An Automatic Detection Method for Charging Batteries in a Battery Swapping Cabinet

Abstract

This invention discloses a method for automatically detecting rechargeable batteries in battery swap cabinets, which relates to the technical field of automatic detection of rechargeable batteries in battery swap cabinets. Through a task-characteristic-driven health prediction model, the invention overcomes the limitations of traditional single-parameter monitoring and enables early warning of damage in high-frequency cycling scenarios. A dynamic charging strategy combines impedance spectroscopy with vibration analysis to accurately identify changes in the battery's internal microstructure. A thermal coupling compensation mechanism effectively eliminates monitoring distortion caused by concurrent charging in multiple compartments, improving detection reliability under complex operating conditions. The system employs a lightweight algorithm architecture, expanding detection dimensions without hardware modification, significantly enhancing the ability to predict sudden failures.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic detection of rechargeable batteries in a power-swap cabinet, and in particular to a method for automatic detection of rechargeable batteries in a power-swap cabinet. Background Art

[0002] The distribution industry has spawned high-frequency battery cycle scenarios, and the daily peak number of battery replacements at takeaway stations has surged; existing battery replacement cabinets are generally equipped with third-generation intelligent detection systems, integrating multi-sensor fusion technology and cloud-based health assessment models, and supporting concurrent management of 50 groups of batteries; the industry's leading solutions have introduced dynamic threshold adjustment mechanisms, such as the SOH health prediction algorithm launched by a certain company in 2023, which corrects the alarm threshold according to the number of cycles; this type of system has better solved the false alarm problem caused by early fixed thresholds, but still faces special challenges in high-frequency scenarios.

[0003] The core flaw of existing solutions lies in the lag in micro-damage detection; actual measured data from a food delivery platform in 2023 showed that 67% of battery bulging failures occurred between two system health warnings; although the latest national standard DB4403-2024 mandates non-contact temperature monitoring, it cannot capture internal structural changes caused by lithium dendrite growth; cutting-edge industry solutions such as a certain automobile manufacturer's BaaS3.0 system predict internal resistance mutations through differential analysis of the charging curve, but have a high misjudgment rate in fast charging scenarios; more seriously, the thermal coupling effect when multiple batteries are charged concurrently causes single-point temperature measurement distortion, resulting in a high missed reporting rate.

[0004] The latest solutions are mainly based on two breakthrough paths: using microphone arrays to capture abnormal noises inside the battery, but the algorithm fails under 50dB of ambient noise; and the multi-frequency impedance spectroscopy technology proposed in the patent application requires the addition of a dedicated detection circuit, which increases costs; some manufacturers have tried to introduce a transfer learning framework to migrate the battery attenuation model of electric vehicles to two-wheeled vehicle scenarios, but the difference in charge and discharge rates leads to amplified prediction deviations; although these solutions are effective in specific scenarios, they have not fundamentally solved the contradiction between the accumulation of microscopic damage caused by high-frequency cycles and the prediction of sudden failures, exposing the adaptation gap between the detection dimension and the business scenario. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides an automatic detection method for rechargeable batteries in a battery swap cabinet to solve the problem that existing detection schemes rely on fixed thresholds and single-dimensional sensor data, making it difficult to cope with the accumulation of microscopic damage and interference from complex thermal environments caused by high-frequency fast charging.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a method for automatically detecting rechargeable batteries in a battery swap cabinet, which includes:

[0009] Step S1: constructing a task coverage model, wherein the model establishes a battery health prediction relationship based on historical charge and discharge task characteristics;

[0010] The task characteristics include a triplet parameter of the standard deviation of the task interval duration, the dispersion of the charge start SOC, and the ambient temperature change rate, wherein the model establishes a nonlinear mapping relationship between the triplet parameter and the capacity decay rate through a gated recurrent unit;

[0011] Step S2, acquiring multi-dimensional operating data of the target battery in real time, where the multi-dimensional data at least includes temperature field distribution, vibration frequency, and impedance spectrum characteristics;

[0012] Step S3, generating a dynamic charging strategy based on the task coverage model, wherein the strategy includes a charging priority queue and a cut-off voltage adjustment rule;

[0013] Step S4, dynamically adjusting charging control parameters based on the multi-dimensional data, wherein the parameters include a charging current gradient and a charging cut-off voltage threshold;

[0014] Step S5: output the detection result including the hidden damage risk level, and the result is associated with the battery swap cabinet position allocation instruction.

[0015] As a preferred solution of the method for automatically detecting rechargeable batteries in a battery swap cabinet according to the present invention, the construction of the task coverage model includes:

[0016] Extracting battery characteristic parameters under different charge and discharge tasks within a preset cycle, the parameters including task interval duration, ambient temperature change rate, and charging current peak distribution;

[0017] Establishing a mapping relationship between task density and capacity decay rate, where task density is defined by the ratio of the number of charge and discharge times per unit time to the task interval duration;

[0018] A dynamic health prediction threshold is generated, and the threshold is adaptively adjusted according to the current task density.

[0019] As a preferred solution of the automatic detection method of rechargeable batteries in a battery swap cabinet described in the present invention, in the process of establishing the mapping relationship between the triplet parameter and the capacity decay rate in step S1, the historical task triplet sequence is input into the gated recurrent unit GRU, and the process includes:

[0020] Define the input vector:

[0021] ,

[0022] in, Indicates time The task feature triplet of represents the standard deviation of task interval duration, Indicates the SOC dispersion at the start of charging, Indicates the rate of change of ambient temperature;

[0023] Update gate calculation:

[0024] ,

[0025] in, represents the update gate vector, is the S-type activation function, Input weight matrix to update gate, is the set of real numbers, For the Input vector at any moment, is the circulant weight matrix, For the Hidden state at all times, To update the gate bias, represents the hidden state dimension;

[0026] Reset gate calculation:

[0027] ,

[0028] in, represents the reset gate vector, Input weight for reset gate, To reset the gate recurrence weight, Bias for reset gate;

[0029] Candidate hidden state calculation:

[0030] ,

[0031] in, is a candidate hidden state, is the hyperbolic tangent activation, Enter weights for candidate states, is the candidate state cycle weight, represents element-wise multiplication, is the candidate bias,

[0032] Hidden state update:

[0033] ,

[0034] in, Indicates the Always hidden state;

[0035] Output mapping and coefficient expansion:

[0036] ,

[0037] in, represents the predicted capacity decay rate, is the output weight vector, is the last hidden state of the sequence, represents the length of the input sequence, is the output bias; ,in, is the global scaling factor, is the basic weight vector.

[0038] As a preferred solution of the method for automatically detecting rechargeable batteries in a battery swap cabinet according to the present invention, the acquisition of multi-dimensional operating data includes:

[0039] In the charging preparation stage, multi-band micro-current excitation is applied to collect the battery impedance spectrum response signal;

[0040] The impedance amplitude and phase information of the characteristic frequency band are extracted through Fourier transform;

[0041] Perform similarity matching on the information and the cloud-based aging feature library to output a lithium dendrite growth risk index;

[0042] The multi-band microcurrent excitation covers a frequency range of 1 Hz to 10 kHz, and adjacent frequency points are distributed at logarithmic intervals.

[0043] As a preferred embodiment of the method for automatically detecting rechargeable batteries in a battery swap cabinet according to the present invention, in step S2, during the charging preparation phase, a multi-band micro-current excitation within the range of 1 Hz–10 kHz is applied to the target battery, with the frequency points distributed on a logarithmic scale, including:

[0044] Frequency selection, the formula is:

[0045] ,

[0046] in, Indicates the Excitation frequency points, represents the minimum frequency, Indicates the maximum frequency, Indicates the total number of frequency points. is the frequency index;

[0047] The excitation amplitude is:

[0048] ,

[0049] in, Indicates the frequency The microcurrent applied at Indicates the rated capacity of the battery 0.05 times,

[0050] Distribution duration:

[0051] ,

[0052] in, For the Frequency excitation duration, The total duration of the scan cycle.

[0053] As a preferred solution of the method for automatically detecting rechargeable batteries in a battery swap cabinet according to the present invention, the generation of the dynamic charging strategy includes:

[0054] Generate a charging priority queue based on battery health and task urgency, and allocate batteries in the queue whose health is lower than a first preset value and whose task urgency is higher than a second preset value to fast charging positions;

[0055] A pulse repair mode is enabled for batteries whose health level is higher than a third preset value, wherein the pulse repair mode includes alternately applying a rated current and a preset ratio reduced current charging phase.

[0056] As a preferred solution of the method for automatically detecting rechargeable batteries in a battery swap cabinet according to the present invention, wherein:

[0057] A task modeling module, configured to construct and update the task coverage model;

[0058] Data acquisition module, including a distributed temperature sensor array, a piezoelectric vibration sensor, and an impedance spectrum analysis unit;

[0059] Strategy generation module, which generates charging control instructions based on real-time task requirements and battery status;

[0060] A parameter adjustment module dynamically adjusts the output current and cut-off voltage of the charging device;

[0061] The result output module maps the detection results to the battery swap cabinet position status indicator.

[0062] As a preferred solution of the automatic detection method of the rechargeable battery of the battery swap cabinet described in the present invention, the data acquisition module further includes:

[0063] The impedance spectrum analysis unit is integrated into the battery swap station charging port and includes a multi-band signal generator and a synchronous sampling circuit.

[0064] The vibration sensor array is arranged on the inner wall of the battery swap compartment and uses MEMS piezoelectric ceramic materials to capture vibration signals in the 20Hz-10kHz frequency range;

[0065] The vibration sensor array performs wavelet packet decomposition on the collected vibration signal, and extracts the sub-frequency band whose energy proportion exceeds a preset threshold in the third layer decomposition coefficient as the characteristic frequency band.

[0066] As a preferred solution of the method for automatically detecting rechargeable batteries in a battery swap cabinet according to the present invention, the strategy generation module further includes:

[0067] The task density calculation unit counts the time distribution characteristics of the delivery tasks associated with the target battery in real time;

[0068] Thermal coupling compensation unit, which corrects the temperature monitoring value of the current bin according to the temperature gradient distribution of adjacent bins;

[0069] The correction includes calculating the temperature gradient vectors of three adjacent warehouses. When the vector modulus exceeds a preset threshold, the actual temperature field distribution inside the warehouse is inverted using the heat conduction partial differential equation.

[0070] As a preferred solution of the automatic detection method of rechargeable batteries in a battery swap cabinet described in the present invention, in which: the thermal coupling compensation subunit of the strategy generation module calculates the temperature gradient vector of the current position through the spatial position and temperature measurement values of the three adjacent positions and compares it with the threshold to decide whether to start the partial differential equation inversion. The process includes:

[0071] set up: ,

[0072] in, Indicates the The position vector of each position, Indicates the The position vector of each position, Indicates the The position vector of each position,

[0073] The corresponding temperature measurement value is:

[0074] ,

[0075] in, Indicates the The temperature of each position is measured. Indicates the The temperature of each position is measured. Indicates the The temperature of each position is measured;

[0076] Calculate position and The gradient vector between:

[0077] ,

[0078] in, Indicates the With the Temperature gradient vector between bins, represents the Euclidean norm of a vector;

[0079] Calculate position and The gradient vector between:

[0080] ,

[0081] in, Indicates the With the Temperature gradient vector between bins;

[0082] The temperature gradient of the current position is obtained by fusing the two gradient vectors:

[0083] ,

[0084] in, Indicates the Temperature gradient vector of each position;

[0085] Calculate the gradient magnitude and compare it with the threshold:

[0086] ,

[0087] in, represents the modulus of the temperature gradient vector, Indicates the preset gradient threshold.

[0088] The present invention has the following beneficial effects: Through a task-characteristic-driven health prediction model, it overcomes the limitations of traditional single-parameter monitoring and enables early warning of damage in high-frequency cycling scenarios. A dynamic charging strategy combines impedance spectroscopy with vibration analysis to accurately identify changes in the battery's internal microstructure. A thermal coupling compensation mechanism effectively eliminates monitoring distortion during concurrent multi-cell charging, improving detection reliability under complex operating conditions. The system utilizes a lightweight algorithm architecture, expanding detection dimensions without hardware modifications, significantly enhancing the ability to predict sudden failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0090] Figure 1This is a flow chart of the automatic detection method for rechargeable batteries in the battery swap cabinet in Example 1. DETAILED DESCRIPTION

[0091] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0092] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0093] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0094] Example 1, with reference to Figure 1 This embodiment provides a method for automatically detecting rechargeable batteries in a battery swap cabinet, comprising the following steps:

[0095] Step S1: constructing a task coverage model, which establishes a battery health prediction relationship based on historical charge and discharge task characteristics;

[0096] The task characteristics include a triplet of parameters: the standard deviation of the task interval duration, the dispersion of the charge start state of charge (SOC), and the rate of change of the ambient temperature. The model establishes a nonlinear mapping relationship between the triplet parameters and the capacity decay rate through a gated recurrent unit.

[0097] The construction of the task coverage model includes:

[0098] Extract battery characteristic parameters under different charge and discharge tasks within a preset cycle, including task interval duration, ambient temperature change rate, and charging current peak distribution;

[0099] Establish a mapping relationship between task density and capacity decay rate. Task density is defined by the ratio of the number of charge and discharge times per unit time to the task interval length.

[0100] Generate dynamic health prediction threshold, which is adaptively adjusted according to the current task density;

[0101] In step S1, during the process of establishing the mapping relationship between triplet parameters and capacity decay rate, the historical task triplet sequence is input into the gated recurrent unit GRU. The process includes:

[0102] Define the input vector:

[0103] ,

[0104] in, Indicates time The task feature triplet of represents the standard deviation of task interval duration, Indicates the SOC dispersion at the start of charging, Indicates the rate of change of ambient temperature;

[0105] Update gate calculation:

[0106] ,

[0107] in, represents the update gate vector, is the S-type activation function, Input weight matrix to update gate, is the set of real numbers, For the Input vector at any moment, is the circulant weight matrix, For the Hidden state at all times, To update the gate bias, represents the hidden state dimension;

[0108] Reset gate calculation:

[0109] ,

[0110] in, represents the reset gate vector, Input weight for reset gate, To reset the gate recurrence weight, Bias for reset gate;

[0111] Candidate hidden state calculation:

[0112] ,

[0113] in, is a candidate hidden state, is the hyperbolic tangent activation, Enter weights for candidate states, is the candidate state cycle weight, represents element-wise multiplication, is the candidate bias,

[0114] Hidden state update:

[0115] ,

[0116] in, Indicates the Always hidden state;

[0117] Output mapping and coefficient expansion:

[0118] ,

[0119] in, represents the predicted capacity decay rate, is the output weight vector, is the last hidden state of the sequence, represents the length of the input sequence, is the output bias; ,in, is the global scaling factor, is the basic weight vector.

[0120] Specifically, the design precisely regulates information flow through update and reset gates, enabling the model to adaptively balance long-term and short-term dependencies. Candidate hidden states combine new and old information to generate rich feature representations, which are then linearly mapped to predict the capacity decay rate. The parameterization, which separates the global scaling factor from the basic weight vector, reduces the number of training parameters, improves generalization, and facilitates online fine-tuning and rapid deployment.

[0121] Step S2, acquiring multi-dimensional operating data of the target battery in real time, where the multi-dimensional data at least includes temperature field distribution, vibration frequency, and impedance spectrum characteristics;

[0122] The acquisition of multi-dimensional operation data includes:

[0123] In the charging preparation stage, multi-band micro-current excitation is applied to collect the battery impedance spectrum response signal;

[0124] The impedance amplitude and phase information of the characteristic frequency band are extracted through Fourier transform;

[0125] The information is matched with the cloud-based aging feature library for similarity, and the lithium dendrite growth risk index is output;

[0126] Multi-band microcurrent excitation covers the frequency range of 1Hz-10kHz, and adjacent frequency points are distributed in logarithmic intervals;

[0127] In step S2, during the charging preparation phase, a multi-band microcurrent excitation within the range of 1 Hz–10 kHz is applied to the target battery, with the frequency points distributed on a logarithmic scale, including:

[0128] Frequency selection, the formula is:

[0129] ,

[0130] in, Indicates the Excitation frequency points, represents the minimum frequency, Indicates the maximum frequency, Indicates the total number of frequency points. is the frequency index;

[0131] The excitation amplitude is:

[0132] ,

[0133] in, Indicates the frequency The microcurrent applied at Indicates the rated capacity of the battery 0.05 times,

[0134] Distribution duration:

[0135] ,

[0136] in, For the Frequency excitation duration, is the total duration of the scanning cycle;

[0137] Specifically, the logarithmic scale distribution provides sampling points evenly in the low-frequency and high-frequency regions, taking into account the detection of physical mechanisms such as interface polarization and ion migration, improving the comprehensiveness of impedance spectrum analysis. The constant 0.05C microcurrent ensures a high signal-to-noise ratio without causing additional damage to the battery. The uniform duration distribution ensures that each frequency point obtains the same integrated energy, reducing measurement deviation. Through this excitation rule, dendrite growth risk signals can be captured in the low-frequency band and interface and electrode processes can be reflected in the high-frequency band.

[0138] Step S3: Generate a dynamic charging strategy based on the task coverage model. The strategy includes a charging priority queue and a cut-off voltage adjustment rule.

[0139] The generation of a dynamic charging strategy includes:

[0140] Generate a charging priority queue based on battery health and task urgency. Batteries in the queue with a health lower than a first preset value and a task urgency higher than a second preset value are assigned to fast charging positions.

[0141] For batteries whose health level is higher than a third preset value, a pulse repair mode is enabled, wherein the mode includes alternating application of a rated current and a preset ratio of reduced current charging phase;

[0142] Step S4, dynamically adjusting charging control parameters based on the multi-dimensional data, the parameters including the charging current gradient and the charging cut-off voltage threshold;

[0143] Step S5: Output the detection result including the hidden damage risk level, and associate the result with the battery swap cabinet position allocation instruction;

[0144] This embodiment also provides a battery automatic detection system for a battery swap cabinet, including:

[0145] A task modeling module, configured to construct and update the task coverage model;

[0146] Data acquisition module, including a distributed temperature sensor array, a piezoelectric vibration sensor, and an impedance spectrum analysis unit;

[0147] The data acquisition module also includes:

[0148] The impedance spectrum analysis unit is integrated into the battery swap station charging port and includes a multi-band signal generator and a synchronous sampling circuit.

[0149] The vibration sensor array is arranged on the inner wall of the battery swap compartment and uses MEMS piezoelectric ceramic materials to capture vibration signals in the 20Hz-10kHz frequency range;

[0150] The vibration sensor array performs wavelet packet decomposition on the collected vibration signal and extracts the sub-band whose energy ratio exceeds the preset threshold in the third-layer decomposition coefficient as the characteristic frequency band;

[0151] Strategy generation module, which generates charging control instructions based on real-time task requirements and battery status;

[0152] The strategy generation module also includes:

[0153] The task density calculation unit counts the time distribution characteristics of the delivery tasks associated with the target battery in real time;

[0154] Thermal coupling compensation unit, which corrects the temperature monitoring value of the current bin according to the temperature gradient distribution of adjacent bins;

[0155] The correction involves calculating the temperature gradient vectors of three adjacent bins. When the vector modulus exceeds a preset threshold, the actual temperature field distribution inside the bin is inverted using the partial differential equation of heat conduction.

[0156] In the thermal coupling compensation subunit of the strategy generation module, the temperature gradient vector of the current position is calculated based on the spatial position and temperature measurements of the three adjacent positions. This is compared with the threshold and a decision is made whether to initiate the partial differential equation inversion. The process includes:

[0157] set up: ,

[0158] in, Indicates the The position vector of each position, Indicates the The position vector of each position, Indicates the The position vector of each position,

[0159] The corresponding temperature measurement value is:

[0160] ,

[0161] in, Indicates the The temperature of each position is measured. Indicates the The temperature of each position is measured. Indicates the The temperature of each position is measured;

[0162] Calculate position and The gradient vector between:

[0163] ,

[0164] in, Indicates the With the Temperature gradient vector between bins, represents the Euclidean norm of a vector;

[0165] Calculate position and The gradient vector between:

[0166] ,

[0167] in, Indicates the With the Temperature gradient vector between bins;

[0168] The temperature gradient of the current position is obtained by fusing the two gradient vectors:

[0169] ,

[0170] in, Indicates the Temperature gradient vector of each position;

[0171] Calculate the gradient magnitude and compare it with the threshold:

[0172] ,

[0173] in, represents the modulus of the temperature gradient vector, Indicates the preset gradient threshold.

[0174] Specifically, this method constructs a temperature gradient vector using a three-point local difference method, taking into account the temperature changes and spatial layout of the front and rear positions. The direction of the gradient vector is obtained by normalizing the difference vectors of adjacent positions, and its size is determined by the ratio of temperature difference to distance. Through average fusion, the sensor noise can be effectively smoothed while retaining the directional information of the thermal coupling effect. The threshold judgment ensures that the high-cost partial differential equation inversion is triggered only when the temperature gradient is abnormal, taking into account both real-time and accuracy. The spatial coordinates are provided by the cabinet design parameters, and the temperature is collected by the embedded sensor array. The entire calculation can be executed at high speed on the microcontroller, and the threshold is preset. It can be tuned according to the thermal conductivity of the material and the environmental heat dissipation conditions, and has strong adaptability and robustness;

[0175] A parameter adjustment module dynamically adjusts the output current and cut-off voltage of the charging device;

[0176] The result output module maps the detection results to the battery swap cabinet position status indicator.

[0177] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for automatically detecting rechargeable batteries in a battery swap cabinet, characterized in that: include, Step S1: constructing a task coverage model, wherein the model establishes a battery health prediction relationship based on historical charge and discharge task characteristics; The task characteristics include a triplet parameter of the standard deviation of the task interval duration, the dispersion of the charge start SOC, and the ambient temperature change rate, wherein the model establishes a nonlinear mapping relationship between the triplet parameter and the capacity decay rate through a gated recurrent unit; Step S2, acquiring multi-dimensional operating data of the target battery in real time, where the multi-dimensional data at least includes temperature field distribution, vibration frequency, and impedance spectrum characteristics; Step S3, generating a dynamic charging strategy based on the task coverage model, wherein the strategy includes a charging priority queue and a cut-off voltage adjustment rule; Step S4, dynamically adjusting charging control parameters based on the multi-dimensional data, wherein the parameters include a charging current gradient and a charging cut-off voltage threshold; Step S5: output the detection result including the hidden damage risk level, and the result is associated with the battery swap cabinet position allocation instruction.

2. The automatic detection method for rechargeable batteries in a battery swap cabinet according to claim 1, characterized in that: The construction of the task coverage model includes: Extracting battery characteristic parameters under different charge and discharge tasks within a preset cycle, the parameters including task interval duration, ambient temperature change rate, and charging current peak distribution; Establishing a mapping relationship between task density and capacity decay rate, where task density is defined by the ratio of the number of charge and discharge times per unit time to the task interval duration; A dynamic health prediction threshold is generated, and the threshold is adaptively adjusted according to the current task density.

3. The automatic detection method for rechargeable batteries in a battery swap cabinet according to claim 2, characterized in that: In step S1, during the process of establishing the mapping relationship between the triplet parameters and the capacity decay rate, the historical task triplet sequence is input into the gated recurrent unit GRU. The process includes: Define the input vector: ， in, Indicates time The task feature triplet of represents the standard deviation of task interval duration, Indicates the SOC dispersion at the start of charging, Indicates the rate of change of ambient temperature; Update gate calculation: ， in, represents the update gate vector, is the S-type activation function, Input weight matrix to update gate, is the set of real numbers, For the Input vector at any moment, is the circulant weight matrix, For the Hidden state at all times, To update the gate bias, represents the hidden state dimension; Reset gate calculation: ， in, represents the reset gate vector, Input weight for reset gate, To reset the gate recurrence weight, Bias for reset gate; Candidate hidden state calculation: ， in, is a candidate hidden state, is the hyperbolic tangent activation, Enter weights for candidate states, is the candidate state cycle weight, represents element-wise multiplication, is the candidate bias, Hidden state update: ， in, Indicates the Always hidden state; Output mapping and coefficient expansion: ， in, represents the predicted capacity decay rate, is the output weight vector, is the last hidden state of the sequence, represents the length of the input sequence, is the output bias; ,in, is the global scaling factor, is the basic weight vector.

4. The automatic detection method for rechargeable batteries in a battery swap cabinet according to claim 1, characterized in that: The acquisition of the multi-dimensional operation data includes: In the charging preparation stage, multi-band micro-current excitation is applied to collect the battery impedance spectrum response signal; The impedance amplitude and phase information of the characteristic frequency band are extracted through Fourier transform; Perform similarity matching on the information and the cloud-based aging feature library to output a lithium dendrite growth risk index; The multi-band microcurrent excitation covers a frequency range of 1 Hz to 10 kHz, and adjacent frequency points are distributed at logarithmic intervals.

5. The automatic detection method for rechargeable batteries in a battery swap cabinet according to claim 4, characterized in that: In step S2, during the charging preparation phase, a multi-band microcurrent excitation within the range of 1 Hz–10 kHz is applied to the target battery, with the frequency points distributed on a logarithmic scale, including: Frequency selection, the formula is: ， in, Indicates the Excitation frequency points, represents the minimum frequency, Indicates the maximum frequency, Indicates the total number of frequency points. is the frequency index; The excitation amplitude is: ， in, Indicates the frequency The microcurrent applied at Indicates the rated capacity of the battery 0.05 times, Distribution duration: ， in, For the Frequency excitation duration, The total duration of the scan cycle.

6. The automatic detection method for rechargeable batteries in a battery swap cabinet according to claim 1, characterized in that: The generation of the dynamic charging strategy includes: Generate a charging priority queue based on battery health and task urgency, and allocate batteries in the queue whose health is lower than a first preset value and whose task urgency is higher than a second preset value to fast charging positions; A pulse repair mode is enabled for batteries whose health level is higher than a third preset value, wherein the pulse repair mode includes alternately applying a rated current and a preset ratio reduced current charging phase.

7. A battery swap cabinet rechargeable battery automatic detection system, based on a battery swap cabinet rechargeable battery automatic detection method according to any one of claims 1 to 6, characterized in that: include: A task modeling module, configured to construct and update the task coverage model; Data acquisition module, including a distributed temperature sensor array, a piezoelectric vibration sensor, and an impedance spectrum analysis unit; Strategy generation module, which generates charging control instructions based on real-time task requirements and battery status; A parameter adjustment module dynamically adjusts the output current and cut-off voltage of the charging device; The result output module maps the detection results to the battery swap cabinet position status indicator.

8. The automatic detection system for rechargeable batteries in a battery swap cabinet according to claim 7, characterized in that: The data acquisition module also includes: The impedance spectrum analysis unit is integrated into the battery swap station charging port and includes a multi-band signal generator and a synchronous sampling circuit. The vibration sensor array is arranged on the inner wall of the battery swap compartment and uses MEMS piezoelectric ceramic materials to capture vibration signals in the 20Hz-10kHz frequency range; The vibration sensor array performs wavelet packet decomposition on the collected vibration signal, and extracts the sub-frequency band whose energy proportion exceeds a preset threshold in the third layer decomposition coefficient as the characteristic frequency band.

9. The automatic detection system for rechargeable batteries in a battery swap cabinet according to claim 8, characterized in that: The strategy generation module also includes: The task density calculation unit counts the time distribution characteristics of the delivery tasks associated with the target battery in real time; Thermal coupling compensation unit, which corrects the temperature monitoring value of the current bin according to the temperature gradient distribution of adjacent bins; The correction includes calculating the temperature gradient vectors of three adjacent warehouses. When the vector modulus exceeds a preset threshold, the actual temperature field distribution inside the warehouse is inverted using the heat conduction partial differential equation.

10. The automatic detection system for rechargeable batteries in a battery swap cabinet according to claim 9, characterized in that: In the thermal coupling compensation subunit of the strategy generation module, the temperature gradient vector of the current position is calculated based on the spatial position and temperature measurement values of the three adjacent positions, and compared with the threshold value to decide whether to start the partial differential equation inversion. The process includes: set up: , in, Indicates the The position vector of each position, Indicates the The position vector of each position, Indicates the The position vector of each position, The corresponding temperature measurement value is: ， in, Indicates the The temperature of each position is measured. Indicates the The temperature of each position is measured. Indicates the The temperature of each position is measured; Calculate position and The gradient vector between: ， in, Indicates the With the Temperature gradient vector between bins, represents the Euclidean norm of a vector; Calculate position and The gradient vector between: ， in, Indicates the With the Temperature gradient vector between bins; The temperature gradient of the current position is obtained by fusing the two gradient vectors: ， in, Indicates the Temperature gradient vector of each position; Calculate the gradient magnitude and compare it with the threshold: ， in, represents the modulus of the temperature gradient vector, Indicates the preset gradient threshold.

Images