Battery safety and life synchronous prediction method and device, equipment and storage medium

Abstract

The invention provides a battery safety and service life synchronous prediction method and device, equipment and a storage medium, and relates to the technical field of battery detection, the method comprises the following steps: obtaining charging and discharging data of a battery in a plurality of application scenes, carrying out data preprocessing on the charging and discharging data, and obtaining a standard data set; constructing a characteristic curve according to the standard data set; performing random masking on the characteristic curve to obtain a mask characteristic curve; constructing a deep neural network model, and training the deep neural network model according to the mask feature curve to obtain an initial prediction model; constructing a target mask feature curve of the battery in the target application scene, and performing optimization adjustment on the initial prediction model according to the target mask feature curve to obtain a target prediction model; and constructing a target characteristic curve of the target battery in the target application scene, predicting the target characteristic curve by using the target prediction model, and synchronously obtaining the safety state, the health degree and the service life of the target battery. According to the invention, the battery prediction precision and coverage can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery detection, and particularly to a method, device, equipment and storage medium for synchronously predicting battery safety and life. Background Art

[0002] With the rapid development of the new energy vehicle industry, lithium-ion batteries have been widely used in the field of power batteries due to their significant advantages such as high energy density, low self-discharge rate, long cycle life, and excellent environmental protection performance. However, during the calendar life and cyclic use of lithium-ion batteries, problems such as capacity attenuation, performance degradation, and deterioration of safety and stability are common, resulting in a continuous decrease in the state of health (SOH) of the battery. In particular, the non-linear attenuation characteristics, sudden failure, and thermal runaway risk of lithium-ion batteries seriously restrict their further popularization and application in power systems, energy storage systems, and electric equipment. Therefore, carrying out research on lithium-ion battery health assessment and diagnosis has important engineering practical significance.

[0003] However, once a lithium-ion battery is installed in a vehicle, the real-time prediction of its state of safety (SOS) and SOH faces many technical challenges. Currently, the prediction research on SOS and SOH mainly adopts an independent analysis method: in terms of SOS prediction, it mainly relies on real-time signals such as voltage, temperature, and current collected by the battery management system, judges by setting thresholds, or extracts the safety characteristics of lithium-ion batteries based on laboratory data, and constructs a correlation model between SOS and safety characteristics for estimation; in terms of SOH prediction, it mainly establishes a mapping relationship between the performance of the battery cell and aging factors such as the number of cycles and calendar time based on the aging data of laboratory battery cells, and then realizes the prediction of SOH. However, the existing methods generally have limitations such as overly simplified models and insufficient utilization of data features, resulting in the prediction accuracy and stability under actual working conditions being difficult to meet the requirements of engineering applications. Summary of the Invention

[0004] In view of the above deficiencies of the prior art, the present invention provides a method, device, equipment and storage medium for synchronously predicting battery safety and life, effectively solving the problem that the prediction accuracy and stability of the existing battery safety and life under actual working conditions are difficult to meet the requirements of engineering applications.

[0005] In a first aspect, the present invention provides a method for synchronously predicting battery safety and life, the method comprising:

[0006] Obtain the charge and discharge data of the battery under multiple application scenarios, and perform data preprocessing on the charge and discharge data to obtain a standard data set;

[0007] Construct a characteristic curve based on the standard data set, where the characteristic curve at least includes a charging segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve;

[0008] Perform random masking on the characteristic curve to obtain a masked characteristic curve;

[0009] Construct a deep neural network model, and train the deep neural network model according to the masked characteristic curve to obtain an initial prediction model;

[0010] Construct a target masked characteristic curve of the battery in the target application scenario, and optimize and adjust the initial prediction model according to the target masked characteristic curve to obtain a target prediction model;

[0011] Construct a target characteristic curve of the target battery in the target application scenario, and use the target prediction model to predict the target characteristic curve to simultaneously obtain the safety status, health degree, and life of the target battery.

[0012] In an alternative embodiment, the constructing the characteristic curve according to the standard data set includes:

[0013] Resample the capacity values of the battery cell according to a preset time interval to obtain a capacity value sequence;

[0014] Resample the charging voltage of the battery cell according to the preset time interval to obtain a charging voltage sequence, and construct a charging segment curve according to the charging voltage sequence and the capacity value sequence;

[0015] Resample the temperature values of the battery cell according to the preset time interval to obtain a temperature value sequence, and construct the temperature segment curve according to the temperature value sequence and the capacity value sequence;

[0016] Resample the incremental capacity of the battery cell according to the preset time interval to obtain an incremental capacity sequence, and construct the incremental capacity curve according to the incremental capacity sequence and the capacity value sequence;

[0017] Resample the differential voltage of the battery cell according to the preset time interval to obtain a differential voltage sequence, and construct the differential voltage curve according to the differential voltage sequence and the capacity value sequence.

[0018] In an alternative embodiment, the deep neural network model includes an input layer, an encoder layer, a decoder layer, a fully connected layer, and an output layer, where:

[0019] The input layer is connected to the encoder layer. The encoder layer stacks multiple attention mechanism modules. The decoder layer stacks two of the attention mechanism modules. The attention mechanism module includes a normalization layer, a multi-head attention layer, a residual layer, and a fully connected layer;

[0020] The fully connected layer includes a first fully connected layer and a second fully connected layer. The first fully connected layer is connected to the encoder layer and the decoder layer. The second fully connected layer is connected to the decoder layer and the output layer.

[0021] In an alternative embodiment, training the deep neural network model according to the masked feature curve to obtain an initial prediction model includes:

[0022] Performing positional encoding on the masked feature curve to obtain positional encoding values;

[0023] Combining the masked feature curve and the positional encoding values to obtain an input curve, and inputting the input curve into the encoder layer through the input layer;

[0024] Using the encoder layer to extract general temporal features, and then inputting them into the decoder layer through the first fully connected layer;

[0025] Using the decoder layer to decode the general temporal features, and obtaining an output curve through the second fully connected layer and the output layer;

[0026] Obtaining the initial prediction model by performing backpropagation driving on the output curve and the input curve through a loss function.

[0027] In an alternative embodiment, the target prediction model includes a first target prediction model and a second target prediction model. Constructing a target masked feature curve of the battery in the target application scenario, and optimizing and adjusting the initial prediction model according to the target masked feature curve to obtain the target prediction model, includes:

[0028] Obtaining the test charge and discharge data of the battery in the target application scenario, and identifying the battery capacity under the charging condition to obtain an identification sequence;

[0029] Constructing a target masked feature curve according to the test charge and discharge data, and inputting the target masked feature curve into the initial prediction model to obtain a first test curve;

[0030] Performing backpropagation driving according to the reconstruction error between the first test curve and the target masked feature curve to obtain a first target prediction model;

[0031] Inputting the target masked feature curve and the identification sequence into the initial prediction model to obtain a second test curve;

[0032] Backpropagation drive is performed according to the reconstruction error between the second test curve and the identification sequence to obtain a second target prediction model.

[0033] In an alternative embodiment, constructing the target characteristic curve of the target battery in the target application scenario, and predicting the target characteristic curve by using the target prediction model to synchronously obtain the safety state, health degree, and life of the target battery, includes:

[0034] Obtain the target charge-discharge data of the target battery, and construct the target characteristic curve according to the target charge-discharge data;

[0035] Input the target characteristic curve into the first target prediction model to obtain a first target prediction curve;

[0036] Calculate the reconstruction error between the first target prediction curve and the target characteristic curve;

[0037] Construct an error sequence according to the reconstruction error, and perform normalization processing on the error sequence to obtain a standard error sequence;

[0038] Determine the target battery with a standard error sequence greater than the target threshold as an abnormal battery, and determine the target battery with a standard error sequence less than or equal to the target threshold as a normal battery.

[0039] In an alternative embodiment, after constructing the target characteristic curve according to the target charge-discharge data, it further includes:

[0040] Input the target characteristic curve into the second target prediction model to obtain a second target prediction curve;

[0041] Determine the capacity integral at the cut-off voltage according to the second target prediction curve, and obtain the maximum available capacity as the battery health degree;

[0042] Continuously obtain the battery health degree for a set threshold number of cycles, construct a battery health curve according to the battery health degree, and input the battery health curve into the second target prediction model to obtain a battery life prediction curve.

[0043] In a second aspect, the present invention provides a device for synchronously predicting battery safety and life, and the device includes:

[0044] A data acquisition module, configured to acquire charge-discharge data of batteries in multiple application scenarios, perform data preprocessing on the charge-discharge data, and obtain a standard data set;

[0045] A curve construction module, configured to construct a feature curve according to the standard data set, where the feature curve at least includes a charging segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve;

[0046] A curve masking module, configured to randomly mask the feature curve to obtain a masked feature curve;

[0047] A model training module, configured to construct a deep neural network model, and train the deep neural network model according to the masked feature curve to obtain an initial prediction model;

[0048] A model adjustment module, configured to construct a target masked feature curve of a battery in a target application scenario, and optimize and adjust the initial prediction model according to the target masked feature curve to obtain a target prediction model;

[0049] A model prediction module, configured to construct a target feature curve of a target battery in a target application scenario, and use the target prediction model to predict the target feature curve, and synchronously obtain the safety state, health degree, and life of the target battery.

[0050] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the battery safety and life synchronous prediction method according to any one of the first aspects of the present invention.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the battery safety and life synchronous prediction method according to any one of the first aspects of the present invention.

[0052] The battery safety and life synchronous prediction method, device, equipment, and storage medium provided by the present invention use massive data for self-supervised learning, and only a small number of samples to be measured are used for fine-tuning, so as to complete multi-task prediction of safety state, health state, and life, greatly reducing the labeling cost and development cycle. Based on the high-fidelity restoration after randomly masking the battery curve by the deep neural network model, self-training is realized. One model can be used for different battery types, application scenarios, and tasks, with high self-adaptability and efficiency, and can quickly match different batteries for synchronous prediction of safety state and life. At the same time, the prediction accuracy and coverage are improved, and parallel estimation of safety state and life is realized. Description of the Drawings

[0053] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0054] Figure 1 It is the first schematic diagram of the battery safety and life synchronous prediction method flow provided by the embodiment of the present invention;

[0055] Figure 2 It is the second schematic diagram of the battery safety and life synchronous prediction method flow provided by the embodiment of the present invention;

[0056] Figure 3 It is the schematic diagram of the charging segment curve in the embodiment of the present invention;

[0057] Figure 4 It is the schematic diagram of the incremental capacity curve in the embodiment of the present invention;

[0058] Figure 5 It is the schematic diagram of the mask feature curve of the charging segment curve in the embodiment of the present invention;

[0059] Figure 6 It is the schematic diagram of the mask feature curve of the incremental capacity curve in the embodiment of the present invention;

[0060] Figure 7 It is the schematic diagram of the structure of the deep neural network model provided by the embodiment of the present invention;

[0061] Figure 8 It is the third schematic diagram of the battery safety and life synchronous prediction method flow provided by the embodiment of the present invention;

[0062] Figure 9 It is the schematic diagram of the convergence process of the loss function during model training provided by the embodiment of the present invention;

[0063] Figure 10 It is the fourth schematic diagram of the battery safety and life synchronous prediction method flow provided by the embodiment of the present invention;

[0064] Figure 11 It is the fifth schematic diagram of the battery safety and life synchronous prediction method flow provided by the embodiment of the present invention;

[0065] Figure 12 It is the schematic diagram of the battery health curve provided by the embodiment of the present invention;

[0066] Figure 13 It is the schematic diagram of the battery life prediction curve provided by the embodiment of the present invention;

[0067] Figure 14 It is a schematic structural diagram of a battery safety and life synchronous prediction device provided by an embodiment of the present invention;

[0068] Figure 15 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention.

[0069] Main element symbol description:

[0070] 200, battery safety and life synchronous prediction device; 210, data acquisition module; 220, curve construction module; 230, curve masking module; 240, model training module; 250, model adjustment module; 250, model prediction module; 300, electronic device; 310, processor; 320, communication interface; 330, memory; 340, communication bus. Specific embodiments

[0071] To make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be further clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be noted that the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present invention.

[0072] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0073] Once a lithium-ion battery is installed in a vehicle, the real-time prediction of its battery safety state (SOS) and battery health state (SOH) faces many technical challenges. The prediction research for SOS and SOH mainly adopts an independent analysis method.

[0074] The SOS prediction mainly relies on the key parameters such as voltage, temperature and current collected by the vehicle battery management system, and judges the safety state by setting thresholds. In addition, based on laboratory data, by capturing the safety characteristics of lithium-ion batteries, a correlation model between SOS and characteristic parameters can be constructed, so as to realize the prediction of the battery safety state. However, these methods are difficult to effectively adapt to the complex working conditions of random charging and discharging during the actual vehicle operation, resulting in the prediction accuracy and coverage not reaching the ideal level. The SOH prediction mainly relies on the aging data of the battery cells in the laboratory environment. By establishing the mapping relationship between the battery cell performance and aging factors such as the number of cycles and calendar time, the prediction of the battery SOH can be realized. However, it is difficult to extract the same aging factors as in the laboratory environment under actual working conditions, resulting in the battery aging model trained based on laboratory working conditions being unable to effectively generalize to the actual application scenario, and its prediction accuracy and stability are difficult to reach the expected level throughout the battery life cycle.

[0075] Embodiment 1

[0076] The embodiment of the present invention provides a method for synchronously predicting battery safety and life, which effectively solves the problem that the prediction accuracy and stability of the existing battery safety and life under actual working conditions are difficult to meet the requirements of engineering applications. Figure 1 It is the first schematic diagram of the process of the method for synchronously predicting battery safety and life provided by the embodiment of the present invention, as Figure 1 shown, the method includes the following steps:

[0077] S100. Obtain the charge and discharge data of the battery under multiple application scenarios, and perform data preprocessing on the charge and discharge data to obtain a standard data set.

[0078] In the embodiment of the present invention, the battery can be a lithium-ion battery, and the application scenarios include but are not limited to vehicle terminals, energy storage terminals, and laboratory scenarios. According to the actual operation data of lithium-ion batteries, obtain the vehicle-terminal charge and discharge data, energy storage-terminal charge and discharge data, and laboratory charge and discharge data of lithium-ion batteries, and offline construct a massive data set of lithium-ion batteries. The high-quality massive data set is mainly used for model pre-training to construct the general time-series characteristics of the charge and discharge behavior of lithium-ion batteries.

[0079] Extract the battery cell voltage data, battery cell temperature data, battery cell current data, and charge and discharge status from the massive data set. Specifically, each battery cell generates a voltage value at each moment, and all the voltage values of the series system composed of multiple battery cells form a voltage list, which is the battery cell voltage data. Similarly, the battery cell temperature data, battery cell current data, and charge and discharge status can be obtained in the same way.

[0080] Data cleaning is performed on abnormal values ​​in the extracted data, which include but are not limited to illegal values, jump values, sampling abnormal values, mutation values, and signal jitter values. Specifically, illegal values ​​are directly eliminated. For example, the reasonable range of battery cell voltage is within 1.8V~4.5V, and values ​​outside the range are considered illegal. Forward filling, average filling, or weighted filling are performed on jump values, mutation values, and signal jitter values, and the event range is eliminated for sampling abnormal values.

[0081] After data cleaning, the operating conditions are classified according to the vehicle status and charging status sent by the battery, mainly divided into the following operating conditions: charging condition, static condition and discharge condition. Specifically, if the battery is in a charging state, it is marked as a charging condition, if the battery is in an uncharged state and a discharge state, it is marked as a discharge condition, and if the battery is in a uncharged and undischarged state, it is marked as a static condition.

[0082] S200 , constructing a characteristic curve according to a standard data set, where the characteristic curve at least includes a charging segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve.

[0083] After data cleaning and working condition division of massive data sets, segment data that meet the charging conditions are selected to construct characteristic curves, which include charging segment curves, temperature segment curves, incremental capacity curves and differential voltage curves. Figure 2 FIG. 2 is a second schematic diagram of the battery safety and life synchronous prediction method process provided by an embodiment of the present invention. Figure 2 As shown, the construction of the characteristic curve specifically includes the following steps:

[0084] S210: Resample the capacity values ​​of the battery cells according to a preset time interval to obtain a capacity value sequence.

[0085] In the embodiment of the present invention, the capacity value is the integral of the cell current over time, and its function expression is as follows:

[0086]

[0087] In the above formula, Q t It represents the capacity value at time t, and i represents the current at any time.

[0088] The cell current data is integrated to obtain the capacity value, and then the capacity value is resampled at a preset time interval to obtain a capacity value sequence ah_seq_list = [Q1, Q2, ..., Q n ], where Q n It represents the capacity value at the nth time interval point. In the embodiment of the present invention, the preset time interval can be set to 30S.

[0089] S220. Resample the charging voltage of the battery cell according to a preset time interval to obtain a charging voltage sequence, and construct a charging segment curve according to the charging voltage sequence and the capacity value sequence.

[0090] Resample the cell voltage data according to a preset time interval, that is, obtain a voltage value every 30S, so as to obtain a charging voltage sequence volt_seq_list = [V1, V2, …, V n , where V n represents the voltage value at the nth time interval point. Figure 3 This is a schematic diagram of the charging segment curve in the embodiment of the present invention. As Figure 3 shown, determine the abscissa of the charging segment curve according to the capacity value sequence, and determine the ordinate of the charging segment curve according to the charging voltage sequence.

[0091] S230. Resample the temperature value of the battery cell according to a preset time interval to obtain a temperature value sequence, and construct a temperature segment curve according to the temperature value sequence and the capacity value sequence.

[0092] Resample the cell temperature data according to a preset time interval, that is, obtain a temperature value every 30S, so as to obtain a temperature sequence temp_seq_list = [T1, T2, …, T n , where T n represents the temperature value at the nth time interval point. Determine the abscissa of the temperature segment curve according to the capacity value sequence, and determine the ordinate of the temperature segment curve according to the temperature sequence.

[0093] S240. Resample the incremental capacity of the battery cell according to a preset time interval to obtain an incremental capacity sequence, and construct an incremental capacity curve according to the incremental capacity sequence and the capacity value sequence.

[0094] In the embodiment of the present invention, first calculate the incremental capacity according to the capacity value and the cell voltage data, and the calculation formula is as follows:

[0095]

[0096] In the above formula, IC represents the incremental capacity, dq represents the capacity difference between the tth moment and the (t - 1)th moment, and dv represents the voltage difference between the tth moment and the (t - 1)th moment.

[0097] Resample the incremental capacity according to a preset time interval, that is, obtain an incremental capacity every 30S, so as to obtain an incremental capacity sequence ica_seq_list = [ICA1, ICA2, …, ICA n , where ICA n represents the incremental capacity value at the nth time interval point. Figure 4It is a schematic diagram of the incremental capacity curve in an embodiment of the present invention. As Figure 4 shown, the abscissa of the incremental capacity curve is determined according to the charging voltage sequence, and the ordinate of the incremental capacity curve is determined according to the incremental capacity sequence.

[0098] S250. Resample the differential voltage of the battery cell according to a preset time interval to obtain a differential voltage sequence, and construct a differential voltage curve according to the differential voltage sequence and the capacity value sequence.

[0099] In an embodiment of the present invention, the differential voltage is calculated according to the capacity value and the cell voltage data, and the calculation formula is as follows:

[0100]

[0101] In the above formula, dva represents the differential voltage, dq represents the capacity difference between the t-th moment and the (t - 1)-th moment, and dv represents the voltage difference between the t-th moment and the (t - 1)-th moment.

[0102] Resample the differential voltage according to a preset time interval, that is, obtain the differential voltage every 30S, so as to obtain a differential voltage sequence dva_seq_list = [dva1, dva2,..., dva n , where dva n represents the differential voltage value at the n-th time interval point. The abscissa of the differential voltage curve is determined according to the differential voltage sequence, and the ordinate of the differential voltage curve is determined according to the incremental capacity sequence.

[0103] S300. Perform random masking on the characteristic curve to obtain a masked characteristic curve.

[0104] In an embodiment of the present invention, random masking is performed on the characteristic curve, and the masking ratio ≥ 70%. Exemplarily, when randomly masking the charging voltage curve, assign a null value to the position of the random id number in the charging voltage sequence volt_seq_list. Figure 5 It is a schematic diagram of the masked characteristic curve of the charging segment curve in an embodiment of the present invention. Figure 6 It is a schematic diagram of the masked characteristic curve of the incremental capacity curve in an embodiment of the present invention. As Figure 5 and Figure 6 shown, after random masking, the charging voltage curve and the incremental capacity curve become sparse masked characteristic curves, that is, Figure 5 and Figure 6 shown by the dashed lines in, only 30% of the random information of a complete charging segment curve is retained.

[0105] Masking the feature curve greatly increases the computational efficiency of model training. After being input into the model for training, the model mines the internal laws of random fragment curves from the sparse visible curves, thereby learning a robust and generalized complete curve, and capturing the strongest temporal correlation features through the encoder, so as to reproduce the entire complete curve, minimizing the final reconstruction error.

[0106] S400. Construct a deep neural network model, train the deep neural network model according to the masked feature curve, and obtain an initial prediction model.

[0107] In the embodiment of the present invention, the deep neural network model is a large model built based on the Transformer model. The Transformer model is mainly used for changes in the visual field. The input is changed from serialized text and word embeddings to curve blocks and curve embeddings, and the positional encoding is also converted into a higher-dimensional spatial encoding. The multi-head self-attention mechanism is used to capture the dependencies between multivariate curves. Therefore, this large model has powerful advantages in capturing long-term temporal features, and the parallel computing architecture greatly improves the computational efficiency. Figure 7 It is a schematic structural diagram of the deep neural network model provided by the embodiment of the present invention, as Figure 7 shown. The deep neural network model includes an input layer, an encoder layer, a decoder layer, a fully connected layer, and an output layer.

[0108] The input of the input layer is the sum of the positional encoding value and the masked feature curve. The encoder layer stacks multiple attention mechanism modules, Transformer blocks. The Transformer block includes a normalization layer, a multi-head attention layer, a residual layer, and a fully connected layer. Its composition structure is normalization layer + multi-head attention layer + residual layer + normalization layer + fully connected layer. The core idea of multi-head attention is to divide the input into multiple heads, each head independently calculates the attention, and then combines the outputs of these heads, so that the model can capture the relationships between curves from different angles. Multi-head attention is calculated by the following formula:

[0109]

[0110] In the above formula, Attention(Q, K, V) represents the attention score, and Q, K, and V represent the weight matrices for linear transformation of the input, which are used to calculate the query Query, the key Key, and the value Value respectively, and d k represents the feature dimension.

[0111] Q, K, and V learn by finding the relationship between the target (curve reconstruction) and the randomly masked input curve, so as to achieve the goal of minimizing the reconstruction error. Q, K, and V are the keys to learning the mutual dependencies of the curve sequence and are the core of the self-attention mechanism.

[0112] The decoder layer stacks two attention mechanism modules, namely Transformer blocks. The fully connected layer includes a first fully connected layer and a second fully connected layer. The first fully connected layer is located in the middle of the encoder layer and the decoder layer, bridging the encoder layer and the decoder layer. The second fully connected layer is located between the decoder layer and the output layer, bridging the decoder layer and the output layer.

[0113] Figure 8 It is the third schematic diagram of the battery safety and life synchronization prediction method flow provided by the embodiments of the present invention. As Figure 8 shown, the model training specifically includes the following steps:

[0114] S410. Perform positional encoding on the masked feature curve to obtain the positional encoding value.

[0115] Since the masked feature curve has a strong temporal logical relationship, it is necessary to perform positional encoding on the input masked feature curve in sequence. The calculation formula for performing positional encoding on the pos-th position (index starting from 0) in the curve and the i-th dimension of the encoding vector is as follows:

[0116]

[0117] In the above formula, PE(pos, i) represents the positional encoding value, pos represents the position index in the sequence, i represents the index of the curve feature dimension, and d mod represents the model feature dimension.

[0118] S420. Obtain the input curve from the masked feature curve and the positional encoding value, and input the input curve into the encoder layer through the input layer.

[0119] In the embodiments of the present invention, an input curve is constructed based on the positional encoding value and the masked feature curve, and the input curve is input into the encoder layer through the input layer.

[0120] S430. Use the encoder layer to extract general temporal features, and then input them into the decoder layer through the first fully connected layer.

[0121] The encoder layer extracts strong and effective general temporal features in the input curve through multiple Transformer blocks, integrates the features output by the encoder through the first fully connected layer, and integrates these local features into global features and enters the decoder layer.

[0122] S440. Use the decoder layer to decode the general temporal features, and obtain the output curve through the second fully connected layer and the output layer.

[0123] The decoder layer decodes the general temporal features in the global features through two Transformer blocks. After the decoded general temporal features are combined and transformed by the second fully connected layer, the output curve is obtained by the output layer.

[0124] S450. The output curve and the input curve are driven by backpropagation through the loss function to obtain the initial prediction model.

[0125] In the embodiment of the present invention, backpropagation is driven through the loss function. The loss function is generated based on the reconstruction error between the input curve and the output curve, and is represented by the mean square error. The expression of the loss function is as follows:

[0126]

[0127] In the above formula, L represents the value of the loss function, x i,j represents the actual value of the input curve at the coordinate (i, j), represents the predicted value of the output curve at the coordinate (i, j), M i,j represents the binary mask matrix, where M(i, j)=1 indicates that the position is occluded, M(i, j)=0 indicates that the position is not occluded, and ∑M i,j is the total number of masked positions, which is used for normalization.

[0128] Figure 9 is a schematic diagram of the convergence process of the loss function during model training provided by the embodiment of the present invention. As Figure 9 shown, this loss function enables the model to focus on learning to infer the occluded part from the temporal curve rather than simply copying the visible part.

[0129] S500. Construct the target mask feature curve of the battery in the target application scenario, and optimize and adjust the initial prediction model according to the target mask feature curve to obtain the target prediction model.

[0130] In the embodiment of the present invention, through the feature sharing characteristic of the encoder of the deep neural network model, the prediction of both SOS and SOH can be completed simultaneously. Since the massive data set of the initial prediction model is trained with data of different battery types and different scenarios, it has the characteristics of a general model. Therefore, the model can be fine-tuned only with a small number of samples in the target application scenario. Figure 10 is the fourth schematic diagram of the battery safety and life synchronization prediction method flow provided by the embodiment of the present invention. As Figure 10 shown, the model fine-tuning specifically includes the following steps:

[0131] S510. Obtain the test charge and discharge data of the battery in the target application scenario, and identify the battery capacity under the charging condition to obtain an identification sequence.

[0132] In an embodiment of the present invention, batteries that need to predict SOS and SOH in a target application scenario are selected to obtain test charge and discharge data. Exemplarily, 280Ah, NCM-lithium ternary batteries in a vehicle scenario are used to obtain test charge and discharge data. At the same time, the battery capacity after each charge of the battery is marked, and the marked data is recorded to obtain a marked sequence. Since battery life prediction is also required later, a long-cycle marked sequence needs to be provided. For example, the charge battery capacity data from 100% SOH to 80% SOH only requires samples of a small number of battery cells and does not require fine-tuning with a large amount of data.

[0133] S520. Construct a target mask feature curve according to the test charge and discharge data, and input the target mask feature curve into the initial prediction model to obtain a first test curve.

[0134] After the test charge and discharge data is processed in the same way as in steps S100 - S300, a target mask feature curve is obtained. The target mask feature curve is input into the encoder layer in the initial prediction model, and the encoder layer is fine-tuned to output a first test curve.

[0135] S530. Perform backpropagation drive according to the reconstruction error between the first test curve and the target mask feature curve to obtain a first target prediction model.

[0136] In an embodiment of the present invention, the backpropagation drive is completed by generating the reconstruction error between the first test curve and the target mask feature curve through a loss function, then importing training model parameters, freezing some parameters, and performing retraining. By finding the minimum reconstruction error, the initial prediction model is trained to the best parameters to obtain a first target prediction model, which is a prediction model for SOS in the target application scenario. The sample size required for this training is small, and the number of adjusted parameters is also small, reducing the cost and training cycle of fine-tuning while improving the training efficiency.

[0137] S540. Input the target mask feature curve and the marked sequence into the initial prediction model to obtain a second test curve.

[0138] Construct a marked sequence curve according to the marked sequence, and jointly input the target mask feature curve and the marked sequence curve into the encoder layer in the initial prediction model, and fine-tune the encoder layer to output a second test curve.

[0139] S550. Perform backpropagation drive according to the reconstruction error between the second test curve and the marked sequence to obtain a second target prediction model.

[0140] In an embodiment of the present invention, the backpropagation drive is completed by generating the reconstruction error between the second test curve and the identification sequence curve through a loss function, and then the training model parameters are imported, some parameters are frozen, and retraining is performed. The regression task is completed through identification, and the initial prediction model with the minimum error is found as the second target prediction model, which is the prediction model of SOH and battery life in the target application scenario.

[0141] S600. Construct the target characteristic curve of the target battery in the target application scenario, and use the target prediction model to predict the target characteristic curve to synchronously obtain the safety status, health degree, and life of the target battery.

[0142] After the model fine-tuning is completed, the first target prediction model can be used to predict the safety status of the target battery, and the second target prediction model can be used to predict the health degree and life of the target battery. Figure 11 It is the fifth schematic diagram of the battery safety and life synchronous prediction method flow provided by the embodiment of the present invention. As Figure 11 shown, the model prediction specifically includes the following steps:

[0143] S610. Obtain the target charge and discharge data of the target battery, and construct the target characteristic curve according to the target charge and discharge data.

[0144] In an embodiment of the present invention, the target battery is an NCM ternary lithium battery. The target charge and discharge data of the target battery is obtained, and after the target charge and discharge data is processed through steps S100 - S200, the target characteristic curve is constructed, and the target characteristic curve includes a target charge segment curve, a target temperature segment curve, a target incremental capacity curve, and a target differential voltage curve.

[0145] S620. Input the target characteristic curve into the first target prediction model to obtain the first target prediction curve.

[0146] In an embodiment of the present invention, the target charge segment curve, the target temperature segment curve, the target incremental capacity curve, and the target differential voltage curve are input into the first target prediction model for reconstruction, and the output first target prediction curve is a voltage-capacity curve.

[0147] S630. Calculate the reconstruction error between the first target prediction curve and the target characteristic curve.

[0148] In an embodiment of the present invention, the reconstruction error between the first target prediction curve and the target characteristic curve is generated through a loss function to obtain the loss function values of all target batteries to be tested.

[0149] S640. Construct an error sequence according to the reconstruction error, and perform normalization processing on the error sequence to obtain a standard error sequence.

[0150] Extract and construct the error sequence sos_loss_list = [loss1, loss2, …, loss] monthly according to the loss function value n , and perform normalization processing on the error sequence to obtain the standard error sequence. The calculation formula for normalization processing is as follows:

[0151]

[0152] In the above formula, sos_loss_scale_list represents the standard error sequence, loss_i represents the loss of any target battery to be measured, loss_min represents the minimum loss of the target battery to be measured, and loss_max represents the maximum loss of the target battery to be measured.

[0153] S650. Determine the target batteries in the standard error sequence that are greater than the target threshold as abnormal batteries, and determine the target batteries in the standard error sequence that are less than or equal to the target threshold as normal batteries.

[0154] In the embodiment of the present invention, the target threshold is set to 0.85. Determine the target batteries in the standard error sequence that are greater than 0.85 as abnormal batteries, and determine the target batteries in the standard error sequence that are less than or equal to 0.85 as normal batteries, so as to quickly quantify and complete the prediction and positioning of the battery safety state, with high accuracy and coverage.

[0155] S660. Input the target feature curve into the second target prediction model to obtain the second target prediction curve.

[0156] In the embodiment of the present invention, input the target charging segment curve, target temperature segment curve, target incremental capacity curve, and target differential voltage curve into the second target prediction model for reconstruction, and the output second target prediction curve is also a voltage-capacity curve.

[0157] S670. Determine the capacity integral when reaching the cut-off voltage according to the second target prediction curve, and obtain the maximum available capacity as the battery health.

[0158] In the embodiment of the present invention, according to the capacity integral when the second target prediction curve reaches the cut-off voltage, that is, the last row of the voltage sequence, obtain the maximum available capacity as the battery health, so as to complete the prediction of the battery health.

[0159] S680. Continuously obtain the battery health of a set threshold number of cycles, construct a battery health curve according to the battery health, and input the battery health curve into the second target prediction model to obtain a battery life prediction curve.

[0160] In the embodiment of the present invention, the set threshold is set to 100, that is, continuously obtain the battery health predictions of 100 cycles, and construct a battery health curve according to the battery healthFigure 12 is a schematic diagram of the battery health curve provided by an embodiment of the present invention. As Figure 12 shown, the root mean square error predicted for the full life cycle sample of the target battery is ≤0.38%. By inputting the battery health curve into the second target prediction model, the reduction of the battery life curve can be completed, and the battery life prediction curve can be obtained, thereby completing the prediction of the battery life. Figure 13 is a schematic diagram of the battery life prediction curve provided by an embodiment of the present invention. As Figure 13 shown, the predicted full life cycle life of the target battery has a high curve reduction degree, with a root mean square error of 0.491%, an average absolute error of 0.318%, and a maximum error of 2.243%.

[0161] The battery safety and life synchronous prediction method provided by an embodiment of the present invention uses a large amount of data for self-supervised learning. Only a small number of samples to be measured are used for fine-tuning, and multi-task prediction of the safety state, health state, and life can be completed, greatly reducing the labeling cost and development cycle. Based on the deep neural network model, after randomly masking the battery curve, high-fidelity reduction is achieved to realize self-training. One model can be used for different battery types, application scenarios, and tasks, with high self-adaptability and efficiency. It can quickly match different batteries for synchronous prediction of the safety state and life, while improving the prediction accuracy and coverage, and realizing parallel estimation of the safety state and life.

[0162] Embodiment 2

[0163] Based on the same technical concept as in Method Embodiment 1 above, an embodiment of the present invention provides a battery safety and life synchronous prediction device. Figure 14 is a schematic structural diagram of the battery safety and life synchronous prediction device provided by an embodiment of the present invention. As Figure 14 shown, the battery safety and life synchronous prediction device 200 includes:

[0164] A data acquisition module 210, configured to acquire charge and discharge data of a battery in multiple application scenarios, perform data preprocessing on the charge and discharge data, and obtain a standard data set;

[0165] A curve construction module 220, configured to construct a feature curve according to the standard data set, where the feature curve at least includes a charging segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve;

[0166] A curve masking module 230, configured to randomly mask the feature curve to obtain a masked feature curve;

[0167] A model training module 240, configured to construct a deep neural network model, and train the deep neural network model according to the masked feature curve to obtain an initial prediction model;

[0168] A model adjustment module 250 is configured to construct a target mask feature curve of a battery in a target application scenario, optimize and adjust an initial prediction model according to the target mask feature curve, and obtain a target prediction model.

[0169] A model prediction module 260 is configured to construct a target feature curve of a target battery in a target application scenario, and predict the target feature curve by using the target prediction model, and synchronously obtain the safety state, health degree and life of the target battery.

[0170] The battery safety and life synchronization prediction device provided by the embodiment of the present invention, after deploying the model to the cloud, only uses a small amount of labeled data, can adapt to different battery types, material systems and application scenarios, realizes a model that is universal and matches different batteries, is applicable to multiple tasks, and in a large number of real vehicle scenario verifications, shows high precision and coverage, can quickly adapt to different application scenarios, and greatly shortens the labor cost and development cycle of new battery adaptation.

[0171] It can be understood that the implementation manners in the battery safety and life synchronization prediction method described in the above embodiment 1 are equally applicable to this embodiment and can achieve the same technical effects, so they will not be repeated here.

[0172] Embodiment 3

[0173] Based on the same concept, the embodiment of the present invention further provides an electronic device. Figure 15 It is a schematic structural diagram of an electronic device provided by the embodiment of the present invention. As Figure 15 shown, the electronic device 300 may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340. Among them, the processor 310, the communication interface 320, and the memory 330 complete mutual communication through the communication bus 340. The processor 310 may call logical instructions in the memory 330 to execute the steps of the battery safety and life synchronization prediction method as described in the above embodiments. For example, it includes:

[0174] S100. Obtain charge and discharge data of a battery in multiple application scenarios, perform data preprocessing on the charge and discharge data, and obtain a standard data set.

[0175] S200. Construct a feature curve according to the standard data set, and the feature curve at least includes a charge segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve.

[0176] S300. Perform random masking on the feature curve to obtain a masked feature curve.

[0177] S400. Construct a deep neural network model, train the deep neural network model according to the masked feature curve, and obtain an initial prediction model.

[0178] S500. Construct the target mask feature curve of the battery in the target application scenario, optimize and adjust the initial prediction model according to the target mask feature curve, and obtain the target prediction model;

[0179] S600. Construct the target feature curve of the target battery in the target application scenario, use the target prediction model to predict the target feature curve, and synchronously obtain the safety status, health degree, and life of the target battery.

[0180] Among them, the processor 310 can be a central processing unit (CPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. chips, or a combination of the above various types of chips.

[0181] In addition, when the logical instructions in the above-mentioned memory 330 can be implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. And the aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, etc., all kinds of media that can store program codes.

[0182] The memory 330 may include a program storage area and a data storage area. The program storage area may store an operating system and application programs required for at least one function. The data storage area may store data created by the processor and the like. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely provided with respect to the processor, and these remote memories may be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0183] Embodiment 4

[0184] Based on the same concept, an embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program includes at least one segment of code. The at least one segment of code can be executed by a master device to control the master device to implement the steps of the battery safety and life synchronization prediction method as described in the above embodiments. For example, it includes:

[0185] S100. Obtain the charge and discharge data of the battery in multiple application scenarios, perform data preprocessing on the charge and discharge data, and obtain a standard data set.

[0186] S200. Construct a feature curve according to the standard data set. The feature curve includes at least a charge segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve;

[0187] S300. Perform random masking on the feature curve to obtain a masked feature curve;

[0188] S400. Construct a deep neural network model, train the deep neural network model according to the masked feature curve, and obtain an initial prediction model;

[0189] S500. Construct a target masked feature curve of the battery in the target application scenario, optimize and adjust the initial prediction model according to the target masked feature curve, and obtain a target prediction model;

[0190] S600. Construct a target feature curve of the target battery in the target application scenario, and use the target prediction model to predict the target feature curve to synchronously obtain the safety state, health degree, and life of the target battery.

[0191] Based on the same technical concept, an embodiment of the present invention further provides a computer program, which is used to implement the above method embodiment when executed by a master device.

[0192] In summary, the battery safety and life synchronization prediction method, device, equipment, and storage medium provided by the present invention utilize massive data for self-supervised learning. By only fine-tuning with a small number of samples to be measured, multi-task prediction of the safety state, health state, and life can be completed, significantly reducing the labeling cost and development cycle. Based on the high-fidelity restoration of the battery curve after random masking by the deep neural network model, self-training is achieved. One model can be universal for different battery types, application scenarios, and tasks, with high adaptability and efficiency. It can quickly match different batteries for synchronous prediction of the safety state and life, while improving the prediction accuracy and coverage, and realizing parallel estimation of the safety state and life. After the model is deployed to the cloud, only a small amount of labeled data is required to adapt to different battery types, material systems, and application scenarios, enabling one model to be universal and match different batteries, applicable to multiple tasks. Moreover, in a large number of real vehicle scenario verifications, it shows high accuracy and coverage, can quickly adapt to different application scenarios, and greatly shortens the labor cost and development cycle for new battery adaptation.

[0193] The computer program can be stored in whole or in part on a computer-readable storage medium packaged together with the processor, or can be stored in whole or in part on a memory not packaged together with the processor.

[0194] Based on the same technical concept, an embodiment of the present invention further provides a processor for implementing the above method embodiment. The above processor can be a chip.

[0195] Referring to "embodiments" herein means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0196] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent of the present invention should be subject to the appended claims.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. However, such modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for synchronously predicting battery safety and lifespan, characterized in that, The method includes: Obtaining the charge and discharge data of the battery under multiple application scenarios, performing data preprocessing on the charge and discharge data, and obtaining a standard data set; Constructing a characteristic curve according to the standard data set, where the characteristic curve at least includes a charging segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve; Performing random masking on the characteristic curve to obtain a masked characteristic curve; Constructing a deep neural network model, training the deep neural network model according to the masked characteristic curve, and obtaining an initial prediction model; Constructing a target masked characteristic curve of the battery in the target application scenario, optimizing and adjusting the initial prediction model according to the target masked characteristic curve, and obtaining a target prediction model; Constructing a target characteristic curve of the target battery in the target application scenario, predicting the target characteristic curve by using the target prediction model, and synchronously obtaining the safety state, health degree, and life of the target battery.

2. The battery safety and life synchronization prediction method according to claim 1, characterized in that The constructing a characteristic curve according to the standard data set includes: Resampling the capacity values of the battery cell according to a preset time interval to obtain a capacity value sequence; Resampling the charging voltage of the battery cell according to the preset time interval to obtain a charging voltage sequence, and constructing a charging segment curve according to the charging voltage sequence and the capacity value sequence; Resampling the temperature values of the battery cell according to the preset time interval to obtain a temperature value sequence, and constructing the temperature segment curve according to the temperature value sequence and the capacity value sequence; Resampling the incremental capacity of the battery cell according to the preset time interval to obtain an incremental capacity sequence, and constructing the incremental capacity curve according to the incremental capacity sequence and the capacity value sequence; Resampling the differential voltage of the battery cell according to the preset time interval to obtain a differential voltage sequence, and constructing the differential voltage curve according to the differential voltage sequence and the capacity value sequence.

3. The battery safety and life synchronization prediction method according to claim 1, wherein The deep neural network model includes an input layer, an encoder layer, a decoder layer, a fully connected layer, and an output layer, where: The input layer is connected to the encoder layer, the encoder layer stacks multiple attention mechanism modules, the decoder layer stacks two attention mechanism modules, and the attention mechanism module includes a normalization layer, a multi-head attention layer, a residual layer, and a fully connected layer; The fully connected layer includes a first fully connected layer and a second fully connected layer, the first fully connected layer is connected to the encoder layer and the decoder layer, and the second fully connected layer is connected to the decoder layer and the output layer.

4. The battery safety and life synchronization prediction method according to claim 3, wherein The training the deep neural network model according to the masked characteristic curve to obtain an initial prediction model includes: Performing position encoding on the masked characteristic curve to obtain a position encoding value; Obtaining an input curve from the masked characteristic curve and the position encoding value, and inputting the input curve into the encoder layer through the input layer; Using the encoder layer to extract general time series features, and then inputting them into the decoder layer through the first fully connected layer; Using the decoder layer to decode the general time series features, and obtaining an output curve through the second fully connected layer and the output layer; The initial prediction model is obtained by driving the output curve and the input curve in reverse replay through a loss function.

5. The battery safety and life synchronization prediction method according to claim 1, characterized in that The target prediction model includes a first target prediction model and a second target prediction model. The construction of the target mask feature curve of the battery in the target application scenario and the optimization and adjustment of the initial prediction model according to the target mask feature curve to obtain the target prediction model includes: Obtain the test charge and discharge data of the battery in the target application scenario, and identify the battery capacity under the charging condition to obtain an identification sequence; Construct a target mask feature curve according to the test charge and discharge data, and input the target mask feature curve into the initial prediction model to obtain a first test curve; Drive in reverse propagation according to the reconstruction error between the first test curve and the target mask feature curve to obtain a first target prediction model; Input the target mask feature curve and the identification sequence into the initial prediction model to obtain a second test curve; Drive in reverse propagation according to the reconstruction error between the second test curve and the identification sequence to obtain a second target prediction model.

6. The battery safety and life synchronization prediction method according to claim 5, wherein The construction of the target feature curve of the target battery in the target application scenario, and the use of the target prediction model to predict the target feature curve to simultaneously obtain the safety state, health degree, and life of the target battery includes: Obtain the target charge and discharge data of the target battery, and construct the target feature curve according to the target charge and discharge data; Input the target feature curve into the first target prediction model to obtain a first target prediction curve; Calculate the reconstruction error between the first target prediction curve and the target feature curve; Construct an error sequence according to the reconstruction error, and perform normalization processing on the error sequence to obtain a standard error sequence; Determine the target battery with a standard error sequence greater than the target threshold as an abnormal battery, and determine the target battery with a standard error sequence less than or equal to the target threshold as a normal battery.

7. The battery safety and life synchronization prediction method according to claim 6, characterized in that After constructing the target feature curve according to the target charge and discharge data, it further includes: Input the target feature curve into the second target prediction model to obtain a second target prediction curve; Determine the capacity integral at the cut-off voltage according to the second target prediction curve, and obtain the maximum available capacity as the battery health degree; Continuously obtain the battery health degree for a set threshold number of cycles, construct a battery health curve according to the battery health degree, and input the battery health curve into the second target prediction model to obtain a battery life prediction curve.

8. A battery safety and life synchronization prediction device, characterized in that, The device includes: A data acquisition module, configured to acquire the charge and discharge data of the battery in multiple application scenarios, and perform data preprocessing on the charge and discharge data to obtain a standard data set; A curve construction module, configured to construct a feature curve according to the standard data set, and the feature curve at least includes a charging segment curve, a temperature segment curve, an incremental capacity curve, and a differential voltage curve; A curve masking module, configured to randomly mask the feature curve to obtain a masked feature curve; A model training module, configured to build a deep neural network model, and train the deep neural network model according to the mask feature curve to obtain an initial prediction model; A model adjustment module, configured to build a target mask feature curve of a battery in a target application scenario, and optimize and adjust the initial prediction model according to the target mask feature curve to obtain a target prediction model; A model prediction module, configured to build a target feature curve of a target battery in a target application scenario, and use the target prediction model to predict the target feature curve, and simultaneously obtain the safety state, health degree, and life of the target battery.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the battery safety and life synchronization prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by the processor, it implements the battery safety and life synchronization prediction method according to any one of claims 1 to 7.

Images