Battery fault monitoring method, system and equipment based on voltage characteristics and medium

Abstract

The invention provides a battery fault monitoring method, system and device based on voltage characteristics and a medium. The monitoring method comprises the following steps: extracting battery data of a plurality of continuous charging / discharging time periods from battery historical data; according to the voltage data of each battery cell, calculating a longitudinal outlier average value and an average voltage ranking of each battery cell in each charging / discharging period; for each battery cell, screening out a plurality of charging / discharging time periods from all the charging / discharging time periods according to the average voltage ranking, forming a longitudinal outlier average value sequence by using the longitudinal outlier average values of all the screened charging / discharging time periods, and forming an average voltage ranking sequence by using the average voltage ranking of all the screened charging / discharging time periods; and inputting the longitudinal outlier average value sequence and the average voltage ranking sequence of each battery cell into a pre-trained monitoring model to generate a monitoring result of each battery cell in the battery. By analyzing whether the battery has a fault or not, the safety of the battery in the using process is improved.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and in particular to battery fault monitoring methods, systems, devices and media based on voltage characteristics. Background Technology

[0002] Batteries, as electrical energy storage components, include but are not limited to lead-acid batteries, lithium-ion batteries (such as lithium iron phosphate batteries, ternary lithium batteries, lithium titanate batteries, etc.), sodium-ion batteries, lithium metal batteries, semi-solid-state batteries, and solid-state batteries, and are widely used in various power consumption scenarios. For example, electric vehicles, electric ships, electric aircraft and other electrified transportation tools, as well as energy storage power stations, data centers, intelligent computing centers and other energy storage systems.

[0003] In practical applications, internal short-circuit faults in batteries are one of the main causes of battery thermal runaway. They are typically caused by lithium dendrite growth or separator defects, with extremely subtle initial symptoms that can easily escalate into severe internal short circuits leading to fires. Because the initial symptoms of internal short-circuit faults are so subtle, they are difficult to diagnose directly using battery parameters. Therefore, there are areas for improvement. Summary of the Invention

[0004] This invention provides a battery fault monitoring method, system, device, and medium based on voltage characteristics, to improve the technical problem that the initial characteristics of short circuit faults in existing batteries are extremely weak, making it difficult to determine them directly through battery parameters.

[0005] This invention proposes a battery fault monitoring method based on voltage characteristics, comprising:

[0006] Battery data from multiple consecutive charging / discharging periods is extracted from historical battery data; the battery data includes the voltage data of each cell in the battery during the charging / discharging period. Based on the voltage data of each cell, calculate the longitudinal outlier average and average voltage ranking of each cell in each charging / discharging period; For each battery cell, multiple charging / discharging periods are selected from all charging / discharging periods based on the average voltage ranking. The longitudinal outlier average of all selected charging / discharging periods is used to form a longitudinal outlier average sequence, and the average voltage ranking of all selected charging / discharging periods is used to form an average voltage ranking sequence. The longitudinal outlier average sequence and average voltage ranking sequence of each cell are input into the pre-trained monitoring model to generate the monitoring results of each cell in the battery.

[0007] In one embodiment of the present invention, the voltage data includes the voltage of each cell at each sampling point; the step of calculating the longitudinal outlier average of each cell in each charging / discharging period based on the voltage data of each cell includes: Calculate the longitudinal outlier value of each cell at each sampling point during each charging / discharging period; Based on the longitudinal outlier value and the number of sampling points of all sampling points for each cell in each charging / discharging period, the longitudinal outlier average value of each cell in each charging / discharging period is calculated. Longitudinal outliers ,satisfy: ; Longitudinal outlier mean ,satisfy: ; in, Represented as the first in the battery Each battery cell, Represented as the first Each charging / discharging period Represented as the first The first charging / discharging period within the [number]th [period] One sampling point, Represented as in the first When the sampling point is _th The voltage of each battery cell Represented as in the first The median voltage of all cells at each sampling point. Represented as the first Number of sampling points within each charging / discharging period.

[0008] In one embodiment of the present invention, the voltage data includes the voltage of each cell at each sampling point; the step of calculating the average voltage ranking of each cell in each charging / discharging period based on the voltage data of each cell includes: At each sampling point during each charging / discharging period, the voltage of all cells is sorted, and the voltage ranking of each cell at each sampling point during each charging / discharging period is recorded. Based on the voltage ranking and number of sampling points of each cell in each charging / discharging period, calculate the average voltage ranking of each cell in each charging / discharging period; Average voltage ranking ,satisfy: ; in, Represented as the first in the battery Each battery cell, Represented as the first Each charging / discharging period Represented as the first The first charging / discharging period within the [number]th [period] One sampling point, Represented as in the first When the sampling point is _th Voltage ranking of individual battery cells Represented as the first Number of sampling points within each charging / discharging period.

[0009] In one embodiment of the present invention, the step of selecting multiple charging / discharging periods from all charging / discharging periods based on average voltage ranking for each battery cell, forming a vertical outlier average sequence from the vertical outlier average of all selected charging / discharging periods, and forming an average voltage ranking sequence from the average voltage ranking of all selected charging / discharging periods includes: For each cell, calculate the difference between its average voltage ranking in each charging / discharging period and the average voltage ranking in the previous charging / discharging period, and record the charging / discharging period with the largest negative difference as the intermediate charging / discharging period; Based on the location of the intermediate charging / discharging period, a preset number of charging / discharging periods are selected sequentially from all charging / discharging periods; The longitudinal outlier averages of all charging / discharging periods will be selected to form a longitudinal outlier average sequence, and the average voltage rankings of all charging / discharging periods will be selected to form an average voltage ranking sequence.

[0010] In one embodiment of the present invention, the charging / discharging period satisfies: ; in, This indicates the starting position of the selected charging / discharging period among all charging / discharging periods; This indicates the position of the intermediate charging / discharging period within all charging / discharging periods; This indicates the preset quantity corresponding to the selected charging / discharging period.

[0011] In one embodiment of the present invention, the monitoring model is trained using the following method: Obtain a training sample set, which includes the longitudinal outlier average sequence and average voltage ranking sequence of multiple battery cells, as well as the sample labels of multiple battery cells. The longitudinal outlier average sequence and average voltage ranking sequence of multiple cells in the training sample set are input into the monitoring model to be trained to obtain the corresponding prediction results; the monitoring model is a one-dimensional convolutional neural network model. Based on the sample labels and prediction results of multiple battery cells, the loss value of the monitoring model loss function is calculated; Based on the loss value, the parameters of the monitoring model to be trained are adjusted to obtain the trained monitoring model.

[0012] In one embodiment of the present invention, the loss function is expressed as: ; in, Represented as the training sample set Average loss value per cell Let be the th in the training sample set Each battery cell, This represents the weighting factor configured for the faulty samples. Represented as the first Sample label for each battery cell, Represented as a logarithmic function; This is represented as the monitoring model predicting the first... The probability that a cell is a faulty cell ranges from 0 to 1. This represents the weighting factor configured for normal samples.

[0013] This invention also proposes a battery fault monitoring system based on voltage characteristics, comprising: An extraction unit is used to extract battery data from multiple consecutive charging / discharging periods from historical battery data; the battery data includes the voltage data of each cell in the battery during the charging / discharging period; The calculation unit is used to calculate the longitudinal outlier average and average voltage ranking of each cell in each charging / discharging period based on the voltage data of each cell. The sequence forming unit is used to select multiple charging / discharging periods from all charging / discharging periods for each cell based on the average voltage ranking, and to form a longitudinal outlier average sequence from the longitudinal outlier average of all selected charging / discharging periods, and to form an average voltage ranking sequence from the average voltage ranking of all selected charging / discharging periods. The generation unit is used to input the longitudinal outlier average sequence and average voltage ranking sequence of each cell into the pre-trained monitoring model to generate the monitoring results of each cell in the battery.

[0014] The present invention also proposes an electronic device, the electronic device comprising: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the electronic device to implement the voltage-characteristic-based battery fault monitoring method as described above.

[0015] The present invention also proposes a computer-readable storage medium, characterized in that it stores a computer program thereon, which, when executed by a computer processor, causes the computer to perform any of the above-described battery fault monitoring methods based on voltage characteristics.

[0016] The beneficial effects of this invention are as follows: The battery fault monitoring method, system, device, and medium based on voltage characteristics proposed in this invention extract battery data from multiple consecutive charging / discharging periods from historical battery data. This battery data includes the voltage data of each cell within the battery during each charging / discharging period. Based on the voltage data of each cell, the longitudinal outlier mean and average voltage ranking of each cell are calculated for each charging / discharging period. Based on the average voltage ranking of each cell, multiple charging / discharging periods are selected from all charging / discharging periods. Based on the longitudinal outlier mean and average voltage ranking of each cell across all selected charging / discharging periods, a corresponding longitudinal outlier mean sequence and average voltage ranking sequence are constructed. Then, the longitudinal outlier mean sequence and average voltage ranking sequence of each cell are input into a pre-trained monitoring model to generate monitoring results for each cell within the battery. This invention utilizes a pre-trained monitoring model to analyze each cell within the battery, enabling early detection of internal short-circuit faults in each cell, thereby improving battery safety during use. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0018] In the attached diagram: Figure 1 A flowchart illustrating a battery fault monitoring method based on voltage characteristics is provided for one embodiment of the present invention. Figure 2 This invention provides an architecture diagram of the monitoring model in a battery fault monitoring method based on voltage characteristics, according to an embodiment of the present invention.

[0019] Figure 3 An architecture diagram of a battery fault monitoring system based on voltage characteristics is provided for one embodiment of the present invention.

[0020] Figure 4 An architecture diagram of an electronic device is provided for one embodiment of the present invention. Detailed Implementation

[0021] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0022] Please see Figures 1 to 4 This invention proposes a battery fault monitoring method, system, device, and medium based on voltage characteristics. It can be applied to battery management or energy management fields, specifically in the field of electric vehicles, to monitor whether the power battery of an electric vehicle has experienced an internal short-circuit fault. This invention uses historical battery data and a pre-trained monitoring model to analyze each cell within the battery, thereby enabling early detection of internal short-circuit faults in individual cells and improving battery safety during use.

[0023] Please see Figure 1 This invention proposes a battery fault monitoring method based on voltage characteristics, which may include the following steps.

[0024] Step S10: Extract battery data from multiple consecutive charging / discharging periods from battery historical data; the battery data includes the voltage data of each cell in the battery during the charging / discharging period.

[0025] Specifically, battery historical data originates from continuous time-series signals collected and uploaded by sensors during actual operation. To define complete charging / discharging periods, these periods can be divided based on physical characteristics reflecting the battery's operating state. For example, when the battery is used in an electric vehicle, the historical data can be divided based on the vehicle's operating state and mileage. Charging / discharging periods are represented as charging and discharging periods, and operating states include charging and discharging states.

[0026] For example, by identifying continuous sampling points from historical battery data where the battery is charging and the electric vehicle's mileage remains constant, each segment of these continuous sampling points can be divided into a corresponding charging period. Conversely, by identifying continuous sampling points from historical battery data where the battery is discharging and the electric vehicle's mileage changes, each segment of these continuous sampling points can be divided into a corresponding discharging period.

[0027] During parking and charging, the electric vehicle's mileage remains constant because it is stationary. Therefore, we can first identify all continuous sampling points where the vehicle is in a parking and charging state with a constant mileage, and divide them into independent charging periods. The time interval between two adjacent charging periods constitutes a discharging period. For example, the charging / discharging periods of battery history data can be divided into: charging period - discharging period - charging period - discharging period - charging period - ...

[0028] For example, in battery history data, consecutive sampling points where the battery is in a parking charging state and the driving range remains unchanged are recorded as an independent charging period. If the battery is subsequently parked for charging again and the driving range increases, it is recorded as another independent charging period.

[0029] In addition, all identified charging periods are sorted and numbered according to their corresponding mileage values ​​from smallest to largest. For example, the charging period with the smallest mileage is numbered as charging period 1. Similarly, discharging periods are also sorted and numbered according to mileage from smallest to largest. Through the above process, the original long-term battery history data is transformed into a series of complete charging and discharging periods arranged in chronological order. Each charging / discharging period contains the voltage data of each cell in the battery during that period, and this voltage data forms the basis for subsequent analysis.

[0030] Step S20: Based on the voltage data of each cell, calculate the longitudinal outlier average and average voltage ranking of each cell in each charging / discharging period.

[0031] Specifically, based on the voltage data of each cell during each charging / discharging period, two key features are calculated: the longitudinal outlier mean and the average voltage ranking.

[0032] The Longitudinal Outlier Average (LOA) quantifies the degree of deviation between the voltage behavior of an individual battery cell and the overall behavior of all cells in the battery during a specific charging / discharging period. For example, the LOA quantifies the degree of deviation between the voltage behavior of an individual battery cell and the median voltage of all cells in the battery during a specific charging / discharging period. A positive LOA may indicate that the overall voltage of that cell is too high. A negative LOA may indicate that the overall voltage of that cell is too low.

[0033] Average Voltage Ranking (AVR) provides another perspective on the relative position of an individual cell within the entire battery pack. AVR can capture long-term trends in cell voltage ranking within a group; for example, a gradually deteriorating cell may exhibit a specific upward or downward pattern in its AVR.

[0034] For each battery cell, the vertical outlier mean and average voltage ranking are calculated for all charging and discharging periods. Then, feature sequences can be constructed for each cell across multiple charging / discharging periods: one is a vertical outlier mean sequence composed of the vertical outliers across multiple charging / discharging periods, and the other is an average voltage ranking sequence composed of the average voltage rankings across multiple charging / discharging periods. The vertical outlier mean sequence and average voltage ranking sequence for each battery cell form the data foundation for subsequent analysis of whether the cell is in a normal state or experiencing an internal short-circuit fault.

[0035] In one embodiment of the present invention, in step S20, the voltage data includes the voltage of each cell at each sampling point; based on the voltage data of each cell, the longitudinal outlier average value of each cell in each charging / discharging period is calculated, which may include the following steps.

[0036] Step S210: Calculate the longitudinal outlier value of each sampling point for each cell during each charging / discharging period.

[0037] Step S211: Based on the longitudinal outlier value and the number of sampling points of all sampling points of each cell in each charging / discharging period, calculate the longitudinal outlier average value of each cell in each charging / discharging period.

[0038] Longitudinal outliers ,satisfy: .

[0039] Longitudinal outlier mean ,satisfy: .

[0040] in, Represented as the first in the battery Each battery cell, Represented as the first Each charging / discharging period Represented as the first The first charging / discharging period within the [number]th [period] One sampling point, Represented as in the first When the sampling point is _th The voltage of each battery cell Represented as in the first The median value of the voltage of all cells at each sampling point. Represented as the first Number of sampling points within each charging / discharging period.

[0041] In one embodiment of the present invention, in step S20, the average voltage ranking of each cell during each charging / discharging period is calculated based on the voltage data of each cell, which may include the following steps.

[0042] Step S220: At each sampling point during each charging / discharging period, sort the voltage of all cells and record the voltage ranking of each cell at each sampling point during each charging / discharging period.

[0043] Step S221: Based on the voltage ranking and number of sampling points of all sampling points of each cell in each charging / discharging period, calculate the average voltage ranking of each cell in each charging / discharging period.

[0044] Average voltage ranking ,satisfy: .

[0045] in, Represented as the first in the battery Each battery cell, Represented as the first Each charging / discharging period Represented as the first The first charging / discharging period within the [number]th [period] One sampling point, Represented as in the first When the sampling point is _th Voltage ranking of individual battery cells Represented as the first Number of sampling points within each charging / discharging period.

[0046] Step S30: For each cell, select multiple charging / discharging periods from all charging / discharging periods based on the average voltage ranking, and form a vertical outlier average sequence by the vertical outlier average of all selected charging / discharging periods, and form an average voltage ranking sequence by the average voltage ranking of all selected charging / discharging periods.

[0047] Specifically, firstly, for each battery cell, based on its longitudinal outlier, several charging / discharging periods with the most representative characteristics or significant abrupt changes can be selected from all charging / discharging periods. Secondly, for each battery cell, the longitudinal outlier sequences of all selected charging / discharging periods are used to form a longitudinal outlier sequence, and the average voltage rankings of all selected charging / discharging periods are used to form an average voltage ranking sequence. The feature sequence pair consisting of the longitudinal outlier sequence and the average voltage ranking sequence for each battery cell is used as input to the subsequent monitoring model.

[0048] It's important to note that the feature sequence pairs, consisting of the longitudinal outlier average sequence and the average voltage ranking sequence, are constructed based on multiple discharge periods or multiple charging periods of the battery cell. This is because charging and discharging are two independent operating states of the battery. Internal short circuit faults manifest as slow charging during charging and rapid discharging during discharging. These two anomalies form stable and identifiable patterns during the charging and discharging periods, respectively. Therefore, analyzing the longitudinal outlier average sequence and the average voltage ranking sequence during either the charging or discharging period provides the ability to independently diagnose internal short circuits within the battery cell.

[0049] In one embodiment of the present invention, in step S30, for each cell, multiple charging / discharging periods are selected from all charging / discharging periods based on the average voltage ranking, and the longitudinal outlier average of all selected charging / discharging periods is formed into a longitudinal outlier average sequence, and the average voltage ranking of all selected charging / discharging periods is formed into an average voltage ranking sequence, which may include the following steps.

[0050] Step S310: For each cell, calculate the difference between its average voltage ranking in each charging / discharging period and the average voltage ranking in the previous charging / discharging period, and record the charging / discharging period with the largest negative difference as the intermediate charging / discharging period.

[0051] Specifically, for each cell in the battery, its overall behavior changes across all charging / discharging periods are analyzed. For each cell, its average voltage ranking in each charging / discharging period is calculated, along with the difference between its average voltage ranking in the immediately preceding charging / discharging period. All differences are iterated through, and the one with the largest negative absolute value is identified. The charging / discharging period corresponding to this difference is selected as the intermediate charging / discharging period for that cell, which typically represents a potential critical point where the cell's voltage consistency undergoes the most significant negative abrupt change.

[0052] Step S320: Based on the location of the intermediate charging / discharging period, select a preset number of charging / discharging periods in sequence from all charging / discharging periods.

[0053] Specifically, after determining the intermediate charging / discharging period, a continuous period of fixed length is extracted based on its position within the entire period. For example, if the preset length is 128 periods, then the middle period is used as the center, and periods are extended forward and backward to form a continuous period set with a total length of 128.

[0054] In one embodiment of the present invention, the selected charging / discharging period satisfies: .

[0055] in, This indicates the starting position of the selected charging / discharging period among all charging / discharging periods; This indicates the position of the intermediate charging / discharging period within all charging / discharging periods; This indicates the preset quantity corresponding to the selected charging / discharging period.

[0056] The monitoring model in this embodiment is a one-dimensional convolutional neural network model, which can process feature sequences of length 128. It can be 128.

[0057] When the total number of all charging or discharging periods of the battery cell is less than the preset number In this case, all charging or discharging periods need to be filled in to meet the requirements. Specifically, before the start of the charging or discharging period of the battery cell, several charging or discharging periods with a value of 0 are added, so that the total number of all charging or discharging periods is greater than the preset number. .

[0058] Step S330: Select the longitudinal outlier average of all charging / discharging periods to form a longitudinal outlier average sequence, and select the average voltage ranking of all charging / discharging periods to form an average voltage ranking sequence.

[0059] Specifically, the longitudinal outlier averages of all charging / discharging periods will be selected to form a longitudinal outlier average sequence, and the average voltage rankings of all charging / discharging periods will be selected to form an average voltage ranking sequence. The feature sequence pair consisting of the longitudinal outlier average sequence and the average voltage ranking sequence for each cell will be used as the input to the subsequent monitoring model.

[0060] Step S40: Input the longitudinal outlier average sequence and average voltage ranking sequence of each cell into the pre-trained monitoring model to generate the monitoring results of each cell in the battery.

[0061] Specifically, for each battery cell, the feature sequence pair consisting of its longitudinal outlier average sequence and average voltage ranking sequence constitutes the input of the monitoring model. The monitoring model analyzes the longitudinal outlier average sequence and average voltage ranking sequence of each battery cell and directly outputs the monitoring result of each battery cell, that is, it determines whether each battery cell is normal or has an internal short circuit fault.

[0062] like Figure 2 As shown, in one embodiment of the present invention, the monitoring model is trained using the following method: Obtain a training sample set, which includes the longitudinal outlier mean sequence and average voltage ranking sequence of multiple battery cells, as well as the sample labels of multiple battery cells.

[0063] The longitudinal outlier average sequence and average voltage ranking sequence of multiple cells in the training sample set are input into the monitoring model to be trained to obtain the corresponding prediction results; the monitoring model is a one-dimensional convolutional neural network model.

[0064] Based on the sample labels and prediction results of multiple battery cells, the loss value of the monitoring model loss function is calculated.

[0065] Based on the loss value, the parameters of the monitoring model to be trained are adjusted to obtain the trained monitoring model.

[0066] The training process of the monitoring model for the training sample set is as follows.

[0067] First, a training sample set is constructed, containing data samples from a large number of battery cells in known states. Each data sample includes the cell's vertical outlier mean sequence, average voltage ranking sequence, and a sample label indicating whether the cell is normal or faulty. The vertical outlier mean sequence and average voltage ranking sequence are constructed using a standardized fixed-length method.

[0068] Secondly, the longitudinal outlier mean sequence and average voltage ranking sequence from the data samples are input into the monitoring model to be trained. This monitoring model, consistent with the one described above, is a one-dimensional convolutional neural network model. The monitoring model calculates for each input data sample and outputs a prediction result, which represents the probability that the corresponding battery cell is in a faulty state.

[0069] Then, the prediction results of the monitoring model are compared with the actual sample labels, and the difference between the two is calculated using a pre-defined loss function. This difference is quantified as a loss value. The loss value reflects the overall error of the model's current prediction.

[0070] Based on the calculated loss value, all trainable parameters in the monitoring model are adjusted in reverse, such as convolutional kernel weights and biases. The goal is to continuously reduce the loss value, that is, to make the monitoring model's predictions increasingly closer to the true labels. This process is iterated multiple times on the training set. Training terminates when the monitoring model performs optimally on an independent validation set or when the stopping condition is met. At this point, a monitoring model with fixed parameters is obtained, which is the monitoring model that can be used for actual diagnostic training.

[0071] For example, to prevent the monitoring model from overfitting to the data samples during training, this scheme introduces an early stopping mechanism. During training iterations, the loss value performance of the monitoring model on an independent validation dataset is continuously monitored. If the validation set loss value does not decrease for 10 consecutive training epochs, the training process will be automatically terminated. At the same time, the current parameters will not be saved; instead, the system will automatically roll back and load the weight parameters corresponding to the monitoring model with the lowest validation set loss value during the entire training process. This will be used as the final trained monitoring model.

[0072] In one embodiment of the present invention, the loss function for a single data sample ,satisfy: .

[0073] Weighting factors of fault samples ,satisfy: .

[0074] Weighting factors for normal samples ,satisfy: .

[0075] in, This represents the weighting factor configured for the faulty samples. The sample label represents a single data sample cell. Represented as a logarithmic function; This represents the probability that a single data sample cell is a faulty cell, as predicted by the monitoring model, and its value ranges from 0 to 1. This represents the weighting factor configured for normal samples.

[0076] Sample Labels A value of 1 indicates that the battery cell is faulty; sample label. A value of 0 indicates that the battery cell is a normal battery cell.

[0077] It should be noted here that a weighting factor for faulty samples is added to the loss function. Weighting factors for normal samples This is addressed by introducing a classification weight mechanism to solve the class imbalance problem, such as when the number of normal samples far exceeds the number of faulty samples. By assigning higher weights to the minority class (faulty samples), a greater penalty can be imposed on misclassification of faulty samples during the loss calculation process, thereby ensuring that the monitoring model focuses more on the accurate identification of faulty samples.

[0078] The average loss function for the training sample set ,satisfy: .

[0079] in, Represented as the training sample set Average loss value per cell Let be the th in the training sample set Each battery cell, This represents the weighting factor configured for the faulty samples. Represented as the first Sample label for each battery cell, Represented as a logarithmic function; This is represented as the monitoring model predicting the first... The probability that a cell is a faulty cell ranges from 0 to 1. This represents the weighting factor configured for normal samples.

[0080] For actual cell diagnostics, the monitoring model's processing procedure is as follows.

[0081] like Figure 2 As shown, in one embodiment of the present invention, the monitoring model is a one-dimensional convolutional neural network model, which consists of a feature extraction module and a classification module. There are two feature extraction modules, and their structures are completely identical.

[0082] The input to the monitoring model is a two-dimensional feature sequence for each battery cell, represented as a 128×2 matrix. The first column is a 128-column sequence of vertical outlier averages, and the second column is a 128-column sequence of average voltage rankings. For each battery cell's two-dimensional feature sequence, it is first input into the feature extraction module of the monitoring model for feature extraction. Then, the extracted features are input into the classification module of the monitoring model. The classification module then monitors the extracted features and outputs the monitoring results.

[0083] The feature extraction module works as follows: like Figure 2 As shown, the feature extraction module consists of a convolutional layer, a normalization layer, an activation layer, and a max pooling layer from top to bottom.

[0084] for Figure 2 The first feature extraction module on the left: In the convolutional layer, a one-dimensional convolutional kernel with a specific span is used to slide along the time axis, simultaneously covering the sequences of the two channels: the longitudinal outlier mean and the average voltage ranking. This allows for the simultaneous extraction of the joint change patterns of the two features in the time dimension and their nonlinear coupling relationship, generating 32 sets of preliminary abstract feature maps.

[0085] In the normalization layer, the feature maps output by the convolutional layer are batch normalized to eliminate differences in data units, accelerate the convergence speed of the training process, and stabilize the training.

[0086] In the activation layer, the normalized features are nonlinearly transformed by the activation function to enhance the model's sensitivity and representation ability to small internal short-circuit fault features.

[0087] In the max pooling layer, max pooling is performed on the activated features, which can reduce the dimensionality of the data and filter out some noise while preserving the key fault signals.

[0088] Then, enter Figure 2 In the second feature extraction module on the right, the above convolution, normalization, activation, and pooling process is repeated to perform deeper feature extraction and abstraction.

[0089] The classification module works as follows: like Figure 2 As shown, the classification module consists of a flattened layer and a fully connected layer from top to bottom.

[0090] In the flattening layer, the multidimensional feature map obtained after two rounds of feature extraction is flattened into a one-dimensional feature vector.

[0091] In the fully connected layer, the one-dimensional vector is fed into the fully connected layer for comprehensive calculation of high-dimensional features and nonlinear mapping.

[0092] To improve the generalization ability of the monitoring model and prevent overfitting to the two-dimensional feature sequences of the battery cell, the classification module introduces a regularization mechanism in the output calculation of the fully connected layer. Next, the output value of the fully connected layer is input into an activation function, which processes the output of the fully connected layer, mapping it to a probability between 0 and 1. Then, the classification module uses the probability output by the activation function as the basis for output judgment, setting 0.5 as the decision threshold. When the output probability is greater than or equal to 0.5, the battery cell is determined to be in a fault state (i.e., an internal short circuit has occurred); otherwise, the battery cell is determined to be in a normal state.

[0093] Please see Figure 3 The present invention proposes a battery fault monitoring system 100 based on voltage characteristics, which may include an extraction unit 110, a calculation unit 120, a sequence forming unit 130 and a generation unit 140.

[0094] The extraction unit 110 is used to extract battery data from multiple consecutive charging / discharging periods from battery historical data; the battery data includes the voltage data of each cell in the battery during the charging / discharging period.

[0095] The calculation unit 120 is used to calculate the longitudinal outlier average and average voltage ranking of each cell in each charging / discharging period based on the voltage data of each cell.

[0096] The sequence forming unit 130 is used to select multiple charging / discharging periods from all charging / discharging periods for each cell based on the average voltage ranking, and to form a longitudinal outlier average sequence from the longitudinal outlier average of all selected charging / discharging periods, and to form an average voltage ranking sequence from the average voltage ranking of all selected charging / discharging periods.

[0097] The generation unit 140 is used to input the longitudinal outlier average sequence and average voltage ranking sequence of each cell into the pre-trained monitoring model to generate the monitoring results of each cell in the battery.

[0098] Please see Figure 4 The present invention proposes an electronic device 200, which may include a memory 210, a processor 220 and a bus, and may also include a computer program stored in the memory 210 and executable on the processor 220, such as a battery fault monitoring program.

[0099] The memory 210 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory, magnetic memory, disk, optical disk, etc. In some embodiments, the memory 210 can be an internal storage unit of the electronic device 200, such as the portable hard drive of the electronic device 200. In other embodiments, the memory 210 can be an external storage device of the electronic device 200, such as a plug-in portable hard drive, smart memory card, secure digital card, flash memory card, etc., equipped on the electronic device 200. Furthermore, the memory 210 can include both internal and external storage units of the electronic device 200. The memory 210 can be used not only to store application software and various types of data installed on the electronic device 200, such as battery fault monitoring code, but also to temporarily store data that has been output or will be output.

[0100] In some embodiments, processor 220 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units, microprocessors, digital processing chips, graphics processors, and various control chips. Processor 220 is the control core of electronic device 200, connecting various components of the entire electronic device 200 via various interfaces and lines. It executes programs or modules (such as battery fault monitoring programs) stored in memory 210 and calls data stored in memory 210 to perform various functions of electronic device 200 and process data.

[0101] The processor 220 executes the operating system of the electronic device 200 and various installed applications. The processor 220 executes the applications to implement the steps in the battery fault monitoring method described above.

[0102] For example, a computer program may be divided into one or more modules, one or more of which are stored in memory 210 and executed by processor 220 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in electronic device 200. For example, the computer program may be divided into an extraction unit 110, a calculation unit 120, a sequence forming unit 130, and a generation unit 140.

[0103] The integrated unit implemented as a software functional module described above can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The software functional module stored in the storage medium includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute some functions of the battery fault monitoring method of the various embodiments of this application.

[0104] In summary, this invention proposes a battery fault monitoring method, system, device, and medium based on voltage characteristics. It extracts battery data from multiple consecutive charging / discharging periods from historical battery data. This battery data includes the voltage data of each cell within the battery during each charging / discharging period. Based on the voltage data of each cell, it calculates the longitudinal outlier mean and average voltage ranking for each cell in each charging / discharging period. Based on the average voltage ranking of each cell, multiple charging / discharging periods are selected from all charging / discharging periods. Based on the longitudinal outlier mean and average voltage ranking of each cell in all selected charging / discharging periods, a corresponding longitudinal outlier mean sequence and average voltage ranking sequence are constructed. Then, the longitudinal outlier mean sequence and average voltage ranking sequence of each cell are input into a pre-trained monitoring model to generate monitoring results for each cell within the battery. This invention utilizes a pre-trained monitoring model to analyze each cell within the battery, enabling early detection of internal short-circuit faults in each cell, thereby improving battery safety during use.

[0105] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A battery fault monitoring method based on voltage characteristics, characterized in that, include: Battery data from multiple consecutive charging / discharging periods is extracted from historical battery data; the battery data includes the voltage data of each cell in the battery during the charging / discharging period. Based on the voltage data of each cell, calculate the longitudinal outlier average and average voltage ranking of each cell in each charging / discharging period; For each battery cell, multiple charging / discharging periods are selected from all charging / discharging periods based on the average voltage ranking. The longitudinal outlier average of all selected charging / discharging periods is used to form a longitudinal outlier average sequence, and the average voltage ranking of all selected charging / discharging periods is used to form an average voltage ranking sequence. The longitudinal outlier average sequence and average voltage ranking sequence of each cell are input into the pre-trained monitoring model to generate the monitoring results of each cell in the battery.

2. The battery fault monitoring method based on voltage characteristics according to claim 1, characterized in that, The voltage data includes the voltage of each cell at each sampling point; the calculation of the longitudinal outlier average for each cell during each charging / discharging period based on the voltage data of each cell includes: Calculate the longitudinal outlier value of each cell at each sampling point during each charging / discharging period; Based on the longitudinal outlier value and the number of sampling points of all sampling points for each cell in each charging / discharging period, the longitudinal outlier average value of each cell in each charging / discharging period is calculated. Longitudinal outliers ,satisfy: ； Longitudinal outlier mean ,satisfy: ； in, Represented as the first in the battery Each battery cell, Represented as the first Each charging / discharging period Represented as the first The first charging / discharging period within the [number]th [period] One sampling point, Represented as in the first When the sampling point is _th The voltage of each battery cell Represented as in the first The median voltage of all cells at each sampling point. Represented as the first Number of sampling points within each charging / discharging period.

3. The battery fault monitoring method based on voltage characteristics according to claim 1, characterized in that, The voltage data includes the voltage of each cell at each sampling point; The calculation of the average voltage ranking of each cell during each charging / discharging period, based on the voltage data of each cell, includes: At each sampling point during each charging / discharging period, the voltage of all cells is sorted, and the voltage ranking of each cell at each sampling point during each charging / discharging period is recorded. Based on the voltage ranking and number of sampling points of each cell in each charging / discharging period, calculate the average voltage ranking of each cell in each charging / discharging period; Average voltage ranking ,satisfy: ； in, Represented as the first in the battery Each battery cell, Represented as the first Each charging / discharging period Represented as the first The first charging / discharging period within the [number]th [period] One sampling point, Represented as in the first When the sampling point is _th Voltage ranking of individual battery cells Represented as the first Number of sampling points within each charging / discharging period.

4. The battery fault monitoring method based on voltage characteristics according to claim 1, characterized in that, For each battery cell, multiple charging / discharging periods are selected from all charging / discharging periods based on average voltage ranking. The longitudinal outlier averages of all selected charging / discharging periods are then used to form a longitudinal outlier average sequence. The average voltage rankings of all selected charging / discharging periods are then used to form an average voltage ranking sequence, including: For each cell, calculate the difference between its average voltage ranking in each charging / discharging period and the average voltage ranking in the previous charging / discharging period, and record the charging / discharging period with the largest negative difference as the intermediate charging / discharging period; Based on the location of the intermediate charging / discharging period, a preset number of charging / discharging periods are selected sequentially from all charging / discharging periods; The longitudinal outlier averages of all charging / discharging periods will be selected to form a longitudinal outlier average sequence, and the average voltage rankings of all charging / discharging periods will be selected to form an average voltage ranking sequence.

5. The battery fault monitoring method based on voltage characteristics according to claim 4, characterized in that, The selected charging / discharging period satisfies: ； in, This indicates the starting position of the selected charging / discharging period among all charging / discharging periods; This indicates the position of the intermediate charging / discharging period within all charging / discharging periods; This indicates the preset quantity corresponding to the selected charging / discharging period.

6. The battery fault monitoring method based on voltage characteristics according to claim 1, characterized in that, The monitoring model was trained using the following method: Obtain a training sample set, which includes the longitudinal outlier average sequence and average voltage ranking sequence of multiple battery cells, as well as the sample labels of multiple battery cells. The longitudinal outlier average sequence and average voltage ranking sequence of multiple cells in the training sample set are input into the monitoring model to be trained to obtain the corresponding prediction results; the monitoring model is a one-dimensional convolutional neural network model. Based on the sample labels and prediction results of multiple battery cells, the loss value of the monitoring model loss function is calculated; Based on the loss value, the parameters of the monitoring model to be trained are adjusted to obtain the trained monitoring model.

7. The battery fault monitoring method based on voltage characteristics according to claim 6, characterized in that, The loss function is expressed as: ； in, Represented as the training sample set Average loss value per cell Let be the th in the training sample set Each battery cell, This represents the weighting factor configured for the faulty samples. Represented as the first Sample label for each battery cell, Represented as a logarithmic function; This is represented as the monitoring model predicting the first... The probability that a cell is a faulty cell ranges from 0 to 1. This represents the weighting factor configured for normal samples.

8. A battery fault monitoring system based on voltage characteristics, characterized in that, include: An extraction unit is used to extract battery data from multiple consecutive charging / discharging periods from historical battery data; the battery data includes the voltage data of each cell in the battery during the charging / discharging period; The calculation unit is used to calculate the longitudinal outlier average and average voltage ranking of each cell in each charging / discharging period based on the voltage data of each cell. The sequence forming unit is used to select multiple charging / discharging periods from all charging / discharging periods for each cell based on the average voltage ranking, and to form a longitudinal outlier average sequence from the longitudinal outlier average of all selected charging / discharging periods, and to form an average voltage ranking sequence from the average voltage ranking of all selected charging / discharging periods. The generation unit is used to input the longitudinal outlier average sequence and average voltage ranking sequence of each cell into the pre-trained monitoring model to generate the monitoring results of each cell in the battery.

9. An electronic device, characterized in that, The electronic device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the electronic device to implement the battery fault monitoring method based on voltage characteristics as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by the computer's processor, causes the computer to perform the battery fault monitoring method based on voltage characteristics as described in any one of claims 1 to 7.

Images