Lithium battery initial fault diagnosis method and device, and electronic device

Abstract

The invention relates to the technical field of battery fault diagnosis, and discloses a lithium battery initial fault diagnosis method and device, and an electronic device, and the method comprises the steps: carrying out the cyclic continuous charging and discharging test of a lithium battery pack comprising a normal battery and a simulated fault battery; collecting a voltage sequence and a corresponding current sequence of a specific charging voltage interval close to the full charging voltage of the lithium battery, and extracting a plurality of characteristic parameters representing the initial fault of the battery from the voltage sequence; constructing a data set with a label based on the characteristic parameters, and training to obtain a model meeting a preset accuracy rate requirement as an initial fault diagnosis model of the lithium battery; and inputting the characteristic parameters of the lithium battery to be detected into the lithium battery initial fault diagnosis model, and outputting a fault state classification result for judging whether the lithium battery has an initial internal short circuit fault or not. The method solves the problems of false alarm and missing alarm and difficult initial fault identification of a traditional method, has the advantages of accurate diagnosis, strong anti-interference and adaptation to complex working conditions, and can efficiently prevent thermal runaway accidents.

Description

Technical Field

[0001] This invention relates to the field of battery fault diagnosis technology, specifically to a method, device, and electronic device for diagnosing initial faults in lithium batteries. Background Technology

[0002] With the rapid development of the clean energy industry, lithium batteries, with their advantages of high energy density and long cycle life, have become the core energy storage equipment in fields such as grid energy storage and new energy power generation. Their operational safety directly determines the stability and reliability of the energy system.

[0003] The development of internal short circuits in lithium-ion batteries can generally be divided into three stages: initial, intermediate, and final. For example... Figure 1 As shown, in the initial stage, due to the large internal short-circuit resistance, the changes in battery voltage and temperature are not obvious, making it difficult to identify the fault in a timely manner. However, as the fault progresses to the middle and later stages, the battery temperature rises sharply and the voltage drops rapidly, easily triggering thermal runaway. Therefore, effective diagnosis of short-circuit faults in their early stages is crucial for preventing thermal runaway.

[0004] Currently, initial fault diagnosis in energy storage systems largely relies on voltage or SOC differences as the basis for judgment, such as detecting micro-short circuits by comparing voltage deviations between series-connected batteries. However, this method is limited by the inconsistencies within the batteries themselves: if the threshold is set too high, the sensitivity for identifying micro-short circuits is insufficient; if the threshold is set too low, it is prone to misjudgments due to differences in battery resistance, capacitance, and other parameters, or external interference, resulting in information redundancy and difficulty in capturing the weak characteristic signals of initial faults. Methods that partially rely on manual experience or simple parameter monitoring are inefficient and inaccurate, failing to meet the real-time diagnostic needs of large-scale energy storage systems, leading to insufficient diagnostic stability and reliability in practical applications. Summary of the Invention

[0005] To address the technical problems of weak initial short-circuit fault signals in lithium batteries, insufficient feature extraction, and low diagnostic accuracy of traditional methods, this invention provides a method, device, and electronic device for diagnosing initial faults in lithium batteries. Based on battery charge and discharge data, potential features with fault characteristics are extracted, and a machine learning model is pre-trained. The trained model is then used to detect battery faults, which can effectively detect initial micro-short-circuit faults in batteries, is less affected by noise, and can be applied to various scenarios.

[0006] In a first aspect, the present invention provides a method for diagnosing initial faults in lithium batteries, comprising: Cyclic continuous charge-discharge tests were conducted on lithium battery packs, including normal batteries and simulated faulty batteries. The voltage sequence and corresponding current sequence of a specific charging voltage range near the full charge voltage of the lithium battery are collected, and multiple characteristic parameters characterizing the initial fault of the battery are extracted from the voltage sequence and current sequence. A labeled dataset is constructed based on the aforementioned feature parameters, and a training set and a test set are created. A preset machine learning model is trained using the training set, and the model performance is evaluated using the test set. The model that meets the preset accuracy requirements is used as the initial fault diagnosis model for lithium batteries. The characteristic parameters of the lithium battery under test are input into the lithium battery initial fault diagnosis model, and the fault state classification result is output. Based on the fault state classification result, it is determined whether the lithium battery has an initial internal short circuit fault.

[0007] The lithium battery initial fault diagnosis method provided in this invention solves the problems of traditional methods relying on threshold design and lacking flexibility. It requires no complex hardware modifications, achieving initial internal short-circuit fault identification solely through data-driven methods. It can capture weak fault signals, avoiding delayed alarms. Adaptable to the actual operating conditions of energy storage systems, it balances comprehensive diagnostics with practicality, enabling early detection of initial micro-short-circuit faults, reducing the risk of thermal runaway from the source, and providing core protection for the safe operation of energy storage systems. It has a wide range of applications and strong practicality.

[0008] In one optional implementation, the characteristic parameters characterizing the initial failure of the battery include at least one of: voltage difference, voltage standard deviation, voltage skewness, average charging current, and capacity decay rate. The dataset is labeled with the label values ​​for normal batteries and the label values ​​for faulty batteries.

[0009] The multi-dimensional feature parameters provided by this invention cover key dimensions such as voltage, current, and capacity, comprehensively characterizing the initial fault features, avoiding the information limitations of single features, and improving fault identification accuracy. The labeled dataset clearly distinguishes between normal and faulty samples, providing clear supervision signals for model training and helping the model accurately learn fault mapping relationships. The feature parameters can be optionally combined and flexibly adjusted according to actual scenarios, balancing diagnostic accuracy and data processing efficiency, effectively reducing external noise interference, and further improving the stability and reliability of fault diagnosis.

[0010] In one optional implementation, the voltage range of the cyclic continuous charge-discharge test is the safe charging voltage range of the lithium battery, and each cycle includes one charge and one discharge at a preset time interval; the simulated fault battery simulates the initial internal short circuit fault of the battery by means of an external short-circuit resistor.

[0011] The safe charging voltage range of this invention ensures stable battery pack operation during testing, preventing damage from overcharging and over-discharging, while also ensuring that the collected data closely reflects actual operating conditions. Preset time intervals enable continuous data acquisition, fully recording the fault evolution process and providing high-quality data support for feature extraction. External short-circuit resistors simulate faults without damaging the battery structure, flexibly adapting to different levels of initial fault scenarios, and exhibiting strong experimental repeatability, ensuring the reliability of model training and validation, while avoiding the safety risks of real fault injection and reducing experimental costs.

[0012] In one alternative implementation, the safe charging voltage range is 2.5V to 3.65V, and the specific charging voltage range near the full charge voltage is 3.35V to 3.5V.

[0013] The 2.5V to 3.65V voltage range provided by this invention precisely matches the characteristics of lithium iron phosphate batteries, ensuring the safety and authenticity of the tests. The specific range of 3.35V-3.5V is close to the full charge voltage; during this stage, the battery's electrochemical reactions are sensitive, and subtle parameter changes caused by initial faults are easily apparent, solving the core problem of the difficulty in capturing hidden initial fault signals. Focusing on key ranges reduces interference from irrelevant operating conditions, improves the specificity of characteristic parameters, makes subsequent model training more efficient, enhances the adaptability of diagnostic methods, and better facilitates their transfer to practical energy storage system applications.

[0014] In one optional implementation, the voltage difference is the difference between the maximum and minimum voltages 10 minutes before the 3.5V cutoff. The voltage standard deviation includes: the voltage standard deviation 10 minutes before the 3.5V cutoff, and the voltage standard deviation in the 3.35V-3.5V range; The voltage deviation includes: the voltage deviation 10 minutes before the 3.5V cutoff, and the voltage deviation in the 3.35V-3.5V range; The average charging current is the average charging current within the range of 3.35V to 3.5V. The capacity decay rate is calculated based on the difference in charging capacity over multiple cycles.

[0015] This invention combines local time-period (10 minutes before 3.5V cutoff) and global range (3.35V-3.5V) voltage characteristics to achieve dual monitoring of "local fluctuations + global trends," accurately amplifying initial fault signals. Average charging current reflects abnormal energy input, and capacity decay rate captures long-term performance degradation trends, complementing voltage characteristics to comprehensively cover the multi-dimensional manifestations of initial short circuits. The characteristic parameters are clearly defined, and the calculation logic is clear, facilitating engineering implementation and effectively distinguishing between normal fluctuations and fault anomalies, improving the accuracy of model diagnosis and reducing false alarms and missed alarms.

[0016] In one optional implementation, the formula for calculating the capacity decay rate is:

[0017] in, This is the total capacity of the battery during its Nth 3.35V-3.5V charging cycle. It is the total charging capacity of the battery during the Nn-th cycle at 3.35V-3.5V, where n represents the number of historical cycles.

[0018] This invention uses a capacity decay rate formula to accurately quantify the gradual capacity decline of a battery caused by an initial short circuit by comparing the current capacity with historical cycle capacity, thus capturing the long-term evolution pattern of faults. The setting of the n-value (historical cycle count) makes the capacity change calculation more targeted, avoiding the random interference of single-cycle data and improving the stability of characteristic parameters. The formula is simple and efficient, requiring no complex hardware support, and can be quickly integrated into data processing workflows, providing the model with highly identifiable long-term fault characteristics and facilitating the accurate identification of initial short-circuit faults.

[0019] In one alternative implementation, the capacity decay rate formula accurately quantifies the gradual capacity reduction caused by the initial short circuit by comparing the current capacity with historical cycle capacity, thus capturing the long-term evolution pattern of the fault. Setting the value of n (the number of historical cycles) makes the capacity change calculation more targeted, avoiding the random interference of single-cycle data and improving the stability of characteristic parameters. This formula is simple and efficient, requires no complex hardware support, and can be quickly integrated into data processing workflows, providing the model with highly identifiable long-term fault characteristics and facilitating the accurate identification of initial short-circuit faults.

[0020] The Xgboost model provided by this invention combines strong feature learning capabilities with high computational efficiency, making it suitable for battery fault binary classification tasks and enabling rapid discovery of the mapping relationship between features and fault states. Grid search automatically optimizes hyperparameters, avoiding the subjectivity and limitations of manual parameter tuning, ensuring optimal model parameter combinations, balancing complexity and generalization ability, and reducing the risk of overfitting. Optimal parameter settings enable the model to maintain high diagnostic performance under different operating conditions, improving the stability and reliability of the method, ensuring the model can accurately identify initial internal short-circuit faults, and meeting practical application requirements.

[0021] Secondly, the present invention provides a lithium battery initial fault diagnosis device, the device comprising: The charge / discharge test module is used to perform cyclic continuous charge / discharge tests on lithium battery packs, including normal batteries and simulated faulty batteries. The feature parameter extraction module is used to collect the voltage sequence and corresponding current sequence of a specific charging voltage range near the full charge voltage of the lithium battery, and extract multiple feature parameters characterizing the initial fault of the battery from the voltage sequence and current sequence. The dataset construction module is used to construct a labeled dataset based on the feature parameters and to divide it into a training set and a test set. The diagnostic model training module is used to train a preset machine learning model using the training set and evaluate the model performance using the test set, and to use the model that meets the preset accuracy requirements as the initial fault diagnosis model for lithium batteries. The fault status diagnosis module is used to input the characteristic parameters of the lithium battery under test into the initial fault diagnosis model of the lithium battery, output the fault status classification result, and determine whether the lithium battery has an initial internal short circuit fault based on the fault status classification result.

[0022] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the lithium battery initial fault diagnosis method described in the first aspect or any corresponding embodiment.

[0023] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the lithium battery initial fault diagnosis method described in the first aspect or any corresponding embodiment thereof.

[0024] Fifthly, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute the lithium battery initial fault diagnosis method described in the first aspect or any corresponding embodiment. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram illustrating the characteristics of lithium batteries at their initial, intermediate, and final stages. Figure 2 This is a flowchart illustrating a method for diagnosing initial faults in lithium batteries according to an embodiment of the present invention. Figure 3 This is a structural block diagram of a lithium battery initial fault diagnosis device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0029] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0030] Current methods for detecting early-stage battery faults mainly rely on detecting differences between batteries. When the difference between a faulty battery and a normal battery exceeds a certain threshold, a fault alarm is issued. However, this method is limited by the manual design of the threshold, lacks flexibility, and is prone to false alarms and missed alarms. At the same time, it cannot identify minor early-stage fault phenomena, and the alarm timing is too late to prevent accidents from happening in time.

[0031] This invention provides an embodiment of a method for diagnosing initial faults in lithium batteries. This method is based on a machine learning model for detecting initial internal short-circuit faults. It extracts potential features with fault characteristics from battery charge and discharge data, pre-trains a machine learning model, and uses the trained model to detect battery faults. This method can effectively detect initial micro-short-circuit faults in batteries, is less affected by noise, and can be applied to various scenarios. It should be noted that the steps shown in the flowchart can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here. Figure 2 This is a flowchart of a lithium battery initial fault diagnosis method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S1: Perform a cyclic continuous charge-discharge test on the lithium battery pack, which includes normal batteries and simulated faulty batteries.

[0032] Specifically, an operational experiment was designed for the lithium battery pack. An external short-circuit resistor was used to simulate an initial internal short-circuit fault in some batteries, while the rest were normal batteries. A continuous charge-discharge cycle test was conducted, recording charge-discharge data for N cycles. Each cycle consisted of one charge and one discharge. The voltage range for the continuous charge-discharge cycle test was within the safe charging voltage range of the lithium battery, and each cycle included one charge and one discharge at preset time intervals.

[0033] This invention simulates initial internal short-circuit faults using an external short-circuit resistor, without damaging the battery structure. The short-circuit resistance value can be flexibly adjusted to adapt to different levels of initial fault scenarios, accurately reproducing the weak electrical characteristics of initial internal short circuits. It eliminates the need for complex fault injection equipment or destructive experimental operations, lowering the experimental threshold and cost. Simultaneously, it avoids the thermal runaway risk that may be caused by real internal short-circuit faults, accumulating initial fault data in a safe and controllable environment, solving the problem of the difficulty in manually triggering and controlling real internal short-circuit faults. Cyclic charge-discharge testing covers the safe voltage range of lithium batteries, continuously recording complete data for N cycles at preset time intervals, including comparative samples of normal and faulty batteries. This provides multi-dimensional, long-term, high-quality data support for extracting initial fault characteristic parameters, avoiding feature loss caused by single operating conditions or fragmented data.

[0034] In one example, the safe charging voltage range for lithium batteries is the same as that for lithium iron phosphate batteries: 2.5V to 3.65V. This avoids damage to the battery from overcharging and over-discharging, ensuring the safe and stable operation of the battery pack during the experiment, and ensuring that the collected data matches the actual operating state of the battery. The corresponding specific charging voltage range near full charge voltage is 3.35V to 3.5V. During this stage, the battery's electrochemical reaction is more sensitive, and the subtle voltage and current changes caused by initial internal short-circuit faults are more easily observed. This solves the problem of the initial fault signals being hidden and difficult to capture, providing a highly identifiable data foundation for feature extraction. By focusing on a specific charging range, interference factors at different charging and discharging stages are reduced, making the diagnostic method unaffected by fluctuations in the overall charging and discharging conditions of the battery pack. This allows it to adapt to diverse operating scenarios in energy storage systems, enhancing the practicality and versatility of the method.

[0035] By setting a preset time interval of 1 minute, the continuity of data collection can be ensured, and the parameter changes during the fault evolution process can be fully recorded. At the same time, data redundancy will not be caused by too short an interval, reducing the difficulty of subsequent data processing and improving the efficiency of model training and diagnosis.

[0036] Step S2: Collect the voltage sequence and corresponding current sequence of a specific charging voltage range near the full charge voltage of the lithium battery, and extract multiple characteristic parameters characterizing the initial fault of the battery from the voltage sequence and current sequence.

[0037] The characteristic parameters for characterizing the initial fault of a battery in this embodiment of the invention include at least one of voltage difference, voltage standard deviation, voltage skewness, average charging current and capacity decay rate. These parameters cover the potential characteristics of the initial fault, avoid the information limitations of a single feature, focus on a specific charging range to reduce the interference of operating condition fluctuations throughout the charging and discharging cycle, and are less affected by individual battery differences and external environmental noise, thereby improving the stability and reliability of subsequent model diagnosis.

[0038] This invention extracts only key interval data instead of full-cycle data, which reduces data redundancy while retaining fault feature information and reducing the computational load for subsequent model training. It balances diagnostic accuracy and efficiency, and the feature parameters can accurately reflect the gradual evolution of the initial internal short circuit, providing highly discriminative training data for the machine learning model. This ensures that the model can identify minor faults in advance and avoids late alarms.

[0039] In one example, N consecutive charge cycles of each battery are extracted, including a voltage sequence V with a charging voltage range of 3.35V to 3.5V and the corresponding current sequence I. Wherein, the voltage sequence V = [ , , ,……, ], I=[ , , ,……, ], where t is the length of the sequence. Based on the above sequence data, relevant features are extracted: (1) The difference between the maximum and minimum voltage 10 minutes before the 3.5V cutoff (this value is based on the best value obtained from the experiment): , where t is the sequence length.

[0040] (2) Standard deviation of voltage in the 10 minutes before 3.5V cutoff: ,in ,V=[ , ,……, ].

[0041] (3) Voltage deviation 10 minutes before 3.5V cutoff: , in ,V=[ , ,……, ].

[0042] (4) Standard deviation of voltage in the 3.35V-3.5V range: ,in ,V=[ , ,……, ].

[0043] (5) Voltage deviation in the 3.35V-3.5V range: , in ,V=[ , ,……, ].

[0044] (6) Average charging current in the 3.35V-3.5V range .

[0045] (7) ,in This is the total charging capacity of the battery during its Nth cycle at 3.35V-3.5V (obtained from the cumulative sum of current). This represents the total charging capacity of the battery during the Nn-th cycle from 3.35V to 3.5V, where n represents the number of historical cycles. In one example, n=3, and when a short circuit fault exists in the series-connected battery pack, the charging capacity decreases with each cycle. This capacity decay rate formula accurately quantifies the gradual capacity reduction caused by the initial short circuit by comparing the current capacity with the historical cycle capacity, capturing the long-term evolution pattern of the fault. Setting the value of n (the number of historical cycles) makes the capacity change calculation more targeted, avoiding the random interference of single-cycle data and improving the stability of characteristic parameters.

[0046] The aforementioned feature parameters extract features from five dimensions: voltage fluctuation amplitude (voltage difference), data dispersion (voltage standard deviation), distribution pattern (voltage skewness), energy input (average charging current), and long-term degradation trend (capacity degradation rate). This comprehensively covers the changes in electrical parameters caused by short-circuit faults in the initial stage, avoiding the information limitations of single-dimensional features and improving fault identification accuracy. Specifically, the voltage difference, standard deviation, and skewness in the 10 minutes before the 3.5V cutoff focus on the fault-sensitive period, amplifying the minute voltage fluctuations caused by the initial short circuit. The global standard deviation, skewness, and average current in the 3.35V-3.5V range reflect the overall anomalies in this critical range. Combining these two features enables dual monitoring of local fluctuations and global trends, solving the problem of concealed initial fault signals. The average charging current reflects abnormal energy input, and the capacity degradation rate, calculated based on the capacity differences across multiple charging cycles, captures the gradual capacity decrease caused by the initial short circuit. Complementing the voltage-related features, this comprehensively reflects the process of fault development from its inception to its progression, ensuring the model identifies faults in advance. The feature extraction process fully explores the key information in the voltage and current sequences, preserving the core fault features while avoiding data redundancy, reducing the computational load of model training, and balancing diagnostic accuracy and efficiency.

[0047] Step S3: Construct a labeled dataset based on the feature parameters, and divide it into a training set and a test set.

[0048] Specifically, in this embodiment of the invention, a labeled dataset is constructed based on the extracted feature parameters described above. Each battery sample Each is an array containing the seven features mentioned above, where j is the total number of samples and each sample... There is a corresponding true value for the label. For the voltage characteristics of a normal battery, the label value is 0, and for the voltage-related characteristics of a faulty battery, the label value is 1.

[0049] The dataset is randomly divided into a training set and a training set in a 7:3 ratio based on the total number of samples. and test set ,in .

[0050] Step S4: Train a preset machine learning model using the training set, and evaluate the model performance using the test set. The model that meets the preset accuracy requirements is used as the initial fault diagnosis model for lithium batteries.

[0051] In this embodiment of the invention, an XGBoost machine learning model is constructed as the initial model. The hyperparameters of the XGBoost model are set, with the main parameters being: `max_depth`: the depth of the tree, affecting model complexity and the risk of overfitting, ranging from 3 to 10; `learning_rate`: the learning rate, controlling the step size of each iteration, ranging from 0.05 to 0.15; and `n_estimators`: the number of boosting trees, i.e., the number of training rounds, ranging from 100 to 1000. The optimal parameters are automatically found using the `gridSearchCV` method, calculating the results for different parameter combinations, and the parameters corresponding to the best results are set and saved.

[0052] training set The Xgboost model is trained to perform a binary classification task. The model takes the aforementioned features as input and outputs a classification prediction of either the battery's normal state (0) or fault state (1). Xgboost constructs a new tree by minimizing the following objective function, including the loss function and regularization term. The general form of the objective function is:

[0053] in: For the first The objective function of the wheel, The loss function measures the residual between the model's predictions and the actual values. For battery fault classification, The binary cross-entropy loss function is... For the first The true value of each sample For the first One sample in The predicted values ​​of the wheel model, For the current number Each model (tree) for samples The predicted value; This is a regularization term used to control the complexity of the model, including the number of leaf nodes and weights, to prevent overfitting. ,in It is the number of leaf nodes in the tree. It is the first The weight of each leaf, It is the regularization coefficient.

[0054] The specific Xgboost model building process is as follows: 1. Initialize the model: Build a constant model, usually the mean of the target variable, to minimize the initial loss of the model.

[0055] 2. Calculate residuals: Calculate the residuals between the current model predictions and the actual values.

[0056] 3. Construct a new tree: Use the residuals as the new target variable to construct a new tree.

[0057] 4. Calculate leaf node weights: Determine the optimal output value for each leaf based on the residual of the sample falling into each leaf.

[0058] 5. Update the model: Multiply the predictions of the new tree by the learning rate and add them to the existing model.

[0059] 6. Repeat the loop: Repeat steps 2-5 continuously, with each new tree working to correct the residual errors from the previous round.

[0060] 7. Final prediction: The prediction results of all trees are weighted and summed to obtain the final output.

[0061] Furthermore, using the test set The performance of the Xgboost model is evaluated, and the output is the sum of the predictions from all trees. When the output accuracy is greater than 90%, the model has good detection capabilities and meets the criteria for being used as an early-stage fault diagnosis model for lithium batteries.

[0062] The Xgboost model used in this invention combines strong feature learning capabilities with fast training characteristics, making it suitable for binary classification tasks in battery fault diagnosis. It can efficiently mine the mapping relationship between features and fault states, significantly improving training efficiency. By automatically searching for the optimal hyperparameter combination through gridSearchCV, it avoids the subjectivity and limitations of manual parameter tuning. At the same time, it limits the range of hyperparameter values, balancing model complexity and generalization ability, reducing the risk of overfitting. The model performance is independently verified using a test set, and an accuracy threshold of over 90% is set to ensure the practical value of the diagnostic model and avoid deploying underperforming models in practical applications, thus ensuring the credibility of fault diagnosis. Furthermore, the model is trained based on multi-dimensional features, combined with regularization mechanisms and large-scale sample learning, making it less susceptible to individual battery differences and external noise interference. It can adapt to the complex operating conditions of energy storage systems and has strong generalization ability.

[0063] Step S5: Input the characteristic parameters of the lithium battery under test into the lithium battery initial fault diagnosis model, output the fault state classification result, and determine whether the lithium battery has an initial internal short circuit fault based on the fault state classification result.

[0064] This step involves applying the trained Xgboost model to predict actual fault scenarios and determine the battery's fault state. The true value for a normal battery is 0, and the true value for a faulty battery is 1.

[0065] In one example, three lithium iron phosphate batteries (numbered B1, B2, and B3) were selected from an energy storage system. Voltage and current sequences for each battery within the 3.35V-3.5V charging range were collected, and seven characteristic parameters (V...) were extracted. diff V std_10 V skew_10 V std V skew I mean Q diff The three feature arrays are then input into the trained Xgboost diagnostic model to generate classification results: B1→0 (normal), B2→1 (fault), and B3→0 (normal). Upon analysis, it was confirmed that B2 does indeed exhibit an initial internal short-circuit fault, demonstrating the model's accurate assessment.

[0066] In actual fault scenarios, this invention requires only input of 7 key feature parameters in the critical interval, and the model quickly outputs classification results without the need for manual analysis of complex data, significantly shortening the diagnosis time. It is adapted to the real-time monitoring needs of energy storage systems and can identify faults in the early stage of short circuits in the battery (before thermal runaway is triggered), allowing maintenance personnel sufficient time to handle the situation and preventing safety accidents such as fires in energy storage systems from the source.

[0067] This embodiment also provides a lithium battery initial fault diagnosis device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated for details already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0068] This embodiment provides a lithium battery initial fault diagnosis device, such as... Figure 3 As shown, it includes: The charge-discharge test module 31 is used to perform cyclic continuous charge-discharge tests on lithium battery packs, including normal batteries and simulated faulty batteries. Feature parameter extraction module 32 is used to collect voltage sequences and corresponding current sequences of a specific charging voltage range near the full charge voltage of the lithium battery, and extract multiple feature parameters characterizing the initial fault of the battery from the voltage sequence and current sequence. The dataset construction module 33 is used to construct a labeled dataset based on the feature parameters and to divide it into a training set and a test set. The diagnostic model training module 34 is used to train a preset machine learning model using the training set and evaluate the model performance using the test set, and to use the model that meets the preset accuracy requirements as the initial fault diagnosis model for lithium batteries. The fault status diagnosis module 35 is used to input the characteristic parameters of the lithium battery under test into the lithium battery initial fault diagnosis model, output the fault status classification result, and determine whether the lithium battery has an initial internal short circuit fault based on the fault status classification result.

[0069] In some optional implementations, the feature parameters characterizing the initial battery fault in the feature parameter extraction module 32 include at least one of the following: voltage difference, voltage standard deviation, voltage skewness, average charging current, and capacity decay rate; the labels of the dataset in the dataset construction module 33 are: label values ​​corresponding to normal batteries and label values ​​corresponding to faulty batteries.

[0070] In some optional implementations, the voltage range of the cyclic continuous charge-discharge test in the charge-discharge test module 31 is the safe charging voltage range of the lithium battery, and each cycle includes one charge and one discharge at a preset time interval; the simulated fault battery simulates the initial internal short circuit fault of the battery by means of an external short-circuit resistor.

[0071] In some alternative implementations, the safe charging voltage range is 2.5V to 3.65V, and the specific charging voltage range near the full charge voltage is 3.35V to 3.5V.

[0072] In some optional implementations, the voltage difference is the difference between the maximum and minimum voltages 10 minutes before the 3.5V cutoff. The voltage standard deviation includes: the voltage standard deviation 10 minutes before the 3.5V cutoff, and the voltage standard deviation in the 3.35V-3.5V range; The voltage deviation includes: the voltage deviation 10 minutes before the 3.5V cutoff, and the voltage deviation in the 3.35V-3.5V range; The average charging current is the average charging current within the range of 3.35V to 3.5V. The capacity decay rate is calculated based on the difference in charging capacity over multiple cycles.

[0073] In some optional implementations, the capacity decay rate is calculated using the following formula:

[0074] in, This is the total capacity of the battery during its Nth 3.35V-3.5V charging cycle. It is the total charging capacity of the battery during the Nn-th cycle at 3.35V-3.5V, where n represents the number of historical cycles.

[0075] In some optional implementations, the machine learning model is an Xgboost model, whose hyperparameters are automatically optimized using a grid search method. The model calculates the results under different parameter combinations and saves the parameters corresponding to the best results.

[0076] The lithium battery initial fault diagnosis device provided in this embodiment of the invention can execute the lithium battery initial fault diagnosis method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0077] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0078] The following is a detailed reference. Figure 4This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0079] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0080] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the lithium battery initial fault diagnosis method of the embodiments of the present invention.

[0081] Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0082] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the lithium battery initial fault diagnosis method shown in the above embodiments is implemented.

[0083] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0084] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A lithium battery initial failure diagnosis method characterized by, include: Cyclic continuous charge-discharge tests were conducted on lithium battery packs, including normal batteries and simulated faulty batteries. The voltage sequence and corresponding current sequence of a specific charging voltage range near the full charge voltage of the lithium battery are collected, and multiple characteristic parameters characterizing the initial fault of the battery are extracted from the voltage sequence and current sequence. A labeled dataset is constructed based on the aforementioned feature parameters, and a training set and a test set are created. A preset machine learning model is trained using the training set, and the model performance is evaluated using the test set. The model that meets the preset accuracy requirements is used as the initial fault diagnosis model for lithium batteries. The characteristic parameters of the lithium battery under test are input into the lithium battery initial fault diagnosis model, and the fault state classification result is output. Based on the fault state classification result, it is determined whether the lithium battery has an initial internal short circuit fault.

2. The method of claim 1, wherein, The characteristic parameters characterizing the initial failure of the battery include at least one of the following: voltage difference, voltage standard deviation, voltage skewness, average charging current, and capacity decay rate. The dataset is labeled with the label values ​​for normal batteries and the label values ​​for faulty batteries.

3. The method of claim 2, wherein, The voltage range of the cyclic continuous charge and discharge test is the safe charging voltage range of the lithium battery. Each cycle includes one charge and one discharge at a preset time interval. The simulated fault battery simulates the initial internal short circuit fault of the battery by using an external short-circuit resistor.

4. The method of claim 3, wherein, The safe charging voltage range is 2.5V to 3.65V, and the specific charging voltage range near the full charge voltage is 3.35V to 3.5V.

5. The method according to claim 4, characterized in that, The voltage difference is the difference between the maximum and minimum voltages 10 minutes before the 3.5V cutoff. The voltage standard deviation includes: the voltage standard deviation 10 minutes before the 3.5V cutoff, and the voltage standard deviation in the 3.35V-3.5V range; The voltage deviation includes: the voltage deviation 10 minutes before the 3.5V cutoff, and the voltage deviation in the 3.35V-3.5V range; The average charging current is the average charging current within the range of 3.35V to 3.5V. The capacity decay rate is calculated based on the difference in charging capacity over multiple cycles.

6. The method of claim 5, wherein, The formula for calculating the capacity decay rate is: wherein, is the total capacity of the battery at the current Nth cycle for 3.35V-3.5V charging, is the total capacity of the battery at the N-nth cycle for 3.35V-3.5V charging, n represents the number of historical cycles.

7. The method according to claim 1, characterized in that, The machine learning model is the Xgboost model, whose hyperparameters are automatically optimized using a grid search method. It calculates the results under different parameter combinations and saves the parameter settings corresponding to the best results.

8. A lithium battery initial fault diagnosis device, characterized in that, The device includes: The charge / discharge test module is used to perform cyclic continuous charge / discharge tests on lithium battery packs, including normal batteries and simulated faulty batteries. The feature parameter extraction module is used to collect the voltage sequence and corresponding current sequence of a specific charging voltage range near the full charge voltage of the lithium battery, and extract multiple feature parameters characterizing the initial fault of the battery from the voltage sequence and current sequence. The dataset construction module is used to construct a labeled dataset based on the feature parameters and to divide it into a training set and a test set. The diagnostic model training module is used to train a preset machine learning model using the training set and evaluate the model performance using the test set, and to use the model that meets the preset accuracy requirements as the initial fault diagnosis model for lithium batteries. The fault status diagnosis module is used to input the characteristic parameters of the lithium battery under test into the initial fault diagnosis model of the lithium battery, output the fault status classification result, and determine whether the lithium battery has an initial internal short circuit fault based on the fault status classification result.

9. An electronic device, characterized in that, include: A memory and a processor are interconnected, the memory stores computer instructions, and the processor executes the computer instructions to perform the lithium battery initial fault diagnosis method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the lithium battery initial fault diagnosis method according to any one of claims 1 to 7.

Images