Method for monitoring safety of lithium ion battery

Abstract

The invention relates to the field of lithium battery monitoring, and particularly discloses a lithium ion battery safety monitoring method, which dynamically calculates battery internal resistance data by acquiring state parameter data of battery charge state, voltage, charge and discharge current and the like in a preset time window from a battery management system. And then, further introducing an intelligent prediction algorithm based on deep learning, and performing time sequence conjoint analysis on the battery internal resistance data and the battery state parameter data so as to predict the expected surface temperature of the current lithium ion battery in a normal state, thereby obtaining the surface temperature of the lithium ion battery. And dynamically evaluating the safety state of the battery and performing corresponding safety early warning based on the deviation between the expected normal surface temperature data of the battery and the actually monitored surface temperature data. According to the method, the multi-dimensional state characteristics of the battery are comprehensively considered, the battery safety monitoring requirements under different working conditions can be met, the limitation of traditional threshold value monitoring is broken through, and more reliable fault early warning is achieved.

Description

Technical Field

[0001] The present application relates to the field of lithium battery monitoring, and more specifically, to a method for monitoring the safety of lithium-ion batteries. Background Art

[0002] Lithium-ion batteries, the core power source for current energy storage and electric vehicles, have a safety that is directly linked to the reliable operation of equipment and the personal safety of users. However, due to the complex electrochemical reaction mechanisms within batteries and the dynamic changes in external operating conditions, batteries can experience thermal runaway during charging and discharging due to abnormally high internal resistance and localized overheating, which can even lead to fire or explosion in severe cases. Therefore, developing reliable lithium-ion battery safety monitoring methods and achieving early warning is crucial to ensuring the safe operation of battery systems.

[0003] Currently, safety monitoring of lithium-ion batteries primarily relies on the battery management system (BMS) monitoring of basic parameters such as voltage, current, and temperature, triggering alarms when these parameters exceed preset thresholds. However, while monitoring basic parameters such as voltage, current, and temperature can partially reflect the battery's operating status, threshold-triggered alarm mechanisms often struggle to accurately capture potential fault signals within the battery under complex operating conditions and noisy backgrounds, posing a certain risk of false alarms and missed reports. Furthermore, such methods often only respond when the battery is already exhibiting significant abnormalities or even approaching thermal runaway, resulting in delayed warning times and a difficulty in effectively preventing safety incidents.

[0004] Therefore, an optimized monitoring method for lithium-ion battery safety is desired. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. An embodiment of the present application provides a method for monitoring the safety of lithium-ion batteries, which dynamically calculates the battery internal resistance data by obtaining state parameter data such as battery state of charge, voltage, and charge and discharge current within a preset time window from a battery management system. Then, an intelligent prediction algorithm based on deep learning is further introduced, and the expected surface temperature of the current lithium-ion battery under normal conditions is predicted by performing a time series joint analysis of the battery internal resistance data and the above battery state parameter data. Then, based on the deviation between the expected normal surface temperature data of the battery and the actually monitored surface temperature data, the safety status of the battery is dynamically evaluated and corresponding safety warnings are issued. This method can adapt to the battery safety monitoring needs under different working conditions by comprehensively considering the multi-dimensional state characteristics of the battery, breaking through the limitations of traditional threshold monitoring and achieving more reliable fault warnings.

[0006] According to one aspect of the present application, a method for monitoring the safety of a lithium-ion battery is provided, comprising:

[0007] Acquire a time series of battery status parameters within a preset time window from a battery management system, wherein the battery status parameters include the state of charge, voltage, and charge and discharge current of the battery;

[0008] Calculating a time series of a battery internal resistance based on the time series of the battery state parameter;

[0009] Inputting the time series of the battery internal resistance and the time series of the battery state parameter into a battery temperature short-time prediction model to obtain a short-time series of current battery surface temperature prediction values;

[0010] Obtaining a short-time series of current battery surface temperature measurement values, and calculating the difference between the short-time series of current battery surface temperature prediction values ​​and the short-time series of current battery surface temperature prediction values ​​to obtain a temperature prediction residual sequence;

[0011] Based on the temperature prediction residual sequence, it is determined whether to issue a battery safety warning.

[0012] Compared with the prior art, the lithium-ion battery safety monitoring method provided by the present application dynamically calculates the battery internal resistance data by obtaining the battery state of charge, voltage, charge and discharge current and other state parameter data within a preset time window from the battery management system. Then, an intelligent prediction algorithm based on deep learning is further introduced to perform a time series joint analysis of the battery internal resistance data and the above-mentioned battery state parameter data to predict the expected surface temperature of the current lithium-ion battery under normal conditions. Then, based on the deviation between the expected normal surface temperature data of the battery and the actual monitored surface temperature data, the safety status of the battery is dynamically evaluated and corresponding safety warnings are issued. By comprehensively considering the multi-dimensional state characteristics of the battery, this method can adapt to the battery safety monitoring needs under different working conditions, break through the limitations of traditional threshold monitoring, and achieve more reliable fault warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0014] Figure 1 Flowchart of a method for monitoring lithium-ion battery safety according to an embodiment of the present application.

[0015] Figure 2 Schematic diagram of data flow of a method for monitoring lithium-ion battery safety according to an embodiment of the present application.

[0016] Figure 3Flowchart of sub-step S2 of the method for monitoring lithium-ion battery safety according to an embodiment of the present application.

[0017] Figure 4 Flowchart of sub-step S3 of the method for monitoring lithium-ion battery safety according to an embodiment of the present application.

[0018] Figure 5 Flowchart of sub-step S32 of the method for monitoring lithium-ion battery safety according to an embodiment of the present application.

[0019] Figure 6 This is a flowchart of sub-step S5 of the method for monitoring lithium-ion battery safety according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0021] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0022] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0024] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0025] In response to the technical problems described in the above background technology, this application proposes a method for monitoring the safety of lithium-ion batteries, which dynamically calculates the battery internal resistance data by obtaining the battery state of charge, voltage, charge and discharge current and other state parameter data within a preset time window from the battery management system. Then, an intelligent prediction algorithm based on deep learning is further introduced, and the expected surface temperature of the current lithium-ion battery under normal conditions is predicted by performing a time series joint analysis of the battery internal resistance data and the above battery state parameter data. Then, based on the deviation between the expected normal surface temperature data of the battery and the actual monitored surface temperature data, the safety status of the battery is dynamically evaluated and corresponding safety warnings are issued. By comprehensively considering the multi-dimensional state characteristics of the battery, this method can adapt to the battery safety monitoring needs under different working conditions, break through the limitations of traditional threshold monitoring, and achieve more reliable fault warnings.

[0026] Figure 1 Flowchart of a method for monitoring lithium-ion battery safety according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for monitoring the safety of lithium-ion batteries according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the method for monitoring the safety of a lithium-ion battery includes the following steps: S1, obtaining a time series of battery status parameters within a preset time window from a battery management system, wherein the battery status parameters include the battery state of charge, voltage, and charge and discharge current; S2, calculating a time series of battery internal resistance based on the time series of the battery status parameters; S3, inputting the time series of the battery internal resistance and the time series of the battery status parameters into a battery temperature short-time prediction model to obtain a short-time series of current battery surface temperature prediction values; S4, obtaining a short-time series of current battery surface temperature measurement values, and calculating the difference between the short-time series and the short-time series of current battery surface temperature prediction values ​​to obtain a temperature prediction residual sequence; S5, determining whether to issue a battery safety warning based on the temperature prediction residual sequence.

[0027] In the above-mentioned method for monitoring the safety of lithium-ion batteries, the step S1 obtains the time series of battery status parameters within a preset time window from the battery management system, and the battery status parameters include the battery's state of charge, voltage, and charge and discharge current. It should be understood that the operating state of the battery is dynamically changing, and its internal electrochemical reaction and thermal behavior are closely related to the current state of charge (SOC), terminal voltage, and charge and discharge current. The three are basic data that characterize the battery's external characteristics and workload. Therefore, the present application collects the battery's state of charge, voltage, and charge and discharge current in real time from the battery management system (BMS) to achieve a comprehensive record of the battery's operating conditions. In the specific implementation process, the BMS continuously collects the real-time voltage, current and estimated SOC value of the lithium-ion battery at a preset sampling frequency (for example, 1Hz) through its internally integrated sensors (such as Hall sensors to measure current, voltage divider circuits with ADCs to measure voltage) and estimation algorithms (such as the Ampere-hour integration method), arranges the collected data points in chronological order to form their own time series, and intercepts the data within the most recent preset time window (for example, the past 5 minutes or 100 sampling points), thereby obtaining high-density, multi-dimensional time series data that can reflect the recent operating history and current status of the battery, providing a necessary and rich original information basis for subsequent battery status analysis, while avoiding the false alarm risk of traditional single-point parameter threshold judgment.

[0028] In practice, the battery current is measured using a Hall-effect sensor. This sensor, renowned for its high precision and excellent linearity, provides reliable current readings without affecting circuit operation. Meanwhile, a voltage divider circuit combined with an analog-to-digital converter (ADC) is used to measure the battery terminal voltage. The voltage divider circuit design must account for the wide battery voltage range to ensure a stable output signal under all operating conditions. The ADC is selected based on its resolution and sampling rate requirements to capture subtle voltage fluctuations. The state of charge (SOC) is typically estimated using the ampere-hour integration method. This method calculates the current SOC by accumulating or subtracting the charge and discharge current over time. While this method is simple and straightforward, it often requires additional correction methods, such as the open-circuit voltage method, to improve accuracy, especially after extended use or rapid charge / discharge cycles.

[0029] To ensure the reliability and accuracy of the collected data, a reasonable sampling frequency must be set. For example, choosing a sampling frequency of 1Hz means that data is collected every second, including the battery's state of charge, voltage, and charge and discharge current. This sampling frequency meets the needs of most application scenarios while minimizing the amount of data generated, which would burden storage and processing. The collected data points are arranged in chronological order to form their own time series. These time series not only contain information about the temporal changes of individual parameters but also reveal the relationships between different parameters, which is crucial for a deeper understanding of battery operating mechanisms.

[0030] Next, to facilitate subsequent analysis and processing, data from the most recent preset time window is captured. This preset time window can be flexibly adjusted based on specific monitoring needs; for example, it can be set to include data from the past 5 minutes or the most recent 100 sampling points. This method yields a dataset that reflects both the battery's recent operating history and its current state. This dataset consists of a high-density, multi-dimensional time series, with each sampling point containing information about the battery's state of charge, voltage, and charge and discharge current. This data structure not only provides a rich foundation of raw information but also helps avoid the risk of false alarms associated with traditional reliance on a single parameter threshold. Under complex operating conditions, relying solely on the threshold of a single parameter to assess battery status may overlook potential issues. However, comprehensively considering the changing trends of multiple parameters can more accurately capture potential abnormalities within the battery.

[0031] Furthermore, during implementation, necessary preprocessing of the collected data is necessary to further enhance data validity and usability. For example, removing noise interference is crucial, as sensors are inevitably affected by environmental factors during operation, which can introduce a certain amount of noise into the data. Appropriate filtering techniques, such as low-pass filters, can effectively smooth the data curve and reduce the effect of random noise on the true signal. Furthermore, strategies must be developed to address potential data loss or outliers. This may involve using interpolation algorithms to fill in missing data points or using statistical principles to identify and remove outliers that significantly deviate from the normal range. Ultimately, these meticulous operations yield high-quality time series data of battery state parameters suitable for in-depth analysis.

[0032] In the above-mentioned method for monitoring the safety of lithium-ion batteries, the step S2 calculates the time series of the battery internal resistance based on the time series of the battery status parameters. It should be understood that the battery internal resistance is a sensitive physical quantity that reflects the battery's health status, aging degree, and early fault phenomena such as micro-short circuits and lithium precipitation that may occur internally. Its changes are often earlier than significant abnormalities in voltage and current. Therefore, in order to obtain physical quantities that can better characterize subtle changes in the battery than macroscopic electrical parameters, this application is based on the principle of the battery's equivalent circuit model and Ohm's law, and indirectly estimates the equivalent internal resistance of the battery by utilizing the acquired battery status parameters. Among them, Figure 3 FIG. 1 is a flow chart of sub-step S2 of the method for monitoring the safety of a lithium-ion battery according to an embodiment of the present application. Figure 3 As shown, the step S2 includes the following steps: S21, based on the state of charge of the battery, looking up the table to obtain the battery open circuit voltage; S22, calculating the absolute difference between the battery open circuit voltage and the battery voltage, and then dividing the absolute difference by the charge and discharge current of the battery to obtain the battery internal resistance.

[0033] Specifically, step S21 involves looking up a table to obtain the battery open circuit voltage based on the battery state of charge (SOC). Specifically, for each sampling point in the time series of the battery state parameters, the corresponding battery open circuit voltage value is first looked up in a pre-calibrated SOC-open circuit voltage (OCV) characteristic curve stored in the BMS memory based on the battery state of charge (SOC) value at that point.

[0034] Specifically, step S22 calculates the absolute difference between the battery open-circuit voltage and the battery voltage, and then divides the absolute difference by the battery's charge and discharge current to obtain the battery internal resistance. That is, the absolute difference between the open-circuit voltage obtained from the table lookup and the battery terminal voltage actually measured by the BMS is calculated. This absolute difference represents the voltage drop generated when current flows through the battery internal resistance. By dividing this absolute difference by the battery charge and discharge current at the corresponding sampling point (to avoid division by zero or small current disturbances, a current threshold is usually set, and calculation is performed only when the absolute value of the current is greater than the threshold), the estimated battery internal resistance at that moment is obtained. This process is repeated for all sampling points within a preset time window to obtain a time series of the battery internal resistance, which more directly reflects the safety-related electrochemical changes that may occur within the battery and provides key indicative information for subsequent predictions of battery heat generation behavior.

[0035] In the above-mentioned method for monitoring the safety of lithium-ion batteries, in step S3, the time series of the battery internal resistance and the time series of the battery state parameters are input into the battery temperature short-time prediction model to obtain a short-time series of the current battery surface temperature prediction value. Specifically, since the heat generation behavior of the lithium-ion battery is closely related to its working state (characterized by SOC, voltage, and current) and internal characteristics (characterized by internal resistance), that is, the temperature response of a healthy battery under specific operating conditions and internal resistance levels has a certain degree of predictability. Therefore, in order to establish a normal surface temperature benchmark that the battery should have under the current operating conditions and healthy state, this application is based on machine learning technology, and learns the complex nonlinear mapping relationship between battery state parameters, internal resistance and surface temperature by constructing a time series prediction model, thereby predicting the normal surface temperature that the battery should have under the current working conditions. Among them, Figure 4 FIG. 1 is a flow chart of sub-step S3 of the method for monitoring the safety of a lithium-ion battery according to an embodiment of the present application. Figure 4 As shown, the step S3 includes the following steps: S31, performing time series modeling analysis on the time series of the battery internal resistance and the time series of the battery state parameters to obtain a time series of the battery multi-dimensional state feature coding vector; S32, performing time series context transfer coding on the time series of the battery multi-dimensional state feature coding vector to obtain a battery multi-dimensional state feature dynamic propagation coding vector; S33, inputting the battery multi-dimensional state feature dynamic propagation coding vector into a temperature prediction decoder based on a fully connected layer to obtain a short-time series of the current battery surface temperature prediction value.

[0036] Specifically, in a specific example of the present application, step S31 is to time-align and combine the time series of the battery internal resistance and the time series of the battery state parameter, and then input them into a time series encoder based on the LSTM model to obtain a time series of the battery multi-dimensional state feature encoding vector. It should be understood that the temperature change of the battery is affected by multiple factors, and a single parameter is difficult to fully characterize its thermal behavior. The battery's state of charge (SOC), voltage, charge and discharge current, and internal resistance are all key factors affecting heat generation and heat dissipation, and have time series dependencies with each other. Therefore, in order to effectively extract the deep-level time series feature representation of the current comprehensive operating state of the battery and lay the foundation for subsequent temperature prediction, the present application is based on the powerful time series modeling capability of the long short-term memory network (LSTM) to perform time series joint encoding of the battery internal resistance, state of charge, voltage, and charge and discharge current, so as to capture the long-term time series dependency between the battery state parameters through its internal memory unit and forget gate mechanism, thereby generating a time series of the battery multi-dimensional state feature encoding vector that can fully reflect the battery operating conditions. Specifically, first, the time series of the battery internal resistance and the time series of the battery state parameters are strictly aligned in terms of timestamps, that is, at each time step, the four parameter values ​​of state of charge, voltage, charge and discharge current, and internal resistance are combined into a multidimensional input vector. Then, the time series composed of the multidimensional input vectors of each time step is input into a time series encoder based on the LSTM model. The LSTM encoder contains multiple LSTM layers, which can selectively memorize and forget historical information through its unique gating mechanism (input gate, forget gate, output gate). At each time step, based on the hidden state of the previous time step and the input vector of the current time step, its internal state is updated, and the hidden state of the current time step is output, thereby effectively capturing the long-term dependencies and nonlinear dynamics in the input sequence, and obtaining a time series of battery multidimensional state feature encoding vectors composed of the hidden states of each time step, providing high-quality intermediate representation for subsequent more refined temperature predictions.

[0037] Specifically, the step S32 performs temporal context transfer encoding on the time series of the battery multi-dimensional state feature encoding vector to obtain a battery multi-dimensional state feature dynamic propagation encoding vector. Specifically, the present application takes into account that the battery operating states at adjacent time points often contain stronger temporal correlations, and although the LSTM model can effectively capture the long-term temporal dependencies of the battery operating states, it has the potential problem of insufficient capture of the interaction of local neighborhood detail features. Therefore, in order to further enhance the ability to understand the temporal context association of battery operating conditions, the present application proposes a temporal context transfer encoding mechanism, which constructs a local neighborhood perception window around the battery multi-dimensional state feature encoding vector of each time step, so as to utilize the temporal context information within the window to enhance the feature expression capability of the battery state at the current time step, and then realizes the dynamic propagation modeling of the battery state characteristics in the time dimension through global temporal context information aggregation, and obtains the battery multi-dimensional state feature dynamic propagation encoding vector. Among them, Figure 5 FIG. 1 is a flow chart of sub-step S32 of the method for monitoring the safety of a lithium-ion battery according to an embodiment of the present application. Figure 5 As shown, the step S32 includes the steps of: S321, determining the window size of the local semantic enhancement perception window of each battery multidimensional state feature coding vector based on the feature distribution of each battery multidimensional state feature coding vector in the time series of the battery multidimensional state feature coding vector; S322, based on all the battery multidimensional state feature coding vectors in the local semantic enhancement perception window of each battery multidimensional state feature coding vector, performing semantic enhancement perception on each battery multidimensional state feature coding vector to obtain a time series of the local semantic enhancement coding vector of the battery multidimensional state feature; S323, performing global time domain transfer coding on the time series of the local semantic enhancement coding vector of the battery multidimensional state feature to obtain the battery multidimensional state feature dynamic propagation coding vector.

[0038] More specifically, step S321 is expressed as follows:

[0039]

[0040] Where exp(·) represents the exponential function operation with e as the base, h i and h j They represent the i-th and j-th battery multi-dimensional state feature encoding vectors in the time series of the battery multi-dimensional state feature encoding vector, ||·|| represents the calculation of the Euclidean norm, τ represents the temperature coefficient, and s ij Indicates h i and h j The feature correlation between is the preset neighborhood, ∈ represents the first smoothing term, γ represents the second smoothing term, k represents the index of the vector, Represents a preset neighborhood The j-th battery multidimensional state feature encoding vector relative to h i The correlation weight coefficient, log2(·) represents the logarithmic function with base 2, ε i Represents vector h i The neighbor feature distribution entropy, w max Indicates the preset maximum window size, w i Indicates h i The window size of the local semantic enhancement perception window corresponding to the battery multi-dimensional state feature encoding vector.

[0041] That is, based on the feature distribution correlation between the battery multi-dimensional state feature encoding vectors, the size of the local semantic enhancement perception window corresponding to each vector is dynamically defined, so that the window can go beyond the fixed structural proximity and accurately cover the most relevant local feature environment at the semantic level, thereby achieving a refined semantic understanding of the local context of the battery operating state in time series modeling. Through this dynamic window setting, the model can adaptively capture the semantically relevant local feature interactions between battery state parameters under different working conditions, avoid the omission of key semantic information or the redundant inclusion of irrelevant information in the fixed window, enhance the semantic accuracy and contextual relevance of the local feature representation, and thus improve the quality of the battery multi-dimensional state feature encoding.

[0042] More specifically, in a specific example of the present application, step S322 includes: inputting all battery multi-dimensional state feature encoding vectors in the local semantic enhancement perception window of the battery multi-dimensional state feature encoding vector into a local neighborhood context enhancement fusion network based on the attention mechanism to obtain the local semantic enhancement encoding vector of the battery multi-dimensional state feature, which is expressed as follows:

[0043]

[0044]

[0045] Among them, softmax(·) represents the normalized exponential function, W q 、W k and W g represents the different weight parameter matrices in the local neighborhood context enhancement fusion network, (·) T represents the transpose of the vector, and d represents h i The characteristic dimension, α ij Indicates h j The corresponding attention weight factor, Sigmoid(·) represents the sigmoid activation function, Indicates h i The corresponding local semantic enhanced encoding vector of the battery multi-dimensional state feature.

[0046] That is, the local neighborhood context enhancement fusion network based on the attention mechanism is used to enhance the fusion of the battery multi-dimensional state feature encoding vectors within the local semantic enhancement perception window. The weights of each feature vector are dynamically allocated through the attention mechanism, focusing on key features with close semantic associations, and suppressing irrelevant or noise information, thereby converting the battery multi-dimensional state feature encoding vector into a battery multi-dimensional state feature local semantic enhancement encoding vector containing richer local semantic information. This not only retains the physical meaning of the battery multi-dimensional state characteristics, but also enhances the capture of key semantic information in the battery operating status through the attention mechanism, and improves the representation accuracy of the feature sequence for local state changes related to potential abnormal heat generation of the battery, thereby providing more accurate feature input for the temperature prediction model and providing a more representational feature basis for subsequent global feature propagation and temperature prediction.

[0047] In particular, here, since the temperature coefficient τ and the first smoothing term ∈ and the second smoothing term γ are introduced in the process of determining the window size of the local semantic enhancement perception window, the weight parameter matrix W in the local neighborhood enhancement fusion is g 、W q and W k For example, it is preferred to simultaneously consider the deformation quantitative representation to enhance the contextual semantic embedding under the predetermined local semantic enhancement perception window. Based on this, in a preferred example of the present application, the step S322 includes: performing dynamic weight adaptation optimization based on the local neighborhood perception field on the weight parameter matrix of the local neighborhood context enhancement fusion network based on the attention mechanism to obtain an updated weight parameter matrix; then, based on the updated weight parameter matrix, performing context enhancement fusion based on the attention mechanism on all battery multidimensional state feature coding vectors in the local semantic enhancement perception window of the battery multidimensional state feature coding vector to obtain the local semantic enhancement coding vector of the battery multidimensional state feature.

[0048] Specifically, first, the strain correlation vector (τ,∈,γ) is linearly interpolated to make its length consistent with the battery multi-dimensional state feature encoding vector, and then the nonlinear curvature mapping is completed through one-dimensional convolution to obtain the strain quantization vector V α , and the weight parameter matrix W g 、W q and W k As a mapping reference, the curvature strain correlation vector is obtained:

[0049] V α1 =V α W g +V α

[0050] V α2 =V αW q +V α

[0051] V α3 =V α W k +V α

[0052] Among them, V α Represents the strain quantization vector, V α1 、V α2 and V α3 Represents the weight parameter matrix W g 、W q and W k The corresponding curvature strain correlation vector.

[0053] Then, considering the direct correlation with the curvature deformation, the main curvature evolution control factor of the weighted deformation response can be simplified, that is, the exogenous high-order nonlinear node perturbation can be ignored. Therefore, the above curvature strain correlation vector is respectively combined with the eigenvector of the weight parameter matrix, that is, V in the following formula eg 、V eq and V ek Associate to modify the weight parameter matrix, expressed as:

[0054] W' g =(V α1 T V eg )⊙W g

[0055] W' q =(V α2 T V eq )⊙W q

[0056] W' k =(V α3 T V ek )⊙W k

[0057] Among them, V eg 、V eq and V ek Respectively represent W g 、W q and W k The eigenvector of ⊙ represents the dot multiplication operation, W' g 、W' q and W' k Respectively represent W g 、W q and W k The corresponding updated weight parameter matrix.

[0058] In this way, the local semantic embedding enhancement based on the weight parameter matrix can take into account the introduction of the deformation field coupling component in the window size determination of the local semantic enhancement perception window, thereby improving the contextual semantic embedding expression effect of the sequence of local semantic enhancement coding vectors of the battery multi-dimensional state features based on the local semantic enhancement perception window.

[0059] More specifically, in a specific example of the present application, step S323 includes: inputting the time series of the local semantic enhancement coding vector of the battery multi-dimensional state feature into the message global transmission coding network based on the Transformer architecture to obtain the dynamic propagation coding vector of the battery multi-dimensional state feature, which is expressed as follows:

[0060]

[0061] in, Represents the time series of local semantic enhanced encoding vectors of multi-dimensional battery state features, and Respectively The first, second and nth battery multi-dimensional state feature local semantic enhancement encoding vectors in Transformer(·) represents the Transformer architecture, v f Represents the dynamic propagation encoding vector of the multi-dimensional state characteristics of the battery.

[0062] That is, by utilizing the self-attention mechanism of the Transformer architecture, the time series of the local semantic enhancement encoding vector of the battery multi-dimensional state characteristics after local semantic enhancement is subjected to global temporal dependency modeling and information interaction integration, so that the features of each time point can dynamically integrate the long-distance correlation information in the entire sequence, breaking through the semantic limitations of the local semantic enhancement perception window, thereby generating a dynamic propagation encoding vector of the battery multi-dimensional state characteristics that combines local details and global temporal propagation laws, which can not only retain the semantic refinement of the local features at each time point, but also integrate the abnormal pattern propagation laws in the global temporal series, providing a more comprehensive state representation basis for temperature prediction.

[0063] Specifically, in step S33, the battery multi-dimensional state feature dynamic propagation coding vector is input into a temperature prediction decoder based on a fully connected layer to obtain a short-time series of the current battery surface temperature prediction value. That is, in order to convert the battery multi-dimensional state feature dynamic propagation coding vector representing the historical evolution characteristics of the battery internal state into a specific numerical prediction of the battery surface temperature within the current short-term time window, the present application uses multiple fully connected layers (also known as densely connected layers or MLPs) as temperature prediction decoders to map the abstract features back to the target space. Specifically, the temperature prediction decoder is trained based on a large amount of historical operating data and temperature labels of healthy lithium-ion batteries to learn the mapping relationship between the operating state and surface temperature of healthy lithium-ion batteries. After training is completed, the fully connected layer uses the weight parameters and bias parameters learned during training to perform a linear transformation on the input dynamic propagation coding vector of the battery's multi-dimensional state features, and introduces nonlinearity through the activation function (such as the ReLU function). After multiple layers of stacking and iterative processing, the high-dimensional abstract features are gradually mapped to low-dimensional representations that are closer to the actual temperature values. Finally, through the k neurons in the output layer, the battery surface temperature prediction values ​​at k time points in the current short-term time window are output, forming a short-term series of temperature prediction values. The short-term series of temperature prediction values ​​represents the change trend and value of the surface temperature of the battery when it is in a healthy state under the current operating conditions, and is an important basis for the formulation of battery thermal management strategies.

[0064] In the above-mentioned method for monitoring the safety of lithium-ion batteries, the step S4 obtains a short-time series of the current battery surface temperature measurement value, and calculates the difference between the short-time series and the current battery surface temperature prediction value to obtain a temperature prediction residual sequence. Specifically, if an early fault occurs inside the battery (such as a micro-short circuit, lithium precipitation, etc.), it is often accompanied by abnormal heat generation, causing the actual temperature to deviate from its level in a healthy state. Therefore, the present application further obtains the actual measurement value of the battery surface temperature at the time point corresponding to the short-time series of the current battery surface temperature prediction value, and performs a point-by-point residual analysis on the two to obtain a temperature prediction residual sequence to quantify the degree of deviation of the actual thermal state of the battery from the ideal healthy thermal state.

[0065] In the above-mentioned method for monitoring the safety of lithium-ion batteries, the step S5 determines whether to issue a battery safety warning based on the temperature prediction residual sequence. It should be understood that the temperature prediction residual sequence effectively filters out the expected temperature fluctuations caused by changes in normal charging and discharging conditions, making the additional temperature rise (i.e., abnormal heat generation) signal caused by unknown or early faults more prominent, providing a more sensitive indicator for fault identification. Based on this, in order to accurately determine whether the battery is in a potential safety risk, the present application is based on the principle of statistical analysis and formulates a warning strategy by analyzing the changing trend and cumulative effect of the temperature prediction residual sequence. Among them, Figure 6 FIG. 5 is a flow chart of sub-step S5 of the method for monitoring the safety of a lithium-ion battery according to an embodiment of the present application. Figure 6 As shown, the step S5 includes the following steps: S51, calculating the residual change rate sequence and the residual cumulative sum based on the temperature prediction residual sequence; S52, in response to any residual change rate in the residual change rate sequence exceeding a preset change rate threshold, or the residual cumulative sum exceeding a preset cumulative sum threshold, issuing a battery safety warning.

[0066] Specifically, step S51 calculates the residual change rate sequence and the residual cumulative sum based on the temperature prediction residual sequence. Specifically, based on the temperature prediction residual sequence, the present application sets two derivative indicators: one is the residual change rate, which is to calculate the ratio of the difference between the residual values ​​of adjacent time points in the temperature prediction residual sequence to the residual value of the previous time point, which is used to capture the speed at which the actual battery temperature deviates from the normal level; the other is the residual cumulative sum, which is to accumulate all the residual values ​​in the temperature prediction residual sequence within the above-mentioned short-term window to measure the cumulative degree of deviation of the actual battery temperature from the normal level.

[0067] Specifically, in step S52, in response to any residual change rate in the residual change rate sequence exceeding a preset change rate threshold, or the residual cumulative sum exceeding a preset cumulative sum threshold, a battery safety warning is performed. Specifically, the present application compares each value in the obtained residual change rate sequence with a preset change rate threshold, and at the same time compares the residual cumulative sum with a preset cumulative sum threshold. If it is monitored that any residual change rate in the residual change rate sequence exceeds its corresponding preset threshold (indicating that the abnormal temperature rise rate is too fast), or when the calculated residual cumulative sum exceeds its preset threshold (indicating that there is a continuous, cumulative abnormal temperature rise), the system determines that there is a safety risk in the battery and triggers a battery safety warning mechanism, such as lighting an alarm light, issuing a sound prompt, limiting the charge and discharge power, or even cutting off the circuit. In this way, early, continuous abnormal heat generation behavior that indicates the risk of thermal runaway can be effectively identified, thereby improving the accuracy and timeliness of the warning.

[0068] In summary, a method for monitoring the safety of lithium-ion batteries based on an embodiment of the present application is illustrated, which dynamically calculates the battery internal resistance data by obtaining state parameter data such as the battery state of charge, voltage, and charge and discharge current within a preset time window from the battery management system. Next, an intelligent prediction algorithm based on deep learning is further introduced, and the expected surface temperature of the current lithium-ion battery under normal conditions is predicted by performing a time series joint analysis of the battery internal resistance data and the above-mentioned battery state parameter data. Then, based on the deviation between the expected normal surface temperature data of the battery and the actual monitored surface temperature data, the safety status of the battery is dynamically evaluated and corresponding safety warnings are issued. This method can adapt to the battery safety monitoring needs under different working conditions by comprehensively considering the multi-dimensional state characteristics of the battery, breaking through the limitations of traditional threshold monitoring and achieving more reliable fault warnings.

[0069] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details of the above embodiments are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0070] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0071] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be encompassed therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0072] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0073] Finally, it should be noted that the above description has been provided for purposes of illustration and description. Furthermore, the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will appreciate that the technical solutions of the present invention may be modified or replaced with equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for monitoring the safety of a lithium-ion battery, characterized in that: include: Acquire a time series of battery status parameters within a preset time window from a battery management system, wherein the battery status parameters include the battery state of charge, voltage, and charge and discharge current; Calculating a time series of a battery internal resistance based on the time series of the battery state parameter; Inputting the time series of the battery internal resistance and the time series of the battery state parameter into a battery temperature short-time prediction model to obtain a short-time series of current battery surface temperature prediction values; Obtaining a short-time series of current battery surface temperature measurement values, and calculating the difference between the short-time series of current battery surface temperature prediction values ​​and the short-time series of the current battery surface temperature prediction values ​​to obtain a temperature prediction residual sequence; Based on the temperature prediction residual sequence, it is determined whether to issue a battery safety warning.

2. The method for monitoring the safety of a lithium-ion battery according to claim 1, wherein: Calculating a time series of a battery internal resistance based on a time series of the battery state parameter includes: Obtaining the battery open circuit voltage by looking up a table based on the state of charge of the battery; The absolute difference between the battery open circuit voltage and the battery voltage is calculated, and then the absolute difference is divided by the charge and discharge current of the battery to obtain the battery internal resistance.

3. The method for monitoring the safety of a lithium-ion battery according to claim 1, wherein: Determining whether to issue a battery safety warning based on the temperature prediction residual sequence includes: Based on the temperature prediction residual sequence, calculating the residual change rate sequence and the residual cumulative sum; In response to any residual change rate in the residual change rate sequence exceeding a preset change rate threshold, or the residual cumulative sum exceeding a preset cumulative sum threshold, a battery safety warning is issued.

4. The method for monitoring the safety of a lithium-ion battery according to claim 1, wherein: Inputting the time series of the battery internal resistance and the time series of the battery state parameter into a battery temperature short-time prediction model to obtain a short-time series of current battery surface temperature prediction values, including: Performing time series modeling analysis on the time series of the battery internal resistance and the time series of the battery state parameter to obtain a time series of a battery multi-dimensional state feature encoding vector; Performing temporal context transfer encoding on the time series of the battery multi-dimensional state feature encoding vector to obtain a battery multi-dimensional state feature dynamic propagation encoding vector; The battery multi-dimensional state feature dynamic propagation coding vector is input into a temperature prediction decoder based on a fully connected layer to obtain a short-time sequence of the current battery surface temperature prediction value.

5. The method for monitoring the safety of a lithium-ion battery according to claim 4, wherein: Performing time series modeling analysis on the time series of the battery internal resistance and the time series of the battery state parameter to obtain a time series of a battery multi-dimensional state feature encoding vector, including: The time series of the battery internal resistance and the time series of the battery state parameter are time-aligned and combined, and then input into a time series encoder based on an LSTM model to obtain a time series of the battery multi-dimensional state feature encoding vector.

6. The method for monitoring the safety of a lithium-ion battery according to claim 5, wherein: Performing temporal context transfer encoding on the time series of the battery multi-dimensional state feature encoding vector to obtain a battery multi-dimensional state feature dynamic propagation encoding vector, including: Determining a window size of a local semantic enhancement perception window of each battery multidimensional state feature coding vector based on a feature distribution of each battery multidimensional state feature coding vector in a time series of the battery multidimensional state feature coding vector; Based on all battery multidimensional state feature coding vectors in the local semantic enhancement perception window of each battery multidimensional state feature coding vector, performing semantic enhancement perception on each battery multidimensional state feature coding vector to obtain a time series of local semantic enhancement coding vectors of battery multidimensional state features; Global time-domain transfer coding is performed on the time series of the local semantic enhancement coding vector of the battery multi-dimensional state feature to obtain the dynamic propagation coding vector of the battery multi-dimensional state feature.

7. The method for monitoring the safety of a lithium-ion battery according to claim 6, wherein: Based on all battery multidimensional state feature coding vectors in the local semantic enhancement perception window of each battery multidimensional state feature coding vector, semantic enhancement perception is performed on each battery multidimensional state feature coding vector to obtain a time series of local semantic enhancement coding vectors of battery multidimensional state features, including: All battery multidimensional state feature encoding vectors in the local semantic enhancement perception window of the battery multidimensional state feature encoding vector are input into the local neighborhood context enhancement fusion network based on the attention mechanism to obtain the local semantic enhancement encoding vector of the battery multidimensional state feature.

8. The method for monitoring the safety of a lithium-ion battery according to claim 7, wherein: Inputting all battery multidimensional state feature encoding vectors in the local semantic enhancement perception window of the battery multidimensional state feature encoding vector into a local neighborhood context enhancement fusion network based on an attention mechanism to obtain the local semantic enhancement encoding vector of the battery multidimensional state feature, including: Performing dynamic weight adaptation optimization based on the local neighborhood perception field on the weight parameter matrix of the local neighborhood context enhancement fusion network based on the attention mechanism to obtain an updated weight parameter matrix; Based on the updated weight parameter matrix, all battery multidimensional state feature coding vectors in the local semantic enhancement perception window of the battery multidimensional state feature coding vector are subjected to context enhancement fusion based on the attention mechanism to obtain the battery multidimensional state feature local semantic enhancement coding vector.

9. The method for monitoring the safety of a lithium-ion battery according to claim 8, wherein: Performing global time-domain transfer coding on the time series of the local semantic enhancement coding vector of the battery multi-dimensional state feature to obtain the dynamic propagation coding vector of the battery multi-dimensional state feature, including: The time series of the local semantic enhancement coding vector of the battery multi-dimensional state feature is input into the message global domain transmission coding network based on the Transformer architecture to obtain the dynamic propagation coding vector of the battery multi-dimensional state feature.

Images