Online detection and sorting method for short circuit fault in power battery

Abstract

The invention discloses a power battery internal short circuit fault on-line detection and sorting method, which comprises the following steps: 1, acquiring basic attribute information of a battery to be detected, inputting a pre-trained context proxy model, and adaptively matching optimal charging and discharging, square wave disturbance and electrochemical impedance spectroscopy test parameters; 2, performing multi-working-condition testing on the battery by using parameters, and synchronously acquiring multi-modal data of voltage, current and impedance spectroscopy; 3, multi-dimensional features such as standing voltage attenuation, relaxation voltage fitting, incremental capacity, dynamic disturbance response and equivalent circuit parameters are extracted, and standardization and dimension reduction processing is performed after fusion; 4, performing unsupervised anomaly detection on the dimension reduction features by adopting an isolated forest algorithm, and judging an internal short circuit fault according to an anomaly score and performing automatic sorting; and 5, monitoring an abnormal rate, a false alarm rate and an omission ratio in continuous batch detection, and triggering a closed-loop calibration process when the deviation exceeds a threshold value. The device is adaptive to different battery models, and high-precision and high-stability online detection and sorting of weak internal short circuits are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new energy vehicle power battery safety detection and production quality control, and is especially suitable for rapid and accurate online detection and automatic sorting of internal short circuit faults of power batteries of different manufacturers and different models on a battery production line. BACKGROUND

[0002] As the core energy storage component of new energy vehicles, the safety and reliability of power batteries are directly related to the safety of vehicle operation. Internal short circuit is one of the most dangerous fault types of power batteries. Early weak internal short circuit may develop into serious thermal runaway during use if not detected in time, causing fire, explosion and other accidents.

[0003] At present, the commonly used internal short circuit detection technology in power battery production and factory detection has many limitations: the conventional charge-discharge and standing detection method depends on the macroscopic voltage change, has low sensitivity to early weak internal short circuit, and has poor adaptability to different types of batteries; the infrared imaging and thermal analysis detection method is greatly affected by environmental factors, cannot meet the online detection needs of high-speed production lines, and is difficult to identify micro short circuits that do not show significant heating; the electrochemical impedance spectroscopy detection method has high sensitivity, but the test period is long and is not suitable for batch online detection; the existing data-driven anomaly detection method uses fixed test steps, and the accuracy decreases when applied across different types, and lacks step optimization and long-term performance calibration mechanism.

[0004] Therefore, there is an urgent need for an online detection and sorting method that can adapt to different battery characteristics, has high sensitivity and high efficiency, to solve the problems of insufficient detection accuracy, poor universality and weak adaptability in the prior art. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art, and to provide a power battery internal short circuit fault online detection and sorting method, so as to realize individualized accurate detection for batteries of different manufacturers and models, significantly improve the identification ability of weak internal short circuits, reduce the false detection and missed detection rate, and meet the online detection needs of the production line.

[0006] In order to achieve the above-mentioned purpose of the present application, the following technical solutions are adopted: The power battery internal short circuit fault online detection and sorting method of the present application is characterized in that the following steps are performed: Step 1: Collect a set of basic attribute information C of different types of power batteries, and obtain a plurality of different sets of step parameters X and their corresponding performance evaluation data for each type of power battery through experimental testing or historical records, thereby forming a training data set for training the context agent model, and obtaining the trained context agent model; Step 2: Obtain the basic attribute information of a batch of power batteries of the same type ,in, This represents the basic attribute information of the m-th power battery, and ,in, This provides manufacturer information for the m-th power battery of the same type. This refers to the model information of the m-th power battery of the same type. This represents the nominal capacity of the m-th power battery of the same type. M represents the quantity of a batch of power batteries of the same type. Step 3: Collect the basic attribute information of a batch of power batteries of the same type. The trained context proxy model is then processed to obtain the optimal step parameters for the M power batteries. This is used to perform multi-condition testing and data acquisition on M power batteries of the same type, thereby obtaining multimodal data of the M power batteries: Step 4: Extract features from the multimodal data of a batch of power batteries of the same type to obtain the multimodal features of the batch of power batteries of the same type; Step 5: Fuse the multimodal features of the m-th power battery in a batch of power batteries of the same type to obtain the high-dimensional feature vector of the m-th power battery. and to After standardization, the standardized feature vector of the m-th power battery is obtained. Then, principal component analysis (PCA) was used to... Dimensionality reduction is performed, and principal components whose variance contribution rates meet preset requirements are retained, thus obtaining the dimension-reduced feature vector of the m-th power battery. ; Step 6: Use the Isolation Forest algorithm to... Perform anomaly detection and calculate the anomaly score of the m-th power battery. and compared with the preset sorting threshold The comparison is performed to determine whether the m-th power battery is an internal short-circuit abnormal battery, so as to achieve automatic sorting; Step 7: When the detected abnormality rate deviation of the power battery exceeds the threshold or the performance index of the context proxy model exceeds the allowable range, update the context proxy model and adjust the sorting threshold. This allows the system to adapt to changes in the high-dimensional feature vectors of the power battery, thereby optimizing the sorting performance.

[0007] The online detection and sorting method for internal short-circuit faults in power batteries described in this invention is characterized in that step one is performed as follows: Step 1.1: Construct any set of process parameters for any type of power battery. ,in, The charging current for any type of power battery. This refers to the discharge current of any type of power battery. The resting time for any type of power battery. For square wave frequency, The amplitude of the square wave excitation current. This is the minimum value in the EIS frequency range. This represents the maximum value within the EIS frequency range. For the number of EIS test frequency points, EIS incentive magnitude; Step 1.2: Construct the optimization objective function of the context proxy model using equation (1). This allows the context proxy model to be trained using the training dataset to update its parameters, resulting in the trained context proxy model. (1) In equation (1), This refers to the basic attribute information of any type of power battery. For any sample, the context proxy model The detection accuracy, For any sample, the context proxy model The test duration, For any sample, the context proxy model Safety constraint penalty items, when Beyond safe space hour, Take a positive value, otherwise, =0; There are 3 weighting coefficients.

[0008] Furthermore, step three is carried out as follows: Step 3.1: Based on optimal work step parameters Optimal charging current and optimal discharge current and optimal discharge current A constant current charge-discharge operation is performed on the m-th power battery, and after the charge-discharge operation ends and the battery has been left to stand for a preset time, the data of the m-th power battery is simultaneously collected. Voltage at time , Current at any moment ; Step 3.2: Based on optimal work step parameters The optimal square wave frequency and the optimal square wave excitation current amplitude A small-amplitude square wave current excitation is applied to the m-th power battery, and the readings of the m-th power battery are recorded. Voltage response at time t ; Step 3.3: Based on optimal work step parameters Minimum value of the optimal EIS frequency range and the maximum value of the optimal EIS frequency range and the optimal number of EIS test frequency points And the optimal EIS incentive magnitude A small current signal is set to excite the m-th power battery, and impedance measurements are performed on the m-th power battery under excitation at several frequency points within a preset frequency range to obtain the optimal test frequency for the m-th power battery. Impedance spectral data .

[0009] Furthermore, step four is carried out as follows: Step 4.1: Calculate the static voltage decay characteristics of the m-th power battery using equation (2). : (2) In equation (1), This indicates the time at which the m-th power battery begins its resting period. voltage, This indicates the time at which the m-th power battery ends its resting period. The voltage; Step 4.2: Use equation (3) to obtain the value of the m-th power battery. Relaxation model fitting characteristics at time points : (3) In equation (2), Let be the steady-state voltage of the m-th power battery. Let be the initial polarization voltage of the m-th power battery. Let m be the time constant of the m-th power battery; For a specific moment; Step 4.3: Extract the incremental capacity characteristics of the m-th power battery using equation (4). Used to extract incremental capacity features peak corresponding voltage And the area under the curve of the m-th power battery within the set voltage range is obtained using equation (5). : (4) (5) In equations (4)-(5), This indicates that the m-th power battery is at time [time missing]. voltage, This indicates that the m-th power battery is at time [time missing]. The current; Indicates the voltage of the power battery. This indicates the lower limit of the set voltage range. This indicates the upper limit of the set voltage range; Step 4.4: Use equation (6) to extract the m-th power battery. Dynamic disturbance response characteristics at time intervals : (6) In equation (5), express The voltage change amplitude, This represents the recovery time constant of the m-th power battery; Step 4.5: Use the Randle equivalent circuit to... By performing a fitting operation, the ohmic internal resistance of the m-th power battery is obtained. Charge transfer resistance Double-layer capacitors and Warburg diffusion resistance coefficient ; Step 4.6: By calculating the average correlation coefficient of the voltage curves of the m-th battery and M-1 other power batteries of the same type during the charging and discharging process, the dynamic consistency characteristics of the power battery are obtained.

[0010] Furthermore, step seven is performed as follows: Step 7.1: Calculate the abnormality rate of the bth batch of power batteries of the same type according to formula (7). : (7) In equation (7), This refers to the number of power batteries identified as abnormal in the bth batch of power batteries of the same type. The total number of tested batteries in the bth batch of power batteries of the same type; Step 7.2: Calculate the cumulative false alarm rate of all batches of power batteries of the same type according to formula (8). : (8) In equation (8), and These represent the number of false positives and true negatives of the same type of power battery in batch b after verification. Indicates the total number of batches of the same type; Step 7.3: Calculate the cumulative missed detection rate of all batches of power batteries of the same type according to formula (9). : (9) In equation (9), and These represent the number of false negatives and true positives of the same type of power battery in the b batch after verification. Step 7.4: Calculate the similar types of near-nearest neighbors using formula (10). Average abnormality rate of batch power batteries ; (10) In equation (10), Indicates the first of the same type Abnormal rate of batch power batteries; This is the preset monitoring time window length; Step 7.5, when Compared with historical benchmarks The difference between them exceeds the anomaly rate offset threshold. or cumulative false alarm rate Greater than the false alarm rate threshold or cumulative false negative rate Greater than the false negative rate threshold At that time, the training dataset is expanded to retrain the context agent model, and the sorting threshold is changed. Then, return to step three to execute; otherwise, directly return to step three to process the next batch of power batteries of the same type.

[0011] The present invention provides an electronic device, including a memory and a processor, characterized in that the memory is used to store a program that supports the processor in executing the online detection and sorting method for short-circuit faults in a power battery, and the processor is configured to execute the program stored in the memory.

[0012] The present invention discloses a computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, performs the steps of the online detection and sorting method for internal short-circuit faults in a power battery.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. High detection sensitivity: By combining dynamic high-frequency square wave excitation with local EIS measurement and extracting multi-dimensional features for fusion analysis, early weak internal short circuits are effectively identified, solving the problem of insufficient sensitivity in existing technologies.

[0014] 2. Strong cross-model adaptability: Based on the context proxy model, the optimal test parameters are automatically matched according to the basic attributes of the battery, realizing the transformation from "fixed steps" to "personalized optimization steps", which significantly improves the universality and detection accuracy of batteries from different manufacturers and models.

[0015] 3. High testing efficiency: Both local EIS and square wave disturbance tests are short-time tests, the total testing time is controllable, the process is compatible with existing production line equipment, and meets the needs of online real-time testing of large batches of batteries.

[0016] 4. Excellent long-term stability: Through a closed-loop performance monitoring and calibration mechanism, key indicators such as anomaly rate and false alarm rate are tracked in real time, and model and threshold optimization are automatically triggered when the threshold is exceeded, which effectively copes with equipment drift and battery batch fluctuations and maintains long-term stable detection performance.

[0017] 5. High degree of automation: It achieves full automation from parameter optimization, data acquisition, anomaly detection, and physical sorting. Employing the unsupervised Isolation Forest algorithm reduces reliance on pre-labeled fault data and, combined with quality traceability, forms a digital closed loop, significantly reducing the cost of manual intervention.

[0018] In summary, this invention achieves high-precision and high-efficiency online detection and sorting of different battery models through multi-condition testing, multi-modal feature fusion, adaptive parameter optimization, and continuous performance calibration. Attached Figure Description

[0019] Figure 1 This is a flowchart of the online detection and sorting process for internal short-circuit fault diagnosis in power batteries according to the present invention; Figure 2 This is a flowchart of the training process for the context proxy model of this invention; Figure 3 This is a schematic diagram illustrating the principle of short-circuit fault detection and sorting in isolated forests according to the present invention. Detailed Implementation

[0020] In this embodiment, an online detection and sorting method for internal short-circuit faults in power batteries is proposed. The test system includes a battery charge-discharge test unit, a dynamic excitation generation unit, an electrochemical impedance measurement unit, a data acquisition and processing unit, and an automatic sorting unit. It supports the synchronous or batch testing of multiple power batteries to meet the high-speed testing requirements of the production line.

[0021] like Figure 1 As shown, this method collects multi-dimensional data through multi-condition testing, extracts multi-modal features, and combines the isolated forest algorithm to achieve anomaly detection. Simultaneously, it introduces an adaptive optimization mechanism for process parameters, coupled with long-term performance monitoring, to achieve personalized and accurate detection and sorting of batteries from different manufacturers and models. The specific implementation steps are as follows: Step 1: Collect a set C of basic attribute information for different types of power batteries, and obtain multiple sets of different process parameter sets X for each type of power battery and their corresponding performance evaluation data through experimental testing or historical records. This forms a training dataset for training the context proxy model, resulting in the trained context proxy model. The training process of this context proxy model is as follows: Figure 2 As shown, the main steps include data collection, training set construction, objective function definition, model iterative training, and performance evaluation. This method integrates historical and experimental data to construct an adaptive mapping mechanism from basic battery properties to optimal process parameters, thereby achieving intelligent parameter matching and optimization for different battery models.

[0022] Step 1.1: Construct any set of process parameters for any type of power battery. ,in, The charging current for any type of power battery. This refers to the discharge current of any type of power battery. The resting time for any type of power battery. For square wave frequency, The amplitude of the square wave excitation current. This is the minimum value in the EIS frequency range. This represents the maximum value within the EIS frequency range. For the number of EIS test frequency points, This represents the EIS incentive magnitude.

[0023] Step 1.2: Construct the optimization objective function of the context proxy model using equation (1). This allows the context proxy model to be trained using the training dataset to update its parameters, resulting in the trained context proxy model. (1) In equation (1), This refers to the basic attribute information of any type of power battery. For any sample, the context proxy model The detection accuracy, For any sample, the context proxy model The test duration, For any sample, the context proxy model Safety constraint penalty items, when Beyond safe space hour, Take a positive value, otherwise, =0; There are 3 weighting coefficients.

[0024] Step 2: Obtain basic attribute information of a batch of power batteries of the same type. The basic attribute information of the m-th power battery is denoted as... ,in, This provides manufacturer information for the m-th power battery of the same type. This refers to the model information of the m-th power battery of the same type. This represents the nominal capacity of the m-th power battery of the same type. M represents the quantity of a batch of power batteries of the same type.

[0025] Step 3: Collect the basic attribute information of a batch of power batteries of the same type. The trained context proxy model is then processed to obtain the optimal step parameters for the M power batteries. This is used to perform multi-condition testing and data acquisition on M power batteries of the same type, thereby obtaining multimodal data of the M power batteries: Step 3.1: Based on optimal work step parameters Optimal charging current and optimal discharge current and optimal discharge current A constant current charge-discharge operation is performed on the m-th power battery, and after the charge-discharge operation ends and the battery has been left to stand for a preset time, the data of the m-th power battery is simultaneously collected. Voltage at time , Current at any moment .

[0026] Step 3.2: Based on optimal work step parameters The optimal square wave frequency and the optimal square wave excitation current amplitude A small-amplitude square wave current excitation is applied to the m-th power battery, and the readings of the m-th power battery are recorded. Voltage response at time t ; Step 3.3: Based on optimal work step parameters Minimum value of the optimal EIS frequency range and the maximum value of the optimal EIS frequency range and the optimal number of EIS test frequency points And the optimal EIS incentive magnitude A small current signal is set to excite the m-th power battery, and impedance measurements are performed on the m-th power battery under excitation at several frequency points within a preset frequency range to obtain the optimal test frequency for the m-th power battery. Impedance spectral data .

[0027] Step 4: Extract features from the multimodal data of a batch of power batteries of the same type to obtain the multimodal features of the batch of power batteries of the same type, including: Step 4.1: Calculate the static voltage decay characteristics of the m-th power battery using equation (2). : (2) In equation (1), This indicates the time at which the m-th power battery begins its resting period. voltage, This indicates the time at which the m-th power battery ends its resting period. The voltage.

[0028] Step 4.2: Use equation (3) to obtain the value of the m-th power battery. Relaxation model fitting characteristics at time points : (3) In equation (2), Let be the steady-state voltage of the m-th power battery. Let be the initial polarization voltage of the m-th power battery. Let m be the time constant of the m-th power battery; For a moment.

[0029] Step 4.3: Extract the incremental capacity characteristics of the m-th power battery using equation (4). Used to extract incremental capacity features peak corresponding voltage And the area under the curve of the m-th power battery within the set voltage range is obtained using equation (5). : (4) (5) In equations (4)-(5), This indicates that the m-th power battery is at time [time missing]. voltage, This indicates that the m-th power battery is at time [time missing]. The current; Indicates the voltage of the power battery. This indicates the lower limit of the set voltage range. This indicates the upper limit of the set voltage range.

[0030] Step 4.4: Use equation (6) to extract the m-th power battery. Dynamic disturbance response characteristics at time intervals : (6) In equation (5), express The voltage change amplitude, This represents the recovery time constant of the m-th power battery.

[0031] Step 4.5: Use the Randle equivalent circuit to... By performing a fitting operation, the ohmic internal resistance of the m-th power battery is obtained. Charge transfer resistance Double-layer capacitors and Warburg diffusion resistance coefficient ; Step 4.6: By calculating the average correlation coefficient of the voltage curves of the m-th battery and M-1 other power batteries of the same type during the charging and discharging process, the dynamic consistency characteristics of the power battery are obtained.

[0032] Step 5: Fuse the multimodal features of the m-th power battery in a batch of power batteries of the same type to obtain the high-dimensional feature vector of the m-th power battery. and to After standardization, the standardized feature vector of the m-th power battery is obtained. Then, principal component analysis (PCA) was used to... Dimensionality reduction is performed, and principal components whose variance contribution rates meet preset requirements are retained, thus obtaining the dimension-reduced feature vector of the m-th power battery. The purpose of dimensionality reduction is to remove potential redundancy and collinearity between features, reduce data noise, and lower the computational complexity and "curse of dimensionality" risk of subsequent algorithms such as Isolation Forest in high-dimensional spaces.

[0033] Step 6: Use the Isolation Forest algorithm to... Perform anomaly detection and calculate the anomaly score of the m-th power battery. and compared with the preset sorting threshold The comparison is performed to determine whether the m-th power battery is an internal short-circuit abnormal battery and to automatically sort it, such as... Figure 3 As shown; Step 6.1: Construct an isolated forest model: using the dimensionality-reduced feature vector set of all M batteries in the same batch. Using the input as input, randomly select features and split points to construct multiple isolated trees (iTrees), forming an isolated forest; Step 6.2, Calculate the anomaly score: For the feature vector of the m-th battery Calculate the path length of the isolated tree in each tree to obtain the average path length. Substitute into the abnormal score formula ,in For normalization factor, , for Harmonic number.

[0034] Step 6.3, Anomaly Detection: Set sorting threshold ,like If the m-th battery is determined to be an internal short-circuit faulty battery, the system immediately triggers a sorting command, controlling the automatic sorting unit to sort it into the faulty product channel. Otherwise, if it is a qualified product, the system controls the automatic sorting unit to guide it to the qualified product channel for subsequent processes.

[0035] Step 7: If a significant shift in the battery anomaly rate is detected or the model performance metrics exceed the allowable range, update the context proxy model and adjust the sorting threshold. To adapt to changes in battery population characteristics and continuously optimize sorting performance: Step 7.1: Calculate the abnormality rate of the bth batch of power batteries of the same type according to formula (7). : (7) In equation (7), This refers to the number of power batteries identified as abnormal in the bth batch of power batteries of the same type. This refers to the total number of tested batteries in the bth batch of power batteries of the same type.

[0036] Step 7.2: Calculate the cumulative false alarm rate of all batches of power batteries of the same type according to formula (8). : (8) In equation (8), and These represent the number of false positives and true negatives of the same type of power battery in batch b after verification. This indicates the total number of batches of the same type.

[0037] Step 7.3: Calculate the cumulative missed detection rate of all batches of power batteries of the same type according to formula (9). : (9) In equation (9), and These represent the number of false negatives and true positives of the same type of power battery in the b batch after verification. Step 7.4: Calculate the similar types of near-terminals using formula (10). Average abnormality rate of batch power batteries ; (10) In equation (10), Indicates the first of the same type Abnormal rate of batch power batteries; This is the preset monitoring time window length.

[0038] Step 7.5, when Compared with historical benchmarks The difference between them exceeds the anomaly rate offset threshold. or cumulative false alarm rate Greater than the false alarm rate threshold or cumulative false negative rate Greater than the false negative rate threshold At that time, the training dataset is expanded to retrain the context agent model, and the sorting threshold is changed. Then, return to step three to execute; otherwise, directly return to step three to process the next batch of power batteries of the same type.

[0039] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the methods described above, and the processor is configured to execute the program stored in the memory.

[0040] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A method for online detection and sorting of internal short-circuit faults in a power battery, characterized in that, The procedure is as follows: Step 1: Collect a set of basic attribute information C for different types of power batteries, and obtain multiple sets of different process parameters X for each type of power battery and their corresponding performance evaluation data through experimental testing or historical records, thereby forming a training dataset for training the context proxy model and obtaining the trained context proxy model. Step 2: Obtain basic attribute information of a batch of power batteries of the same type. ,in, This represents the basic attribute information of the m-th power battery, and ,in, This provides manufacturer information for the m-th power battery of the same type. This refers to the model information of the m-th power battery of the same type. This represents the nominal capacity of the m-th power battery of the same type. M represents the quantity of a batch of power batteries of the same type. Step 3: Collect the basic attribute information of a batch of power batteries of the same type. The trained context proxy model is then processed to obtain the optimal step parameters for the M power batteries. This is used to perform multi-condition testing and data acquisition on M power batteries of the same type, thereby obtaining multimodal data of the M power batteries: Step 4: Extract features from the multimodal data of a batch of power batteries of the same type to obtain the multimodal features of the batch of power batteries of the same type; Step 5: Fuse the multimodal features of the m-th power battery in a batch of power batteries of the same type to obtain the high-dimensional feature vector of the m-th power battery. and to After standardization, the standardized feature vector of the m-th power battery is obtained. Then, principal component analysis (PCA) was used to... Dimensionality reduction is performed, and principal components whose variance contribution rates meet preset requirements are retained, thus obtaining the dimension-reduced feature vector of the m-th power battery. ; Step 6: Use the Isolation Forest algorithm to... Perform anomaly detection and calculate the anomaly score of the m-th power battery. and compared with the preset sorting threshold The comparison is performed to determine whether the m-th power battery is an internal short-circuit abnormal battery, so as to achieve automatic sorting; Step 7: When the detected abnormality rate deviation of the power battery exceeds the threshold or the performance index of the context proxy model exceeds the allowable range, update the context proxy model and adjust the sorting threshold. This allows the system to adapt to changes in the high-dimensional feature vectors of the power battery, thereby optimizing the sorting performance.

2. The method for online detection and sorting of internal short-circuit faults in a power battery according to claim 1, characterized in that, Step one is to proceed as follows: Step 1.1: Construct any set of process parameters for any type of power battery. ,in, The charging current for any type of power battery. This refers to the discharge current of any type of power battery. The resting time for any type of power battery. The frequency of the square wave is... The amplitude of the square wave excitation current. This is the minimum value in the EIS frequency range. This represents the maximum value within the EIS frequency range. For the number of EIS test frequency points, EIS incentive magnitude; Step 1.2: Construct the optimization objective function of the context proxy model using equation (1). This allows the context proxy model to be trained using the training dataset to update its parameters, resulting in the trained context proxy model. (1) In equation (1), This refers to the basic attribute information of any type of power battery. For any sample, the context proxy model The detection accuracy, For any sample, the context proxy model The test duration, For any sample, the context proxy model Safety constraint penalty items, when Beyond safe space hour, Take a positive value, otherwise, =0; There are 3 weighting coefficients.

3. The method for online detection and sorting of internal short-circuit faults in a power battery according to claim 2, characterized in that, Step three is to proceed as follows: Step 3.1: Based on optimal work step parameters Optimal charging current and optimal discharge current and optimal discharge current A constant current charge-discharge operation is performed on the m-th power battery, and after the charge-discharge operation ends and the battery has been left to stand for a preset time, the data of the m-th power battery is simultaneously collected. Voltage at time , Current at any moment ; Step 3.2: Based on optimal work step parameters The optimal square wave frequency and the optimal square wave excitation current amplitude A small-amplitude square wave current excitation is applied to the m-th power battery, and the readings of the m-th power battery are recorded. Voltage response at time t ; Step 3.3: Based on optimal work step parameters Minimum value of the optimal EIS frequency range and the maximum value of the optimal EIS frequency range and the optimal number of EIS test frequency points And the optimal EIS incentive magnitude A small current signal is set to excite the m-th power battery, and impedance measurements are performed on the m-th power battery under excitation at several frequency points within a preset frequency range to obtain the optimal test frequency for the m-th power battery. Impedance spectral data .

4. The method for online detection and sorting of internal short-circuit faults in a power battery according to claim 3, characterized in that, Step four is to proceed as follows: Step 4.1: Calculate the static voltage decay characteristics of the m-th power battery using equation (2). : (2) In equation (1), This indicates the time at which the m-th power battery begins its resting period. voltage, This indicates the time at which the m-th power battery ends its resting period. The voltage; Step 4.2: Use equation (3) to obtain the value of the m-th power battery. Relaxation model fitting characteristics at time points : (3) In equation (2), Let be the steady-state voltage of the m-th power battery. Let be the initial polarization voltage of the m-th power battery. Let m be the time constant of the m-th power battery; For a specific moment; Step 4.3: Extract the incremental capacity characteristics of the m-th power battery using equation (4). Used to extract incremental capacity features peak corresponding voltage And the area under the curve of the m-th power battery within the set voltage range is obtained using equation (5). : (4) (5) In equations (4)-(5), This indicates that the m-th power battery is at time [time missing]. voltage, This indicates that the m-th power battery is at time [time missing]. The current; Indicates the voltage of the power battery. This indicates the lower limit of the set voltage range. This indicates the upper limit of the set voltage range; Step 4.4: Use equation (6) to extract the m-th power battery. Dynamic disturbance response characteristics at time intervals : (6) In equation (5), express The voltage change amplitude, This represents the recovery time constant of the m-th power battery; Step 4.5: Use the Randle equivalent circuit to... By performing a fitting operation, the ohmic internal resistance of the m-th power battery is obtained. Charge transfer resistance Double-layer capacitors and Warburg diffusion resistance coefficient ; Step 4.6: By calculating the average correlation coefficient of the voltage curves of the m-th battery and M-1 other power batteries of the same type during the charging and discharging process, the dynamic consistency characteristics of the power battery are obtained.

5. The method for online detection and sorting of internal short-circuit faults in a power battery according to claim 1, characterized in that, Step seven is to proceed as follows: Step 7.1: Calculate the abnormality rate of the bth batch of power batteries of the same type according to formula (7). : (7) In equation (7), This refers to the number of power batteries identified as abnormal in the bth batch of power batteries of the same type. The total number of tested batteries in the bth batch of power batteries of the same type; Step 7.2: Calculate the cumulative false alarm rate of all batches of power batteries of the same type according to formula (8). : (8) In equation (8), and These represent the number of false positives and true negatives of the same type of power battery in batch b after verification. Indicates the total number of batches of the same type; Step 7.3: Calculate the cumulative missed detection rate of all batches of power batteries of the same type according to formula (9). : (9) In equation (9), and These represent the number of false negatives and true positives of the same type of power battery in the b batch after verification. Step 7.4: Calculate the similar types of near-nearest neighbors using formula (10). Average abnormality rate of batch power batteries ; (10) In equation (10), Indicates the first of the same type Abnormal rate of batch power batteries; This is the preset monitoring time window length; Step 7.5, when Compared with historical benchmarks The difference between them exceeds the anomaly rate offset threshold. or cumulative false alarm rate Greater than the false alarm rate threshold or cumulative false negative rate Greater than the false negative rate threshold At that time, the training dataset is expanded to retrain the context agent model, and the sorting threshold is changed. Then, return to step three to execute; otherwise, directly return to step three to process the next batch of power batteries of the same type.

6. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the online detection and sorting method for short-circuit faults in a power battery as described in any one of claims 1-5, and the processor is configured to execute the program stored in the memory.

7. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is run by the processor, it executes the steps of the online detection and sorting method for internal short-circuit faults in power batteries as described in any one of claims 1-5.

Images