A lithium battery thermal runaway monitoring method and related device

Abstract

The application discloses a lithium battery thermal runaway monitoring method and related device, and relates to the field of lithium battery thermal runaway monitoring. The method comprises the following steps: acquiring a multi-perception parameter time sequence of a target lithium battery; the multi-perception parameter time sequence comprises a battery temperature time sequence, a battery voltage time sequence, a battery gas concentration time sequence and a battery deformation time sequence; inputting the multi-perception parameter time sequence into a monitoring model to obtain the space-time characteristics of each perception parameter; the monitoring model comprises a convolutional neural network, an LSTM network and an attention mechanism module connected in series; and monitoring the thermal runaway state of the lithium battery according to the space-time characteristics of each perception parameter and the corresponding dynamic threshold. The application can monitor the thermal runaway state of the lithium battery in real time and accurately, can predict the occurrence of thermal runaway in advance, and thus can timely send an early warning signal.

Description

Technical Field

[0001] This application relates to the field of lithium battery thermal runaway monitoring, and in particular to a lithium battery thermal runaway monitoring method and related apparatus. Background Technology

[0002] Lithium-ion batteries possess advantages such as high energy density, long lifespan, low self-discharge, and low pollution. They are widely used in new energy vehicles, 3C digital products, energy storage power stations, and aerospace. With the rapid development of the new energy vehicle and energy storage industries, the scale of lithium-ion battery usage continues to expand. Global attention to new energy sources is deepening, and lithium-ion battery energy storage technology is developing rapidly, demonstrating enormous market potential. However, under conditions such as overcharging and overheating, lithium-ion batteries are prone to thermal runaway, leading to fires or even explosions. The safety of lithium-ion batteries is gradually becoming a major obstacle to their further development. Therefore, early warning and prevention of thermal runaway in lithium-ion batteries is crucial for solving safety issues. Summary of the Invention

[0003] The purpose of this application is to provide a method and related device for monitoring thermal runaway of lithium batteries, which can monitor the thermal runaway state of lithium batteries in real time and accurately, predict the occurrence of thermal runaway in advance, and issue early warning signals in a timely manner.

[0004] To achieve the above objectives, this application provides the following solution:

[0005] In a first aspect, this application provides a method for monitoring thermal runaway in lithium batteries, including:

[0006] Obtain time series of multiple sensing parameters of the target lithium battery; the time series of multiple sensing parameters includes battery temperature time series, battery voltage time series, battery internal gas concentration time series, and battery deformation time series;

[0007] The time series of multiple sensing parameters are input into the monitoring model to obtain the spatiotemporal characteristics of each sensing parameter; the monitoring model includes a convolutional neural network, an LSTM network and an attention mechanism module connected in series.

[0008] The thermal runaway state of lithium batteries is monitored based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold.

[0009] Secondly, this application provides a lithium battery thermal runaway monitoring device, comprising:

[0010] The parameter acquisition module is used to acquire the time series of multiple sensing parameters of the target lithium battery; the time series of multiple sensing parameters includes the battery temperature time series, the battery voltage time series, the battery internal gas concentration time series, and the battery deformation time series.

[0011] The spatiotemporal feature extraction module is used to input the time series of multiple sensing parameters into the monitoring model to obtain the spatiotemporal features of each sensing parameter; the monitoring model includes a convolutional neural network, an LSTM network and an attention mechanism module connected in series;

[0012] The thermal runaway state monitoring module is used to monitor the thermal runaway state of lithium batteries based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold.

[0013] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described lithium battery thermal runaway monitoring method.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described lithium battery thermal runaway monitoring method.

[0015] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described lithium battery thermal runaway monitoring method.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0017] This application provides a method and related apparatus for monitoring thermal runaway in lithium batteries, acquiring time series of multiple sensing parameters of the target lithium battery. These time series include battery temperature, battery voltage, internal gas concentration, and battery deformation. The time series are input into a monitoring model to obtain the spatiotemporal characteristics of each sensing parameter. The monitoring model includes a convolutional neural network, an LSTM network, and an attention mechanism module connected in series. The thermal runaway state of the lithium battery is monitored based on the spatiotemporal characteristics of each sensing parameter and its corresponding dynamic threshold. This application captures key features generated by electrolyte decomposition through multi-parameter fusion and performs joint analysis to predict the occurrence of thermal runaway in advance. Furthermore, this application utilizes a convolutional neural network to extract the spatial distribution features of each parameter and uses an LSTM network and attention mechanism module to extract and optimize the temporal features of each parameter. This allows for the capture of long-term trends in battery state, effectively capturing the spatiotemporal dependence during thermal runaway. After joint detection, the sensitivity is improved by 3.2 times, and abnormal patterns in μV-level voltage fluctuations can be identified, significantly improving the timeliness of early warning. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an application environment diagram of a lithium battery thermal runaway monitoring method according to an embodiment of this application;

[0020] Figure 2 A schematic flowchart of a lithium battery thermal runaway monitoring method provided in an embodiment of this application;

[0021] Figure 3 A schematic diagram illustrating the technical concept of a lithium battery thermal runaway monitoring method provided in an embodiment of this application;

[0022] Figure 4 This is a schematic diagram of the structure of a CNN-LSTM neural network analysis layer provided in an embodiment of this application;

[0023] Figure 5 This is a schematic diagram of the cloud control platform system structure provided in one embodiment of this application;

[0024] Figure 6 A functional module schematic diagram of a lithium battery thermal runaway monitoring device provided in another embodiment of this application;

[0025] Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] The main technologies for early warning of thermal runaway in lithium batteries at present are: (1) Fiber optic sensing technology: Some studies have used multimode fiber optic sensors implanted inside the battery to directly measure the internal temperature and monitor the core internal parameters in real time. It can simultaneously measure temperature, pressure, gas composition, etc., and withstand high temperature and high pressure environments of 1000℃. It solves the problems of signal lag and crosstalk of traditional external sensors, greatly improves accuracy, and can effectively predict and prevent thermal runaway of lithium-ion batteries. However, fiber optic sensing technology has high manufacturing costs. The manufacturing of fiber optic sensors requires professional technology and equipment, resulting in high costs. In long-distance applications, the deployment cost is high. The packaging process is complex. The fiber optic grating itself is very fragile and needs to be protected so that it can work stably in practical applications. However, the packaging process is complex and requires precise control of process parameters, otherwise it will affect the performance and life of the fiber optic sensor.

[0028] (2) Gas sensor detection method: This method utilizes gas sensors to detect specific gases released during battery thermal runaway. For example, some studies have used MQ-2 smoke (methane) sensors, MQ-7 carbon monoxide sensors, and MQ-8 hydrogen sensors as signal acquisition devices to measure the voltage values ​​output by the sensors under different conditions and to implement audible and visual alarm functions. Other studies have built an experimental platform for the runaway and gas detection of lithium iron phosphate single-cell batteries to conduct online detection of six characteristic gases of thermal runaway (hydrogen, carbon monoxide, carbon dioxide, hydrogen chloride, hydrogen fluoride, and sulfur dioxide). The results showed that hydrogen was detected first, 13 minutes before ignition, demonstrating a certain degree of predictability. However, the gas sensor detection method has a limited response speed. Although gas sensors have certain advantages over temperature sensors in early thermal runaway detection, their response speed is still relatively slow and may not be able to capture early signs of thermal runaway in time. In some applications with extremely high real-time requirements, this limitation in response speed may lead to missing the optimal emergency response opportunity.

[0029] (3) Internal Resistance Measurement Method: Early signs of battery thermal runaway, such as gas generation, bulging, and micro-short circuits, can cause changes in cell impedance. Online impedance measurement can provide early warnings. For example, some studies have compared and analyzed five typical lithium-ion battery internal resistance measurement methods: hybrid pulse power characteristic method, DC internal resistance test method, AC injection method, DC discharge method, and electrochemical impedance spectroscopy (EIS). These studies introduced the relationship between internal resistance and battery life, battery state, and battery safety warnings. Other studies have used EIS as a basis to achieve real-time detection of battery internal impedance using single-frequency impedance at 1Hz and 100Hz. The results were verified under different charging rates, ambient temperatures, and health conditions. The characteristic impedance showed good stability, and the final warning level of this method can issue a safety warning more than 5 minutes before thermal runaway. However, internal resistance measurement methods, such as DC discharge and AC injection, have certain limitations. DC discharge may cause some damage to the battery and can only measure large-capacity batteries. Although AC injection causes less damage to the battery, the measurement equipment is usually more complex and expensive, and the measurement accuracy may be affected by ripple current and harmonic current interference.

[0030] (4) Voltage and Current Monitoring Method: The Battery Management System (BMS) monitors battery voltage, current, and other signals in real time. When abnormal changes occur in voltage and current, they can serve as one of the bases for early warning of thermal runaway. However, relying solely on external parameter monitoring cannot fully and accurately reflect the electrochemical changes inside the battery, so it is often used in combination with other monitoring methods. Voltage and Current Monitoring Method: The BMS mainly assesses the battery state by monitoring parameters such as voltage, current, and temperature, but changes in these parameters may lag behind actual changes in the battery's internal state, making it impossible to capture early signals of thermal runaway in a timely manner.

[0031] In response, this application provides a lithium battery thermal runaway monitoring method and related device, which can monitor the state of lithium batteries in real time and accurately, predict the occurrence of thermal runaway in advance, and issue early warning signals in a timely manner, providing sufficient time for taking corresponding safety measures and effectively reducing the safety risks caused by lithium battery thermal runaway.

[0032] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] The lithium battery thermal runaway monitoring method provided in this application embodiment can be applied to, for example... Figure 1The application environment shown is illustrated. The terminal communicates with the server via a network. The data storage system stores the data the server needs to process. The data storage system can be set up independently, integrated into the server, or located in the cloud or on another server. The terminal can send the time series of multiple sensing parameters of the target lithium battery (including battery temperature time series, battery voltage time series, battery internal gas concentration time series, and battery deformation time series) to the server. After receiving the time series of multiple sensing parameters of the target lithium battery, the server inputs the time series of multiple sensing parameters into a monitoring model to obtain the spatiotemporal characteristics of each sensing parameter. The monitoring model includes a convolutional neural network, an LSTM network, and an attention mechanism module connected in series. The lithium battery thermal runaway state is monitored based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold. The server can provide feedback on the obtained lithium battery thermal runaway state to the terminal. Furthermore, in some embodiments, the lithium battery thermal runaway monitoring method can also be implemented independently by the server or the terminal. For example, the terminal can directly perform lithium battery thermal runaway monitoring based on the time series of multiple sensing parameters of the target lithium battery, or the server can obtain the time series of multiple sensing parameters of the target lithium battery from the data storage system and perform lithium battery thermal runaway monitoring.

[0034] The terminal can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. The server can be a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0035] In one exemplary embodiment, such as Figures 2 to 4 As shown, a method for monitoring thermal runaway in lithium batteries is provided. This method is executed by a computer device, specifically a terminal or server, or both. In this embodiment, the method is applied to... Figure 1 The following steps, 101 to 103, are used as an example to illustrate the process of using a server in the example.

[0036] Step 101: Obtain the time series of multiple sensing parameters of the target lithium battery; the time series of multiple sensing parameters includes the battery temperature time series, battery voltage time series, battery internal gas concentration time series and battery deformation time series.

[0037] The system employs multi-parameter sensing, acquiring temperature, voltage, strain, and gas concentration information, and deploying various types of sensors. Distributed temperature sensors are cleverly embedded in the gaps between battery modules, with their spacing carefully adjusted to ensure it does not exceed 10 cm. This sensor achieves an accuracy of ±0.5℃ and a sampling frequency of 1Hz. MEMS gas sensors are positioned at the vents on the top of the battery compartment to detect carbon monoxide, hydrogen, and volatile organic compounds (VOCs), with a sensitivity of at least 10 ppm. Furthermore, a strain gauge array is tightly attached to the battery casing surface, achieving a deformation detection accuracy of ±0.1 mm. A high-frequency voltage acquisition module is connected to the positive and negative terminals of the battery, with a sampling rate of at least 1 kHz and an accuracy of ±1 nV. In addition, a self-calibration mechanism is included; the sensors automatically perform zero-point calibration every 24 hours, effectively eliminating long-term drift errors and ensuring the accuracy and stability of data acquisition.

[0038] Step 102: Input the time series of multiple sensing parameters into the monitoring model to obtain the spatiotemporal characteristics of each sensing parameter; the monitoring model includes a convolutional neural network, an LSTM network and an attention mechanism module connected in series.

[0039] Step 103: Monitor the thermal runaway state of the lithium battery based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold.

[0040] By implementing steps 101 to 103 above, this application captures key features generated by the decomposition of battery electrolyte through multi-parameter fusion and performs joint analysis to predict the occurrence of thermal runaway in advance. Furthermore, this application utilizes a convolutional neural network to extract the spatial distribution features of each parameter and uses an LSTM network and attention mechanism module to extract and optimize the temporal features of each parameter. This enables the capture of long-term trends in battery state, effectively capturing the spatiotemporal dependence during thermal runaway. After joint detection, the sensitivity is improved by 3.2 times, and abnormal patterns in μV-level voltage fluctuations can be identified, significantly improving the timeliness of early warning.

[0041] Furthermore, the method proposed in this application exhibits high robustness in extreme environments. The sensor array is calibrated across a full temperature range of -40℃ to 85℃, and the humidity compensation algorithm covers 10-95% RH. Under vibration conditions (5Grms), the fiber optic strain monitoring error is <1.5%, demonstrating excellent anti-interference capability and adaptability. Conventional thermal runaway monitoring technologies tend to experience increased errors or even failure under extreme temperature, humidity, or strong interference conditions. This application's technology overcomes the limitations of existing technologies in complex operating conditions, ensuring monitoring accuracy and response reliability. This technology can operate stably in environments ranging from frigid to tropical and high-humidity areas, ensuring accurate acquisition of thermal runaway signals. This provides a solid data foundation for early warning of thermal runaway.

[0042] In another exemplary embodiment of this application, in step 102, data preprocessing is first performed, using cubic spline interpolation to align data at different sampling rates (e.g., 1Hz temperature and 10kHz voltage). The isolated forest algorithm is used to identify faulty sensor data. Upon detecting abnormal data from the primary sensor (i.e., deeming it faulty or unreliable), the system automatically switches data acquisition or control to a pre-set backup sensor or data node to ensure the continuity and reliability of system functions, while simultaneously eliminating outliers.

[0043] The input data is selected as a multi-channel time series input, including temperature (1 channel), voltage (3 channels: cell / module / system), gas concentration (3 channels: CO / H2 / VOCs), and deformation (1 channel). Next, feature extraction is performed. In the lithium battery thermal runaway monitoring scenario, the data has multi-dimensional spatiotemporal characteristics. The spatial dimension includes spatial correlation features such as temperature field distribution and gas diffusion gradient. The temporal dimension includes temporal evolution patterns such as voltage fluctuations and deformation accumulation. Temporal features can extract the temperature rise rate, voltage fluctuation standard deviation, and gas concentration time integral. Frequency domain features involve performing wavelet packet transform on the voltage signal to calculate the energy proportion of each sub-band. Spatial features are extracted by using convolutional kernels to extract the spatial gradient distribution of the battery module's temperature field. The CNN-LSTM neural network achieves accurate modeling of the thermal runaway chain reaction through a collaborative mechanism of CNN extracting spatial features and LSTM capturing temporal dependencies.

[0044] A bidirectional LSTM network is constructed to capture dynamic evolution. The bidirectional LSTM layer has 128 hidden units and a sliding window length of 32, capturing forward and backward dependencies. The gating mechanism is optimized by adding a temperature threshold to the input gate to suppress irrelevant inputs in low-temperature environments. The forget gate uses the Sigmoid function to control the information decay rate. Its output range is (0, 1), mapping input values ​​to this interval. The output of the bidirectional LSTM network passes through a fully connected layer with 64 nodes. Dropout is a regularization technique used to prevent overfitting in neural networks. Its dropout probability is set to 0.3. The Sigmoid function is a commonly used activation function; its mathematical expression is... An attention mechanism is introduced to automatically weight important feature channels and output the results.

[0045] Based on the above, in step 102, the time series data of the multiple sensing parameters are input into the monitoring model to obtain the spatiotemporal characteristics of each sensing parameter, specifically including:

[0046] (2-1) Input the battery temperature time series, battery voltage time series, battery internal gas concentration time series and battery deformation time series into the convolutional neural network to obtain the spatial distribution characteristics of battery temperature, battery voltage, battery internal gas concentration and battery deformation.

[0047] (2-2) Input the spatial distribution characteristics of battery temperature, battery voltage, gas concentration in battery and battery deformation into the LSTM network to obtain the time-series characteristics of battery temperature, battery voltage, gas concentration in battery and battery deformation.

[0048] (2-3) Input the battery temperature time series characteristics, battery voltage time series characteristics, battery internal gas concentration time series characteristics and battery deformation time series characteristics into the attention mechanism module to obtain the optimized battery temperature time series characteristics, optimized battery voltage time series characteristics, optimized battery internal gas concentration time series characteristics and optimized battery deformation time series characteristics.

[0049] A fusion data processing engine based on Convolutional-Long Short-Term Memory (CNN-LSTM) forms the core intelligent hub of state recognition. Through a collaborative processing mechanism, it achieves deep feature mining and dynamic evolution modeling of multi-source heterogeneous data. By using multi-channel time-series input data and combining CNN and LSTM, the spatiotemporal features contained in the data are fully extracted, thereby achieving accurate modeling of the thermal runaway chain reaction of lithium batteries. In this process, CNN is responsible for extracting spatial features through three layers of one-dimensional convolutions, each containing 32 filters with a kernel size of 3. A fully connected layer is added at the end of the CNN module for dimensionality compression. Each convolutional layer is followed by a pooling layer to reduce feature dimensionality and enhance feature robustness. LSTM captures temporal dependencies, with 128 hidden units, employing a bidirectional LSTM (Bi-LSTM) structure to simultaneously consider forward and backward temporal dependencies, followed by a Dropout layer to improve the ability to identify abnormal patterns. The collaborative work of CNN and LSTM effectively enhances the system's predictive ability.

[0050] To further optimize the feature extraction process, this system introduces a self-attention mechanism. In the network structure, the attention mechanism is added after the LSTM layer and applied to all time steps of the LSTM's last layer output. Specifically, the time-series features output by the LSTM layer are passed to the attention layer, which calculates the weight for each time step and generates a weighted context vector. This context vector integrates the information most contributing to the current prediction throughout the entire time series, enhancing the model's ability to perceive critical moments. This mechanism allows the network to highlight changes in key regions, assigning higher weights to features more important to the current task while ignoring less important information, thus significantly improving the accuracy and effectiveness of feature extraction. Through the attention mechanism, the model can automatically focus on key time points where abnormal signals appear, improving the ability to identify weak anomalies. With this optimized model architecture, the system can not only accurately capture early signs of lithium battery thermal runaway but also flexibly adjust monitoring strategies according to different operating conditions and environmental conditions, ensuring efficient and timely early warning and response.

[0051] This application enhances early signal capture capabilities. Existing technologies rely on single temperature or voltage thresholds, typically triggering alarms only after thermal runaway has occurred, making it difficult to capture early characteristics. This application, however, uses multi-parameter fusion to capture key features such as electrolyte decomposition and performs joint analysis, solving the problem of delayed warnings caused by existing technologies analyzing only a single time series and ignoring crucial information such as the spatial distribution of the temperature field. This application employs CNN primarily for spatial feature extraction, utilizing convolutional layers to extract local spatial information from the input data. LSTM, a temporal model, is mainly used to capture temporal dependencies, enabling the capture of long-term trends in battery state. The combination of these two effectively captures the spatiotemporal dependencies during thermal runaway. Joint detection improves sensitivity by 3.2 times, identifying abnormal patterns in μV-level voltage fluctuations and significantly enhancing warning timeliness.

[0052] In another exemplary embodiment of this application, step 103, monitoring the thermal runaway state of the lithium battery based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold, specifically includes:

[0053] (3-1) Determine the thermal runaway state of the lithium battery based on the comparison results of the optimized battery temperature time sequence characteristics and the corresponding dynamic threshold and the battery temperature spatial distribution characteristics.

[0054] (3-2) Determine the thermal runaway state of the lithium battery based on the comparison results of the optimized battery voltage time sequence characteristics and the corresponding dynamic threshold and the battery voltage spatial distribution characteristics.

[0055] (3-3) The thermal runaway state of the lithium battery is determined based on the comparison results of the time sequence characteristics of the gas concentration in the battery after optimization and the corresponding threshold, and the spatial distribution characteristics of the gas concentration in the battery.

[0056] (3-4) Determine the thermal runaway state of the lithium battery based on the comparison results of the optimized battery deformation time sequence characteristics and the corresponding threshold and the battery deformation spatial distribution characteristics.

[0057] As an example, the optimized battery temperature time series characteristic is the battery temperature rise rate; the optimized battery voltage time series characteristic is the voltage deviation; the optimized battery internal gas concentration time series characteristic is the gas concentration change trend over time; and the optimized battery deformation time series characteristic is the deformation change trend over time.

[0058] The expression for the dynamic threshold corresponding to the optimized battery voltage timing characteristics is:

[0059] T voltage =a v ·(1-b v ·(100-SOH))

[0060] Among them, T voltage Indicates the voltage deviation threshold; a v Indicates the benchmark deviation threshold (e.g., initially set to 5%); b v SOH impact coefficient, original value is 0.01 (for every 1% decrease, the threshold decreases by 1%); SOH represents the battery health status.

[0061] The expression for the dynamic threshold corresponding to the optimized battery temperature time-series characteristics is:

[0062] T temp =a t ·(1-b t ·(T env -T0))

[0063] Among them, T temp Indicates the battery temperature rise rate threshold; a t Indicates the reference temperature rise rate threshold (e.g., 2.0 °C / min); b t T0 represents the ambient temperature influence coefficient (0.015); T0 represents the reference ambient temperature (e.g., 25℃).

[0064] The expression for the threshold corresponding to the time-series characteristics of gas concentration inside the battery after optimization is as follows:

[0065] T gas =a g ·(1-b g VOCs baseline )

[0066] Among them, T gas This indicates the gas concentration threshold, specifically the alarm threshold for combustible gas (such as H2) concentration; VOCs baseline Indicates the background concentration of volatile organic compounds; a gIndicates the reference gas alarm threshold (e.g., 50 ppm); b g Indicates the VOC background adjustment factor;

[0067] The expression for the threshold corresponding to the optimized battery deformation time-series characteristics is:

[0068] T strain =α·ε yield

[0069] Among them, T strain Indicates the deformation threshold; ε yield The yield strain of the battery casing / encapsulation material is typically 2000–5000 με; α represents the safety factor, which is 0.5 and is used for early warning.

[0070] This application employs a dynamic threshold decision-making system, while existing technologies use fixed thresholds, which are prone to false alarms due to environmental interference. This invention dynamically adjusts the threshold based on SOC (State of Charge), SOH (State of Health), and ambient temperature, enabling more accurate identification of abnormal signals in complex and dynamic working environments. The dynamic threshold decision-making system is a method that adaptively adjusts alarm thresholds based on the battery's health status and the surrounding environmental conditions. This approach allows for more accurate assessment of whether the battery is at risk of thermal runaway or other abnormalities, significantly reducing the false alarm rate. The threshold is not fixed but changes in real-time based on SOH and environmental conditions. For example, as the battery ages, the temperature rise may be faster, and the system will appropriately lower the alarm threshold; at lower ambient temperatures, higher temperature fluctuations may be allowed. The dynamic threshold is the alarm trigger threshold for a specific monitoring parameter (such as voltage deviation, temperature rise rate, or gas concentration).

[0071] The dynamic threshold decision-making system enables the system to respond more sensitively to potential anomalies by setting dynamic thresholds for key parameters. After introducing this mechanism, the false alarm rate of the system was reduced to 1.2% in charge-discharge cycle testing, a 92% reduction compared to traditional methods, significantly improving the accuracy and reliability of early warning.

[0072] In another exemplary embodiment of this application, the lithium battery thermal runaway monitoring method further includes: providing an early warning based on the lithium battery thermal runaway state; specifically:

[0073] (1) Determine the local overheating area based on the spatial distribution characteristics of battery temperature; when the optimized battery temperature time sequence characteristics exceed the corresponding dynamic threshold, determine the warning area based on the identified local overheating area to trigger the warning.

[0074] For temperature data, CNNs extract the distribution features of the temperature field through convolutional layers, enabling them to identify differences in heat distribution within the battery module and the entire system, and to pinpoint localized overheating areas. LSTMs, on the other hand, focus on analyzing the temporal trends of temperature changes, especially the rate of temperature rise (e.g., rapid temperature increases exceeding 2°C / min). This time-series analysis capability allows LSTMs to accurately capture patterns in temperature changes and promptly determine whether to trigger an early warning.

[0075] (2) Determine the voltage spatial distribution pattern based on the voltage spatial distribution characteristics; when the optimized battery voltage timing characteristics exceed the corresponding dynamic threshold, determine the warning area in combination with the voltage spatial distribution pattern to trigger the warning.

[0076] For voltage data, voltages in battery systems typically exhibit spatial variations (e.g., differences between individual cells and modules, and between the entire system). CNNs effectively identify spatial voltage distribution patterns in this process, providing crucial foundational data. LSTMs, through analysis of voltage temporal fluctuations, particularly during rapid voltage drops, can identify potential early signs of thermal runaway and issue timely warnings.

[0077] (3) Determine the spatial distribution of gas concentration based on the spatial distribution characteristics of gas concentration in the battery; when the time sequence characteristics of gas concentration in the optimized battery exceed the corresponding threshold, determine the warning area in combination with the spatial distribution of gas concentration to trigger the warning.

[0078] In gas analysis, CNNs extract the spatial distribution of gas concentration through convolutional layers, helping to identify the diffusion gradient of gas in battery modules or systems. Different gases (such as CO, H2, and VOCs) may have different diffusion patterns in different regions, and this spatial feature helps to locate abnormal areas. LSTMs are responsible for capturing the trend of gas concentration changes over time. In particular, when the gas concentration continues to rise, LSTMs can effectively identify this cumulative effect and provide support for early warning mechanisms.

[0079] (4) Determine the deformation spatial distribution based on the characteristics of battery deformation spatial distribution; when the optimized battery deformation time sequence characteristics exceed the corresponding threshold, determine the warning area in combination with the battery deformation spatial distribution to trigger the warning.

[0080] For strain data, CNNs extract spatial features of deformation through convolution operations, effectively identifying local deformation or expansion phenomena in battery modules. Especially during thermal runaway, deformation is often accompanied by gas release and temperature increases. LSTMs, on the other hand, analyze the temporal data of deformation to determine if it exhibits a continuous increasing trend, which is usually a warning signal that the battery is entering thermal runaway.

[0081] like Figure 5As shown, data is transmitted to the cloud via sensors and IoT modules, then analyzed and processed to trigger corresponding early warning rules. For visualization, the cloud platform uses dashboards, charts, and color coding to display different levels of warning status and updates data in real time. In addition, the platform provides historical data query functions and a map view displaying the location and status of multiple batteries for quick problem identification. It also provides detailed battery parameter change curves to support customers in fully understanding battery status. The warning levels are divided into four levels based on the development trend and the degree of potential harm: minor faults, low risk, and high risk. Level 1 response targets minor faults; when the voltage deviation of a single cell reaches the voltage deviation threshold, an adaptive balancing strategy is triggered to delay battery degradation. The adaptive balancing strategy refers to identifying "weak cells" during battery management based on real-time voltage monitoring, historical data, and current charge / discharge trends. It dynamically decides whether to perform balancing and which balancing method to use to delay battery degradation and improve safety. An active balancing method is used, transferring energy from high-voltage cells to low-voltage cells using bidirectional DC-DC converters, for example, transferring 0.2A per second for 15 seconds to reach the target difference. Level 2 response addresses low-risk situations. When the module's temperature rise rate reaches a threshold, localized cooling (liquid / air cooling) is initiated to limit current. Level 3 response addresses high-risk situations where gas concentration begins to increase, causing deformation, typically reaching a strain rate of approximately 4000 με. The faulty module is isolated, and the fire suppression system is prepared for activation. Level 4 response addresses critical thermal runaway conditions, where the temperature exceeds 80°C and continues to rise, accompanied by a sudden voltage drop. The main circuit is disconnected, and total flooding fire suppression is triggered.

[0082] This application employs a composite sensor for multi-physical quantity acquisition and transmits the data to the cloud via an IoT module. It analyzes and models historical data from lithium-ion batteries, comparing extracted time-series features with corresponding thresholds to predict alarms. Data visualization includes a real-time monitoring interface, color-coding battery status, pop-up alert windows, notifications to mobile phones or email addresses, report generation, trend analysis charts, and historical data backtracking functionality. The map view displays the location and status of all battery devices connected to the platform for rapid problem localization. When a thermal runaway event occurs, a corresponding early warning mechanism is triggered to ensure timely implementation of safety measures.

[0083] This application adopts a three-tier architecture of "perception layer - analysis layer - decision layer" to achieve closed-loop management from data acquisition to proactive prevention and control. The perception layer collects battery and environmental parameters in real time through a multi-dimensional sensor array. The analysis layer utilizes a multi-parameter fusion analysis engine based on convolutional-long short-term memory neural networks. The decision layer employs a four-level early warning triggering dynamic response mechanism.

[0084] The sensing layer architecture of this application integrates multi-dimensional parameters such as temperature, voltage, current, capacity, gas composition, and deformation, and constructs an early thermal runaway feature database based on these parameters. Through a dynamic threshold decision-making system, the system can accurately identify weak abnormal signals, providing an early warning time 5-10 minutes earlier than existing technologies (verified by test data). This achieves early and accurate warning of lithium battery thermal runaway, effectively solving the problem of warning lag in existing technologies. Traditional single-parameter threshold monitoring (such as monitoring only temperature or voltage) struggles to capture early thermal runaway features, often resulting in irreversible thermal runaway by the time a warning is issued.

[0085] This application employs a dynamic adaptive monitoring mechanism that automatically adjusts monitoring thresholds based on historical battery data and real-time operating conditions, reducing the false alarm rate by over 60%. This improvement effectively solves the problem of false alarms easily triggered by significant interference from operating conditions (such as charging / discharging temperature rise and ambient temperature fluctuations) on a single parameter. Furthermore, this system is compatible with various battery types and application scenarios, such as electric vehicles and energy storage power stations, demonstrating broad applicability.

[0086] This application constructs a four-level early warning and response mechanism. Level 1 response addresses minor faults, triggering an adaptive balancing strategy to delay battery degradation. Level 2 response addresses potential risks, initiating localized cooling (liquid cooling / air cooling) and current limiting. Level 3 response addresses high-risk situations, isolating the faulty module and preparing to activate the fire suppression system. Level 4 response addresses the critical state of thermal runaway, cutting off the main circuit and triggering total flooding fire suppression. This mechanism effectively solves the problem of rigid response mechanisms in traditional systems. Traditional systems only trigger a single alarm, lacking a tiered response strategy, making it difficult to match the prevention and control needs of different stages of thermal runaway.

[0087] This application utilizes long-term multi-parameter trend analysis to identify battery health degradation 3-6 months in advance, predict battery health status, and identify potential failure risks. This enables predictive maintenance and provides strong support for it. Even after 2000 aging cycles, it maintains a 93% early warning accuracy rate, while traditional methods only maintain an accuracy rate of around 65%. This not only improves the safety and lifespan of lithium batteries but also reduces maintenance costs, enhancing overall economic efficiency and safety.

[0088] This application has the following technical effects:

[0089] 1. Multi-physics collaborative sensing:

[0090] By combining gas diffusion dynamics models with cell deformation field analysis, the limitations of traditional single-parameter monitoring are overcome, enabling the simultaneous capture of early signals such as electrolyte decomposition and internal short circuits, providing a more comprehensive basis for early warning of thermal runaway.

[0091] 2. Spatiotemporal feature fusion algorithm:

[0092] By employing CNN to extract the spatial gradient distribution of the temperature field, utilizing LSTM to capture the temporal evolution of gas concentration, and combining an attention mechanism to dynamically focus on key features, the early warning sensitivity was improved by 3.2 times, significantly enhancing the ability to identify early features of thermal runaway.

[0093] 3. Dynamic threshold decision-making system:

[0094] Based on adaptive threshold adjustment of battery state of health (SOH) and environmental conditions, the false alarm rate is reduced to 1.2%, a reduction of 92.4% compared to traditional methods.

[0095] 4. For minor faults, trigger adaptive balancing strategies (such as adjusting charge / discharge curves and starting auxiliary cooling) to delay battery degradation.

[0096] 5. Tiered active defense mechanism:

[0097] A closed-loop control system is formed from early warning to response, and the thermal runaway suppression time is shortened to the 200ms level, meeting stringent safety standards such as UL9540A.

[0098] This application also provides an application scenario in which the above-described lithium battery thermal runaway monitoring method is applied. Specifically, the lithium battery thermal runaway monitoring method provided in this embodiment can be applied in a lithium battery thermal runaway monitoring scenario. This scenario includes a data acquisition stage (for acquiring time series of multi-sensor parameters of the target lithium battery), a state identification stage (for determining the thermal runaway state of the lithium battery based on the acquired data), and an early warning stage (for providing early warning based on the state identification results). The lithium battery thermal runaway monitoring method provided in this embodiment belongs to the state identification stage.

[0099] Based on the same inventive concept, this application also provides a lithium battery thermal runaway monitoring device for implementing the lithium battery thermal runaway monitoring method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the lithium battery thermal runaway monitoring device provided below can be found in the limitations of the lithium battery thermal runaway monitoring method described above, and will not be repeated here.

[0100] In one exemplary embodiment, such as Figure 6 As shown, a lithium battery thermal runaway monitoring device is provided, comprising:

[0101] The parameter acquisition module M1 is used to acquire the time series of multiple sensing parameters of the target lithium battery; the time series of multiple sensing parameters includes the battery temperature time series, the battery voltage time series, the gas concentration time series inside the battery time series, and the battery deformation time series.

[0102] The spatiotemporal feature extraction module M2 is used to input the time series of multiple sensing parameters into the monitoring model to obtain the spatiotemporal features of each sensing parameter; the monitoring model includes a convolutional neural network, an LSTM network and an attention mechanism module connected in series.

[0103] The thermal runaway state monitoring module M3 is used to monitor the thermal runaway state of lithium batteries based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold.

[0104] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores lithium-ion battery thermal runaway monitoring data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a lithium-ion battery thermal runaway monitoring method.

[0105] Those skilled in the art will understand that Figure 7 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0106] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0107] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0108] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0109] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0110] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0111] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for monitoring thermal runaway in lithium batteries, characterized in that, include: Acquire time series of multiple sensing parameters of the target lithium battery; The multi-sensor parameter time series includes battery temperature time series, battery voltage time series, battery internal gas concentration time series, and battery deformation time series; The time series of multiple sensing parameters are input into the monitoring model to obtain the spatiotemporal characteristics of each sensing parameter; the monitoring model includes a convolutional neural network, an LSTM network and an attention mechanism module connected in series. The thermal runaway state of the lithium battery is monitored based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold. Specifically, the time series data of multiple sensing parameters are input into the monitoring model to obtain the spatiotemporal characteristics of each sensing parameter, including: The battery temperature time series, battery voltage time series, battery internal gas concentration time series, and battery deformation time series are input into a convolutional neural network to obtain the spatial distribution characteristics of battery temperature, battery voltage, battery internal gas concentration, and battery deformation. The spatial distribution characteristics of battery temperature, battery voltage, gas concentration inside the battery, and battery deformation are input into the LSTM network to obtain the time-series characteristics of battery temperature, battery voltage, gas concentration inside the battery, and battery deformation. The battery temperature time series characteristics, battery voltage time series characteristics, battery internal gas concentration time series characteristics, and battery deformation time series characteristics are input into the attention mechanism module to obtain optimized battery temperature time series characteristics, optimized battery voltage time series characteristics, optimized battery internal gas concentration time series characteristics, and optimized battery deformation time series characteristics.

2. The lithium battery thermal runaway monitoring method according to claim 1, characterized in that, The thermal runaway state of lithium batteries is monitored based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold, specifically including: The thermal runaway state of the lithium battery is determined based on the comparison results of the optimized battery temperature time-series characteristics and the corresponding dynamic threshold, as well as the battery temperature spatial distribution characteristics. The thermal runaway state of the lithium battery is determined based on the comparison results between the optimized battery voltage time-series characteristics and the corresponding dynamic threshold, and the battery voltage spatial distribution characteristics. The thermal runaway state of the lithium battery is determined based on the comparison results of the optimized temporal characteristics of gas concentration in the battery with the corresponding threshold and the spatial distribution characteristics of gas concentration in the battery. The thermal runaway state of the lithium battery is determined based on the comparison results of the optimized battery deformation time sequence characteristics and the corresponding threshold, as well as the battery deformation spatial distribution characteristics.

3. The lithium battery thermal runaway monitoring method according to claim 2, characterized in that, The optimized battery temperature time-series characteristic is the battery temperature rise rate; the optimized battery voltage time-series characteristic is the voltage deviation; the optimized battery internal gas concentration time-series characteristic is the gas concentration change trend over time; and the optimized battery deformation time-series characteristic is the deformation change trend over time.

4. The lithium battery thermal runaway monitoring method according to claim 3, characterized in that, The expression for the dynamic threshold corresponding to the optimized battery voltage timing characteristics is: in, Indicates the voltage deviation threshold; Indicates the reference deviation threshold; : SOH influence coefficient; SOH represents the battery health status; The expression for the dynamic threshold corresponding to the optimized battery temperature time-series characteristics is: in, Indicates the battery temperature rise rate threshold; Indicates the reference temperature rise rate threshold; Indicates the influence coefficient of ambient temperature; Indicates the reference ambient temperature; The expression for the threshold corresponding to the time-series characteristics of gas concentration inside the battery after optimization is as follows: in, Indicates the gas concentration threshold; Indicates the background concentration of volatile organic compounds; Indicates the reference gas alarm threshold; Indicates the VOC background adjustment factor; The expression for the threshold corresponding to the optimized battery deformation time-series characteristics is: in, Indicates the deformation threshold; α represents the yield strain of the battery casing / encapsulation material; α represents the safety factor.

5. The lithium battery thermal runaway monitoring method according to claim 2, characterized in that, The lithium battery thermal runaway monitoring method further includes: providing early warning based on the lithium battery thermal runaway state; specifically: Local overheating areas are determined based on the spatial distribution characteristics of battery temperature. When the optimized battery temperature timing characteristics exceed the corresponding dynamic threshold, the warning area is determined by combining the identified local overheating area to trigger the warning. Determine the spatial distribution pattern of voltage based on its spatial distribution characteristics; When the optimized battery voltage timing characteristics exceed the corresponding dynamic threshold, the warning area is determined by combining the voltage spatial distribution pattern to trigger a warning. Determine the spatial distribution of gas concentration based on the spatial distribution characteristics of gas concentration inside the battery; When the time-series characteristics of gas concentration in the optimized battery exceed the corresponding threshold, the warning area is determined by combining the spatial distribution of gas concentration to trigger a warning. Determine the spatial distribution of deformation based on the spatial distribution characteristics of battery deformation; When the optimized battery deformation timing characteristics exceed the corresponding threshold, the warning area is determined by combining the battery deformation spatial distribution to trigger a warning.

6. A lithium battery thermal runaway monitoring device, characterized in that, include: The parameter acquisition module is used to acquire the time series of multiple sensing parameters of the target lithium battery; the time series of multiple sensing parameters includes the battery temperature time series, the battery voltage time series, the battery internal gas concentration time series, and the battery deformation time series. The spatiotemporal feature extraction module is used to input the time series of multiple sensing parameters into the monitoring model to obtain the spatiotemporal features of each sensing parameter; the monitoring model includes a convolutional neural network, an LSTM network and an attention mechanism module connected in series; Specifically, the time series data of multiple sensing parameters are input into the monitoring model to obtain the spatiotemporal characteristics of each sensing parameter, including: The battery temperature time series, battery voltage time series, battery internal gas concentration time series, and battery deformation time series are input into a convolutional neural network to obtain the spatial distribution characteristics of battery temperature, battery voltage, battery internal gas concentration, and battery deformation. The spatial distribution characteristics of battery temperature, battery voltage, gas concentration inside the battery, and battery deformation are input into the LSTM network to obtain the time-series characteristics of battery temperature, battery voltage, gas concentration inside the battery, and battery deformation. The battery temperature time series characteristics, battery voltage time series characteristics, battery internal gas concentration time series characteristics, and battery deformation time series characteristics are input into the attention mechanism module to obtain the optimized battery temperature time series characteristics, optimized battery voltage time series characteristics, optimized battery internal gas concentration time series characteristics, and optimized battery deformation time series characteristics. The thermal runaway state monitoring module is used to monitor the thermal runaway state of lithium batteries based on the spatiotemporal characteristics of each sensing parameter and the corresponding dynamic threshold.

7. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the lithium battery thermal runaway monitoring method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the lithium battery thermal runaway monitoring method according to any one of claims 1-5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the lithium battery thermal runaway monitoring method according to any one of claims 1-5.

Images