Fault detection method and detection device for lithium ion battery and lithium ion battery pack, and electronic equipment

Abstract

The fault detection method comprises the following steps: acquiring a plurality of sample data groups and a plurality of corresponding reference voltage data of a normal lithium ion battery; training a deep learning algorithm by applying the sample data to obtain a first mapping relation and a second mapping relation respectively; and calculating a voltage prediction error threshold based on the voltage prediction values of the first mapping relation and the second mapping relation. And acquiring a plurality of test data groups and a plurality of corresponding real voltage data of the to-be-tested lithium ion battery, calculating to obtain a first error, a second error and a voltage prediction error by using the first mapping relation and the second mapping relation, and determining a fault judgment result of the to-be-tested lithium ion battery through a preset logic relation.

Description

[Technical field]

[0001] The present application relates to the technical field of lithium-ion batteries, and in particular to a fault detection method, a detection device and an electronic device for lithium-ion batteries and lithium-ion battery packs. [Background technology]

[0002] Lithium-ion batteries have been widely used due to their high energy density and long cycle life. However, the frequent occurrence of safety accidents related to thermal runaway (TR) has caused people to worry about the safety of lithium-ion batteries. The spread of local hot spots triggered by early internal short circuit (ISC) failure throughout the battery is one of the main causes of TR.

[0003] Therefore, before ISC develops into TR, effective detection of lithium-ion batteries is of great significance for the safe and reliable use of lithium-ion batteries. [Summary of the invention]

[0004] The inventors of the present application discovered in the process of developing lithium-ion battery fault detection that the electrochemical reaction mechanism inside lithium-ion batteries is very complex and the battery parameter changes caused by ISC are difficult to accurately capture. Some existing threshold-based or model-based ISC diagnosis and detection methods have many defects and deficiencies and cannot meet the needs of practical applications.

[0005] To this end, the inventor of the present application provides a first aspect of a fault detection method for a lithium-ion battery pack. The fault detection method includes: obtaining multiple sample data groups of normal lithium-ion batteries and multiple corresponding reference voltage data; each sample data group includes: sample voltage data, sample current data, sample temperature data collected by the normal lithium-ion battery within a preset time period, and simulated voltage data simulated by a pre-constructed electrochemical mechanism model, one of the reference voltage data is the voltage data collected by the normal lithium-ion battery within a preset time period after the end time of the corresponding sample data group; using the sample voltage data, the sample current data and the sample temperature data as input information, and using the reference voltage data as output information, training a preset deep learning algorithm to obtain a first mapping relationship; using the simulated voltage data, the sample current data and the sample temperature data as input information, and using the reference voltage data as output information, training the deep learning algorithm to obtain a second mapping relationship; based on the reference voltage data and the first A first error threshold is calculated based on a sampled voltage prediction value output by a first mapping relationship, and a second error threshold is calculated based on the reference voltage data and the simulated voltage prediction value output by the second mapping relationship; the first error threshold and the second error threshold are integrated into a voltage prediction error threshold by a preset numerical integration method; a plurality of test data groups of the lithium-ion battery to be tested and a plurality of corresponding real voltage data are collected; each test data group includes: voltage data, current data and temperature data collected by the lithium-ion battery to be tested within a preset time period; a first error is calculated based on the real voltage data and the first voltage prediction value calculated and output by the first mapping relationship, and a second error is calculated based on the real voltage data and the second voltage prediction value calculated and output by the second mapping relationship; the first error and the second error are integrated into a voltage prediction error by the numerical integration method; according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold, a fault judgment result of the lithium-ion battery to be tested is determined by a preset logical relationship.

[0006] The above fault detection method flexibly predicts voltage sequences of different time lengths by adjusting the length of input data when training and learning the mapping relationship, and has the characteristics of high voltage prediction accuracy. Moreover, based on the predicted voltage, short-circuit faults in the battery can be diagnosed more accurately and quickly, and high diagnostic accuracy can be obtained under various working conditions, achieving early detection and early warning.

[0007] In combination with the first aspect, in a possible implementation, the acquiring of multiple sample data groups and multiple corresponding reference voltage data of a normal lithium-ion battery specifically includes: acquiring sample voltage data, sample current data, and sample temperature data of the normal lithium-ion battery that change with operating time under different operating conditions; acquiring simulation voltage data of the normal lithium-ion battery under the different operating conditions through a pre-constructed electrochemical mechanism model; taking the start time of the operating time as the starting point, and dividing the sample voltage data, the sample current data, the sample temperature data, and the simulation voltage data into multiple data segments according to a preset time step; the time step is equal to the length of the preset time period; combining the data segments of the sample voltage data, the sample current data, the sample temperature data, and the simulation voltage data of the same time period into one sample data group; taking the end time of each sample data group as the starting point, acquiring the voltage data of the lithium-ion battery within a preset time period after the end time of the sample data group as the reference voltage data corresponding to the sample data group.

[0008] In combination with the first aspect, or any one of the above-mentioned possible implementations of the first aspect, in another possible implementation, the electrochemical mechanism model is an electrochemical mechanism model that includes thermal field simulation, including: single particle model, quasi-two-dimensional model, multi-dimensional multi-field electrochemical model and other models constructed by electrochemical theory.

[0009] In combination with the first aspect, or any one of the above possible implementations of the first aspect, in another possible implementation, the deep learning algorithm includes: a convolutional neural network, a fully connected neural network, a recurrent neural network, a long short-term memory neural network, an attention mechanism, and a Transformer.

[0010] In combination with the first aspect, or any of the foregoing possible implementations of the first aspect, in yet another possible implementation, the first error threshold is calculated by the following formula: Wherein, N is the length of the preset time period, B is the correlation coefficient, C is the charge or discharge rate of the battery, E1 is the first error threshold, U FP1 is the predicted value of the sampled voltage, U F is the reference voltage data; the second error threshold is calculated by the following formula: Wherein, N is the length of the preset time period, B is the correlation coefficient, C is the charge or discharge rate of the battery, E2 is the second error threshold, and U FP2 is the predicted value of the simulated voltage, U F is the reference voltage data; the preset numerical integration method is: weighted average.

[0011] In combination with the first aspect, or any one of the above possible implementations of the first aspect, in another possible implementation, the determining the fault judgment result of the lithium-ion battery to be tested through a preset logical relationship according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold specifically includes: when at least two of the first error, the second error and the voltage prediction error are less than or equal to the voltage prediction error threshold, determining that the lithium-ion battery to be tested has not failed; when at least two of the first error, the second error and the voltage prediction error are greater than the voltage prediction error threshold, determining that the lithium-ion battery to be tested has failed; when it is determined that the lithium-ion battery to be tested has failed, determining the fault classification of the lithium-ion battery to be tested according to the values ​​of the first error, the second error and the voltage prediction error exceeding the voltage prediction error threshold.

[0012] In a second aspect, an embodiment of the present application provides a method for detecting a fault of a lithium-ion battery pack. The lithium-ion battery pack includes a plurality of lithium-ion batteries, and the method for detecting a fault includes: using the above-described method for detecting a fault of a lithium-ion battery to obtain a fault judgment result of each of the lithium-ion batteries in the lithium-ion battery pack to be tested; based on the method for detecting a fault of a lithium-ion battery, analyzing the differences between different lithium-ion batteries in the lithium-ion battery pack; and determining the fault state of the lithium-ion battery pack to be tested according to the fault judgment result of each of the lithium-ion batteries and the differences between different lithium-ion batteries.

[0013] In combination with the second aspect, in a possible implementation, the fault detection method based on the lithium-ion battery analyzes the differences between different lithium-ion batteries in the lithium-ion battery pack, specifically including: obtaining historical detection data of one target lithium-ion battery among the multiple lithium-ion batteries; using the historical detection data to retrain the first mapping relationship and the second mapping relationship in the fault detection method of the lithium-ion battery; based on the retrained first mapping relationship and the second mapping relationship, respectively calculating the voltage prediction error corresponding to each lithium-ion battery; the fault state of the lithium-ion battery pack to be tested is determined according to the fault judgment result of each lithium-ion battery and the differences between different lithium-ion batteries, specifically including: in the When the fault judgment result of the target lithium-ion battery is that a fault occurs, and the difference between different lithium-ion batteries does not meet the standard, the fault state of the lithium-ion battery group to be tested is determined to be a secondary alarm state; when the fault judgment result of the target lithium-ion battery is that a fault occurs, and the difference between different lithium-ion batteries meets the standard, the fault state of the lithium-ion battery group to be tested is determined to be a primary alarm state; wherein, when the voltage prediction error corresponding to each of the remaining batteries is greater than a preset first value, and when the voltage prediction error corresponding to the target lithium-ion battery calculated by using the first mapping relationship and the second mapping relationship retrained with historical detection data of other lithium-ion batteries is greater than a preset second value, it is determined that the difference between different lithium-ion batteries does not meet the standard.

[0014] In a third aspect, the present application provides a fault detection device for a lithium-ion battery. The fault detection device includes: a sample data acquisition module, which is used to acquire multiple sample data groups and multiple corresponding reference voltage data of normal lithium-ion batteries; each sample data group includes: sample voltage data, sample current data, sample temperature data collected by the normal lithium-ion battery within a preset time period, and simulated voltage data simulated by a pre-constructed electrochemical mechanism model, and one of the reference voltage data is the voltage data collected by the normal lithium-ion battery within a preset time period after the end time of the corresponding sample data group; a mapping relationship training module, which is used to train a preset deep learning algorithm with the sample voltage data, the sample current data and the sample temperature data as input information, and the reference voltage data as output information, to obtain a first mapping relationship; the deep learning algorithm is trained with the simulated voltage data, the sample current data and the sample temperature data as input information, and the reference voltage data as output information, to obtain a second mapping relationship; a threshold calculation module, which is used to calculate the output sample voltage based on the reference voltage data and the first mapping relationship. A prediction value is calculated, a first error threshold is calculated, and a second error threshold is calculated based on the reference voltage data and the simulated voltage prediction value calculated and output by the second mapping relationship; the first error threshold and the second error threshold are integrated into a voltage prediction error threshold by a preset numerical integration method; a test data acquisition module is used to collect multiple test data groups and multiple corresponding real voltage data of the lithium-ion battery to be tested; each test data group includes: voltage data, current data and temperature data collected by the lithium-ion battery to be tested within a preset time period; an error calculation module is used to calculate a first error based on the real voltage data and the first voltage prediction value calculated and output by the first mapping relationship, and calculate a second error based on the real voltage data and the second voltage prediction value calculated and output by the second mapping relationship; the first error and the second error are integrated into a voltage prediction error by the numerical integration method; a judgment module is used to determine the fault judgment result of the lithium-ion battery to be tested according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold through a preset logical relationship.

[0015] In a fourth aspect, the present application provides an electronic device, which includes a processor and a memory, wherein a non-volatile storage medium is stored in the memory, and when the non-volatile storage medium is called by the processor, the processor executes the above-mentioned fault detection method.

[0016] The beneficial effects of the related devices provided in the second, third and fourth aspects of the present application can refer to the beneficial effects of the technical solution of the first aspect, and will not be repeated here.

Brief Description of the Drawings

[0017] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0018] Figure 1 A method flow chart of a lithium-ion battery fault detection method provided in an embodiment of the present application;

[0019] Figure 2 A flow chart of a method for obtaining sample data and reference voltage provided in an embodiment of the present application;

[0020] Figure 3 A flow chart of a method for determining a fault judgment result of a lithium-ion battery provided in an embodiment of the present application;

[0021] Figure 4 A method flow chart of a fault detection method for a lithium-ion battery pack provided in an embodiment of the present application;

[0022] Figure 5 A flow chart of a method for analyzing differences between different lithium-ion batteries in a lithium-ion battery pack provided in an embodiment of the present application;

[0023] Figure 6 A flow chart of a method for determining a fault judgment result of a lithium-ion battery pack provided in an embodiment of the present application;

[0024] Figure 7 A flow chart of a method for determining whether differences between different lithium-ion batteries meet standards provided in an embodiment of the present application;

[0025] Fig. 8A A method flow chart of the mapping relationship training phase of the fault detection method provided in an embodiment of the present application;

[0026] Figure 8B A method flow chart of the test use phase of the fault detection method provided in an embodiment of the present application;

[0027] Fig. 9 A schematic diagram of sample data and reference voltage data provided in an embodiment of the present application;

[0028] Fig.10 A comparison diagram of a predicted voltage value calculated using the first mapping relationship ALG1 and a real voltage value acquired by acquisition provided in an embodiment of the present application;

[0029] Fig.11 A comparison diagram of a predicted voltage value calculated using the second mapping relationship ALG2 and a real voltage value obtained by acquisition provided in an embodiment of the present application;

[0030] Fig.12 A schematic diagram of a logical judgment relationship for determining a fault state of a lithium-ion battery pack according to a voltage prediction error provided in an embodiment of the present application;

[0031] Fig.13 A functional block diagram of a fault detection device provided in an embodiment of the present application;

[0032] Fig.14 A schematic diagram of a control terminal provided in an embodiment of the present application. [Specific implementation method]

[0033] In order to facilitate the understanding of the present application, the present application is described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that when an element is described as "fixed to" another element, it can be directly on another element or there can be one or more centered elements therebetween. When an element is described as "connected" to another element, it can be directly connected to another element or there can be one or more centered elements therebetween. The orientation or positional relationship indicated by the terms "upper", "lower", "inner", "outer", "bottom" and the like used in this specification is based on the orientation or positional relationship shown in the accompanying drawings, only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second", "third" and the like are used only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0034] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used in this specification and in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0035] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0036] Typical lithium-ion battery fault detection methods are mainly divided into threshold-based methods and model-based methods. Among them, the threshold-based method refers to a detection method that sets the warning threshold of key battery parameters (such as voltage, temperature, internal resistance, etc.) and triggers and provides corresponding fault detection results when the measured parameters exceed or fall below the preset threshold.

[0037] Model-based methods refer to those detection methods that establish a mathematical model to describe the battery behavior, compare the actual measured data with the model prediction results, and determine whether there is a fault in the lithium-ion battery through parameter identification and state estimation.

[0038] In the process of implementing this application, the applicant found that the threshold-based method takes a long time to complete the fault diagnosis of lithium-ion batteries, and due to the dynamic characteristics of the battery, the judgment basis will change accordingly under different working conditions and life cycles, making it difficult to set an accurate threshold. Although the model-based method has a higher accuracy of detection results, its recognition capabilities in many aspects such as diagnosis time, diagnosis accuracy, and maximum identification of internal short-circuit equivalent resistance are still limited and cannot meet the needs of actual detection.

[0039] The fault detection method of the lithium-ion battery provided in the embodiment of the present application uses simulated voltage data, sampled real voltage data, current data and temperature data as training data to train the selected deep learning model to obtain a first mapping relationship that can respectively reflect the voltage, current, temperature at the current moment and the voltage at the future moment, and a second mapping relationship that reflects the simulated voltage and the real voltage. Based on the predicted voltage obtained by the first mapping relationship and the second mapping relationship, the short circuit fault in the battery is diagnosed more accurately and more quickly, and a high diagnostic accuracy is obtained under various operating conditions, achieving early detection and early warning.

[0040] Moreover, when training and learning the first mapping relationship and the second mapping relationship, it is convenient to flexibly predict voltage sequences of different time lengths by adjusting the length of input data, so that the predicted voltage has better accuracy.

[0041] Figure 1 A method flow chart of a fault detection method for a lithium-ion battery provided in an embodiment of the present application. The fault detection method is roughly divided into a mapping relationship training phase and a test use phase. When performing actual lithium-ion battery fault detection, after the fault detection method obtains the required first mapping relationship and second mapping relationship, the steps involved in the mapping relationship training phase are no longer repeatedly executed, but the first mapping relationship and second mapping relationship that have been stored are directly called, and the steps of the test use phase are executed to complete the fault detection.

[0042] In some embodiments, Figure 1 As shown, the mapping relationship training phase of the fault detection method includes the following steps:

[0043] S11, obtaining a plurality of sample data groups of normal lithium-ion batteries and a plurality of corresponding reference voltage data.

[0044] The sample data set refers to a combination of multiple data used to record and describe the working state of the lithium-ion battery within a specific time period, which is used as training data and is further used in the subsequent model training process.

[0045] In this embodiment, each sample data group includes: sample voltage data, sample current data, sample temperature data of the lithium-ion battery collected within a preset time period, and simulated voltage data simulated by a pre-built electrochemical mechanism model.

[0046] Specifically, the sample voltage data is the actual change characteristics of the lithium-ion battery voltage over time. The sample current data is the actual charge and discharge current of the lithium-ion battery. The sample temperature data is the change of the temperature of the lithium-ion battery over time. The simulation voltage data is the theoretical value of the lithium-ion battery voltage calculated by the electrochemical mechanism model.

[0047] The reference voltage data refers to the continuous observation data after the sample data group is collected. It is used as voltage reference data and as a reference for the predicted value output by the model. A reference voltage data is the voltage data collected by a normal lithium-ion battery within a preset time period after the end time of the corresponding sample data group.

[0048] S12: Using the sample voltage data, the sample current data, and the sample temperature data as input information, and using the reference voltage data as output information, train a deep learning algorithm to obtain a first mapping relationship.

[0049] The "first mapping relationship" is a time series prediction model based on historical data. By taking sample voltage data, sample current data and sample temperature data as input and reference voltage data as expected output, it can learn the voltage evolution law of lithium-ion batteries under different working conditions, thereby achieving accurate prediction of future voltage change trends.

[0050] S13: Using the simulation voltage data, the sample current data, and the sample temperature data as input information, and using the reference voltage data as output information, the deep learning algorithm is trained to obtain a second mapping relationship.

[0051] Among them, the second mapping relationship is obtained by taking the prediction results of the simulation theoretical model (simulated voltage data) together with the operating condition information (current and temperature data) as input and training with the reference voltage data as the expected output, which is a model reflecting the conversion relationship between theoretical prediction and actual measurement.

[0052] Specifically, the preset deep learning algorithm includes but is not limited to: convolutional neural network, fully connected network, recurrent neural network, long short-term memory neural network, attention mechanism and Transformer, etc.

[0053] In the above steps S12 and S13, the reliability and accuracy of the lithium-ion battery status assessment are effectively improved through the mutual complementation and coordination between the two different mapping relationships.

[0054] In some embodiments, different types of deep learning algorithms are selected and used in the above steps S12 and S13. In other embodiments, the same type of deep learning algorithm is used in steps S12 and S13.

[0055] S14, calculating a first error threshold based on the reference voltage data and the sampled voltage prediction value output by the first mapping relationship.

[0056] The first error threshold represents the deviation range between the voltage prediction value output by the first mapping relationship and the actual measured real voltage value, which can be used as an indicator for evaluating the prediction accuracy of the actual working state of the battery.

[0057] Specifically, the first error threshold is calculated by the following formula (1):

[0058]

[0059] Wherein, N is the length of the preset time period, B is the correlation coefficient, C is the charge or discharge rate of the battery, E1 is the first error threshold, U FP1 is the predicted value of the sampled voltage, U F It is the reference voltage data.

[0060] S15, calculating a second error threshold based on the reference voltage data and the simulated voltage prediction value output by the second mapping relationship.

[0061] The second error threshold is the deviation range between the voltage value predicted by the electrochemical mechanism model and the actual voltage value measured, which reflects the ability of the electrochemical mechanism model to explain and simulate the actual behavior of lithium-ion batteries.

[0062] Specifically, the second error threshold is calculated by the following formula (2):

[0063]

[0064] Wherein, N is the length of the preset time period, B is the correlation coefficient, C is the charge or discharge rate of the battery, E2 is the second error threshold, and U FP2 is the predicted value of the simulated voltage, U F It is the reference voltage data.

[0065] S16. Integrate the first error threshold and the second error threshold into a voltage prediction error threshold by a preset numerical integration method.

[0066] Among them, "numerical integration method" refers to a data processing method that combines multiple independent numerical indicators into a comprehensive indicator through a specific mathematical operation method. It specifically adopts simple arithmetic mean, weighted arithmetic mean, geometric mean, harmonic mean and other methods.

[0067] Specifically, the preset numerical integration method is weighted averaging. The weighted averaging method is an integration method that allocates different weight coefficients to the first error threshold and the second error threshold to reflect the importance and influence of the two in the voltage prediction error threshold.

[0068] By integrating the first error threshold and the second error threshold into a unified voltage prediction error threshold, a more comprehensive evaluation standard can be obtained by comprehensively considering the prediction errors in two different dimensions. Moreover, by reasonably allocating weights, the advantages and limitations of the first error threshold and the second error threshold are balanced, thereby improving the reliability and accuracy of the voltage prediction error threshold.

[0069] Please continue reading Figure 1 , the test use phase of the fault detection method includes the following steps:

[0070] S21, collecting multiple test data groups and multiple corresponding real voltage data of the lithium-ion battery to be tested.

[0071] The test data set is a combination of multiple data used to record and describe the working status of the lithium-ion battery to be tested within a specific time period. It is the initial collection data used for subsequent further fault diagnosis. The real voltage data refers to the continued observation data after the test data set is collected.

[0072] In this embodiment, each test data group includes: voltage data, current data and temperature data of the lithium-ion battery to be tested collected within a preset time period. The real voltage data is collected in the same manner as the reference voltage data described above, so that the test data group and the corresponding real voltage data have the same data structure as the sample data group and the corresponding reference voltage data.

[0073] S22: Calculate a first error based on the real voltage data and a first voltage prediction value calculated and output by the first mapping relationship.

[0074] Among them, the first error reflects the accuracy of predicting future voltage changes based on historical data, and is one of the indicators for determining whether the battery working state is abnormal.

[0075] Specifically, the first error is calculated in the same manner as the first error threshold value calculated in step S14. In other words, the first error is also calculated using the above formula (1) by substituting the first voltage prediction value and the actual voltage data.

[0076] S23: Calculate a second error based on the real voltage data and a second voltage prediction value calculated and outputted by a second mapping relationship.

[0077] The second error reflects the matching degree between the simulation theoretical model and the actual lithium-ion battery behavior, and reflects whether the lithium-ion battery deviates from the normal physical properties. It can also be used as an indicator of whether the battery working state is abnormal.

[0078] Specifically, the second error is calculated in the same manner as the second error threshold value calculated in step S15. In other words, the second error is also calculated by substituting the second voltage prediction value and the real voltage data into the above formula (2).

[0079] S24. Integrate the first error and the second error into a voltage prediction error through a numerical integration method.

[0080] Among them, the voltage prediction error is a comprehensive indicator that combines the first error based on timing prediction and the second error based on simulation theoretical model prediction. It utilizes the complementary advantages of the two prediction methods to obtain a more comprehensive error evaluation result and provide a more reliable basis for fault judgment.

[0081] S25. Determine a fault judgment result of the lithium-ion battery to be tested according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold through a preset logical relationship.

[0082] Among them, the "logical relationship" includes several judgment rules at multiple levels. Based on the comparison results between the first error, the second error, the voltage prediction error calculated in the above steps S22 to S24, and the voltage prediction error threshold value determined in advance in step S16, a series of multi-level judgment rules are used to comprehensively consider different indicators to form and determine the final fault judgment result of the lithium-ion battery.

[0083] The fault detection method provided in the embodiment of the present application flexibly predicts voltage sequences of different time lengths by adjusting the length of input data when training and learning the mapping relationship, and provides a predicted voltage value with higher accuracy.

[0084] In addition, the corresponding logical judgment is made based on the error of the predicted voltage value, and the short-circuit fault in the battery is diagnosed under a variety of different working conditions, so that high-precision diagnosis results can be obtained quickly, and early detection and early warning can be achieved, which well meets the needs of practical applications.

[0085] Figure 2 is a flow chart of a method for obtaining a sample data set and corresponding reference voltage data provided by an embodiment of the present application. Figure 2 As shown, the method for obtaining multiple sample data groups and multiple corresponding reference voltage data includes the following steps:

[0086] S111. Collect sample voltage data, sample current data, and sample temperature data of a normal lithium-ion battery that changes with operating time under different operating conditions.

[0087] Among them, "operating conditions" refers to the combination of test conditions and charge and discharge conditions of lithium-ion batteries during testing. Specifically, the test conditions are not limited to specific temperature, charge and discharge rate, ambient humidity, irradiation and other conditions, and the charging methods include constant current charging, constant current and constant voltage charging, multi-stage constant current charging, pulse charging, etc.

[0088] "Normal lithium-ion battery" refers to a battery that has not failed and is used as a reference for battery failure judgment. The specific type of lithium-ion battery, positive electrode material or negative electrode material, etc. are selected according to the actual needs.

[0089] "Run time" refers to the time interval from a specific start time t0 to the end time of data collection. For example, the start time t0 is the time when the lithium-ion battery starts to charge or discharge.

[0090] In addition, in order to ensure the richness and comprehensive coverage of training data and enable the trained model to better adapt to different usage scenarios, data under multiple different operating conditions are collected. In other words, the operating condition is also one of the parameters in the sample data.

[0091] S112. Obtain simulated voltage data of a normal lithium-ion battery under different operating conditions through a pre-constructed electrochemical mechanism model.

[0092] Among them, "simulated voltage data" is a theoretical voltage value calculated by an electrochemical mechanism model. There is a corresponding relationship between the simulation data under these different operating conditions and the sample voltage data collected in the above step S111. In other words, for each sample voltage data, the simulation voltage data under the same operating condition and the same time node is obtained by calculating the electrochemical mechanism model.

[0093] Specifically, the electrochemical mechanism model is an electrochemical mechanism model that includes thermal field simulation. “Electrochemical mechanism model that includes thermal field simulation” refers to a theoretical model that can simultaneously consider the electrochemical reaction process and the heat transfer process. This model is based on the basic working principle of lithium-ion batteries and simulates various physical and chemical processes inside the battery by solving a series of partial differential equations that describe electrochemical reactions, material transport, and heat transfer.

[0094] In actual application, the electrochemical mechanism model including thermal field simulation includes but is not limited to: simplifying the electrode into a single representative particle, mainly considering the solid phase diffusion process, with a small amount of calculation and suitable for low-rate working conditions, a quasi-two-dimensional model (Pseudo Two-Dimensional Model, P2D) that combines the concentration and potential distribution in the thickness direction of the electrode to describe the transmission process inside the battery, and further introduces temperature field calculation on the basis of the P2D model. Multi-dimensional and multi-field electrochemical models that can simulate the coupled response of batteries under various complex working conditions, as well as other suitable models formed by electrochemical theory.

[0095] S113, taking the start time of the operation time as the starting point, according to the preset time step, the sample voltage data, the sample current data, the sample temperature data and the simulation voltage data are divided into multiple data segments, and the data segments of the sample voltage data, the sample current data, the sample temperature data and the simulation voltage data in the same time period are combined into a sample data group.

[0096] The "time step" indicates the time interval length of a data segment. The time step is set to an appropriate length according to needs and different considerations (e.g., battery dynamic response characteristics, data sampling frequency, and prediction accuracy requirements, etc.). For example, 1 second, 10 seconds, 100 seconds, 1000 seconds, or other appropriate time lengths.

[0097] "Same time period" indicates the time correspondence between different data segments. For example, a data segment of a sample voltage data records the battery voltage data collected in a certain time interval. The data segments of sample current data, sample temperature data and simulated voltage data in the same time period refer to the current data, temperature data collected in the same time interval and the simulated voltage data calculated by the electrochemical mechanism model simulation.

[0098] Therefore, each sample data group contains complete information about the working status of the lithium-ion battery within a specific time period, including both the actual collected data and the corresponding theoretical calculated values, and can provide structured training data.

[0099] S114, taking the end time of each sample data group as the starting point, collecting the voltage data of the lithium-ion battery within a preset time period after the end time of the sample data group as reference voltage data corresponding to the sample data group.

[0100] The “end time” is both the end point of the acquisition time of a certain sample data group and the starting point of the acquisition of the reference voltage data, thereby establishing a connection relationship between the sample data group and its corresponding reference voltage data in the time series.

[0101] The reference voltage data is the actual voltage value collected within a preset time period after the end of the sample data group, which is used as a judgment basis for verifying data and accuracy. Specifically, the preset time period for collecting the reference voltage data is set to have an equal time length as the above time step, which is convenient for subsequent comparative analysis.

[0102] In other embodiments, multiple test data groups and multiple corresponding real voltage data of the lithium-ion battery to be tested are obtained by:

[0103] First, the voltage data, current data and sample temperature of the lithium-ion battery under test during a period of operation.

[0104] Then, taking the start time of the operation as the starting point, the voltage data, current data and temperature data are divided into multiple data segments according to the preset time step, and the data segments of the voltage data, current data and temperature data in the same time period are combined into a test data group.

[0105] Finally, taking the end time of each test data group as the starting point, the voltage data of the lithium-ion battery to be tested within a preset time period after the end time of the test data group is collected as the real voltage data corresponding to the test data group.

[0106] In other words, multiple test data groups and multiple corresponding real voltage data can be collected in a similar manner to executing the above steps S111, S113 and S114, thereby forming structured input information required for the first mapping relationship and the second mapping relationship.

[0107] Figure 3 A flow chart of a method for determining a fault judgment result of a lithium-ion battery to be tested provided in an embodiment of the present application. Figure 3 As shown, the method for determining the fault judgment result includes the following steps:

[0108] S251. Determine a difference between a first error and a voltage prediction error threshold, a difference between a second error and the voltage prediction error threshold, and a difference between the voltage prediction error and the voltage prediction error threshold.

[0109] Among them, the difference can quantify the degree of deviation of various error indicators relative to the voltage prediction error threshold. The difference between the first error and the voltage prediction error threshold reflects the difference between the lithium-ion battery to be tested and the normal lithium-ion battery as the standard under the prediction of the first mapping relationship. The difference between the second error and the voltage prediction error threshold reflects the difference between the lithium-ion battery to be tested and the normal lithium-ion battery as the standard under the prediction of the second mapping relationship. The difference between the voltage prediction error and the voltage prediction error threshold is a comprehensive evaluation indicator that takes into account the results of the first mapping relationship and the second mapping relationship.

[0110] S252: Determine whether at least two of the first error, the second error and the voltage prediction error are less than or equal to the voltage prediction error threshold. If so, execute step S253; if not, execute step S254.

[0111] Among them, when only one error indicator exceeds the voltage prediction error threshold, it may be caused by inaccurate model prediction or temporary interference factors, while when two or all error indicators exceed the voltage prediction error threshold, it is more certain that the current lithium-ion battery under test is different from the normal lithium-ion battery. Such a judgment mechanism effectively reduces the misjudgment rate and improves the reliability of fault diagnosis.

[0112] S253: Determine that the lithium-ion battery to be tested has no fault.

[0113] When there is no difference in the battery behaviors such as the electrochemical characteristics and working performance between the lithium-ion battery to be tested and the normal lithium-ion battery, it is considered that the lithium-ion battery to be tested should still be normal and no possible fault has occurred inside.

[0114] S254: Determine that the lithium-ion battery to be tested fails.

[0115] When there are differences in the electrochemical characteristics and working performance of the lithium-ion battery to be tested and normal lithium-ion batteries, it is considered that the lithium-ion battery to be tested has failed or is abnormal.

[0116] In other embodiments, after determining that the lithium-ion battery to be tested has failed, the fault classification of the lithium-ion battery to be tested is further determined based on the first error, the second error, and the value of the voltage prediction error exceeding the voltage prediction error threshold, thereby more carefully displaying and presenting the current fault state of the lithium-ion battery to be tested.

[0117] For example, when the exceeding values ​​(i.e., the absolute value of the difference) are all less than 50% of the voltage prediction error threshold, it is displayed as a mild fault, indicating that the battery performance is deteriorating. When the exceeding values ​​are between 50% and 100% of the voltage prediction error threshold, it is displayed as a moderate fault, indicating that the battery performance is significantly degraded. When the exceeding values ​​are greater than 100% of the voltage prediction error threshold, it is displayed as a serious fault, indicating that there is a safety hazard in the battery.

[0118] Based on the lithium-ion battery fault detection method provided in one or more of the above embodiments, the embodiment of the present application further applies it to the fault detection of a lithium-ion battery pack including multiple lithium-ion batteries to determine the current state of the lithium-ion battery pack.

[0119] Figure 4 A method for detecting a fault in a lithium-ion battery pack provided in an embodiment of the present application. In some embodiments, Figure 4 As shown, the fault detection method of the lithium-ion battery pack includes the following steps:

[0120] S31, determining a fault judgment result of each lithium-ion battery in the lithium-ion battery pack to be tested.

[0121] After the training phase is completed and the first mapping relationship, the second mapping relationship and the voltage prediction error threshold are determined, the fault judgment result of each lithium-ion battery is calculated and obtained by executing the above steps S21 to S23.

[0122] Specifically, the fault judgment result is a simple binary judgment result, that is, the lithium-ion battery has a fault or the lithium-ion battery has not a fault.

[0123] S32, analyzing differences between different lithium-ion batteries in the lithium-ion battery pack.

[0124] The “differences between different lithium-ion batteries” refer to the inconsistencies in electrochemical characteristics, working conditions, performance, etc. of each lithium-ion battery in the same lithium-ion battery pack. In the present application, the differences between different lithium-ion batteries are quantified and described using the lithium-ion battery fault detection method described in one or more of the above embodiments.

[0125] S33, determining the fault state of the lithium-ion battery pack to be tested according to the fault judgment result of each lithium-ion battery and the difference between different lithium-ion batteries.

[0126] When judging the fault status of a lithium-ion battery pack, the two dimensions of the fault judgment result of a single lithium-ion battery and the difference between different lithium-ion batteries are combined to form an effective complementary and cross-validation mechanism to make up for the blind spots that may exist in a single indicator. In this way, the problems existing in the entire lithium-ion battery pack can be more accurately distinguished, providing comprehensive and reliable fault diagnosis results.

[0127] Figure 5 A flow chart of a method for analyzing differences between different lithium-ion batteries in a lithium-ion battery pack provided in an embodiment of the present application. In some embodiments, Figure 5 As shown, the difference analysis method includes the following steps:

[0128] S321. Obtain historical detection data of one target lithium-ion battery among N lithium-ion batteries.

[0129] The target lithium-ion battery is one of the N lithium-ion batteries in the lithium-ion battery pack. N is a positive integer, indicating the number of lithium-ion batteries in the lithium-ion battery pack. The value of N is determined by the lithium-ion battery pack actually tested, for example, 3, 4, 5 or other positive integers.

[0130] S322: Use historical detection data of the target lithium-ion battery to retrain the first mapping relationship and the second mapping relationship used in the fault detection method to obtain new first mapping relationship and second mapping relationship.

[0131] Retraining refers to the process of processing historical detection data in an appropriate data processing manner to form structured training data in the same form as step S11, and then re-executing steps S12 and S13 to update the model parameters in the first mapping relationship and the second mapping relationship.

[0132] It can be understood that after such retraining, the new first mapping relationship and the new second mapping relationship obtained represent the battery behavior of the target lithium-ion battery.

[0133] S323: Based on the new first mapping relationship and the new second mapping relationship, the voltage prediction error corresponding to each lithium-ion battery is calculated respectively through the steps of the test use phase in the above-mentioned fault detection method.

[0134] The first error, second error and voltage prediction error calculated by the other lithium-ion batteries based on the new first mapping relationship and the second mapping relationship actually also represent the difference between the other lithium-ion batteries and the target lithium-ion battery. In other words, the first error, second error and voltage prediction error are used to quantify or describe the difference between the two lithium-ion batteries.

[0135] S324: After reselecting a new target lithium-ion battery from the remaining N-1 lithium-ion batteries, re-execute steps S321 to S323 until a voltage prediction error of each lithium-ion battery relative to other different lithium-ion batteries is obtained.

[0136] The above steps S321 to S324 describe that each lithium-ion battery is used as a target lithium-ion battery in turn by traversing one by one, and different first mapping relationships and second mapping relationships are trained, so as to obtain the differences between the lithium-ion batteries in the complete lithium-ion battery group.

[0137] This approach avoids the one-sidedness that may be caused by a single reference standard, and can more accurately reflect the real differences between different lithium-ion batteries, and reflect the uniqueness and degree of difference of a certain lithium-ion battery in a lithium-ion battery pack.

[0138] Figure 6 A flow chart of a method for determining a fault state of a lithium-ion battery pack to be tested provided in an embodiment of the present application. Figure 6 As shown, the method for determining the fault state of the lithium-ion battery pack to be tested includes the following steps:

[0139] S331. Select one of the N lithium-ion batteries as the lithium-ion battery to be determined.

[0140] The process of determining the fault state of the lithium-ion battery pack to be tested is carried out by testing the lithium-ion batteries one by one. N lithium-ion batteries are tested and the results are judged in turn. The lithium-ion battery currently selected for judgment is called the "lithium-ion battery to be judged".

[0141] S332, determining the fault judgment result of the lithium-ion battery to be judged. If the fault judgment result is that a fault occurs, executing step S333; if the fault judgment result is that no fault occurs, executing step S336.

[0142] The fault judgment result is a binary result, including two types: the lithium-ion battery fails and the lithium-ion battery does not fail. Specifically, the method of steps S251 to S254 described above is used to determine the specific fault judgment result.

[0143] S333: Determine whether the difference between the lithium-ion battery to be determined and other lithium-ion batteries meets the standard. If yes, execute step S334; if no, execute step S335.

[0144] Among them, "standard" is a set of pre-set logical judgment rules for judging and measuring differences. The standard consists of one or more judgment levels and integrates a variety of different indicators.

[0145] S334, determining that the fault state of the lithium-ion battery pack to be tested is a first-level alarm state.

[0146] S335, determining that the fault state of the lithium-ion battery pack to be tested is a level 2 alarm state.

[0147] Among them, "Level 1 alarm state" and "Level 2 alarm state" are the classification of the fault state of lithium-ion battery packs, indicating the degree of fault of the current lithium-ion battery pack. In other words, when the fault judgment result of a lithium-ion battery alone is that a fault has occurred, and there is a significant difference between the lithium-ion battery and other lithium-ion battery packs, it indicates that the lithium-ion battery has a greater risk, and the alarm state of the lithium-ion battery pack is set to a higher level of Level 2 alarm state.

[0148] When the fault judgment result of a certain lithium-ion battery is that there is no fault, but there is a significant difference between the lithium-ion battery and other lithium-ion battery packs, it indicates that there is an imbalance and difference problem inside the lithium-ion battery pack. The alarm state of the lithium-ion battery pack to be tested is set to a relatively low level one alarm state.

[0149] S336. Among the remaining N-1 lithium-ion batteries, reselect one as the lithium-ion battery to be determined, and execute steps S332 to S335.

[0150] S337: When the fault judgment results of the N lithium-ion batteries are all that no fault has occurred, determine that the lithium-ion battery pack to be tested has no fault.

[0151] Among them, after traversing all lithium-ion batteries and determining that there is little difference between these lithium-ion batteries and they are all in normal operation without failure, it indicates that the status of the lithium-ion battery pack is not faulty or normal, and no alarm state needs to be issued.

[0152] Figure 7 A flow chart of a method for determining whether the difference between a lithium-ion battery to be determined and other lithium-ion batteries meets the standard provided in an embodiment of the present application. Figure 7 As shown, the method for determining whether the difference meets the standard includes the following steps:

[0153] S3331. Calculate the voltage prediction error of the lithium-ion battery to be judged and other lithium-ion batteries using the first mapping relationship and the second mapping relationship that have been retrained with the historical detection data of the lithium-ion battery to be judged.

[0154] S3332: When it is determined that the voltage prediction error of the lithium-ion battery is less than a preset first value, and whether the voltage prediction error corresponding to each of the remaining lithium-ion batteries is greater than a preset second value, execute step S3333.

[0155] The first value and the second value are both preset values, which are set by technicians according to actual needs and are presented in any suitable form, including but not limited to percentages or absolute values.

[0156] The first value is a standard for measuring whether the predicted voltage error is small, and the second value is a standard for measuring whether the predicted voltage error is large. In other words, when the lithium-ion battery to be judged is the target lithium-ion battery, the voltage prediction error calculated by the lithium-ion battery to be judged is small, while the voltage errors calculated by the remaining lithium-ion batteries are large. Specifically, the value of the first value is smaller than the second value.

[0157] S3333. Recalculate the voltage prediction errors of the lithium-ion battery to be judged and other lithium-ion batteries using the first mapping relationship and the second mapping relationship that have been retrained with other different lithium-ion battery historical detection data.

[0158] S3334: When the voltage prediction error of the lithium-ion battery to be determined is greater than the second value, and the voltage prediction errors of other lithium-ion batteries are less than the first value, determine that the differences between different lithium-ion batteries do not meet the standard.

[0159] Among them, the above steps S3332 and S3334 set judgment conditions in two directions, thus forming a more rigorous difference measurement standard. Only when the judgment conditions of steps S3332 and S3334 are met at the same time, it is confirmed that the differences between different lithium-ion batteries do not meet the standards, that is, it is relatively certain that there are obvious differences between different lithium-ion batteries. Otherwise, it is considered that the differences between different lithium-ion batteries meet the standards and there are no obvious differences between them.

[0160] Such two-way verification can verify each other, reduce the risk of misjudgment and ensure the stability and credibility of the difference judgment results.

[0161] The following provides a specific example that fully illustrates and describes the above fault detection method in detail. The specific example shows in detail the complete fault detection process of obtaining from the sample data group, training to obtain the first mapping relationship and the second mapping relationship, determining the fault judgment result of a single lithium-ion battery, and then comprehensively determining the fault state of the lithium-ion battery group based on the fault judgment result of the single lithium-ion battery and the differences between different lithium-ion batteries.

[0162] Fig. 8A A method flow chart of the mapping relationship training phase of the fault detection method provided in an embodiment of the present application. Figure 8B This is a method flow chart of the test use phase of the fault detection method provided in the embodiment of the present application. Fig. 8AAs shown, the steps of the mapping relationship training phase include:

[0163] 1) Train to obtain the first mapping relationship and the second mapping relationship:

[0164] S101, subjecting a normal lithium-ion battery to charge and discharge cycles under different temperatures, humidity, charging schemes, and discharging schemes, to obtain voltage (U), current (I), and temperature (T) under different aging states and charging and discharging conditions.

[0165] The charging scheme includes, but is not limited to, constant current charging, constant current constant voltage charging, multi-stage constant current charging, and pulse charging. The normal lithium-ion battery is a suitable type of lithium-ion battery selected according to the scene requirements.

[0166] S102, constructing a quasi-two-dimensional model including thermal field simulation, and obtaining the simulated battery voltage (U) of step S101 under the same charge and discharge conditions through the quasi-two-dimensional model including thermal field simulation. S ).

[0167] S103, starting from the start time of operation, dividing U, I, and T during the charging and discharging process of the lithium-ion battery into a data segment every 30 seconds.

[0168] Among them, Fig. 9 As shown, the data segments of U, I, and T in each same time period are combined into a data group, which is marked as X1.

[0169] S104, starting from the end time of each X1, and then continuing to collect the voltage data of the lithium-ion battery for the next 30 seconds (U F ).

[0170] Please continue to read Fig. 9 , the voltage data U collected 30 seconds after each X1 F Combined into one data set, labeled Y.

[0171] S105, outputting the electrochemical mechanism model simulation including the thermal field simulation S Divide into corresponding data segments according to the same time step as each X1.

[0172] Please continue to read Fig. 9 , U of each data fragment S Denoted as X2.

[0173] S106, based on the above data sets X1 and Y, with U, I, T as input, U F As an output, a long short-term memory neural network algorithm with a self-attention mechanism is trained to obtain a first mapping relationship ALG1.

[0174] The first mapping relationship ALG1 reflects the mapping relationship between the voltage, current, temperature at the current moment and the voltage at the future moment. Therefore, based on the first mapping relationship ALG1, the future sampled voltage prediction value UFP1 of the lithium-ion battery is calculated.

[0175] S107, based on the above data sets X2 and X1, using a quasi-two-dimensional model including thermal field simulation to simulate the voltage U S , I, T as input, with U F As output, a dual algorithm consisting of a long short-term memory neural network with a self-attention mechanism and a fully connected neural network algorithm is trained to obtain a second mapping relationship ALG2.

[0176] The second mapping relationship ALG2 reflects the mapping relationship between the simulated voltage and the real voltage at the future moment. Therefore, based on the second mapping relationship ALG2, the future voltage UFP2 simulated by the electrochemical mechanism model is calculated and obtained.

[0177] For example, Fig.10 A comparison chart of the predicted voltage value calculated using the first mapping relationship ALG1 and the actual voltage value obtained by acquisition is shown. Fig.11 The comparison chart of the predicted voltage value calculated using the second mapping relationship ALG2 and the actual voltage value obtained by acquisition is shown. Fig.10 and Fig.11 As shown, the error between the predicted voltage value and the actual voltage value of the two mapping relationships is very small, and has high accuracy.

[0178] 2) Determine the voltage prediction error threshold:

[0179] S201. Calculate the first error threshold using the following formula (1):

[0180]

[0181] Wherein, N is the length of the data segment, B is the correlation coefficient (set to 1 by default), and C is the charge or discharge rate of the battery.

[0182] S202: Calculate the second error threshold using the following formula (2):

[0183]

[0184] Wherein, N is the length of the data segment, B is the correlation coefficient (set to 1 by default), and C is the charge or discharge rate of the battery.

[0185] S203 , performing weighted averaging on the first error threshold E1 and the second error threshold E2 to obtain a voltage prediction error threshold E3 .

[0186] like Figure 8B As shown, the test use phase includes the following steps:

[0187] 3) Detection of lithium-ion batteries:

[0188] S301, subjecting the lithium-ion battery to be tested to charge and discharge cycles under different temperatures, humidity, charging schemes and discharging schemes, respectively, to obtain the voltage (U GZ ), current (I GZ ) and temperature (T GZ )data.

[0189] Among them, the charging scheme includes constant current charging, constant current constant voltage charging, multi-stage constant current charging, pulse charging, etc. The lithium-ion battery to be tested can be selected according to the needs of the application scenario to use lithium-ion batteries with different faults such as lithium deposition, thermal runaway, internal short circuit, external short circuit, etc. to verify and test the reliability and accuracy of the fault detection method.

[0190] S302 : Using the same method as steps S101 to S105 , obtain fault test data having the same data structure as the training data.

[0191] S303: Based on the acquired fault test data, the first mapping relationship ALG1 and the second mapping relationship ALG2 are used to calculate and obtain the first voltage prediction value U fp3 and the second voltage prediction value U fp4 .

[0192] S304: respectively calculate the first voltage prediction value U fp3 and the second voltage prediction value U fp4 Substituting into the above equations (1) and (2), the corresponding first error E is calculated. GZ1 and the second error E GZ2 .

[0193] S305, for the first error E GZ1 and the second error E GZ2 Perform weighted averaging to obtain the voltage prediction error E GZ3 .

[0194] S306, at the first error E GZ1 , the second error E GZ2 and voltage prediction error E GZ3 When the values ​​of any two of the above are greater than the voltage prediction error threshold E3, it is determined that the lithium-ion battery currently being tested has a fault.

[0195] If only one value is greater than the voltage prediction error threshold E3 or all three values ​​are less than the voltage prediction error threshold E3, it is determined that the battery behavior and characteristics of the lithium-ion battery currently being tested are consistent with those of a normal lithium-ion battery and no fault occurs.

[0196] S307 , after determining that the lithium-ion battery currently being tested has a fault, classify the fault according to the value exceeding the voltage prediction error threshold E3 .

[0197] Specifically, the fault classification result of the lithium-ion battery currently being tested is output by the corresponding battery management system or energy management system, etc., and the corresponding battery classification fault state is presented and displayed to the user.

[0198] 4) Detection of lithium-ion battery packs:

[0199] In this embodiment, a lithium-ion battery pack is taken as an example, which is composed of three lithium-ion batteries connected in series. Figure 8B , the testing process for lithium-ion battery packs includes:

[0200] S401, using the above steps S301 to S307, respectively and sequentially calculate the first error, the second error and the voltage prediction error of three lithium-ion battery cells.

[0201] Among them, Fig.12 As shown, the first error, the second error and the voltage prediction error of the first lithium-ion battery are marked as E GZB11 、E GZB12 and E GZB13 The first error, second error, and voltage prediction error of the second lithium-ion battery are marked as E GZB21 、E GZB22 and E GZB23 The first error, second error, and voltage prediction error of the third lithium-ion battery are marked as E GZB31 、E GZB32 and E GZB33 .

[0202] S402 , retraining the first mapping relationship ALG1 and the second mapping relationship ALG2 using the historical operation data of the first lithium-ion battery, the second lithium-ion battery, and the third lithium-ion battery in turn to obtain three new first mapping relationships ALG1 and second mapping relationships ALG2 .

[0203] S403 , using the new first mapping relationship ALG1 and the first mapping relationship ALG2 in sequence, and performing the same steps S03 to S305 as above, to calculate and obtain the voltage prediction errors of the three lithium-ion batteries.

[0204] Please continue to read Fig.12 Based on the first mapping relationship ALG1 and the second mapping relationship ALG2 obtained by retraining the historical operation data of the first lithium-ion battery, the voltage prediction errors of the first lithium-ion battery, the second lithium-ion battery, and the third lithium-ion battery are respectively marked as E B11 , E B12 , E B13 .

[0205] Based on the first mapping relationship ALG1 and the second mapping relationship ALG2 obtained by retraining the historical operation data of the second lithium-ion battery, the voltage prediction errors of the first lithium-ion battery, the second lithium-ion battery, and the third lithium-ion battery are respectively marked as E B21 , E B22 , E B23 .

[0206] Based on the first mapping relationship ALG1 and the second mapping relationship ALG2 obtained by retraining the historical operation data of the third lithium-ion battery, the voltage prediction errors of the first lithium-ion battery, the second lithium-ion battery, and the third lithium-ion battery are respectively marked as E B31 , E B32 , E B33 .

[0207] S404, combining the error values ​​calculated from the above steps S401 to S403, determining the state of the lithium-ion battery pack to be tested and locating the faulty lithium-ion battery.

[0208] Please continue to read Fig.12 The status of the lithium-ion battery pack includes: the first lithium-ion battery is not faulty, the second lithium-ion battery is not faulty, the third lithium-ion battery is not faulty, the first level alarm state, and the second level alarm state. The logical judgment relationship between the status of the lithium-ion battery pack and the positioning result of the faulty lithium-ion battery and the above-mentioned multiple error values ​​is as follows: Fig.12 shown.

[0209] Specifically, for one of the lithium-ion batteries, if at least two of the first error, the second error and the voltage prediction error are greater than the voltage prediction error threshold E3, it is determined that the lithium-ion battery has failed. Otherwise, it is determined that the lithium-ion battery has not failed.

[0210] Further, when it is determined that a lithium-ion battery fails, whether to issue a secondary alarm state is determined by the following two conditions. On the one hand, if the voltage prediction error of the other two lithium-ion batteries calculated based on the lithium-ion battery is large (for example, greater than the aforementioned second value), the voltage prediction error of the lithium-ion battery is small (for example, less than the aforementioned first value). On the other hand, when the other two lithium-ion batteries are used as the reference, the voltage prediction error of the lithium-ion battery calculated based on the other two lithium-ion batteries is large (for example, greater than the aforementioned second value), while the voltage prediction error distribution of the other lithium-ion batteries is the same, all of which are small errors (for example, less than the aforementioned first value).

[0211] When the above two judgment conditions are met, the second level alarm state is issued. On the contrary, when the above judgment conditions are not met, the first level alarm state is issued.

[0212] Based on the lithium ion battery fault detection method provided in the above embodiment, the present application embodiment further provides a lithium ion battery fault detection device. Fig.13 This is a functional block diagram of a fault detection device provided in an embodiment of the present application.

[0213] In some embodiments, Fig.13 As shown, the fault detection device 10 includes: a sample data acquisition module 11, a mapping relationship training module 12, a threshold calculation module 13, a test data collection module 14, an error calculation module 15 and a judgment module 16.

[0214] The sample data acquisition module 11 is used to acquire multiple sample data groups and multiple corresponding reference voltage data of normal lithium-ion batteries; each sample data group includes: sample voltage data, sample current data, sample temperature data collected by the normal lithium-ion battery within a preset time period, and simulated voltage data simulated by a pre-constructed electrochemical mechanism model, and one of the reference voltage data is the voltage data collected by the normal lithium-ion battery within a preset time period after the end time of the corresponding sample data group. The mapping relationship training module 12 is used to train a preset deep learning algorithm with the sample voltage data, the sample current data and the sample temperature data as input information, and the reference voltage data as output information to obtain a first mapping relationship; and to train the deep learning algorithm with the simulated voltage data, the sample current data and the sample temperature data as input information, and the reference voltage data as output information to obtain a second mapping relationship. The threshold calculation module 13 is used to calculate the first error threshold based on the reference voltage data and the sampled voltage prediction value calculated and output by the first mapping relationship, and to calculate the second error threshold based on the reference voltage data and the simulated voltage prediction value calculated and output by the second mapping relationship; the first error threshold and the second error threshold are integrated into the voltage prediction error threshold through a preset numerical integration method. The test data acquisition module 14 is used to collect multiple test data groups and multiple corresponding real voltage data of the lithium-ion battery to be tested; each test data group includes: voltage data, current data and temperature data collected by the lithium-ion battery to be tested within a preset time period. The error calculation module 15 is used to calculate the first error based on the real voltage data and the first voltage prediction value calculated and output by the first mapping relationship, and to calculate the second error based on the real voltage data and the second voltage prediction value calculated and output by the second mapping relationship; the first error and the second error are integrated into the voltage prediction error through the numerical integration method. The judgment module 16 is used to determine the fault judgment result of the lithium-ion battery to be tested according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold through a preset logical relationship.

[0215] The fault detection device flexibly predicts voltage sequences of different time lengths by adjusting the length of input data, and has the characteristics of high voltage prediction accuracy. Moreover, based on the predicted voltage, the short circuit fault in the battery can be diagnosed more accurately and quickly, and a high diagnostic accuracy can be obtained under various working conditions, realizing early detection and early warning.

[0216] In some embodiments, the sample data acquisition module 11 is specifically used to: collect sample voltage data, sample current data and sample temperature data of the normal lithium-ion battery that change with the operating time under different operating conditions; obtain the simulated voltage data of the normal lithium-ion battery under the different operating conditions through a pre-constructed electrochemical mechanism model; take the start time of the operating time as the starting point, and divide the sample voltage data, the sample current data, the sample temperature data and the simulated voltage data into multiple data segments according to a preset time step; the time step is equal to the length of the preset time period; the data segments of the sample voltage data, the sample current data, the sample temperature data and the simulated voltage data of the same time period are combined into a sample data group; take the end time of each sample data group as the starting point, collect the voltage data of the lithium-ion battery within a preset time period after the end time of the sample data group as the reference voltage data corresponding to the sample data group.

[0217] Specifically, the electrochemical mechanism model is an electrochemical mechanism model that includes thermal field simulation, including: single particle model, quasi-two-dimensional model and multi-dimensional multi-field electrochemical model. The deep learning algorithm includes: convolutional neural network, fully connected network, recurrent neural network, long short-term memory neural network, attention mechanism and Transformer.

[0218] In some embodiments, the threshold calculation module 13 is specifically used to: calculate the first error threshold by formula (1), calculate the second error threshold by formula (2), and integrate the first error threshold and the second error threshold into the voltage prediction error threshold by weighted averaging.

[0219] In some embodiments, the judgment module 16 is specifically used to: determine that the lithium-ion battery to be tested has not failed when at least two of the first error, the second error and the voltage prediction error are less than or equal to the voltage prediction error threshold; determine that the lithium-ion battery to be tested has failed when at least two of the first error, the second error and the voltage prediction error are greater than the voltage prediction error threshold; when it is determined that the lithium-ion battery to be tested has failed, determine the fault classification of the lithium-ion battery to be tested according to the values ​​of the first error, the second error and the voltage prediction error exceeding the voltage prediction error threshold.

[0220] It should be noted that the fault detection device provided in the embodiment of the present application can be executed by any suitable type of electronic computing platform. The electronic computing platform can implement or perform one or more of the above functions with electronic hardware, computer software program or a combination of the two. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0221] Fig.14 A schematic diagram of a control terminal provided in an embodiment of the present application. The control terminal can be used to perform one or more steps of the fault detection method described in one or more embodiments above. The control terminal is implemented by various forms of digital computers, such as workstations, servers, blade servers, mainframe computers, and other appropriate electronic computing platforms.

[0222] It should be noted that Fig.14 The components, their connections and relationships, and their functions shown are for illustration only and are not intended to limit the specific implementation of the control terminal.

[0223] like Fig.14 As shown, the control terminal 20 includes a processor 21, a memory 22, a storage device 23, a high-speed interface 25 connected to the memory 22 and multiple high-speed expansion ports 24, and a low-speed interface 27 connected to the low-speed expansion port 26 and the storage device 23.

[0224] Each of the processor 21, memory 22, storage device 23, high-speed interface 25, high-speed expansion port 24, and low-speed interface 27 are interconnected using various buses and mounted on a common motherboard or other suitable means.

[0225] The processor 21 processes computer program instructions stored in the memory 22 or on the storage device 23 for executing computer program instructions for displaying graphic information on an external input / output device (eg, a display device 28 coupled to the high-speed interface 25).

[0226] In some embodiments, multiple processors and / or multiple buses are also used together with multiple memories and multiple types of memories. Also, multiple electronic devices are connected, each of which provides part of the necessary operations (e.g., as a server group, blade server group, or multi-processor system).

[0227] The memory 22 stores information within the control terminal and is one or more volatile memory units, non-volatile memory units, or another form of computer-readable media, such as a magnetic disk or optical disk.

[0228] The storage device 23 can provide mass storage for the control terminal. It contains computer readable media, such as a floppy disk device, a hard disk device, an optical disk device or a tape device, a flash memory or other similar solid-state memory device, or a device array, including devices in a storage area network or other configurations.

[0229] The computer program instructions are stored in an information carrier. When the computer program instructions are executed by one or more processing devices (such as the processor 21), the fault detection method described in one or more embodiments above is implemented.

[0230] The high-speed interface 25 manages bandwidth-intensive operations for controlling the terminal, while the low-speed interface 27 manages less bandwidth-intensive operations. In some embodiments, the high-speed interface 25 is coupled to the memory 22, the display 28 (e.g., via a graphics processor or accelerator), and the high-speed expansion port 24 that accepts various expansion cards. The low-speed interface 27 is coupled to the storage device 23 and the low-speed expansion port 26.

[0231] The low-speed expansion port 26 includes a communication port (eg, USB, Ethernet, wireless Ethernet) and can be coupled to one or more input / output devices, such as a keyboard, pointing device, scanner, or networking device such as a switch (eg, through a network adapter).

[0232] For example, Fig.14 As shown, in order to provide interaction with the user, the control terminal 20 also has a display device 28 (e.g., a cathode ray tube or a liquid crystal display monitor) for displaying information to the user and a pointing device 29 (e.g., a mouse) for the user to provide input to the computer. Other types of interactive devices can also be used to provide interaction with the user;

[0233] Of course, the feedback provided to the user is any form of sensory feedback (eg, visual feedback, auditory feedback, or tactile feedback); and input from the user is received in any form, including acoustic, voice, or tactile input.

[0234] The fault detection method described in one or more embodiments of the present application is implemented in a digital electronic circuit, an integrated circuit, a specially designed ASIC (application-specific integrated circuit), computer hardware, firmware, software and / or a combination thereof. It includes an embodiment in one or more computer programs, which are executed and / or interpreted on a programmable system, wherein the programmable system includes at least one programmable processor, which can be dedicated or general-purpose, coupled to receive data and instructions from a storage system, at least one input device, and at least one output device, and send data and instructions to the storage system, at least one input device, and at least one output device.

[0235] These computer programs (also referred to as programs, software, software applications or codes) include machine instructions for programmable processors and are implemented in high-level procedural and / or object-oriented programming languages ​​and / or in assembly / machine languages. In the present embodiment, the above-mentioned machine-readable medium and computer-readable medium refer to any computer program product, device and / or device (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The above-mentioned machine-readable signals refer to any signals for providing machine instructions and / or data to a programmable processor.

[0236] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Under the concept of the present application, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other changes in different aspects of the present application as described above, which are not provided in detail for the sake of simplicity. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A lithium-ion battery fault detection method, characterized in that: include: Acquire multiple sample data groups of normal lithium-ion batteries and multiple corresponding reference voltage data; Each sample data group includes: sample voltage data, sample current data, sample temperature data collected by the normal lithium-ion battery within a preset time period, and simulated voltage data simulated by a pre-built electrochemical mechanism model, and one reference voltage data is voltage data collected by the normal lithium-ion battery within a preset time period after the end time of the corresponding sample data group; Taking the sample voltage data, the sample current data and the sample temperature data as input information and taking the reference voltage data as output information, training a deep learning algorithm to obtain a first mapping relationship; Taking the simulation voltage data, the sample current data and the sample temperature data as input information and taking the reference voltage data as output information, training a deep learning algorithm to obtain a second mapping relationship; Calculating a first error threshold based on the reference voltage data and the sampled voltage prediction value calculated and output by the first mapping relationship, and calculating a second error threshold based on the reference voltage data and the simulated voltage prediction value calculated and output by the second mapping relationship; Integrating the first error threshold and the second error threshold into a voltage prediction error threshold by a preset numerical integration method; Collect multiple test data groups and multiple corresponding real voltage data of the lithium-ion battery to be tested; each test data group includes: voltage data, current data and temperature data collected from the lithium-ion battery to be tested within a preset time period; Calculating a first error based on the real voltage data and a first voltage prediction value calculated and outputted by the first mapping relationship, and calculating a second error based on the real voltage data and a second voltage prediction value calculated and outputted by the second mapping relationship; By using the numerical integration method, the first error and the second error are integrated into a voltage prediction error; According to the first error, the second error, the voltage prediction error and the voltage prediction error threshold, a fault judgment result of the lithium-ion battery to be tested is determined through a preset logical relationship.

2. The fault detection method according to claim 1, characterized in that: The step of obtaining a plurality of sample data groups of normal lithium-ion batteries and a plurality of corresponding reference voltage data specifically includes: Collecting sample voltage data, sample current data, and sample temperature data of the normal lithium-ion battery that changes with operating time under different operating conditions; Obtaining simulated voltage data of the normal lithium-ion battery under the different operating conditions through a pre-constructed electrochemical mechanism model; Taking the start time of the running time as the starting point, the sample voltage data, the sample current data, the sample temperature data and the simulation voltage data are respectively divided into a plurality of data segments according to a preset time step; the time step is equal to the length of the preset time period; The data segments of the sample voltage data, the sample current data, the sample temperature data and the simulation voltage data in the same time period are combined into one sample data group; Taking the end time of each sample data group as a starting point, voltage data of the lithium-ion battery within a preset time period after the end time of the sample data group is collected as reference voltage data corresponding to the sample data group.

3. The fault detection method according to claim 1, characterized in that: The electrochemical mechanism model is an electrochemical mechanism model including thermal field simulation, including: a single particle model, a quasi-two-dimensional model and a multi-dimensional multi-field electrochemical model.

4. The fault detection method according to claim 1, characterized in that: The deep learning algorithm includes: convolutional neural network, fully connected network, recurrent neural network, long short-term memory neural network, attention mechanism and transformer.

5. The fault detection method according to claim 1, characterized in that: The first error threshold is calculated by the following formula: Wherein, N is the length of the preset time period, B is the correlation coefficient, C is the charge or discharge rate of the battery, E1 is the first error threshold, U FP1 is the predicted value of the sampled voltage, U F is the reference voltage data; The second error threshold is calculated by the following formula: Wherein, N is the length of the preset time period, B is the correlation coefficient, C is the charge or discharge rate of the battery, E2 is the second error threshold, and U FP2 is the predicted value of the simulated voltage, U F is the reference voltage data; The preset numerical integration method is: weighted average.

6. The fault detection method according to claim 1, characterized in that: The method of determining the fault judgment result of the lithium-ion battery to be tested according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold through a preset logical relationship specifically includes: When at least two of the first error, the second error and the voltage prediction error are less than or equal to the voltage prediction error threshold, determining that the lithium-ion battery to be tested has no fault; When at least two of the first error, the second error and the voltage prediction error are greater than the voltage prediction error threshold, determining that the lithium-ion battery to be tested fails; When it is determined that the lithium-ion battery to be tested fails, the fault classification of the lithium-ion battery to be tested is determined according to the first error, the second error, and the value of the voltage prediction error exceeding the voltage prediction error threshold.

7. A method for detecting a fault of a lithium-ion battery pack, the lithium-ion battery pack comprising a plurality of lithium-ion batteries, characterized in that: The method comprises: Using the lithium-ion battery fault detection method according to any one of claims 1 to 6, respectively obtaining a fault judgment result of each lithium-ion battery in the lithium-ion battery pack to be tested; Analyzing differences between different lithium-ion batteries in the lithium-ion battery pack based on the lithium-ion battery fault detection method; The fault state of the lithium-ion battery pack to be tested is determined according to the fault judgment result of each lithium-ion battery and the difference between different lithium-ion batteries.

8. The fault detection method according to claim 7, characterized in that: The fault detection method based on the lithium-ion battery analyzes the differences between different lithium-ion batteries in the lithium-ion battery pack, specifically including: Acquire historical detection data of one target lithium-ion battery among the plurality of lithium-ion batteries; Retraining the first mapping relationship and the second mapping relationship in the lithium-ion battery fault detection method using the historical detection data; Based on the retrained first mapping relationship and the second mapping relationship, respectively calculating the voltage prediction error corresponding to each lithium-ion battery; Determining the fault state of the lithium-ion battery pack to be tested according to the fault judgment result of each lithium-ion battery and the difference between different lithium-ion batteries specifically includes: When the fault judgment result of the target lithium-ion battery is that a fault occurs, and the difference between different lithium-ion batteries does not meet the standard, determining that the fault state of the lithium-ion battery pack to be tested is a secondary alarm state; When the fault judgment result of the target lithium-ion battery is that a fault occurs, and the difference between different lithium-ion batteries meets the standard, determining that the fault state of the lithium-ion battery pack to be tested is a first-level alarm state; Among them, when the voltage prediction error corresponding to the target lithium-ion battery calculated using the first mapping relationship and the second mapping relationship retrained with the historical detection data of other lithium-ion batteries is greater than the preset second value, and when the voltage prediction error corresponding to the remaining lithium-ion batteries calculated using the first mapping relationship and the second mapping relationship retrained with the historical detection data of the target lithium-ion battery is greater than the preset second value, it is determined that the differences between different lithium-ion batteries do not meet the standard.

9. A lithium-ion battery fault detection device, characterized in that: include: A sample data acquisition module, used to acquire multiple sample data groups of normal lithium-ion batteries and multiple corresponding reference voltage data; Each sample data group includes: sample voltage data, sample current data, sample temperature data collected by the normal lithium-ion battery within a preset time period, and simulated voltage data simulated by a pre-built electrochemical mechanism model, and one reference voltage data is voltage data collected by the normal lithium-ion battery within a preset time period after the end time of the corresponding sample data group; a mapping relationship training module, configured to train a preset deep learning algorithm using the sample voltage data, the sample current data, and the sample temperature data as input information and the reference voltage data as output information to obtain a first mapping relationship; and to train the deep learning algorithm using the simulation voltage data, the sample current data, and the sample temperature data as input information and the reference voltage data as output information to obtain a second mapping relationship; a threshold calculation module, configured to calculate a first error threshold based on the sampled voltage prediction value calculated and outputted by the reference voltage data and the first mapping relationship, and to calculate a second error threshold based on the simulated voltage prediction value calculated and outputted by the reference voltage data and the second mapping relationship; and to integrate the first error threshold and the second error threshold into a voltage prediction error threshold by a preset numerical integration method; A test data acquisition module is used to acquire multiple test data groups and multiple corresponding real voltage data of the lithium-ion battery to be tested; each test data group includes: voltage data, current data and temperature data acquired by the lithium-ion battery to be tested within a preset time period; an error calculation module, configured to calculate a first error based on the real voltage data and a first voltage prediction value calculated and outputted by the first mapping relationship, and to calculate a second error based on the real voltage data and a second voltage prediction value calculated and outputted by the second mapping relationship; and to integrate the first error and the second error into a voltage prediction error by using the numerical integration method; The judgment module is used to determine the fault judgment result of the lithium-ion battery to be tested according to the first error, the second error, the voltage prediction error and the voltage prediction error threshold through a preset logical relationship.

10. An electronic device, characterized in that: include: A processor and a memory, wherein a non-volatile storage medium is stored in the memory, and when the non-volatile storage medium is called by the processor, the processor executes the lithium-ion battery fault detection method according to any one of claims 1 to 6.

Images