of the drone battery according to the stress-current coupling model; The third unit is used to trigger the active balancing repair mode of the intelligent charging management cabinet when the degree of internal damage exceeds a preset threshold. It promotes the uniform deposition of active substances inside the drone battery by generating mechanical vibrations in the charging compartment, and guides the current redistribution inside the drone battery through the alternating magnetic field generated by the three-axis magnetic induction coil, thereby achieving dynamic repair and balance of the drone battery performance.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images