Battery abnormity type determination method and device and electronic equipment

Abstract

The invention discloses a battery abnormity type determination method and apparatus, and an electronic device. The method comprises the following steps: acquiring multiple pieces of voltage data of a target battery; according to the plurality of voltage signal sequences, determining a plurality of frequency domain signal components corresponding to the target battery in the plurality of predetermined electric quantity intervals; determining correlation coefficients of the plurality of voltage signal sequences and the plurality of corresponding frequency domain signal components; determining a target voltage characteristic of the target battery according to the plurality of correlation coefficients; and determining a target abnormality type corresponding to the target battery according to the target voltage characteristic. According to the method and the device, the technical problem that the determined abnormal type of the battery is not accurate due to the fact that the determined prediction model is low in precision in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method, device and electronic device for determining a battery abnormality type. Background Art

[0002] Determining the type of battery anomaly provides a key basis for accurately locating the root cause of the problem, which is crucial for taking timely and effective measures, ensuring battery performance and safety, and extending battery life. Currently, methods based on dimensionality reduction and feature selection are primarily used to determine prediction models for battery anomaly types. However, these methods often produce low-precision prediction models, leading to inaccurately determined battery anomaly types.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a method, device, and electronic device for determining a battery abnormality type, to at least solve the technical problem in related technologies of inaccurately determining a battery abnormality type due to low accuracy of a determined prediction model.

[0005] According to one aspect of an embodiment of the present invention, a method for determining a battery abnormality type is provided, comprising: acquiring multiple voltage data of a target battery, wherein the multiple voltage data include voltage signal sequences corresponding to multiple predetermined power intervals of the target battery during charging; determining, based on the multiple voltage signal sequences, multiple frequency domain signal components corresponding to the target battery in the multiple predetermined power intervals, wherein the corresponding multiple frequency domain signal components are signal components corresponding to the corresponding voltage signal sequence in multiple predetermined frequency domain intervals; determining correlation coefficients between the multiple voltage signal sequences and the corresponding multiple frequency domain signal components; determining, based on the multiple correlation coefficients, a target voltage characteristic of the target battery; and determining, based on the target voltage characteristic, a target abnormality type corresponding to the target battery.

[0006] Optionally, the target voltage characteristics of the target battery are determined based on multiple correlation coefficients, including: determining the target signal components corresponding to the multiple predetermined power intervals based on the correlation coefficients between the multiple voltage signal sequences and the corresponding multiple frequency domain signal components; determining the voltage frequency domain characteristics corresponding to the multiple predetermined power intervals based on the target signal components corresponding to the multiple predetermined power intervals; determining the voltage time domain characteristics corresponding to the multiple predetermined power intervals based on the voltage signal sequences corresponding to the multiple predetermined power intervals; and determining the target voltage characteristics of the target battery based on the voltage frequency domain characteristics and voltage time domain characteristics corresponding to the multiple predetermined power intervals.

[0007] Optionally, determining the target abnormality type corresponding to the target battery based on the target voltage characteristic includes: retrieving an abnormality type determination model corresponding to the target voltage characteristic, wherein the abnormality type determination model is provided with target parameters, and the target parameters are determined based on the sample abnormality types corresponding to a plurality of sample batteries and the sample voltage characteristics corresponding to the plurality of sample batteries; determining a plurality of abnormality type probability values ​​corresponding to the target battery based on the target voltage characteristic and the abnormality type determination model, wherein the plurality of abnormality type probability values ​​are respectively used to indicate that the battery abnormality type corresponding to the target battery is a probability value corresponding to a predetermined battery abnormality type; determining the target abnormality type corresponding to the target battery based on the plurality of abnormality type probability values.

[0008] Optionally, before calling the abnormality type determination model corresponding to the target voltage feature, it also includes: determining an initial determination model, wherein the initial determination model is set with initial parameters; calling the target loss function corresponding to the initial determination model, wherein the target loss function includes a data loss term and a parameter loss term, the data loss term is used to determine the loss value between the battery abnormality type determined according to the initial parameters and the corresponding real abnormality type, and the parameter loss term is used to determine the loss value of the battery abnormality type determined according to the first voltage feature, the first voltage feature is the voltage feature determined by selecting the voltage feature input to the initial determination model according to the initial parameters; based on the sample voltage features corresponding to multiple sample batteries, the sample abnormality types corresponding to the multiple sample batteries, the initial determination model and the target loss function, determine the initial loss value corresponding to the initial parameter; when the initial loss value is less than a predetermined loss threshold, determine the initial parameter as the target parameter; based on the target parameter and the initial determination model, determine the abnormality type determination model.

[0009] Optionally, the target voltage characteristics of the target battery are determined based on multiple correlation coefficients, including: when the target battery includes multiple single cells, determining the second voltage characteristics corresponding to the multiple single cells respectively based on the multiple correlation coefficients; and determining the target voltage characteristics of the target battery based on the second voltage characteristics corresponding to the multiple single cells respectively.

[0010] Optionally, based on the voltage signal sequences corresponding to the multiple predetermined power intervals, the voltage time domain characteristics of the target battery corresponding to the multiple predetermined power intervals are determined, including: based on the voltage signal sequences corresponding to the multiple predetermined power intervals, determining the multiple voltage values ​​of the target battery corresponding to the multiple predetermined power intervals; based on the multiple voltage values ​​of the target battery corresponding to the multiple predetermined power intervals, determining the multiple statistical indicator values ​​of the target battery corresponding to the multiple predetermined power intervals; based on the multiple statistical indicator values, determining the voltage time domain characteristics of the target battery corresponding to the multiple predetermined power intervals.

[0011] Optionally, after determining the target abnormality type corresponding to the target battery based on the target voltage characteristics, it also includes: determining the actual abnormality type of the target battery; when the target abnormality type is inconsistent with the actual abnormality type, determining a correction coefficient to determine the battery abnormality type based on the correction coefficient.

[0012] According to one aspect of an embodiment of the present invention, a battery abnormality type determination device is provided, comprising: an acquisition module for acquiring multiple voltage data of a target battery, wherein the multiple voltage data include voltage signal sequences corresponding to multiple predetermined power intervals of the target battery during charging; a first determination module for determining, based on the multiple voltage signal sequences, multiple frequency domain signal components corresponding to the target battery in the multiple predetermined power intervals, wherein the corresponding multiple frequency domain signal components are signal components corresponding to the corresponding voltage signal sequences in multiple predetermined frequency domain intervals; a second determination module for determining correlation coefficients between the multiple voltage signal sequences and the corresponding multiple frequency domain signal components; a third determination module for determining a target voltage characteristic of the target battery based on the multiple correlation coefficients; and a fourth determination module for determining a target abnormality type corresponding to the target battery based on the target voltage characteristic.

[0013] According to one aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement any of the above-mentioned methods for determining a battery abnormality type.

[0014] According to one aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute any of the above-mentioned methods for determining a battery abnormality type.

[0015] In an embodiment of the present invention, a method is adopted in which multiple voltage data of a target battery are obtained, wherein the multiple voltage data include voltage signal sequences corresponding to multiple predetermined power intervals of the target battery during charging; multiple frequency domain signal components corresponding to the target battery in the multiple predetermined power intervals are determined based on the multiple voltage signal sequences, wherein the corresponding multiple frequency domain signal components are signal components corresponding to the corresponding voltage signal sequence in the multiple predetermined frequency domain intervals; correlation coefficients between the multiple voltage signal sequences and the corresponding multiple frequency domain signal components are determined; a target voltage characteristic of the target battery is determined based on the multiple correlation coefficients; and a target abnormality type corresponding to the target battery is determined based on the target voltage characteristic. By determining the target voltage characteristic of the target battery, the purpose of determining the target abnormality type corresponding to the target battery based on the target voltage characteristic is achieved. Since the multiple frequency domain signal components of the target vehicle battery in the multiple predetermined power intervals and the correlation coefficients corresponding to the multiple frequency domain signal components with the battery data are determined, the target voltage characteristic that can represent the frequency domain characteristic of the voltage signal of the target vehicle battery is determined, thereby adding frequency domain feature analysis to the determination of the battery abnormality type, improving the accuracy of the prediction, and thus solving the technical problem in the related art of inaccurately determining the battery abnormality type due to the low accuracy of the determined prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 is a flow chart of a method for determining a battery abnormality type according to an embodiment of the present invention;

[0018] Figure 2 is a flow chart of a method for determining a battery abnormality type provided by an optional embodiment of the present invention;

[0019] Figure 3 It is a structural block diagram of a device for determining a battery abnormality type according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] Example 1

[0023] According to an embodiment of the present invention, an embodiment of a method for determining a battery abnormality type is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0024] Figure 1 FIG. 1 is a flow chart of a method for determining a battery abnormality type according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0025] Step S102 : acquiring a plurality of voltage data of the target battery, wherein the plurality of voltage data includes voltage signal sequences corresponding to a plurality of predetermined power intervals during the charging process of the target battery.

[0026] In step S102 provided in this application, multiple voltage data of the target battery are obtained.

[0027] Among them, the target battery is involved. The target battery refers to the battery that needs to be detected for abnormality type to determine whether there is an abnormality and the type of abnormality, such as the power battery pack in an electric vehicle or a hybrid vehicle.

[0028] Among them, voltage data is involved. Voltage data refers to the voltage value sequence collected in the target battery, which can reflect the voltage behavior of the battery under different working conditions.

[0029] The predetermined power range is a range pre-set within the state of charge (SOC) range of the battery during the battery charging process.

[0030] Among them, a voltage signal sequence is involved. The voltage signal sequence refers to a series of voltage readings of the target battery recorded over time within a predetermined power range, that is, the voltage time series data of the target battery within the predetermined power range.

[0031] In this step, when performing battery anomaly detection on the target battery, it is necessary to first obtain the voltage data of the target battery. That is, during the charging process of the target battery, for multiple preset predetermined power intervals, the voltage time series data corresponding to the target battery is collected to obtain the voltage signal sequences corresponding to the target battery in the multiple predetermined power intervals. Through this step, by collecting voltage data in multiple predetermined power intervals, it is possible to capture the subtle characteristic changes in the battery voltage at different states of charge, thereby improving the accuracy of the voltage characteristics subsequently determined. At the same time, obtaining voltage data in multiple predetermined power intervals can comprehensively determine the voltage change characteristics of the battery during the charging process, revealing the dynamic characteristics of the battery as the SOC changes, and providing a richer data foundation for subsequent anomaly detection.

[0032] In step S104 , a plurality of frequency domain signal components corresponding to the target battery in a plurality of predetermined power intervals are determined based on the plurality of voltage signal sequences, wherein the corresponding plurality of frequency domain signal components are signal components corresponding to the corresponding voltage signal sequence in a plurality of predetermined frequency domain intervals.

[0033] In step S104 provided in the present application, a plurality of frequency domain signal components corresponding to a plurality of predetermined power intervals of the target battery are determined.

[0034] In this step, the voltage data is converted from the time domain to the frequency domain through frequency domain conversion methods such as Fourier transform, empirical mode decomposition (EMD) and wavelet transform. The original time domain voltage signal is decomposed into a series of signal components with different frequencies, and multiple frequency domain signal components corresponding to the target battery in multiple predetermined power intervals are determined. Each frequency domain signal component represents the frequency information of the voltage signal sequence of the target battery in the corresponding predetermined power interval.

[0035] Through this step, the voltage signal is converted from the time domain to the frequency domain, and the frequency domain signal component of the battery voltage is extracted, which can capture the spectral characteristics of the voltage signal. The frequency domain signal component can reveal hidden patterns or anomalies in the battery voltage signal, so that the battery status assessment is expanded from a single time series analysis to a joint time-frequency analysis, which increases the dimension of data analysis and helps to more comprehensively understand the operating status of the battery.

[0036] Step S106 : determining correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components.

[0037] In step S106 provided in the present application, correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components are determined.

[0038] Among them, the correlation coefficient is involved, which refers to a statistical indicator that measures the degree of linear correlation between multiple determined frequency domain signal components and their corresponding voltage signal sequences.

[0039] In this step, the correlation coefficients between multiple voltage signal sequences and corresponding multiple frequency domain signal components are determined. The value range of the correlation coefficient is usually between -1 and 1. The closer the value is to 1 or -1, the stronger the correlation between the voltage signal sequence and the corresponding frequency domain signal component. A value close to 0 indicates that there is almost no linear relationship between the voltage signal sequence and the corresponding frequency domain signal component.

[0040] Through this step, multiple correlation coefficients are determined, and the correlation between multiple frequency domain signal components and the voltage signal sequence is quantified. Since the frequency domain signal components contain important frequency characteristics of battery abnormalities, quantifying the correlation between multiple frequency domain signal components and the voltage signal sequence can identify the frequency domain components most correlated with voltage changes, improve the accuracy of subsequent selection of frequency domain components, focus on the frequency domain signal components that are most valuable for detecting battery abnormality types, improve model accuracy, reduce the dimension of data input, and reduce calculation complexity.

[0041] In step S108 , a target voltage characteristic of the target battery is determined based on the plurality of correlation coefficients.

[0042] In step S108 provided in the present application, a target voltage characteristic of the target battery is determined.

[0043] Among them, the target voltage characteristics are involved. The target voltage characteristics refer to the voltage attribute characteristics that can most effectively characterize the target battery state, including the time domain characteristics and frequency domain characteristics of the target battery voltage.

[0044] In this step, first, by determining the correlation coefficients between multiple voltage signal sequences and corresponding multiple frequency domain signal components, the frequency domain characteristics of the target battery in the corresponding predetermined power range are determined. Then, the time domain characteristics of the corresponding predetermined power range are determined. The frequency domain characteristics and time domain characteristics of multiple predetermined power ranges are combined to determine the target voltage characteristics of the target battery.

[0045] Through this step, based on multiple correlation coefficients, the target voltage characteristics that are highly correlated with the actual state of the target battery are determined, providing accurate voltage characteristic information for the subsequent determination of the battery abnormality type. The target voltage characteristics can reveal the potential failure mode of the battery and improve the detection accuracy of the battery abnormality type.

[0046] Step S110 : determining a target abnormality type corresponding to a target battery according to a target voltage characteristic.

[0047] In step S110 provided in the present application, a target abnormality type corresponding to the target battery is determined.

[0048] Among them, the target abnormality type is involved, and the target abnormality type refers to the battery abnormality type of the determined target battery.

[0049] In this step, based on the target voltage characteristics, the abnormality type with the highest similarity to the target battery voltage characteristics is determined from multiple predetermined battery abnormality types to obtain the target abnormality type of the target battery. This step monitors and analyzes the target battery voltage characteristics, identifies abnormal battery conditions, and provides timely warnings for battery maintenance and troubleshooting, reducing safety incidents and economic losses caused by battery failures.

[0050] Through the above steps S102-S110, a plurality of voltage data of the target battery can be obtained, wherein the plurality of voltage data include voltage signal sequences corresponding to a plurality of predetermined power intervals of the target battery during charging; based on the plurality of voltage signal sequences, a plurality of frequency domain signal components corresponding to the target battery in the plurality of predetermined power intervals are determined, wherein the corresponding plurality of frequency domain signal components are signal components corresponding to the corresponding voltage signal sequence in the plurality of predetermined frequency domain intervals; correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components are determined; based on the plurality of correlation coefficients, a target voltage characteristic of the target battery is determined; and based on the target voltage characteristic, a target abnormality type corresponding to the target battery is determined. By determining the target voltage characteristic of the target battery, the purpose of determining the target abnormality type corresponding to the target battery based on the target voltage characteristic is achieved. By determining the plurality of frequency domain signal components of the target vehicle battery in the plurality of predetermined power intervals and the correlation coefficients corresponding to the plurality of frequency domain signal components with the battery data, a target voltage characteristic that can represent the frequency domain characteristic of the voltage signal of the target vehicle battery is determined, thereby adding frequency domain characteristic analysis to the determination of the battery abnormality type, improving the accuracy of the prediction, and thereby solving the technical problem in the related art of inaccurately determining the battery abnormality type due to the low accuracy of the determined prediction model.

[0051] As an optional embodiment, the target voltage characteristics of the target battery are determined based on multiple correlation coefficients, including: determining the target signal components corresponding to multiple predetermined power intervals based on the correlation coefficients between multiple voltage signal sequences and the corresponding multiple frequency domain signal components; determining the voltage frequency domain characteristics of the target battery corresponding to the multiple predetermined power intervals based on the target signal components corresponding to the multiple predetermined power intervals; determining the voltage time domain characteristics of the target battery corresponding to the multiple predetermined power intervals based on the voltage signal sequences corresponding to the multiple predetermined power intervals; and determining the target voltage characteristics of the target battery based on the voltage frequency domain characteristics and voltage time domain characteristics corresponding to the multiple predetermined power intervals.

[0052] In this embodiment, specific steps of determining the target voltage characteristics of the target battery based on multiple correlation coefficients are described.

[0053] Here, a target signal component is involved, and the target signal component refers to a signal component whose correlation coefficient with the corresponding voltage signal sequence is greater than a predetermined threshold among the corresponding multiple frequency domain signal components.

[0054] Among them, the voltage frequency domain characteristics are involved. The voltage frequency domain characteristics refer to the frequency domain characteristic description of the target battery voltage determined based on the target signal component in the corresponding predetermined power range.

[0055] Among them, the voltage time domain feature is involved. The voltage time domain feature refers to the characteristic description extracted from the voltage signal sequence, which can reflect the change of the target battery voltage over time in the corresponding predetermined power range.

[0056] In this step, after obtaining multiple frequency domain signal components and the correlation coefficient between each frequency domain signal component and the corresponding voltage signal sequence through methods such as frequency domain modal decomposition, the frequency domain signal component with a significant linear relationship with the voltage signal sequence is selected based on the correlation coefficient to obtain the target signal component. The target signal component can better reflect the essential characteristics of the voltage. Next, for each predetermined power interval, the voltage frequency domain characteristics are reconstructed from the selected frequency domain signal components, and the voltage time domain characteristics are directly determined from the voltage signal sequence. Finally, all selected voltage time domain characteristics and voltage frequency domain characteristics are combined to form the target voltage characteristics of the target battery.

[0057] Through this step, based on the correlation coefficient between the voltage signal sequence and the corresponding multiple frequency domain signal components, the frequency domain signal components that are irrelevant or have little correlation with the battery voltage change are removed, thereby reducing the data dimension, and determining the target signal component with the greatest correlation with the original voltage signal sequence, and then determining the voltage frequency domain characteristics. At the same time, combined with the voltage time domain characteristics corresponding to the predetermined power interval, the frequency domain and time domain characteristics are fused to determine the target voltage characteristics, further improving the accuracy of the battery abnormality type subsequently determined based on the target voltage characteristics.

[0058] As an optional embodiment, determining a target abnormality type corresponding to a target battery based on a target voltage characteristic includes: retrieving an abnormality type determination model corresponding to the target voltage characteristic, wherein the abnormality type determination model is provided with target parameters, and the target parameters are determined based on sample abnormality types corresponding to a plurality of sample batteries and sample voltage characteristics corresponding to a plurality of sample batteries; determining a plurality of abnormality type probability values ​​corresponding to the target battery based on the target voltage characteristic and the abnormality type determination model, wherein the plurality of abnormality type probability values ​​are respectively used to indicate that the battery abnormality type corresponding to the target battery is a probability value corresponding to a predetermined battery abnormality type; and determining a target abnormality type corresponding to the target battery based on the plurality of abnormality type probability values.

[0059] In this embodiment, specific steps of determining a target abnormality type corresponding to a target battery based on a target voltage characteristic are described.

[0060] This involves an abnormality type determination model, which identifies and predicts the abnormality type of a target battery based on the input target voltage characteristics. The abnormality type determination model optimizes model parameters through training to accurately classify battery abnormalities.

[0061] Among them, the target parameters are involved. The target parameters refer to the model parameters that are optimized after training to improve the prediction ability of the model. They determine how the abnormality type determines how the model processes the input target voltage characteristics to output the prediction results.

[0062] Among them, sample batteries are involved. Sample batteries refer to batteries used to train abnormal type determination models, including batteries with various abnormal conditions and normal batteries.

[0063] Among them, the sample abnormality type is involved, and the sample abnormality type refers to the abnormality type corresponding to the sample battery.

[0064] Here, a sample voltage feature is involved, and the sample voltage feature refers to a target voltage feature extracted from a voltage signal of a sample battery.

[0065] Among them, the abnormal type probability value is involved, and the abnormal type probability value refers to the probability that the target battery predicted by the abnormal type determination model belongs to each predetermined abnormal type.

[0066] In this step, first, the abnormality type determination model that matches the target voltage feature is called from the model library. The abnormality type determination model is set with target parameters. The abnormality type determination model has been trained to learn the relationship between the sample voltage features of a large number of sample batteries and their sample abnormality types, so that it can accurately predict the abnormality type corresponding to the target battery based on the input target voltage feature. The target voltage feature is input into the abnormality type determination model, and the abnormality type determination model will output a series of abnormality type probability values. Each abnormality type probability value corresponds to a predetermined battery abnormality type, reflecting the probability of the target battery being under the predetermined battery abnormality type. According to the abnormality type probability value output by the model, the most likely abnormality type of the target battery, that is, the target abnormality type, is determined by the set decision rule (such as selecting the abnormality type with the highest probability value).

[0067] Through this step, the abnormality type probability value can be accurately output according to the voltage characteristics of the target battery, and the possibility of the target battery being in a predetermined battery abnormality type can be predicted. It not only provides abnormality type information, but also quantifies the risk level of each abnormality type, providing data support for battery maintenance and management. The entire process is highly automated, reducing dependence on manual expert judgment, reducing operation and maintenance costs, and improving the efficiency and consistency of abnormality detection.

[0068] As an optional embodiment, before calling the abnormality type determination model corresponding to the target voltage feature, it also includes: determining an initial determination model, wherein the initial determination model is set with initial parameters; calling the target loss function corresponding to the initial determination model, wherein the target loss function includes a data loss term and a parameter loss term, the data loss term is used to determine the loss value between the battery abnormality type determined based on the initial parameters and the corresponding true abnormality type, and the parameter loss term is used to determine the loss value of the battery abnormality type determined based on the first voltage feature, the first voltage feature is a voltage feature determined by selecting the voltage feature input to the initial determination model based on the initial parameters; determining the initial loss value corresponding to the initial parameter based on the sample voltage features corresponding to multiple sample batteries, the sample abnormality types corresponding to multiple sample batteries, the initial determination model and the target loss function; when the initial loss value is less than the predetermined loss threshold, determining the initial parameter as the target parameter; determining the abnormality type determination model based on the target parameter and the initial determination model.

[0069] In this embodiment, specific steps of determining an abnormality type determination model are described.

[0070] Among them, the initial determination model is involved. The initial determination model refers to the baseline model set at the initial stage of abnormal type determination model training, which includes untrained initial parameters.

[0071] Among them, the initial parameters are involved. The initial parameters refer to the weights and biases and other parameters that are randomly initialized before the training of the abnormality type determination model begins. They will be optimized with the goal of minimizing the target loss function during the training process.

[0072] Among them, the target loss function is involved. The target loss function refers to the function that measures the difference between the predicted result and the actual label during the training process of the abnormal type determination model. It consists of data loss terms and parameter loss terms.

[0073] Among them, the data loss term is involved. The data loss term refers to a function term that measures the gap between the anomaly type predicted by the anomaly type determination model and the actual anomaly type.

[0074] Among them, parameter loss terms are involved. Parameter loss terms refer to function terms added to achieve model dimensionality reduction and feature selection constraints, such as the loss terms of model sparsity and dimensionality reduction constraints.

[0075] Among them, the first voltage feature is involved. The first voltage feature refers to a feature subset screened from the target voltage feature through initial parameters, which is used to preliminarily evaluate the model's sensitivity and predictive ability to the input feature.

[0076] Among them, a predetermined loss threshold is involved. The predetermined loss threshold refers to the loss value standard set in advance during the training process of the anomaly type determination model. When the loss value of the anomaly type determination model is lower than the predetermined loss threshold, the model is considered to have reached an acceptable performance level.

[0077] In this step, a model structure suitable for battery anomaly type prediction is first selected, such as a sparse reduced-rank multinomial logistic regression model, and initial parameters are set for it to obtain an initial determination model. Next, a target loss function is constructed, which includes a data loss term and a parameter loss term. The data loss term is used to evaluate the difference between the model's prediction and the actual anomaly type, while the parameter loss term is used to limit the complexity of the model parameters. The sample voltage characteristics of multiple groups of sample batteries are input into the initial determination model. The target loss function is used to measure the loss between the predicted anomaly type determined by the initial determination model and the known anomaly type label. The initial loss value of the initial determination model for predicting anomaly types under the current initial parameter state is determined. If the initial loss value is below a predetermined loss threshold, the current initial parameters are considered the target parameters, and the model achieves the predetermined performance level. Otherwise, the model parameters are adjusted until the model loss value falls below the predetermined loss threshold. Finally, based on the target parameters and the structure of the initial determination model, an optimized anomaly type determination model is determined for actual anomaly type prediction.

[0078] Through this step, the target loss function with data loss terms and parameter loss terms is called up, and the model parameters of the initial determination model are continuously optimized according to the loss value. While ensuring the accuracy of the determined battery abnormality type, the dimensions of the model parameters and voltage characteristics are reduced, and the complexity of the calculation is reduced.

[0079] As an optional embodiment, the target voltage characteristics of the target battery are determined based on multiple correlation coefficients, including: when the target battery includes multiple single cells, determining the second voltage characteristics corresponding to the multiple single cells respectively based on the multiple correlation coefficients; and determining the target voltage characteristics of the target battery based on the second voltage characteristics corresponding to the multiple single cells respectively.

[0080] In this embodiment, specific steps of determining the target voltage characteristics of the target battery based on multiple correlation coefficients are described.

[0081] Among them, the second voltage feature is involved, which refers to a target voltage feature extracted from the voltage signal sequence of each single cell and represents the potential failure mode of each single cell.

[0082] In this step, for each single cell in the target battery, the correlation coefficient between the voltage signal sequence corresponding to the single cell and the modal component obtained after its frequency domain modal decomposition is determined, and the modal component with the highest correlation is selected to reconstruct the voltage information. Combined with the time domain voltage characteristics, the second voltage characteristic of the single cell is determined. After the second voltage characteristic is obtained for each single cell, multiple second voltage characteristics are aggregated to form a target voltage characteristic that describes the voltage characteristics of the entire target battery. Through this step, the mutual influence between the cells in the battery pack is taken into account. By integrating the second voltage characteristics of all cells, a more comprehensive description of the battery status can be formed, which is conducive to the identification of the overall abnormal pattern.

[0083] As an optional embodiment, based on the voltage signal sequences corresponding to the multiple predetermined power intervals, the voltage time domain characteristics of the target battery corresponding to the multiple predetermined power intervals are determined, including: based on the voltage signal sequences corresponding to the multiple predetermined power intervals, determining the multiple voltage values ​​of the target battery corresponding to the multiple predetermined power intervals; based on the multiple voltage values ​​of the target battery corresponding to the multiple predetermined power intervals, determining the multiple statistical indicator values ​​of the target battery corresponding to the multiple predetermined power intervals; based on the multiple statistical indicator values, determining the voltage time domain characteristics of the target battery corresponding to the multiple predetermined power intervals.

[0084] In this embodiment, specific steps of determining the voltage time-domain characteristics corresponding to a target battery in a plurality of predetermined power intervals are described.

[0085] Among them, statistical indicator values ​​are involved. Statistical indicator values ​​refer to the determined statistics that describe the time-domain characteristics of voltage, such as the mean, median, mode, standard deviation, skewness, kurtosis, maximum value, minimum value, etc. corresponding to multiple voltage values, which are used to quantify the central trend, degree of dispersion and distribution form of voltage data.

[0086] In this step, within each predetermined power interval, corresponding multiple voltage values ​​are extracted from the voltage signal sequence. Statistical analysis is performed on the voltage values ​​within each predetermined power interval to calculate a series of statistical index values, such as average voltage, voltage standard deviation, etc. These indicators can quantify the distribution and variation pattern of the voltage signal within the state of charge range. Based on the above statistical indicators, a set of time domain characteristics representing the voltage behavior of the battery in different SOC intervals is determined. Through this step, statistical analysis is performed on the voltage values ​​within each predetermined power interval, which can analyze the voltage signal from multiple dimensions, extract the time series characteristics of the target battery voltage, and improve the accuracy of the target battery abnormality type determined subsequently.

[0087] As an optional embodiment, after determining the target abnormality type corresponding to the target battery based on the target voltage characteristics, it also includes: determining the actual abnormality type of the target battery; when the target abnormality type is inconsistent with the actual abnormality type, determining a correction coefficient to determine the battery abnormality type based on the correction coefficient.

[0088] In this embodiment, the specific steps of determining the correction coefficient are explained.

[0089] Among them, the correction coefficient is involved, which refers to the parameter used to adjust the deviation between the predicted result and the actual abnormal type.

[0090] In this step, after the model predicts the target battery's anomaly type, it further verifies the accuracy of the prediction. If the prediction differs from the actual situation, a correction factor is introduced. This dynamically adjusts the output based on the deviation between the predicted result and the actual anomaly type, making the prediction more accurate. By comparing the predicted target anomaly type with the actual anomaly type in this step, the source of the prediction error can be identified. The correction factor can then be used to make targeted adjustments, and self-correction can be performed based on actual feedback, thereby further improving prediction accuracy, adapting to different battery types, operating conditions, and environmental conditions, and enhancing the system's generalization capabilities.

[0091] Based on the above embodiment and optional embodiment, an optional implementation manner is provided, which is described in detail below.

[0092] Due to the complex internal coupling mechanisms of battery packs, the dynamic and changeable operating environment, and the nonlinearity of fault evolution paths, potential failure modes of power batteries are often hidden and spread rapidly, posing significant challenges to early warning. In high-dimensional situations where the data sample size is smaller than the feature dimension, most intelligent algorithm models fail and are unable to select important features that influence the model. Fitting an effective model with good predictive and interpretable performance is very difficult. To reduce model complexity and improve model interpretability, many studies in related technologies mainly use two effective methods, dimensionality reduction and feature selection, to obtain sparse reduced-rank models. However, mainstream feature selection methods often use penalized likelihood methods based on convex norms to perform this process, such as the sparse rule operator (L1 norm). Due to the defects of convex norm penalties such as estimation bias and over-selection, the model accuracy is not high and the prediction results are poor.

[0093] In view of this, an optional embodiment of the present invention provides a method for warning vehicle battery anomalies using an improved multinomial logistic regression model. This method uses frequency domain modal decomposition to extract frequency domain feature information at different levels based on time domain signal features. The combined time and frequency domain information enables multi-dimensional analysis of fault evolution patterns. Furthermore, an improved sparse reduced-rank multinomial logistic regression model is proposed based on time-frequency domain derivative data. Multinomial logistic regression is a very useful classification model within a generalized linear model that can be used to predict the probabilities of multiple possible categories. Applying low-rank constraints and non-convex sparsity constraints to the coefficient matrix simultaneously achieves dimensionality reduction and variable selection. Matrix rank reduction can extract key information from the original feature space, making the information more concentrated. Feature selection can focus on important fault features and explore features closely related to the fault. Due to the computational difficulty of solving non-convex norm penalties, an iterative active set algorithm based on dual conditions is developed to efficiently and accurately solve the variable selection problem under non-convex norm penalties. This early warning method based on a sparse reduced-rank multinomial logistic regression model constructed based on time-frequency domain feature derivative data not only improves the accuracy and efficiency of early warnings, but also significantly reduces the need for manual intervention and reduces operation and maintenance costs.

[0094] First, the vehicle battery's raw feature data is obtained and subjected to frequency-domain modal decomposition. Statistical features are then derived from the time and frequency domains. A sparse reduced-rank multinomial logistic regression model is then constructed based on the derived feature data. The optimal model parameters are determined using a dual active set iterative algorithm. Finally, the optimal model is used to classify and warn of vehicle battery anomalies. This algorithm can predict battery failure signs in advance, reducing failure rates and operational costs. Figure 2 This is a flow chart of a method for determining a battery abnormality type provided by an optional embodiment of the present invention. Figure 2 As shown, the battery abnormality type determination method provided by the optional embodiment of the present invention specifically includes vehicle battery data acquisition and category labeling; data processing and frequency domain modal decomposition; statistical feature derivation and data set construction; sparse reduced-rank multinomial logistic regression model construction; dual effective set iterative algorithm for model calculation and solution; model prediction and abnormal vehicle warning.

[0095] The following specifically describes the detailed steps of a method for determining a battery abnormality type provided by an optional embodiment of the present invention.

[0096] S1. Acquire multiple voltage data of a target battery.

[0097] Because the battery charging process can better reflect its own characteristics and potential failure modes, a longer charging process is selected based on the SOC range, and the cell voltage is segmented by a certain SOC interval length. The battery voltage data of the target battery is obtained from the longer charging process within the range of the starting SOC less than 30 and the ending SOC greater than 90. The cell voltage is then segmented by an SOC interval length of 10. This way, cell voltage data (same as the voltage signal sequence described above) is obtained for eight SOC intervals (same as the predetermined charge intervals described above): 30-40, 40-50, 50-60, 60-70, 70-80, 80-90, and 90-100.

[0098] S2. Determine, based on the multiple voltage signal sequences, multiple frequency domain signal components corresponding to the target battery in multiple predetermined power ranges.

[0099] Empirical mode decomposition (EMD modal decomposition) is performed on the voltage of each cell in each SOC interval of the target battery to obtain modal components at different frequency domain levels. Each cell voltage can be decomposed into modal components (the same as the above-mentioned frequency domain signal components) at several frequency domain levels (the same as the above-mentioned predetermined frequency range).

[0100] S3. Determine correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components.

[0101] Calculate the correlation coefficients between the modal components and the voltage.

[0102] S4. Determine a target voltage characteristic of the target battery based on the multiple correlation coefficients.

[0103] Each monomer in each interval can reconstruct a frequency-domain voltage feature. The features are sorted from largest to smallest by correlation coefficient, and the modal components corresponding to the top three correlation coefficients are selected for reconstruction. This means the three modal components are added and averaged, and the modal component with the highest correlation coefficient (the same as the target signal component) is selected for voltage reconstruction. This way, each monomer in each interval obtains a corresponding reconstructed voltage frequency-domain feature (the same as the voltage frequency-domain feature described above), ultimately forming the voltage frequency-domain features for all monomers in all intervals.

[0104] Each cell in each interval has two voltage characteristics, time domain and frequency domain, and then statistical characteristics are derived. For the time domain and frequency domain voltages of each cell in each interval, multiple statistical indicators are calculated, such as maximum value, minimum value, average value, median, standard deviation, skewness, kurtosis and range. If there are multiple charging segments, the statistical characteristics of the corresponding SOC interval segments are averaged. In this way, each cell in each interval will have 8 statistical derivative features in the time domain (the same as the above voltage time domain features) and 8 statistical derivative features in the frequency domain, and then these features are arranged in order of intervals and cells to form the data set dimension. Assuming that a battery pack consists of 100 cells, the statistical derivative features are arranged in the order of cells in the 8 SOC intervals, then the total feature dimension of the data set is: 8*100*8*2.

[0105] S5. Determine the target abnormality type corresponding to the target battery based on the target voltage characteristics.

[0106] Since there are many derived statistical features in S4, the sample size may be smaller than the feature dimension, that is, the high-dimensional problem of the data. The general multi-classification model will fail. Therefore, it is necessary to increase dimensionality reduction and feature selection constraints. The multinomial logistic regression model (the same as the above-mentioned anomaly type determination model) is added with matrix rank constraints and equation restrictions of non-convex norm penalty terms to achieve rank reduction and important feature selection. In this way, a sparse reduced-rank multinomial logistic regression model is constructed. The model automatically solves the high-dimensional problem of the data and performs important feature selection.

[0107] The following describes the detailed steps for building an anomaly type determination model.

[0108] A1. Obtain sample abnormality types corresponding to each sample battery and sample voltage characteristics corresponding to each of the multiple sample batteries.

[0109] Data on vehicles that have experienced various types of faults and have been repaired is collected from a big data platform. A big data cloud platform is a cloud platform built using big data technology to process, store power battery information, and provide after-sales services. Battery data is uploaded to the cloud platform at a fixed frequency from the BMS (Battery Management System). First, self-test data (PS), severe undervoltage, voltage inconsistency, and other fault categories for repaired vehicle batteries are collected from the fault repair database. Then, charging data from the cloud platform's battery operation database is collected for each vehicle for the seven days prior to the reported fault. This data includes raw characteristics such as SOC, current, voltage, temperature, insulation group value, and mileage. Because a battery pack consists of single cells, each frame of voltage and temperature data is in array format. These vehicle batteries are labeled with the corresponding fault sample category. Finally, seven days of data from a certain number of normal vehicle batteries are collected from the cloud platform's operation database and labeled as normal samples. Based on the fault type and reporting time, the relevant vehicle charging data is obtained and labeled.

[0110] A2. Determine the initial determination model.

[0111] The multinomial logistic regression model is used to establish the initial determination model. The multinomial logistic regression model is a type of generalized linear model. Therefore, it has many similarities with the multiple linear regression model. Both have the basic form of linear functions. The difference is that the response variable of the multinomial logistic regression model is a discrete 0-1 label and obeys a multinomial distribution. It models the probabilities of multiple possible categories.

[0112] A3. Retrieve the target loss function corresponding to the initial determined model.

[0113] Assume that there are sample data corresponding to n sample batteries: (X, Y) = {(X i ,Y i ),i=1,…,n}, where X i represents the p-dimensional feature of the i-th sample (same as the above sample voltage feature), Y i Represents the category label of the i-th sample (same as the above sample anomaly type), and the q-dimensional category label vector has only one element that is 1, and the rest are 0. Assume that the probability that the i-th sample belongs to the k-th category (P ik )for:

[0114] P ik =P(Y ik =1|X i. )

[0115] Then the general negative log-likelihood loss function of multinomial logistic regression (L(C; X, Y), the same as the target loss function above) can be written as:

[0116]

[0117] Among them, C is the initial parameter corresponding to the initial determination model, Y ik is the label of the k-th category sample belonging to the i-th sample battery, C k is the model parameter matrix corresponding to the k-class sample label, C l is the model parameter matrix corresponding to the class I sample label.

[0118] A4. Based on the sample voltage characteristics corresponding to the multiple sample batteries and the sample abnormality types corresponding to the multiple sample batteries, the model and target loss function are initially determined to determine the target parameters.

[0119] To achieve dimensionality reduction, the rank of C is restricted to a lower rank. To select important features and increase model interpretability, the number of non-zero rows in the coefficient matrix C is restricted to select the corresponding important features, that is, a norm penalty is imposed on the coefficient matrix. The improved sparse reduced-rank multinomial logistic regression model thus has the following optimization problem:

[0120] r(C)≤r,‖C‖ 2,0 =s

[0121] Among them, r(C) is the rank of the parameter matrix, r is the predetermined rank threshold, s is the quantity threshold of the first voltage feature, and the optimal coefficient matrix C can be solved subsequently.

[0122] Because the coefficient matrix has both low-rank constraints and non-convex sparse constraints, the interplay of these two constraints poses a significant challenge in developing effective algorithms to solve the optimization problem. Since the model's rank constraints and non-convex penalty terms are difficult to solve simultaneously, we developed a novel dual active set iterative solution algorithm to estimate and solve the sparse reduced-rank model, calculate the model's optimal parameters, and present key features.

[0123] The non-convex constraint solving algorithm in the regression problem is improved to the current multi-classification problem, and the dual active set iterative algorithm is used to solve the model. The following is an example of a program using the dual active set iterative algorithm to solve the model provided by an optional embodiment of the present invention:

[0124] Data: class label Y, predictor variable X, sparsity s, and current estimate (B, V)

[0125] Result: Initialize k=0, B (0) =B,

[0126] while B (k) not converged do

[0127]

[0128] set k = k + 1;

[0129] end

[0130] Among them, B: is the model parameter matrix, which is the core object of algorithm iterative optimization and solution. After updating, the estimated result is obtained Used to characterize the relationship between X and Y. V: is an auxiliary parameter matrix, such as in matrix decomposition and multi-factor models, it builds the model structure together with B. A: is often used to represent a selection set of features, such as screening out a set of feature indexes that contribute greatly to the model and need to be retained. is the final selection result. I: the complement of A. k: the number of iteration steps / rounds, marking the current iteration stage of the algorithm, starting from 0 and gradually increasing (k = 0, 1, 2...) until the convergence condition is met. Γ(Γj, Γ (k)(e.g., Δj): gradient / residual variable that assists in updating. Δj: importance metric for parameters / features. P(·): prediction / projection function. is the corrected label. tn is a scalar coefficient used in equilibrium calculations. () C is the complement of the set. S is the number of the subset currently being processed. The superscript T of a matrix / vector indicates the transpose operation, which swaps the rows and columns of the matrix / vector. The subscript j indicates the dimension index, for example, Xj represents the jth column of the feature matrix X, which has p columns, and Bj represents the jth row of the parameter matrix B. The subscripts A and I indicate the set index.

[0131] For orthogonal constraint problems, the orthogonal problem solving algorithm can be used.

[0132] Combining the above non-convex sparse constraints and orthogonal constraints, the coefficient matrix C is expressed as the product of two low-rank matrices B and V, that is:

[0133] C=BV T

[0134] Among them, B is a p*r dimensional matrix, and V is a q*r dimensional matrix.

[0135] The original p-dimensional features are reduced to r linear combinations, which can be interpreted as unobservable latent factors. The low-rank constraints on the coefficient matrix are converted to orthogonal constraints on V. Through the above transformation, the optimization problem now is to estimate the low-rank matrices B and V, which can be completed through block-by-block iteration. The block-by-block iteration formula is as follows:

[0136] ‖B‖ 2,0 =s,

[0137] V T V=I r

[0138] Among them, B (m+1) 、V (m+1) are the low-rank matrices B and V obtained by the m+1th iteration, V m is the matrix obtained by the mth iteration of the low-rank matrix V, I r is the identity matrix.

[0139] The following is an example of a program for calculating model coefficients provided by an optional embodiment of the present invention:

[0140] Data: class labels Y, predictor variables X, rank r, sparsity s, and

[0141] Result:

[0142] Initialize m=0, V (0) for The eigenvectors corresponding to the first r eigenvalues;

[0143] whileL(B (m) V (m)T ;X,Y)in(2,4)not converged do

[0144] run algorithm 2.1with s and(B (m) ,V (m) );

[0145]

[0146] Set m=m+1

[0147] end

[0148] in, The corrected target matrix is ​​given by Definition, where 1 N*(q-1) is a matrix of all 1s (dimensions N×(q-1), where N is the sample size and q is the number of associated categories. The corrected label. The estimated results of the model parameter matrix, B is the core coefficient matrix, V is the auxiliary low-rank matrix, is the final estimated value obtained after iterative optimization. Composite parameter matrix. m: Iteration round, marking the current iteration stage (starting from 0 and increasing). L(·): loss function. algorithm2.1: Sub-algorithm, is another set of parameter update logic defined, input sparsity s and the parameters of the current iteration (B (m) ,V (m) ), output updated intermediate results for nested iterative optimization. (m) : The predicted value matrix of the mth iteration. S (m+1) : Auxiliary update matrix. t: scalar weight coefficient.

[0149] Singular value decomposition (SVD): Matrix decomposition operation, decomposing the matrix Z into Among them: U Z 、V Z It is an orthogonal matrix (column vectors are orthogonal to each other), containing left and right singular vectors; D Z Is a diagonal matrix whose diagonal elements are the singular values ​​of Z (non-negative, in descending order).

[0150] Based on the optimal model coefficients obtained by A4 (the same as the above target parameters), the sample category probability can be calculated (and given important features, and then the vehicle battery is operated online at a certain frequency, and the abnormal category probability (the same as the above abnormal type probability value) of the vehicle battery (the same as the above target battery) can be calculated.

[0151] The calculation formula for the abnormal type probability value is:

[0152]

[0153] Based on the probability value of the abnormality type, the target abnormality type of the target battery is predicted, and the signs of battery failure can be predicted in advance when there is no obvious abnormality in the vehicle battery.

[0154] Through the above optional implementation, at least the following beneficial effects can be achieved:

[0155] (1) Realize multi-dimensional decoupling and reconstruction of battery failure characteristics in the time and frequency domain feature space, effectively breaking through the limitations of the traditional threshold judgment paradigm;

[0156] (2) Innovatively introduce sparse rank reduction into the multinomial logistic regression model, which significantly compresses the high-dimensional feature dimensions while retaining key degradation information and selects important features closely related to the fault;

[0157] (3) When the battery is actually observed to be operating stably, the potential evolution pattern can be extracted to predict the battery failure category in advance, reduce the failure rate and operation and maintenance costs, and improve the user experience.

[0158] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0159] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0160] Example 2

[0161] According to an embodiment of the present invention, a device for implementing the above-mentioned method for determining the battery abnormality type is also provided. Figure 3 FIG. 1 is a structural block diagram of a device for determining a battery abnormality type according to an embodiment of the present invention. Figure 3 As shown, the apparatus includes: an acquisition module 302, a first determination module 304, a second determination module 306, a third determination module 308 and a fourth determination module 310. The apparatus will be described in detail below.

[0162] An acquisition module 302 is used to acquire multiple voltage data of a target battery, wherein the multiple voltage data include voltage signal sequences corresponding to multiple predetermined power intervals of the target battery during charging; a first determination module 304 is connected to the acquisition module 302 and is used to determine multiple frequency domain signal components corresponding to the target battery in multiple predetermined power intervals based on the multiple voltage signal sequences, wherein the corresponding multiple frequency domain signal components are signal components corresponding to the corresponding voltage signal sequences in multiple predetermined frequency domain intervals; a second determination module 306 is connected to the first determination module 304 and is used to determine correlation coefficients between the multiple voltage signal sequences and the corresponding multiple frequency domain signal components; a third determination module 308 is connected to the second determination module 306 and is used to determine the target voltage characteristics of the target battery based on the multiple correlation coefficients; a fourth determination module 310 is connected to the third determination module 308 and is used to determine the target abnormality type corresponding to the target battery based on the target voltage characteristics.

[0163] It should be noted here that the above-mentioned acquisition module 302, first determination module 304, second determination module 306, third determination module 308 and fourth determination module 310 correspond to steps S102 to S110 in implementing the battery abnormality type determination method. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0164] Example 3

[0165] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute the instructions to implement any of the above methods for determining a battery abnormality type.

[0166] Example 4

[0167] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute any of the above-mentioned methods for determining a battery abnormality type.

[0168] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0169] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0170] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0171] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0172] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0173] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0174] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for determining a battery abnormality type, characterized in that: include: Acquire multiple voltage data of a target battery, wherein the multiple voltage data include voltage signal sequences corresponding to multiple predetermined power intervals during the charging process of the target battery; Determining, based on the plurality of voltage signal sequences, a plurality of frequency domain signal components corresponding to the target battery in the plurality of predetermined power intervals, wherein the corresponding plurality of frequency domain signal components are signal components corresponding to the corresponding voltage signal sequence in the plurality of predetermined frequency domain intervals; Determining correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components; determining a target voltage characteristic of the target battery based on a plurality of correlation coefficients; A target abnormality type corresponding to the target battery is determined based on the target voltage characteristic.

2. The method according to claim 1, characterized in that Determining the target voltage characteristic of the target battery based on the multiple correlation coefficients includes: Determining target signal components corresponding to the plurality of predetermined power intervals respectively according to correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components; Determining, based on target signal components corresponding to the plurality of predetermined power intervals, voltage frequency domain characteristics corresponding to the target battery in the plurality of predetermined power intervals; Determining voltage time-domain characteristics of the target battery corresponding to the plurality of predetermined power intervals, respectively, based on voltage signal sequences corresponding to the plurality of predetermined power intervals; The target voltage characteristics of the target battery are determined according to the voltage frequency domain characteristics and the voltage time domain characteristics respectively corresponding to a plurality of predetermined power intervals.

3. The method according to claim 1, characterized in that The determining, based on the target voltage characteristic, a target abnormality type corresponding to the target battery includes: Retrieving an abnormality type determination model corresponding to the target voltage feature, wherein the abnormality type determination model is set with a target parameter, and the target parameter is determined according to sample abnormality types corresponding to a plurality of sample batteries and sample voltage features corresponding to the plurality of sample batteries; Determining, based on the target voltage characteristic and the abnormality type determination model, a plurality of abnormality type probability values ​​corresponding to the target battery, wherein the plurality of abnormality type probability values ​​are respectively used to indicate a probability value that the battery abnormality type corresponding to the target battery is a predetermined battery abnormality type; A target abnormality type corresponding to the target battery is determined according to the multiple abnormality type probability values.

4. The method according to claim 3, characterized in that Before calling the abnormality type determination model corresponding to the target voltage feature, the method further includes: determining an initial determination model, wherein the initial determination model is set with initial parameters; Retrieving a target loss function corresponding to the initial determination model, wherein the target loss function includes a data loss term and a parameter loss term, the data loss term being used to determine a loss value between a battery abnormality type determined based on the initial parameters and a corresponding true abnormality type, and the parameter loss term being used to determine a loss value for determining the battery abnormality type based on a first voltage feature, the first voltage feature being a voltage feature determined by selecting a voltage feature input to the initial determination model based on the initial parameters; Determining an initial loss value corresponding to the initial parameter based on sample voltage characteristics corresponding to a plurality of sample batteries, sample abnormality types corresponding to the plurality of sample batteries, the initial determination model, and the target loss function; When the initial loss value is less than a predetermined loss threshold, determining the initial parameter as a target parameter; The abnormality type determination model is determined according to the target parameter and the initial determination model.

5. The method according to claim 1, wherein Determining the target voltage characteristic of the target battery based on the multiple correlation coefficients includes: In a case where the target battery includes a plurality of single cells, determining second voltage characteristics corresponding to the plurality of single cells respectively according to the plurality of correlation coefficients; The target voltage characteristic of the target battery is determined according to the second voltage characteristics respectively corresponding to the plurality of single batteries.

6. The method according to claim 2, characterized in that The determining, based on the voltage signal sequences corresponding to the plurality of predetermined power intervals, the voltage time-domain characteristics of the target battery corresponding to the plurality of predetermined power intervals, includes: Determining, based on voltage signal sequences corresponding to the plurality of predetermined power intervals, a plurality of voltage values ​​of the target battery corresponding to the plurality of predetermined power intervals; Determining, based on a plurality of voltage values ​​of the target battery corresponding to the plurality of predetermined power intervals, a plurality of statistical indicator values ​​corresponding to the target battery in the plurality of predetermined power intervals; The voltage time-domain characteristics corresponding to the target battery in the plurality of predetermined power intervals are determined according to the plurality of statistical indicator values.

7. The method according to any one of claims 1 to 6, characterized in that After determining the target abnormality type corresponding to the target battery according to the target voltage characteristic, the method further includes: determining an actual abnormality type of the target battery; In a case where the target abnormality type is inconsistent with the actual abnormality type, a correction coefficient is determined to determine the battery abnormality type according to the correction coefficient.

8. A device for determining battery abnormality type, characterized in that: include: an acquisition module, configured to acquire a plurality of voltage data of a target battery, wherein the plurality of voltage data includes voltage signal sequences corresponding to a plurality of predetermined power intervals during the charging process of the target battery; a first determining module, configured to determine, based on a plurality of voltage signal sequences, a plurality of frequency domain signal components corresponding to the target battery in the plurality of predetermined power intervals, wherein the corresponding plurality of frequency domain signal components are signal components corresponding to the corresponding voltage signal sequence in the plurality of predetermined frequency domain intervals; A second determining module is used to determine correlation coefficients between the plurality of voltage signal sequences and the corresponding plurality of frequency domain signal components; a third determining module, configured to determine a target voltage characteristic of the target battery based on a plurality of correlation coefficients; The fourth determining module is configured to determine a target abnormality type corresponding to the target battery according to the target voltage characteristic.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the battery abnormality type determination method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the battery abnormality type determination method according to any one of claims 1 to 7.

Images