A lithium battery short circuit early warning method based on deep learning

Abstract

The application belongs to the technical field of lithium ion battery fault diagnosis, and discloses a lithium battery short circuit early warning method based on deep learning, which comprises the following steps: activating the battery through charge-discharge cycle test; constructing an internal short circuit fault data set caused by overcharging; performing data cleaning and data normalization operation on the data set for training a deep learning model; labeling the actual short circuit time of the processed data according to the data label sequence collected by the sensor; dividing the data set into a training set and a test set; performing regression analysis on the data set by using a neural network; testing the lithium ion battery under mixed working conditions, fine-tuning the trained deep learning model with a small amount of data, and obtaining a detection method suitable for actual complex working conditions. The application can accurately and quickly predict the time when the battery monomer may occur internal short circuit, provide accurate fault information for the battery management system, and has safety and reliability.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery fault diagnosis technology, and in particular to a lithium battery short-circuit early warning method based on deep learning. Background Technology

[0002] In recent years, lithium-ion batteries have become a hot research topic in energy storage applications due to their high energy density. To achieve real-time monitoring and precise control of the charge and discharge state of lithium-ion batteries, a Battery Management System (BMS) is typically installed. However, due to the complexity and harshness of their operating environment, slow lithium dendrite growth may occur inside the battery under overcharge or low-temperature charging conditions, leading to serious safety hazards. Especially in many practical operating conditions, lithium-ion batteries require continuous charge and discharge operations; internal short circuits caused by lithium dendrites not only significantly reduce battery performance but may also pose a threat to user safety.

[0003] In existing technologies, most early warning and monitoring methods for lithium plating short circuits rely on methods such as forced disassembly of the battery, X-ray imaging, or nuclear magnetic resonance. However, these methods are difficult to implement on batteries in operation and lack the necessary detection conditions in most application scenarios. Currently, research on early warning of non-destructive lithium plating short circuits in lithium-ion batteries remains insufficient, especially in practical applications. Conventional detection methods can only collect parameters such as current, voltage, and temperature of the battery through sensors for condition diagnosis. In terms of research methods, existing methods are mostly based on electrochemical models to achieve early warning of lithium plating short circuits. However, due to the complexity of electrochemical models and limitations in computation, it is difficult to effectively port them to hardware platforms under actual operating conditions. Furthermore, short circuits often occur instantaneously, and existing algorithms struggle to provide accurate early warnings of early lithium plating short circuits.

[0004] Therefore, there is an urgent need for a deep learning-based short-circuit warning method for lithium batteries to improve the safety and practicality of lithium-ion batteries. Summary of the Invention

[0005] The purpose of this invention is to provide a deep learning-based method for short-circuit early warning of lithium batteries. By utilizing the voltage, current, and temperature characteristics of the battery and limited sensor data, it can accurately and quickly predict the time when an internal short circuit may occur in a single battery cell, providing accurate fault information to the battery management system (BMS), thereby enabling corresponding fault-tolerant control and significantly improving the safety and reliability of the BMS.

[0006] To achieve the above objectives, this invention provides a deep learning-based short-circuit warning method for lithium batteries, comprising the following steps:

[0007] Step S1: Activate the battery by conducting a charge-discharge cycle test;

[0008] Step S2: Construct a dataset of internal short-circuit faults caused by overcharging;

[0009] Step S3: Perform data cleaning and normalization operations on the above internal short-circuit fault dataset for training the deep learning model.

[0010] Step S4: Label the processed data according to the data tag order collected by the sensor to indicate the actual short-circuit time;

[0011] Step S5: Divide the dataset into a training set and a test set;

[0012] Step S6: Perform regression analysis on the dataset using a neural network;

[0013] Step S7: Test the lithium-ion battery under mixed operating conditions, and fine-tune the trained deep learning model with a small amount of data to obtain a detection method suitable for actual complex operating conditions.

[0014] Preferably, in step S1, the battery is activated by charge-discharge cycle testing according to the operating mode and operating environment of the lithium-ion battery, so that the state of the lithium-ion battery is close to the actual operating conditions.

[0015] Preferably, in step S2, under the actual operating conditions of the lithium-ion battery, an induced overcharging experiment is conducted to trigger the lithium plating process inside the battery, and the current, voltage, and temperature characteristics of the battery are collected to construct a dataset of internal short-circuit faults caused by overcharging.

[0016] Preferably, in step S3, the internal short-circuit fault dataset undergoes data cleaning and normalization operations for training the deep learning model. The specific process is as follows:

[0017] Step S31: Clean the data using the 3x standard deviation method, and detect and process outliers;

[0018] This involves removing data points that deviate from the normal range due to sensor malfunction; and using linear interpolation to fill in missing data points.

[0019] Step S32: Based on step S31, the internal short-circuit fault dataset is normalized using the max-min normalization method, as shown below:

[0020]

[0021] Where X is the value of a certain state component of the current sample; X min It is the minimum value of this state component for all samples; X max It is the maximum value of this state component for all samples.

[0022] Preferably, in step S4, the current, voltage, and temperature data of the constant current charging condition in the internal short-circuit fault dataset are extracted, and the actual time when the internal short circuit is about to occur at each sampling data point is calculated, as shown below:

[0023]

[0024] Among them, T real N is the actual time of the short circuit in the current battery state; N is the sampling sequence label in the current battery state; N start It is the sampling sequence label at the start of charging, and f refers to the sampling frequency.

[0025] Preferably, in step S5, the training set and the test set account for 70% and 30% of the total data, respectively; wherein the distribution of actual short-circuit time in different sets should be the same.

[0026] Preferably, in step S6, a neural network is used to perform regression analysis on the dataset, and the specific process is as follows:

[0027] Step S61: The input data of the deep learning neural network are the current, voltage, and temperature data of the battery, and the output data is the predicted time of short circuit.

[0028] Step S62: A neural network is built using 5 fully connected layers, where ReLU is used as the activation function between each fully connected layer. The dimensions from the input layer to the output layer are 3, 128, 64, 32, 16, and 1, respectively. The network uses Adam as the optimizer and selects MSE to measure the difference between the predicted failure time and the actual failure time. The learning rate is set to 0.003.

[0029] Step S63: Use the constructed neural network to perform regression analysis on the dataset.

[0030] Preferably, in step S7, a mixed-condition test is performed based on step S1, and the specific process is as follows:

[0031] Step S71: First, perform constant current charging, setting the charging current to 500mA, and charge until the lithium-ion battery voltage is 4.0V, then let it stand for 1 minute.

[0032] Step S72: Constant current discharge, the discharge current is set to 500mA, discharge until the lithium-ion battery voltage is 3.5V; Step S74: Let stand for 1 minute.

[0033] Step S73, repeat steps S71 and S72 three times;

[0034] Step S74: Constant current charging, the charging current is set to 500mA, and the lithium-ion battery voltage is charged to 4.9V.

[0035] Preferably, the validation data is processed, and the neural network model is fine-tuned using mixed working condition test data, with 10% of the data used for fine-tuning training of the model and the remaining 90% of the data used for model validation.

[0036] Therefore, the present invention employs the aforementioned deep learning-based lithium battery short-circuit early warning method, which has the following beneficial effects:

[0037] (1) This invention fully utilizes the advantage of data-driven methods, which do not require the construction of precise models, and can accurately predict the time of internal short circuit caused by lithium plating in lithium-ion battery cells.

[0038] (2) Unlike traditional detection methods such as violent disassembly, X-ray imaging or nuclear magnetic resonance, this invention achieves fault early warning by monitoring the external characteristics of the battery (such as current, voltage and temperature), which significantly reduces the cost of fault early warning.

[0039] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0040] Figure 1 This is a flowchart of a lithium battery short-circuit early warning method based on deep learning according to the present invention;

[0041] Figure 2 Schematic diagram of the experimental topology in this embodiment of the invention;

[0042] Figure 3 A schematic diagram of a deep learning model in an embodiment of the present invention. Detailed Implementation

[0043] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] like Figure 1 As shown, a deep learning-based short-circuit warning method for lithium batteries includes the following steps:

[0045] Step S1: Activate the battery by conducting a charge-discharge cycle test;

[0046] Step S2: Construct a dataset of internal short-circuit faults caused by overcharging;

[0047] Step S3: Perform data cleaning and normalization operations on the above internal short-circuit fault dataset for training the deep learning model.

[0048] Step S4: Label the processed data according to the data tag order collected by the sensor to indicate the actual short-circuit time;

[0049] Step S5: Divide the dataset into a training set and a test set;

[0050] Step S6: Perform regression analysis on the dataset using a neural network;

[0051] Step S7: Test the lithium-ion battery under mixed operating conditions, and fine-tune the trained deep learning model with a small amount of data to obtain a detection method suitable for actual complex operating conditions.

[0052] Example

[0053] Step S1: Based on the operating conditions and environment of the lithium-ion battery, activate the battery through charge-discharge cycle testing to bring the lithium-ion battery state close to the actual operating conditions.

[0054] Step S2: Under the actual operating conditions of lithium-ion batteries, an induced overcharging experiment is conducted to trigger the lithium plating process inside the battery. The current, voltage, and temperature characteristics of the battery are collected to construct a dataset of internal short-circuit faults caused by overcharging.

[0055] Step S3: Perform data cleaning and normalization operations on the above internal short-circuit fault dataset for training the deep learning model.

[0056] Step S31: Clean the data using the 3x standard deviation method, and detect and process outliers; remove data points that deviate from the normal range due to sensor malfunction; and use linear interpolation to fill in missing data points.

[0057] Step S32: Based on step S31, the internal short-circuit fault dataset is normalized using the max-min normalization method, as shown below:

[0058]

[0059] Where X is the value of a certain state component of the current sample; X min It is the minimum value of this state component for all samples; X max It is the maximum value of this state component for all samples.

[0060] Step S4: Label the actual short-circuit time of the data processed in step S3 according to the data tag order collected by the sensor.

[0061] Extract the current, voltage, and temperature data of the constant current charging condition from the internal short-circuit fault dataset, and calculate the actual time when an internal short circuit is about to occur at each sampling data point, as shown below:

[0062]

[0063] Among them, Treal N is the actual time of the short circuit in the current battery state; N is the sampling sequence label in the current battery state; N start It is the sampling sequence label at the start of charging, and f refers to the sampling frequency.

[0064] Step S5: Divide the dataset into a training set and a test set, requiring the training set and test set to account for 70% and 30% of the total data, respectively; the distribution of actual short-circuit time in different sets should be the same.

[0065] Step S6: Use a neural network to perform regression analysis on the dataset.

[0066] Step S61: The input data of the deep learning neural network are the current, voltage and temperature data of the battery, and the output data is the predicted time of short circuit.

[0067] Step S62: A neural network is built using 5 fully connected layers, where ReLU is used as the activation function between each fully connected layer. The dimensions from the input layer to the output layer are 3, 128, 64, 32, 16, and 1, respectively. The network uses Adam as the optimizer and MSE is selected to measure the difference between the predicted failure time and the actual failure time. The learning rate is set to 0.003.

[0068] Step S63: Use the constructed neural network to perform regression analysis on the dataset, such as... Figure 3 As shown.

[0069] Step S7: Test the lithium-ion battery under mixed operating conditions, and fine-tune the deep learning model trained in step S3 with a small amount of data to verify the robustness of the short-circuit warning method and obtain a detection method suitable for actual complex operating conditions.

[0070] like Figure 2 As shown, a battery experimental platform was built, and mixed-condition testing was carried out based on step S1. The specific process is as follows:

[0071] Step S71: First, perform constant current charging, setting the charging current to 500mA, and charge until the lithium-ion battery voltage is 4.0V, then let it stand for 1 minute.

[0072] Step S72: Constant current discharge, the discharge current is set to 500mA, discharge until the lithium-ion battery voltage is 3.5V. Step S74: Let stand for 1 minute.

[0073] Step S73, repeat steps S71 and S72 three times.

[0074] Step S74: Constant current charging, the charging current is set to 500mA, and the lithium-ion battery voltage is charged to 4.9V.

[0075] The validation data undergoes the data processing step S3, where mixed-condition test data is used to fine-tune the neural network model. 10% of the data is used for model fine-tuning training, and the remaining 90% is used for model validation.

[0076] Therefore, this invention employs a deep learning-based lithium battery short-circuit early warning method. By utilizing the battery's voltage, current, and temperature characteristics and limited sensor data, it fully leverages the advantage of data-driven methods that do not require the construction of precise models. This allows for accurate and rapid prediction of the potential time of internal short circuits in individual battery cells, providing accurate fault information to the battery management system (BMS). This enables corresponding fault-tolerant control and significantly improves the safety and reliability of the BMS. Unlike traditional methods such as forced disassembly, X-ray imaging, or MRI, this invention achieves fault early warning by monitoring the battery's external characteristics (such as current, voltage, and temperature), significantly reducing the cost of fault early warning.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A short-circuit early warning method for lithium batteries based on deep learning, characterized in that, Includes the following steps: Step S1: Activate the battery by conducting a charge-discharge cycle test; Based on the operating conditions and environment of lithium-ion batteries, the batteries are activated through charge-discharge cycle tests to bring the lithium-ion battery state closer to actual operating conditions. Step S2: Construct a dataset of internal short-circuit faults caused by overcharging; Under actual operating conditions of lithium-ion batteries, an induced overcharging experiment was conducted to trigger the lithium plating process inside the battery. The current, voltage, and temperature characteristics of the battery were collected to construct a dataset of internal short-circuit faults caused by overcharging. Step S3: Perform data cleaning and normalization operations on the above internal short-circuit fault dataset for training the deep learning model. Step S4: Label the processed data according to the data tag order collected by the sensor to indicate the actual short-circuit time; Extract the current, voltage, and temperature data of the constant current charging condition from the internal short-circuit fault dataset, and calculate the actual time when an internal short circuit is about to occur at each sampling data point, as shown below: ； in, T real It is the actual time when a short circuit occurs under the current battery conditions; N It is the sampling sequence label of the current battery state; N start It is the sampling sequence label at the start of charging. f Refers to the sampling frequency; Step S5: Divide the dataset into a training set and a test set; Step S6: Perform regression analysis on the dataset using a neural network. The specific process is as follows: Step S61: The input data of the deep learning neural network are the current, voltage, and temperature data of the battery, and the output data is the predicted time of short circuit. Step S62: A neural network is built using 5 fully connected layers, where ReLU is used as the activation function between each fully connected layer. The dimensions from the input layer to the output layer are 3, 128, 64, 32, 16, and 1, respectively. The network uses Adam as the optimizer and selects MSE to measure the difference between the predicted failure time and the actual failure time. The learning rate is set to 0.

003. Step S63: Perform regression analysis on the dataset using the constructed neural network; Step S7: Test the lithium-ion battery under mixed operating conditions, and fine-tune the trained deep learning model with a small amount of data to obtain a detection method suitable for actual complex operating conditions. The specific process is as follows: Step S71: First, perform constant current charging, setting the charging current to 500mA, and charge until the lithium-ion battery voltage is 4.0V, then let it stand for 1 minute. Step S72: Constant current discharge, set the discharge current to 500mA, discharge until the lithium-ion battery voltage is 3.5V, and let stand for 1 minute; Step S73, repeat steps S71 and S72 three times; Step S74: Constant current charging, the charging current is set to 500mA, and the lithium-ion battery voltage is charged to 4.9V.

2. The lithium battery short-circuit early warning method based on deep learning according to claim 1, characterized in that, In step S3, the internal short-circuit fault dataset undergoes data cleaning and normalization operations for training the deep learning model. The specific process is as follows: Step S31: Clean the data using the 3x standard deviation method, and detect and process outliers; This involves removing data points that deviate from the normal range due to sensor malfunction; and using linear interpolation to fill in missing data points. Step S32: Based on step S31, the internal short-circuit fault dataset is normalized using the max-min normalization method, as shown below: ； in, X It is the value of a certain state component of the current sample; X min It is the minimum value of this state component for all samples; X max It is the maximum value of this state component for all samples.

3. The lithium battery short-circuit early warning method based on deep learning according to claim 1, characterized in that: In step S5, the training set and the test set account for 70% and 30% of the total data, respectively; the distribution of actual short-circuit time in different sets should be the same.

4. The lithium battery short-circuit early warning method based on deep learning according to claim 1, characterized in that: The validation data was processed, and the neural network model was fine-tuned using mixed operating condition test data. 10% of the data was used for fine-tuning training of the model, and the remaining 90% of the data was used for model validation.

Images