Fault detection method of lithium ion battery

Abstract

The invention relates to the field of lithium battery detection, and particularly discloses a lithium ion battery fault detection method, which comprises the following steps of: firstly, acquiring total current data of a battery pack in real time, and judging the current running state of a battery system according to the total current of the battery pack and the duration time; and when the battery system is in a standing state, calculating the voltage drop rate and the temperature rise rate of each single battery in a preset time window, and comparing the voltage drop rate and the temperature rise rate with corresponding thresholds to realize self-discharge detection and fault identification. And when the battery system is in a charging or discharging state, a deep learning algorithm is further introduced, temperature-pressure coupling time sequence fluctuation characteristics of each single battery are extracted, and temperature-pressure coupling time sequence fluctuation state consistency analysis is performed on each single battery, so that intelligent detection of faults of the single batteries in the charging and discharging process is realized. According to the method, the abnormal temperature and pressure fluctuation of the single battery in the charging and discharging process can be effectively captured, so that early warning and accurate positioning of a fault are realized.

Description

Technical Field

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

[0002] As a core component in the current energy storage field, lithium-ion batteries, with their high energy density and long cycle life, are widely used in new energy vehicles, energy storage power stations, portable electronic devices and other fields. However, due to factors such as complex internal chemical reactions and a volatile external environment, lithium-ion batteries are prone to failures such as overcharge, overdischarge, thermal runaway, and abnormal self-discharge during long-term charge and discharge cycles. These failures not only reduce battery performance and service life, but may also cause serious safety accidents such as fire and explosion.

[0003] Currently, common lithium-ion battery fault detection methods mostly focus on monitoring the thresholds of single physical quantities such as voltage, current, and temperature. For example, they use post-alarms based on abnormal voltage drops or sudden temperature rises. These methods are not only susceptible to noise interference but also fail to fully reflect the true health status of the battery. In addition, with the increase in battery energy density and system complexity, performance differences between single cells can easily lead to cascading failures such as thermal runaway, placing higher demands on lithium-ion battery fault detection. Traditional methods of assessing battery health based on single parameter threshold judgment or static consistency analysis (such as analyzing the voltage dispersion between single cells) are susceptible to load fluctuations in dynamic conditions, resulting in high false positive and missed detection rates.

[0004] Therefore, an optimized fault detection method for lithium-ion batteries is desired. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a fault detection method for lithium-ion batteries, which collects the total current data of the battery pack in real time, and judges the current operating status of the battery system according to the size and duration of the total current of the battery pack. Furthermore, when the battery system is in a static state, the voltage drop rate and temperature rise rate of each single cell in a preset time window are calculated, and compared with the corresponding threshold value to achieve self-discharge detection and fault identification. When the battery system is in a charging or discharging state, a deep learning algorithm is further introduced to extract the temperature-pressure coupling timing fluctuation characteristics of each single cell, and perform a temperature-pressure coupling timing fluctuation state consistency analysis on each single cell relative to the battery pack as a whole, thereby achieving intelligent detection of single cell faults during the charging and discharging process. This method can effectively capture abnormal temperature and pressure fluctuations of single cells during the charging and discharging process, thereby achieving early warning and precise positioning of faults.

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

[0007] Collecting total current data of the battery pack in real time, and judging the operating state of the battery system based on the total current data of the battery pack, wherein the operating state of the battery system includes a static state, a charging state, and a discharging state;

[0008] Real-time collection of the voltage and temperature of each single cell in the battery pack to obtain a set of single cell voltage time series data and a set of single cell temperature time series data;

[0009] In response to the battery system operating state being a static state, performing a self-discharge test on each single battery in the battery pack to determine whether the single battery has a fault;

[0010] In response to the battery system operating state being a charging state or a discharging state, a global dynamic response consistency analysis based on temperature-pressure coupling characteristics is performed on each single battery in the battery pack to determine whether each single battery has a fault.

[0011] Compared with the prior art, the lithium-ion battery fault detection method provided by the present application actually collects the total current data of the battery pack in real time, and determines the current operating status of the battery system based on the magnitude and duration of the total current of the battery pack. Furthermore, when the battery system is in a static state, the voltage drop rate and temperature rise rate of each single cell within a preset time window are calculated and compared with the corresponding threshold value to achieve self-discharge detection and fault identification. When the battery system is in a charging or discharging state, a deep learning algorithm is further introduced to extract the temperature-pressure coupling timing fluctuation characteristics of each single cell, and perform a temperature-pressure coupling timing fluctuation state consistency analysis on each single cell relative to the battery pack as a whole, thereby achieving intelligent detection of single cell faults during the charging and discharging process. This method can effectively capture abnormal temperature and pressure fluctuations of single cells during the charging and discharging process, thereby achieving early warning and precise positioning of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] 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.

[0013] Figure 1 Flowchart of a lithium-ion battery fault detection method according to an embodiment of the present application.

[0014] Figure 24 is a flowchart of sub-step S4 of the fault detection method for a lithium-ion battery according to an embodiment of the present application.

[0015] Figure 3 Schematic diagram of data flow of sub-step S4 of the fault detection method for a lithium-ion battery according to an embodiment of the present application.

[0016] Figure 4 4 is a flowchart of sub-step S43 of the lithium-ion battery fault detection method according to an embodiment of the present application.

[0017] Figure 5 4 is a flowchart of sub-step S431 of the lithium-ion battery fault detection method according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] 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.

[0019] 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.

[0020] 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.

[0021] 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.

[0022] 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.

[0023] In response to the technical problems described in the above background technology, this application proposes a lithium-ion battery fault detection method that collects total battery pack current data in real time and determines the current operating status of the battery system based on the magnitude and duration of the total battery pack current. Furthermore, when the battery system is in a static state, the voltage drop rate and temperature rise rate of each single cell within a preset time window are calculated and compared with corresponding thresholds to achieve self-discharge detection and fault identification. When the battery system is in a charging or discharging state, a deep learning algorithm is further introduced to extract the temperature-pressure coupling timing fluctuation characteristics of each single cell and perform a consistency analysis of the temperature-pressure coupling timing fluctuation state of each single cell relative to the entire battery pack, thereby achieving intelligent detection of single cell faults during the charging and discharging process. This method can effectively capture abnormal temperature and pressure fluctuations of single cells during the charging and discharging process, thereby achieving early warning and precise location of faults.

[0024] Figure 1 FIG. 1 is a flow chart of a method for detecting a fault of a lithium-ion battery according to an embodiment of the present application. Figure 1 As shown, the lithium-ion battery fault detection method includes the following steps: S1, real-time acquisition of total current data of the battery pack, and judging the operating state of the battery system based on the total current data of the battery pack, wherein the operating state of the battery system includes a static state, a charging state, and a discharging state; S2, real-time acquisition of the voltage and temperature of each single cell in the battery pack to obtain a set of single cell voltage time series data and a set of single cell temperature time series data; S3, in response to the operating state of the battery system being a static state, performing self-discharge detection on each single cell in the battery pack to determine whether each single cell has a fault; S4, in response to the operating state of the battery system being a charging state or a discharging state, performing a global dynamic response consistency analysis based on temperature-pressure coupling characteristics on each single cell in the battery pack to determine whether each single cell has a fault.

[0025] In the above-mentioned fault detection method for lithium-ion batteries, the step S1 collects the total current data of the battery pack in real time, and judges the operating state of the battery system based on the total current data of the battery pack, and the operating state of the battery system includes the static state, the charging state and the discharging state. It should be understood that lithium-ion batteries have significant differences in their internal electrochemical reaction mechanisms, thermal behavior and failure mode characteristics under different working states (such as static, charging, and discharging). For example, the self-discharge anomaly is more likely to be prominent in the static state, while the performance difference and thermal runaway risk between single cells in the charging and discharging state are more significant. If a unified detection strategy is adopted, it may lead to misjudgment or missed detection. Therefore, in order to accurately adapt the fault characteristics of lithium-ion batteries under different working conditions, the present application first dynamically determines the operating state of the battery system by monitoring the size and duration of the total current of the battery pack in real time based on the direct characterization ability of the current signal for the load characteristics. Specifically, the static current threshold interval [-I o ,I o ], where I o If the total current data of the battery pack is within the static current threshold range and the duration exceeds the preset time threshold, the battery system operation state is determined to be the static state; if the total current data of the battery pack is greater than I o If the duration exceeds the preset time threshold, the battery system is judged to be in the charging state; if the total current data of the battery pack is less than -I o If the duration exceeds a preset time threshold, the battery system is deemed to be in a discharging state. This approach not only avoids misclassification due to sudden load changes associated with traditional single-parameter threshold methods, but also provides accurate operating condition context for subsequent differentiated fault detection, improving the robustness of state recognition.

[0026] In the above-mentioned lithium-ion battery fault detection method, step S2 collects the voltage and temperature of each single cell in the battery pack in real time to obtain a set of single cell voltage time series data and a set of single cell temperature time series data. It should be understood that the present application takes into account that the single cell is the basic unit of the battery pack, and its voltage and temperature are the most direct and important physical quantities to characterize its internal electrochemical state, health status, and potential abnormalities. Voltage can reflect the battery's state of charge, internal resistance change, and open circuit voltage abnormality, while temperature can indicate the balance between heat generation and loss within the battery. Abnormal temperature rise is often an early signal of faults such as internal short circuit, overcharge, and over-discharge. Therefore, in order to construct a multi-dimensional battery health status characterization system, the present application, based on the electrochemical-thermal coupling mechanism, synchronously collects the voltage and temperature time series data of each single cell to form a set of single cell voltage time series data and a set of single cell temperature time series data covering the entire battery pack, thereby providing raw data with high temporal and spatial resolution for subsequent differential analysis under static and dynamic conditions, solving the problem of missing time series information in traditional static consistency analysis.

[0027] During implementation, a high-precision sensor network must be deployed to monitor the status of each individual battery cell. Voltage measurements are typically performed using voltage sensing circuits integrated into the battery management system (BMS) or independent voltage sensors. These devices can capture millivolt-level changes with exceptional precision, ensuring accurate data even under complex operating conditions. Furthermore, considering the potential for electromagnetic interference in practical applications, proper shielding measures are crucial. For example, cables with excellent anti-interference properties should be selected for wiring, and signal transmission paths should be kept as short as possible to minimize the impact of external factors on measurement results.

[0028] Meanwhile, for temperature monitoring, thermistors or thermocouples are among the most commonly used tools. These sensors offer fast response speeds, excellent stability, and the ability to maintain high measurement accuracy over a wide temperature range. The choice of installation location is also crucial; ideally, it should be as close as possible to the core area of ​​the battery to more accurately reflect internal temperature fluctuations. However, in practice, direct access to the core of the battery is often difficult due to factors such as space constraints. In such cases, mathematical models are needed to perform compensation calculations and indirectly infer the accurate temperature value. Infrared temperature measurement technology can also be used as a supplementary measure, particularly in locations where traditional contact sensors are difficult to install.

[0029] The frequency of data acquisition is also a factor that requires careful consideration. In order to capture all the details of the battery state evolution over time, a sufficiently high sampling rate must be guaranteed. Generally speaking, a sampling interval of dozens or even hundreds of times per second is considered to be an ideal range. However, this also means that the system will face enormous data processing pressure, so it is necessary to optimize the algorithm structure and improve computing efficiency. On the one hand, real-time processing capabilities can be improved through hardware accelerators such as FPGAs or GPUs; on the other hand, it is necessary to develop efficient compression encoding schemes that retain the key features of the original data while significantly reducing storage requirements.

[0030] To ensure the authenticity and reliability of collected data, a comprehensive calibration mechanism is necessary. This includes not only regular calibration of the sensors themselves but also inspections of the entire data transmission chain. During the calibration process, comparison tests are performed using known standards to promptly identify and correct deviations. Especially when environmental conditions change significantly, relevant parameter settings should be adjusted promptly to avoid measurement errors caused by external factors. Furthermore, the introduction of adaptive filtering algorithms is an effective way to improve data quality. By analyzing and processing the time domain signal, high-frequency noise and low-frequency drift components are removed, resulting in smoother and clearer voltage and temperature curves.

[0031] In the above-mentioned fault detection method for lithium-ion batteries, in step S3, in response to the battery system operating state being in a static state, a self-discharge test is performed on each single cell in the battery pack to determine whether the single cell has a fault. It should be understood that since the external current is zero or extremely small when the battery system is in a static state, the voltage change and temperature change of the single cell are mainly determined by the self-discharge reaction inside the single cell. Certain latent faults (such as micro-short circuits, a small amount of electrolyte leakage causing aggravated local reactions, etc.) will cause abnormal self-discharge rates, resulting in excessively rapid voltage drops and possibly accompanied by a slight abnormal temperature rise. Therefore, for a battery system in a static state, the present application calculates the single cell voltage drop rate and the single cell temperature rise rate of each single cell within a preset time window based on the set of single cell voltage time series data and the set of single cell temperature time series data, so as to perform self-discharge detection to determine its health status. Specifically, after determining that the battery is in a static state, the real-time collected single-cell voltage time series data and single-cell temperature time series data are used to calculate the voltage drop slope (ΔV / Δt) and temperature rise slope (ΔT / Δt) of each single cell within a preset time window and compare them with the corresponding threshold values ​​(such as μ±3σ) based on historical health data statistics. When the single-cell voltage drop rate of each single cell exceeds the preset voltage drop rate threshold or the single-cell temperature rise rate exceeds the preset temperature rise rate threshold, the corresponding single cell is determined to be faulty. In this way, potential internal defects can be proactively checked during the battery's non-operating period, effectively improving the ability to detect early, hidden faults, thereby preventing faults from worsening.

[0032] In the above-mentioned fault detection method for lithium-ion batteries, in step S4, in response to the battery system operating state being a charging state or a discharging state, a global dynamic response consistency analysis based on the temperature-pressure coupling characteristics is performed on each single cell in the battery pack to determine whether the each single cell has a fault. It should be understood that in the dynamic working process of charging or discharging the battery system, the voltage and temperature response of the single cell are not only affected by the external load, but also reflect its health status parameters such as internal impedance, capacity, polarization characteristics and heat generation characteristics. This makes the healthy battery group show similar temperature-pressure coupling response characteristics under the same excitation, while the voltage platform and temperature change curve of the single cell in poor health or with early faults will deviate significantly from the normal state, resulting in significant temperature-pressure response differences between the single cells in the battery pack. Therefore, in order to achieve early warning of faults under high dynamic loads, this application is based on the group consistency theory and introduces a deep learning algorithm to perform a global dynamic response consistency analysis on each single cell based on the temperature-pressure coupling timing fluctuation characteristics to identify single cells that are inconsistent with the overall response of the battery pack. Among them, Figure 24 is a flowchart of sub-step S4 of the fault detection method for a lithium-ion battery according to an embodiment of the present application. Figure 3 FIG. 1 is a data flow diagram of sub-step S4 of the fault detection method for lithium-ion batteries according to an embodiment of the present application. Figure 2 and Figure 3 As shown, the step S4 includes the steps of: S41, respectively extracting the temperature-pressure coupling characteristics of each single cell to obtain a set of temperature-pressure coupling timing fluctuation feature vectors of the single cell; S42, extracting the temperature-pressure coupling timing fluctuation feature vector of the single cell to be analyzed from the set of temperature-pressure coupling timing fluctuation feature vectors of the single cell; S43, using the temperature-pressure coupling timing fluctuation feature vector of the single cell to be analyzed as a query vector, performing dynamic response characteristic consistency analysis on the query vector and the set of temperature-pressure coupling timing fluctuation feature vectors of the single cell to be analyzed to obtain a temperature-pressure coupling feature query response coding vector of the single cell to be analyzed; S44, inputting the temperature-pressure coupling feature query response coding vector of the single cell to be analyzed into a classifier-based fault diagnosis model to determine whether the single cell to be analyzed has a fault.

[0033] Specifically, in a specific example of the present application, step S41 includes: regularizing the voltage time series data and temperature time series data of the single cell according to the time dimension and parameter dimension into a single cell temperature-pressure coupling time series data matrix, and then performing temperature-pressure coupling feature extraction based on two-dimensional convolutional coding to obtain the temperature-pressure coupling time series fluctuation feature vector of the single cell. Specifically, because the voltage and temperature of lithium-ion batteries under dynamic charge and discharge conditions are affected by multiple factors such as internal electrochemical reactions, ohmic heat, polarization heat, and entropy heat, and the two are mutually coupled, abnormal parameter fluctuations in a single dimension often cannot fully reveal the nature of the fault. Therefore, in order to effectively explore the synergistic response relationship between the voltage and temperature of lithium-ion batteries during the charging and discharging process and comprehensively characterize the health status of single cells, this application further aligns the voltage and temperature time series data of each single cell by time, and constructs a T×2 single cell temperature and pressure coupling time series data matrix in the time and parameter dimensions (T is the time step), which is then input into a two-dimensional convolutional neural network. By utilizing the powerful local feature extraction capability of the convolutional neural network, the high-order spatiotemporal correlation response between the voltage and temperature of the single cell is extracted through multi-layer convolution and pooling operations, and the fault-sensitive features in the dynamic interaction between voltage and temperature are learned. Finally, a fixed-dimensional single cell temperature and pressure coupling time series fluctuation feature vector is output. In this way, the temperature and pressure response mode of each single cell under dynamic working conditions can be accurately captured, significantly improving the accuracy and robustness of fault detection.

[0034] Specifically, the step S42 extracts the temperature-pressure coupling timing fluctuation feature vector of the single cell to be analyzed from the set of the temperature-pressure coupling timing fluctuation feature vectors of the single cell. It should be understood that the failure of the single cell in the battery pack is usually manifested as its temperature-pressure characteristics deviating from the group consistency. Therefore, in order to evaluate the health status of each single cell in the battery pack one by one, so as to accurately locate the potential faulty unit, the present application selects the temperature-pressure coupling timing fluctuation feature vector of the single cell that currently needs to be analyzed from the set of the temperature-pressure coupling timing fluctuation feature vectors of the single cell in a certain order (for example, according to the physical number of the battery or a preset detection sequence), as the temperature-pressure coupling timing fluctuation feature vector of the single cell to be analyzed, so as to clearly define the analysis object, so that the subsequent consistency analysis process can focus on processing the information of the specific single cell, and ensure the pertinence and accuracy of the diagnosis.

[0035] Specifically, in step S43, the temperature-pressure coupling time series fluctuation characteristic vector of the single cell to be analyzed is used as a query vector, and a dynamic response characteristic consistency analysis is performed on the query vector and the set of the temperature-pressure coupling time series fluctuation characteristic vectors of the single cell to be analyzed to obtain a temperature-pressure coupling characteristic query response coding vector of the single cell to be analyzed. Specifically, in order to quantify the degree of difference between the single cell to be analyzed and its peers (i.e., other single cells in the battery pack) in the temperature-pressure dynamic response characteristics, so as to effectively identify abnormal individuals, this application is based on the theory that the deviation from the consistency of group behavior indicates abnormality, and uses the temperature-pressure coupling time series fluctuation characteristic vector of the single cell to be analyzed as a query vector. By designing a dynamic response characteristic consistency analysis mechanism, it is compared with the set of the temperature-pressure coupling time series fluctuation characteristic vectors of the single cell to be analyzed to measure the consistency level of the temperature-pressure response characteristics of the single cell to be analyzed relative to the overall temperature-pressure response characteristics of the battery pack. Among them, Figure 4 FIG. 4 is a flow chart of sub-step S43 of the fault detection method of the lithium-ion battery according to an embodiment of the present application. Figure 4 As shown, the step S43 includes the steps of: S431, calculating the intrinsic semantic correlation factor between any two single-cell temperature-pressure coupling timing fluctuation feature vectors in the set of the single-cell temperature-pressure coupling timing fluctuation feature vectors to obtain the intrinsic optimization embedding matrix of the single-cell temperature-pressure coupling timing fluctuation feature; S432, mapping the query vector to the feature space of the intrinsic optimization embedding matrix of the single-cell temperature-pressure coupling timing fluctuation feature to obtain an aligned query vector; S433, performing query response consistency encoding on the aligned query vector and the set of the single-cell temperature-pressure coupling timing fluctuation feature vectors to obtain the query response encoding vector of the single-cell temperature-pressure coupling feature to be analyzed.

[0036] More specifically, step S431 calculates the intrinsic semantic correlation factor between any two cell temperature-pressure coupling time series fluctuation feature vectors in the set of cell temperature-pressure coupling time series fluctuation feature vectors to obtain the intrinsic optimization embedding matrix of the cell temperature-pressure coupling time series fluctuation feature. Figure 5 FIG. 4 is a flow chart of sub-step S431 of the fault detection method of the lithium-ion battery according to an embodiment of the present application. Figure 5 As shown, the step S431 includes the steps of: S4311, calculating the intrinsic semantic association factor between any two single-cell temperature-pressure coupling timing fluctuation feature vectors in the set of the single-cell temperature-pressure coupling timing fluctuation feature vectors and performing normalization processing to obtain a single-cell temperature-pressure coupling timing fluctuation feature intrinsic semantic association weight matrix composed of multiple normalized intrinsic semantic association factors; S4312, based on the single-cell temperature-pressure coupling timing fluctuation feature intrinsic semantic association weight matrix, performing feature space reconstruction on the set of the single-cell temperature-pressure coupling timing fluctuation feature vectors to obtain the single-cell temperature-pressure coupling timing fluctuation feature intrinsic optimization embedding matrix.

[0037] In a specific example of the present application, step S4311 calculates the intrinsic semantic correlation factor between any two single-cell temperature-pressure coupling timing fluctuation feature vectors in the set of single-cell temperature-pressure coupling timing fluctuation feature vectors and performs normalization processing to obtain a single-cell temperature-pressure coupling timing fluctuation feature intrinsic semantic correlation weight matrix composed of multiple normalized intrinsic semantic correlation factors, which is expressed as follows:

[0038]

[0039] Among them, Softamx(·) represents the normalized exponential function, tanh(·) represents the hyperbolic tangent function, and (·) T represents the transpose of the vector, k i and k j They represent the i-th and j-th cell temperature-pressure coupling time-series fluctuation characteristic vectors in the set of cell temperature-pressure coupling time-series fluctuation characteristic vectors, respectively. γ represents the weight matrix, d represents the characteristic scale of the temperature-pressure coupling time series fluctuation characteristic vector of the single battery, Γ i,j Represents the eigenvalue of the (i, j) position in the endogenous semantic association weight matrix of the temperature-pressure coupling time series fluctuation characteristics of the single cell, that is, k i and k j The normalized endogenous semantic correlation factor between them.

[0040] That is, by quantifying the intrinsic semantic associations between the temperature-pressure coupling time-series fluctuation feature vectors of each single cell, a weight system capable of characterizing the group response characteristics is constructed, thereby capturing the synergistic relationships and contextual dependencies of different single cells in the dynamic changes of temperature and pressure. Specifically, through the calculation and normalization of endogenous semantic association factors, not only can the implicit topological structure between the temperature-pressure coupling time-series fluctuation feature vectors of single cells be excavated, but this association can also be converted into a quantifiable weight matrix, namely the endogenous semantic association weight matrix of the temperature-pressure coupling time-series fluctuation features of single cells. This provides a semantic foundation for subsequent feature space reconstruction, and further accurately measures the degree of semantic deviation between the analyzed cells and the group in the global dynamic response consistency analysis.

[0041] In a specific example of the present application, step S4312, based on the intrinsic semantic association weight matrix of the single cell temperature-pressure coupling timing fluctuation feature, performs feature space reconstruction on the set of the single cell temperature-pressure coupling timing fluctuation feature vectors to obtain the intrinsic optimization embedding matrix of the single cell temperature-pressure coupling timing fluctuation feature, which is expressed as follows:

[0042]

[0043] Where Γ represents the intrinsic semantic association weight matrix of the temperature-pressure coupled temporal fluctuation characteristics of the single cell battery, represents matrix multiplication, Diag(·) represents the construction of a diagonal matrix, exp(·) represents the exponential function operation with base e, ||·||2 represents the L2 norm, and E represents the endogenous optimization embedding matrix of the temperature-pressure coupled timing fluctuation characteristics of the single battery.

[0044] That is, based on the topological structure of the endogenous semantic association weight matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell battery, the temperature-pressure coupling timing fluctuation feature vector of the single cell battery is mapped to a unified feature space containing group semantic associations, so that the reconstructed feature vector can not only retain the temperature-pressure timing fluctuation characteristics of the single cell itself, but also be embedded in the dynamic response relationship of the battery pack as a whole, forming an endogenous optimized embedding matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell battery with context-awareness, and can effectively characterize the semantic neighborhood relationship of the single cell battery in the group, providing a structured feature expression for the subsequent dynamic response consistency analysis of the temperature-pressure coupling characteristics.

[0045] In particular, in a preferred example of the present application, the step S4312 includes: first, performing local topological stability optimization on the intrinsic semantic association weight matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell battery to obtain the optimized intrinsic semantic association weight matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell battery; then, based on the optimized intrinsic semantic association weight matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell battery, reconstructing the feature space of the set of temperature-pressure coupling timing fluctuation feature vectors of the single cell battery to obtain the intrinsic optimized embedding matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell battery. Here, in the process of calculating the intrinsic semantic association factor between any two temperature-pressure coupling timing fluctuation feature vectors of the single cell battery, the temperature-pressure coupling timing fluctuation feature vector k of the single cell battery is firstly converted into the intrinsic semantic association factor of the temperature-pressure coupling timing fluctuation feature vector k of the single cell battery. i and k j As the starting and ending points of the trajectory in the graph model architecture, the trajectory representation term W γ To express the topological structure of the trajectory, and use the hyperbolic tangent function tanh to extract the topological complex features, and then divide it by the spatial distance Only the phase feature expression is retained.

[0046] Thus, in the calculation process of the endogenous optimization embedding matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell, ΓΓ T The phase accumulation reciprocity mechanism can be applied to effectively capture the phase context dependency while maintaining the intrinsic phase configuration. In order to improve the universality of the phase context dependency under the condition of spatial structure correlation, that is, to enhance the interference tolerance to the mismatch of local structure dependency, the gradient differential of the hyperbolic tangent function tanh can be further calculated, that is:

[0047]

[0048] in, Indicates the calculation of partial derivatives.

[0049] And then it is weighted by a predetermined weight coefficient and then used in the calculation of the normalized endogenous semantic association factor, namely:

[0050]

[0051] Where, ω is the predetermined weighting coefficient, Γ' i,j It is the eigenvalue of the position (i, j) in the intrinsic semantic association weight matrix of the temperature-pressure coupling timing fluctuation characteristics of the optimized single cell battery.

[0052] In this way, we can use the gradient diffusion phenomenon of the hyperbolic tangent function tanh, that is, when the input value is too large or too small, the gradient approaches zero, to resist the adverse effects of local structural dependency mismatch, and further substitute the calculation of the endogenous semantic correlation factor into the intrinsic optimization embedding matrix of the key space intrinsic relationship dynamic graph topology expression of the temperature-pressure coupling timing fluctuation characteristics of the single cell. At the same time, by introducing the differential gradient of the hyperbolic tangent function tanh as the geometric phase of the path topology pattern representation, after phase exchange, it acts on the key vector k i The statistical characteristics of ||k i || 2, the topological stability can be ensured by phase accumulation and exchange, thereby improving the topological representation effect of the internal relationship graph of the feature space of the endogenous optimization embedding matrix of the temperature-pressure coupling timing fluctuation characteristics of the single cell.

[0053] More specifically, in step S432, the query vector is mapped to the feature space of the endogenously optimized embedding matrix of the temperature-pressure coupled timing fluctuation characteristics of the single cell to obtain an aligned query vector, which is expressed as:

[0054]

[0055] Among them, v q Represents the query vector, v' q represents the alignment query vector.

[0056] Specifically, by mapping the query vector to the feature space of the intrinsically optimized embedding matrix of the cell temperature-pressure coupling timing fluctuation characteristics, the query vector and the cell temperature-pressure coupling timing fluctuation feature vector are placed in the same semantic dimension, ensuring comparability within the feature space of the group dynamic response. The resulting aligned query vector can be analyzed for semantic compatibility with the group feature vector in the unified feature space. Subsequent matching calculations accurately quantify the degree of deviation of the cell under analysis from the overall temperature-pressure coupling timing fluctuation of the battery pack, effectively improving the accuracy of cell fault identification during charging and discharging, and avoiding diagnostic bias caused by inconsistent feature spaces.

[0057] More specifically, in a specific example of the present application, step S433 includes: first, based on the endogenous optimization embedding matrix of the single cell temperature-pressure coupling timing fluctuation characteristics, performing feature adaptive fine-tuning on each single cell temperature-pressure coupling timing fluctuation characteristic vector in the set of single cell temperature-pressure coupling timing fluctuation characteristic vectors to obtain a set of optimized single cell temperature-pressure coupling timing fluctuation characteristic vectors, which is expressed as follows:

[0058]

[0059] Among them, k' iRepresents the i-th single cell temperature-pressure coupling time series fluctuation characteristic vector in the set of optimized single cell temperature-pressure coupling time series fluctuation characteristic vectors.

[0060] That is, through the synergistic effect of feature space reconstruction and semantic association weights, the noise interference and local abnormal fluctuations in the temperature-pressure coupling timing fluctuation feature vector of a single cell battery are eliminated, so that the generated set of optimized temperature-pressure coupling timing fluctuation feature vectors of the single cell battery can effectively identify fault characteristics that deviate from the group pattern in the global dynamic response consistency analysis, which not only retains the temperature-pressure timing fluctuation characteristics of the single cell battery itself, but also conforms to the temperature-pressure coupling dynamic response law of the battery pack as a whole.

[0061] Then, the matching degree between the alignment query vector and each optimized single cell temperature-pressure coupling timing fluctuation feature vector in the set of optimized single cell temperature-pressure coupling timing fluctuation feature vectors is calculated to obtain a query response encoding vector of the temperature-pressure coupling feature of the single cell to be analyzed composed of multiple matching degrees, which is expressed as follows:

[0062]

[0063] V=[Sim(k'1,v' q ),Sim(k'2,v' q ),...,Sim(k' i ,v' q ),...,Sim(k' n ,v' q )]

[0064] Where Sim(·), ·) represents the calculation of matching degree, ||·|| represents the calculation norm, Tr(·) represents the trace operation of the matrix, λ is the weight hyperparameter, and V represents the encoding vector of the query response to the temperature-pressure coupling feature of the single cell to be analyzed.

[0065] That is, the consistency level of the temperature-pressure coupling timing fluctuation characteristics of the single cell to be analyzed and the group characteristics is further converted into a computable numerical representation, so that the generated temperature-pressure coupling characteristic query response encoding vector of the single cell to be analyzed can accurately quantify the semantic fit difference between the temperature-pressure timing fluctuation of the single cell to be analyzed and the battery pack as a whole, so that the fault diagnosis model can identify abnormally deviated single cells based on the dynamic response law of the group, effectively improving the accuracy of early fault warning during the charging and discharging process, avoiding misjudgment and missed detection due to isolated feature analysis, and realizing dynamic and precise positioning of single cell faults.

[0066] Specifically, step S44 involves inputting the query response code vector for the thermal-pressure coupling characteristics of the battery to be analyzed into a classifier-based fault diagnosis model to determine whether the battery to be analyzed is faulty. It should be understood that the query response code vector for the thermal-pressure coupling characteristics of the battery to be analyzed reflects the degree of consistency in the thermal-pressure dynamic response characteristics of the battery to be analyzed with those of other batteries in the battery pack. As input to the fault diagnosis model, it effectively guides the model in classifying the health status of the battery to be analyzed. Specifically, the classifier-based fault diagnosis model, based on a deep neural network architecture and trained using a large amount of historical data, effectively learns the population deviation patterns of thermal-pressure coupling characteristics of batteries in different health states, thereby accurately determining whether the battery to be analyzed is faulty. During the diagnosis phase, the query response code vector for the battery to be analyzed, generated in real time, is input into the trained classifier. Based on the feature distribution of the query response code vector for the battery to be analyzed, the classifier accurately maps it into a preset state category space and determines whether the battery to be analyzed is faulty. In this way, real-time monitoring and intelligent early warning of the health status of the batteries in the battery pack can be achieved, enhancing the safety and reliability of the battery system.

[0067] In summary, a fault detection method for lithium-ion batteries based on an embodiment of the present application is illustrated, which collects the total current data of the battery pack in real time, and determines the current operating state of the battery system based on the magnitude and duration of the total current of the battery pack. Furthermore, when the battery system is in a static state, the voltage drop rate and temperature rise rate of each single cell within a preset time window are calculated and compared with the corresponding threshold value to achieve self-discharge detection and fault identification. When the battery system is in a charging or discharging state, a deep learning algorithm is further introduced to extract the temperature-pressure coupling timing fluctuation characteristics of each single cell, and perform a temperature-pressure coupling timing fluctuation state consistency analysis on each single cell relative to the battery pack as a whole, thereby achieving intelligent detection of single cell faults during the charging and discharging process. This method can effectively capture abnormal temperature and pressure fluctuations of single cells during the charging and discharging process, thereby achieving early warning and precise positioning of faults.

[0068] 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.

[0069] 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.

[0070] 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.

[0071] 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.

[0072] 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 lithium-ion battery fault detection method, characterized in that: include: Collecting total current data of the battery pack in real time, and judging the operating state of the battery system based on the total current data of the battery pack, wherein the operating state of the battery system includes a static state, a charging state, and a discharging state; Real-time collection of the voltage and temperature of each single cell in the battery pack to obtain a set of single cell voltage time series data and a set of single cell temperature time series data; In response to the battery system operating state being a static state, performing a self-discharge test on each single battery in the battery pack to determine whether the single battery has a fault; In response to the battery system operating state being a charging state or a discharging state, a global dynamic response consistency analysis based on temperature-pressure coupling characteristics is performed on each single battery in the battery pack to determine whether each single battery has a fault.

2. The method for detecting the remaining capacity of a lithium-ion battery according to claim 1, wherein: The battery system operating state is determined based on the total current data of the battery pack, where the battery system operating state includes a static state, a charging state, and a discharging state, including: Set the static current threshold range [-I o ,I o ], where I o is a positive number approaching zero; If the total current data of the battery pack is within the static current threshold range and the duration exceeds the preset time threshold, it is determined that the operating state of the battery system is a static state; If the total current data of the battery pack is greater than I o If the duration exceeds a preset time threshold, the battery system is determined to be in a charging state; If the total current data of the battery pack is less than -I o If the duration exceeds a preset time threshold, it is determined that the battery system is in a discharging state.

3. The method for detecting the remaining capacity of a lithium-ion battery according to claim 1, wherein: In response to the battery system operating state being a stationary state, performing a self-discharge test on each single battery in the battery pack to determine whether the single battery has a fault, including: Based on the set of single cell voltage time series data and the set of single cell temperature time series data, respectively calculating the cell voltage drop rate and the cell temperature rise rate of each single cell within a preset time window; When the cell voltage drop rate of each cell exceeds a preset voltage drop rate threshold or the cell temperature rise rate exceeds a preset temperature rise rate threshold, it is determined that the corresponding cell has a fault.

4. The method for detecting the remaining capacity of a lithium-ion battery according to claim 1, wherein: In response to the battery system operating state being a charging state or a discharging state, performing a global dynamic response consistency analysis based on temperature-pressure coupling characteristics on each single battery in the battery pack to determine whether each single battery has a fault, including: Extracting the temperature-pressure coupling characteristics of each of the single cells respectively to obtain a set of temperature-pressure coupling time series fluctuation characteristic vectors of the single cells; Extracting the temperature-pressure coupling time-series fluctuation characteristic vector of the single cell to be analyzed from the set of temperature-pressure coupling time-series fluctuation characteristic vectors of the single cell; Using the temperature-pressure coupling time-series fluctuation feature vector of the single cell to be analyzed as a query vector, performing a dynamic response characteristic consistency analysis on the query vector and a set of the temperature-pressure coupling time-series fluctuation feature vectors of the single cell to obtain a temperature-pressure coupling feature query response coding vector of the single cell to be analyzed; The temperature-pressure coupling feature query response encoding vector of the single battery to be analyzed is input into a classifier-based fault diagnosis model to determine whether the single battery to be analyzed has a fault.

5. The lithium-ion battery fault detection method according to claim 4, characterized in that: Extracting the temperature-pressure coupling characteristics of each single battery to obtain a set of temperature-pressure coupling time series fluctuation feature vectors of the single battery, including: The voltage time series data and temperature time series data of the single cell are regularized into a single cell temperature-pressure coupling time series data matrix according to the time dimension and the parameter dimension, and then a temperature-pressure coupling feature extraction based on two-dimensional convolution coding is performed to obtain the temperature-pressure coupling time series fluctuation feature vector of the single cell.

6. The lithium-ion battery fault detection method according to claim 5, characterized in that: Performing a dynamic response characteristic consistency analysis on the query vector and the set of the single cell temperature-pressure coupling time series fluctuation characteristic vectors to obtain a query response encoding vector of the single cell temperature-pressure coupling characteristic to be analyzed, including: Calculating the intrinsic semantic correlation factor between any two single-cell temperature-pressure coupling time-series fluctuation feature vectors in the set of the single-cell temperature-pressure coupling time-series fluctuation feature vectors to obtain an intrinsic optimization embedding matrix of the single-cell temperature-pressure coupling time-series fluctuation feature; Mapping the query vector to the feature space of the endogenous optimization embedding matrix of the temperature-pressure coupled timing fluctuation characteristics of the single cell to obtain an aligned query vector; Perform query response consistency coding on the set of the alignment query vector and the single cell temperature-pressure coupling time series fluctuation feature vector to obtain the single cell temperature-pressure coupling feature query response coding vector to be analyzed.

7. The lithium-ion battery fault detection method according to claim 6, characterized in that: Calculating the intrinsic semantic correlation factor between any two single-cell temperature-pressure coupling time-series fluctuation feature vectors in the set of single-cell temperature-pressure coupling time-series fluctuation feature vectors to obtain an intrinsic optimization embedding matrix of the single-cell temperature-pressure coupling time-series fluctuation feature, including: Calculating an intrinsic semantic correlation factor between any two battery cell temperature-pressure coupling time series fluctuation feature vectors in the set of battery cell temperature-pressure coupling time series fluctuation feature vectors and performing normalization processing to obtain a battery cell temperature-pressure coupling time series fluctuation feature intrinsic semantic correlation weight matrix composed of a plurality of normalized intrinsic semantic correlation factors; Based on the intrinsic semantic association weight matrix of the single cell temperature-pressure coupling timing fluctuation feature, a feature space reconstruction is performed on the set of the single cell temperature-pressure coupling timing fluctuation feature vectors to obtain the intrinsic optimized embedding matrix of the single cell temperature-pressure coupling timing fluctuation feature.

8. The lithium-ion battery fault detection method according to claim 7, characterized in that: Based on the intrinsic semantic association weight matrix of the single cell temperature-pressure coupling time series fluctuation feature, a feature space reconstruction is performed on the set of the single cell temperature-pressure coupling time series fluctuation feature vectors to obtain the intrinsic optimization embedding matrix of the single cell temperature-pressure coupling time series fluctuation feature, including: Performing local topological stability optimization on the intrinsic semantic association weight matrix of the temperature-pressure coupling time series fluctuation characteristics of the single cell battery to obtain an optimized intrinsic semantic association weight matrix of the temperature-pressure coupling time series fluctuation characteristics of the single cell battery; Based on the optimized endogenous semantic association weight matrix of the single cell temperature-pressure coupling timing fluctuation feature, the set of the single cell temperature-pressure coupling timing fluctuation feature vectors is reconstructed in feature space to obtain the endogenous optimized embedding matrix of the single cell temperature-pressure coupling timing fluctuation feature.

9. The lithium-ion battery fault detection method according to claim 8, characterized in that: Performing query response consistency coding on the alignment query vector and the set of the single cell temperature-pressure coupling time series fluctuation feature vectors to obtain the single cell temperature-pressure coupling feature query response coding vector to be analyzed, including: Based on the endogenous optimization embedding matrix of the single cell temperature-pressure coupling timing fluctuation characteristics, each single cell temperature-pressure coupling timing fluctuation characteristic vector in the set of single cell temperature-pressure coupling timing fluctuation characteristic vectors is adaptively fine-tuned to obtain a set of optimized single cell temperature-pressure coupling timing fluctuation characteristic vectors; The matching degree between the alignment query vector and each optimized single cell temperature-pressure coupling timing fluctuation feature vector in the set of optimized single cell temperature-pressure coupling timing fluctuation feature vectors is calculated to obtain a single cell temperature-pressure coupling feature query response encoding vector to be analyzed composed of multiple matching degrees.

Images