Battery module thermal runaway detection system and detection method

Abstract

The invention discloses a battery module thermal runaway detection method and a battery module thermal runaway detection system. The detection method comprises the following steps: acquiring a sound signal of a battery module; based on the sound signals, a first type of sound events related to opening of a battery pressure release valve and a second type of sound events related to electric arcs generated in a battery module are recognized, and the occurrence time and the occurrence position of the first type of sound events and the occurrence position of the second type of sound events are recorded; and based on the specific relationship between the occurrence time and the occurrence position of the first type of sound event and the second type of sound event, judging whether the first type of sound event and the second type of sound event are coupled at the same or adjacent position in a preset time window so as to determine the thermal runaway risk level of the battery module. By jointly judging the two key events of pressure release valve opening and arc generation, the thermal runaway state of the battery module with high combustion or explosion risk can be identified more accurately and more timely, and effective detection of the thermal runaway of the battery module is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of battery module thermal runaway detection, and in particular to a battery module thermal runaway detection system and detection method. Background Art

[0002] A battery module typically consists of a housing, battery cells, and a control unit, with the cells and control unit located within the housing. When a battery cell experiences thermal runaway, it can rupture its pressure relief valve, generating large amounts of gas and leaking electrolyte, potentially leading to fire or even explosion. Effective detection solutions for thermal runaway in battery modules are currently unavailable. Summary of the Invention

[0003] The purpose of this method is to overcome the above-mentioned defects or problems existing in the background technology and provide a battery module thermal runaway detection system and detection method, which can effectively detect battery module thermal runaway.

[0004] To achieve the above purpose, this method adopts the following technical solutions:

[0005] Technical Solution 1: A battery module thermal runaway detection method, which is used to determine the thermal runaway risk level of the battery module, comprising: obtaining a sound signal using at least one sound sensor arranged inside or near the battery module; identifying a first type of sound event related to the opening of a cell pressure relief valve in the battery module based on the sound signal, and recording its occurrence time and location; identifying a second type of sound event related to an electric arc generated in the battery module based on the sound signal, and recording its occurrence time and location; determining the thermal runaway risk level of the battery module based on the relationship between the occurrence time and location of the first type of sound event and the second type of sound event.

[0006] Based on at least one of the technical components, amplitude distribution and spectral shape of technical solution one, and comparing the extracted acoustic features with the preset acoustic feature data of the first type of sound event, if the two match, it is identified as the first type of sound event; the step of identifying the second type of sound event related to the electric arc generated inside the battery cell includes: extracting the acoustic features of the sound signal, the acoustic features including at least one of the frequency components, amplitude distribution and spectral shape, and comparing the extracted acoustic features with the preset acoustic feature data of the second type of sound event, if the two match, it is identified as the second type of sound event.

[0007] Technical Solution 3 based on Technical Solution 1: The step of determining the thermal runaway risk level of the battery module includes: judging whether they occur successively or simultaneously within a preset time window, and judging whether the occurrence location of the first type of sound event and the occurrence location of the second type of sound event are the same location or adjacent locations.

[0008] Technical Solution 4 based on Technical Solution 3: After identifying the occurrence of the first type of sound event, within the preset time window, the recognition sensitivity of the second type of sound event is improved for the location that is the same as or adjacent to the location where the first type of sound event occurs.

[0009] Technical Solution 5 based on Technical Solution 4: When the at least one sound sensor includes multiple sound sensors, the step of determining the occurrence location of the first type of sound event and / or the second type of sound event includes: using the arrival time difference of the sound signals collected by the multiple sound sensors to locate the sound source.

[0010] Technical Solution 6 based on Technical Solution 5: Before using the arrival time difference of multiple sound signals to locate the sound source, it also includes a step of locating the sound source to one of the multiple stacking areas pre-divided in the battery module by comparing the amplitudes of the sound signals collected by the multiple sound sensors; each of the stacking areas includes multiple battery cells.

[0011] Technical solution seven based on technical solution one: before or during identifying the first type of sound event and the second type of sound event based on the sound signal, it also includes a step of performing noise filtering processing on the acquired sound signal.

[0012] Technical Solution 8 based on Technical Solution 1: When the at least one sound sensor is controlled by multiple control units respectively to obtain the sound signal, the step of identifying the first type of sound event and the second type of sound event based on the sound signal includes: when at least one control unit identifies the occurrence of the first type of sound event and / or the second type of sound event, the other control units synchronously perform sound recognition analysis, and judge whether the first type of sound event and / or the second type of sound event occurs based on the sound recognition results of the control units occupying a predetermined proportion.

[0013] Technical Solution 9 based on Technical Solution 1: When identifying the occurrence of the first type of sound event and / or the second type of sound event, obtain the temperature data and electrical parameter data of the battery module, and when the temperature data shows a temperature change related to the opening of the battery cell pressure relief valve, confirm the occurrence of the first type of sound event, and when the electrical parameter data shows a change related to the generation of an arc, confirm the occurrence of the second type of sound event.

[0014] In addition, the present invention also provides technical solution ten: a battery module thermal runaway detection system, comprising: at least one sound sensor, arranged inside or near the battery module to be detected; and at least one control unit, communicatively connected to the sound sensor, and using the sound sensor to obtain a sound signal; the control unit is configured to be suitable for identifying, based on the sound signal, a first type of sound event related to the opening of the battery cell pressure relief valve in the battery module and a second type of sound event related to the generated arc, and recording their respective occurrence time and location, and determining the thermal runaway risk level of the battery module based on the relationship between the occurrence time and location of the first type of sound event and the second type of sound event.

[0015] From the above description of the present method, it can be seen that compared with the prior art, the present method has the following beneficial effects:

[0016] Technical Solution 1 provides a battery module thermal runaway detection method. This method utilizes at least one acoustic sensor positioned within or near the battery module to acquire an acoustic signal. Based on this acoustic signal, the method then identifies a first-type acoustic event associated with the opening of a battery cell explosion-proof valve and a second-type acoustic event associated with an arc generated within the battery module. The event records the time and location of each of these two events. Ultimately, the method determines the battery module's thermal runaway risk level based on the correlation between the recorded first-type and second-type acoustic events in terms of their occurrence time and location.

[0017] In the prior art, although detecting the sound of the explosion-proof valve opening of the battery cell alone can indicate that the internal pressure of the battery cell may increase abnormally due to thermal runaway, this event alone is not sufficient to fully assess the immediate fire risk. Because the opening of the explosion-proof valve mainly indicates the existence of internal pressure accumulation and potential leakage of electrolyte or combustible gas, which forms the material basis for combustion, but if there is no effective ignition source, the possibility of directly triggering large-scale combustion is relatively low. Similarly, detecting the arc sound generated inside the battery cell alone can indicate that there may be an electrical short circuit or other forms of abnormal discharge inside the battery cell, which constitutes a potential ignition source. However, if the sealing structure of the battery cell is intact at this time, the explosion-proof valve is not opened, and there is a lack of sufficient combustible materials to contact the arc, then the risk of a serious thermal runaway event (such as fire or explosion) directly caused by the arc is also relatively limited.

[0018] After in-depth analysis, the inventors determined that the risk of severe thermal runaway in a battery module, leading to combustion or explosion, depends largely on the simultaneous presence and interaction of combustible materials (such as leaked electrolyte or internally generated flammable gases, whose presence is often associated with the opening of explosion-proof valves) and effective ignition sources (such as electric arcs) in time and space. Therefore, when these two events exhibit a close correlation in time and space, such as when an arc is detected near or at the same location shortly after the explosion-proof valve is opened, it strongly indicates that the conditions for combustion are in place, and the risk of thermal runaway increases dramatically. Based on the above-mentioned method steps, this technical solution more accurately assesses the actual fire risk, effectively distinguishes between operating conditions where there is only pressure anomaly, only electrical fault, and the combination of the two to form a high-risk combustion condition, and avoids the risk misjudgment (over- or under-assessment) that may be caused by separate detection; and, by capturing the combination of these two key events, it can identify impending serious thermal runaway events earlier than single event monitoring, buying valuable time for taking preventive and control measures; in addition, the reliability of detection is improved, and by correlating and analyzing the acoustic representations of two different physical processes (pressure release and arc discharge), false alarms and missed alarms caused by the randomness or interference of a single signal source are reduced.

[0019] More importantly, this technical solution does not simply identify and distinguish the sound of the pressure relief valve opening and the sound of the arc generation separately, but rather organically combines the two, making them an integral and complete technical means for determining the thermal runaway risk level of the battery module. First, the identification of the first type of sound event, namely the opening of the battery cell explosion-proof valve, establishes the prerequisite for the leakage or accumulation of combustible materials (such as electrolyte vapor or decomposition products) within the battery cell. On this basis, the identification of the second type of sound event, namely the arc generation, closely couples and correlates the signs of combustible materials with the signs of ignition sources in time and space, thus forming a basis for accurately determining the thermal runaway risk. Without the effective identification of either event, and the lack of temporal and spatial correlation between the two events, it is impossible to accurately determine the specific thermal runaway state of the current battery module. This synergistic effect overcomes the one-sidedness of single-event detection, making risk detection no longer a simple superposition of isolated events, but a comprehensive judgment based on the logic of event development, thereby achieving a more comprehensive and in-depth detection of the battery module's safety status. Overall, this detection method can effectively detect and judge the thermal runaway of battery modules, determine the thermal runaway risk of battery modules early, and facilitate the determination of necessary measures based on actual conditions.

[0020] In Technical Solution 2, the method of identifying the first type of sound events and the second type of sound events is further defined. By utilizing multi-dimensional acoustic features, such as the energy distribution of sound signals in different frequency bands, the dynamic changes in amplitude, and the overall shape of the spectrum, and accurately comparing them with the pre-calibrated or learned acoustic feature data representing specific fault events, the target sound events (explosion-proof valve opening sound and arc sound) can be effectively distinguished from complex background noise and other possible interference sounds with similar simple time domain or frequency domain characteristics, thereby greatly reducing the probability of misidentification and missed identification, and ensuring the accuracy and reliability of sound event identification.

[0021] Technical Solution 3 further defines how to determine the level of thermal runaway risk based on the relationship between the time and location of the first and second type sound events. When these two events are closely correlated in time and space—for example, if an arc signal is detected at or very near the opening position of an explosion-proof valve very shortly after it opens—this indicates a high probability of combustible material and an ignition source intersecting, and the risk of thermal runaway has reached an extremely high level. This specific logical judgment effectively distinguishes operating conditions of varying degrees of danger, avoiding risk misjudgments caused by failing to accurately grasp the critical connection between the two events, thereby ensuring timely warnings of the most dangerous situations.

[0022] In the fourth technical solution, after the occurrence of the first type of sound event (the opening of the explosion-proof valve of the battery cell) is identified, within the preset time window, the recognition sensitivity of the second type of sound event (arc) is improved for the location that is the same as or adjacent to the location of the occurrence of the first type of sound event. Once the explosion-proof valve opening event, which is an important precursor to thermal runaway, is detected, the system will immediately focus the monitoring focus or the sensitivity of the analysis algorithm (for example, by lowering the recognition threshold, adopting a more sophisticated feature analysis method, or increasing the sampling frequency, etc.) on the confirmed potential fault area. This adaptive sensitivity adjustment mechanism can more effectively detect arc signals that may appear subsequently, with weaker signal strength or shorter duration, which may be ignored or delayed under conventional monitoring sensitivity. Therefore, this technical solution helps to detect the concurrence of two events indicating a sharp increase in the risk of fire earlier and more reliably, which gains valuable warning time for taking emergency response measures and further strengthens the advantages of collaborative detection.

[0023] Technical Solution 5 further defines a specific method for determining the location of a sound event. By utilizing the differences in the time required for sound signals to propagate to sensors at different known locations in space, the three-dimensional or two-dimensional coordinates of the sound source (i.e., the opening of an explosion-proof valve or the generation of an arc) can be relatively accurately calculated. The information on the location of the sound event obtained through this method is the basis for determining the level of thermal runaway risk in the aforementioned technical solution. Only by accurately locating the locations of the two events can we reliably determine whether they occurred at the same or adjacent locations, thereby accurately assessing the thermal runaway risk indicated by the synergistic effect of these two events.

[0024] In Technical Solution 6, before using the time difference of arrival (TDA) of sound signals for sound source localization, the sound source is first roughly localized to one of multiple pre-demarcated stacking areas within the battery module by comparing the amplitudes of the sound signals collected by multiple sound sensors. Each stacking area contains multiple battery cells. This initial regional localization step improves the efficiency of the overall sound source localization process, thereby supporting subsequent collaborative judgment based on precise positioning. The amplitude of sound signals typically decays with distance. Therefore, by comparing the signal strengths received by different sensors, the approximate area of ​​the sound source can be relatively quickly determined. This preliminary localization result effectively narrows the search range for subsequent precise calculations based on TDA, thereby reducing computational complexity. This improves the real-time performance of positioning, especially in environments with a large number of sensors or limited computing resources. Furthermore, in certain environments with poor signal quality or severe multipath interference, the amplitude-based preliminary regional judgment may provide a more reliable initial estimate or constraint for the TDA-based calculation method, thereby improving the success rate and ultimate accuracy of overall positioning and ensuring the reliability of collaborative judgment based on positional relationships.

[0025] In technical solution seven, the step of performing noise filtering on the acquired sound signal before or during the identification of the first type of sound event and the second type of sound event based on the sound signal is defined. By improving the quality of the input signal, the accuracy and reliability of the subsequent sound event identification are significantly guaranteed. The actual operating environment of the battery module often has various types of background noise, such as the running sound of the cooling fan, the electromagnetic noise generated by the operation of the power electronic device, and the mechanical vibration of the surrounding environment. These noise components will be mixed with the target sound signal (explosion-proof valve opening sound and arc sound) and may even drown out the weak target signal. By adopting an appropriate noise filtering algorithm in advance or during the identification process, these interfering noises can be effectively suppressed or removed, thereby significantly improving the signal-to-noise ratio of the target sound signal. This makes the subsequent acoustic feature extraction more accurate and the feature comparison more reliable, ultimately reducing the misjudgment or omission of the first or second type of sound events due to noise interference, thereby improving the performance and reliability of the entire collaborative detection method.

[0026] In Technical Solution 8, when at least one of the control units identifies a Class I or Class II sound event, the other control units will simultaneously perform sound recognition analysis and ultimately determine whether the event actually occurred based on the recognition results of the control units occupying a predetermined proportion. By introducing this collaborative analysis of multiple control units and decision logic based on specific proportion confirmation, it is possible to effectively avoid overall detection errors of Class I or Class II sound events caused by single sensor failures, strong transient interference in local areas, or accidental misjudgments in the algorithms of single control units, thereby improving the stable operation capability of key sound event recognition and the accuracy of the final judgment results.

[0027] Technical Solution 9 defines the steps of further acquiring the temperature data and electrical parameter data of the battery module when the first type of sound event and / or the second type of sound event is identified, and using these non-acoustic data to confirm the sound event. Although acoustic detection can capture the occurrence of specific events, it may be interfered with by similar sounds in some complex situations. By cross-checking the acoustic detection results with the temperature data that directly reflects the internal thermal state of the battery cell and the electrical parameters that directly reflect the working state of the electrical system, the interference of false sound signals caused by non-fault factors can be effectively eliminated.

[0028] Technical Solution 10 provides a battery module thermal runaway detection system. The control unit in this system can monitor the operating status of the battery module in real time. By simultaneously detecting and correlating the two key events of explosion-proof valve opening and arc generation, it overcomes the limitations of detecting any one event alone in accurately assessing the thermal runaway risk. This enables timely detection of potential high-risk thermal runaway states indicated by these two events, thereby ensuring the safe and stable operation of the battery cell energy storage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present method, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present method. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0030] Figure 1 This is a module schematic diagram of a battery module thermal runaway detection system according to Example 1 of the present invention;

[0031] Figure 2 Schematic diagram of the steps of the battery module thermal runaway detection method involved in Example 2 of the present invention. DETAILED DESCRIPTION

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present method in conjunction with the accompanying drawings. Obviously, the described embodiments are preferred embodiments of the present method and should not be considered as excluding other embodiments. Based on the embodiments of the present method, all other embodiments obtained by ordinary technicians in this field without inventive work are within the scope of protection of the present method.

[0033] In the claims, description and above-mentioned drawings of the present method, unless otherwise clearly defined, the use of terms such as "first", "second" or "third" is for the purpose of distinguishing different objects rather than for describing a specific order.

[0034] In the claims, description and drawings of the present method, when the terms "include", "have" and their variations are used, they are intended to mean "including but not limited to".

[0035] Example 1

[0036] Example 1 of the present invention relates to a battery module thermal runaway detection system, referring to Figure 1 The system is mainly used to accurately assess the thermal runaway risk level of the battery module by monitoring the sound events related to the opening of the battery pressure relief valve and the generation of arcs, and combining the temperature and electrical parameter information of the battery module.

[0037] The battery module thermal runaway detection system is used to perform thermal runaway risk detection on at least one battery module to be detected, and includes: at least one sound sensor, arranged inside or near the battery module to be detected; and at least one control unit, which is communicatively connected to the sound sensor and uses the sound sensor to obtain a sound signal; the control unit is configured to be suitable for identifying, based on the sound signal, a first type of sound event related to the opening of the battery cell pressure relief valve in the battery module and a second type of sound event related to the generated arc, and recording their respective occurrence time and location, and determining the thermal runaway risk level of the battery module based on the relationship between the occurrence time and location of the first type of sound event and the second type of sound event.

[0038] The battery module to be inspected can be a rechargeable battery module used in various energy storage devices or electric vehicles, such as a lithium-ion battery module. This battery module typically consists of multiple single cells connected in series and parallel, encapsulated within a module housing. Importantly, each or some of the single cells is equipped with a pressure relief valve. This pressure relief valve is a safety device, typically a mechanical pressure-relieving structure, such as a spring-loaded valve or a diaphragm with a preset rupture pressure. When internal pressure accumulates to a preset opening threshold due to abnormal operating conditions such as overheating, overcharging, or an internal short circuit within a single cell, the valve of the pressure relief valve is pushed open by the internal high-pressure gas or its diaphragm ruptures, forming a pressure relief channel. The high-pressure gas within the cell, along with any entrained electrolyte vapor, is ejected through this pressure relief channel into the battery module housing or directly to the outside of the module. The process of this pressure relief valve opening and ejecting gas is accompanied by a characteristic popping sound or a continuous high-pressure gas flow sound, which constitutes the physical sound source of the first type of sound event described in the present invention. In addition, arcs may also be generated inside the battery module due to electrical faults (such as internal micro-short circuits, poor connector contact, creepage caused by insulation aging, etc.), such as between single cells, between cell poles and connectors, or due to insulation damage. The arc discharge process usually produces a unique "buzzing", "crackling" or popping sound, which constitutes the physical sound source of the second type of sound event described in the present invention. The physical structure of the battery module, such as the arrangement of the battery cells, the material and structure of the module shell, the internal support parts, the cooling system layout, etc., all provide a specific physical environment basis for the subsequent optimization of the arrangement of sound sensors and the design of the sound source localization algorithm.

[0039] The sound sensor is used to collect the sound signals generated inside or near the battery module in real time. In this embodiment, the sound sensor preferably adopts a micro-electromechanical system (MEMS) microphone with high sensitivity, wide frequency response range, compact size, low power consumption and good consistency. MEMS microphones are adopted because they are easy to integrate inside the battery module with limited space and can effectively capture the subtle sound characteristics of the opening of the pressure relief valve and the generation of arcs. The arrangement and number of sound sensors are crucial to the performance of the detection system. The sensors should be arranged at key positions inside the battery module, or close to the inner wall of the module shell to ensure that the sound signals generated by any potential fault point inside the module (i.e., the possible opening position of the pressure relief valve or the position where the arc occurs) can be effectively collected. For example, for a typical rectangular battery module, it can be considered to arrange a sound sensor at each of the four corner points inside it and near the geometric center, thereby forming a three-dimensional array containing at least five sensors. More specific layout schemes can be optimized based on the size of the battery module, the density of the internal battery cells, and the structural characteristics. For example, some sensors can be installed on specially designed brackets that are closer to potential fault sources (such as areas where battery cell pressure relief valves are densely arranged or near high-voltage connectors) in order to obtain better signal pickup effects. A properly configured sound sensor array can not only improve the capture rate of weak sound events, but also provide the necessary multi-point acoustic information for the subsequent precise positioning of sound sources based on algorithms such as time difference of arrival (TDOA). Each sound sensor is connected to the control unit through wired methods such as shielded cables to minimize the impact of electromagnetic interference generated by the battery module during operation on the quality of sound signal transmission.

[0040] The control unit is the core processing component of the entire thermal runaway detection system, responsible for executing all or part of the detection method steps described in detail in the subsequent embodiments. In hardware implementation, the control unit can be one or more high-performance microcontrollers (MCUs), such as those based on the ARM Cortex-M series or higher-level Cortex-A series processors; or, to meet the needs of complex acoustic signal processing and real-time algorithm operations, a dedicated chip or system-on-chip (SoC) with an integrated digital signal processor (DSP). The control unit typically integrates the following main functional modules or connects them through an external bus: one or more central processing units (CPUs) or microprocessors (MPUs) for running the overall control logic, event judgment algorithms, and risk assessment models; one or more digital signal processors (DSPs) or dedicated hardware accelerators (e.g., processing units implemented in field-programmable gate arrays (FPGAs)) for efficiently executing computationally intensive acoustic signal processing algorithms, such as fast Fourier transforms (FFTs), digital filtering, and acoustic feature extraction; memory modules, including non-volatile memory (e.g., Flash memory) for storing the operating system, application firmware, a preset acoustic feature database, and historical event records, and volatile memory (e.g., RAM) for data caching and processing intermediate algorithm results during program runtime; dedicated audio interface circuits for connecting to the sound sensor array, such as the IS (Inter-IC Sound) bus interface, the PDM (Pulse Density Modulation) interface, and its corresponding codec (CODEC) or digital signal conversion circuit; and interface modules for communicating with the battery management system (BMS) and other external monitoring systems of the battery module, such as the CAN (Controller Area Network) interface. The control unit includes a network controller and transceiver, an Ethernet physical layer interface (PHY) and controller, and an RS interface. Furthermore, it includes a power management module that provides stable power to the entire control unit. In some large or modular battery energy storage systems, the control unit can also be implemented in a distributed manner. For example, multiple sub-control units work together, sharing acoustic data and preliminary processing results via an internal high-speed communication bus, and executing the collaborative confirmation logic described in subsequent embodiments.

[0041] The associated information acquisition function in this system mainly refers to the control unit establishing a communication connection with the battery management system (BMS) of the battery module and following a preset communication protocol (for example, a specific CAN message protocol or Modbus protocol) to obtain the temperature data and electrical parameter data of the battery module monitored and managed in real time by the BMS system. Specifically, the control unit can obtain real-time temperature data of key monitoring points inside the battery module from the BMS. These key monitoring points may include areas near the surface of the single cell, near the pole of the cell, near the pressure relief valve, and other areas inside the module that are prone to heat accumulation or sensitive to temperature changes. At the same time, the control unit can also obtain the macroscopic electrical parameters of the battery module from the BMS, such as total voltage, total current, insulation resistance status, etc., as well as microscopic electrical parameters, such as the voltage of each or a specific group of single cells, the voltage balance status of the battery string, etc. These non-acoustic data provided by the BMS are the basis for implementing the multimodal data fusion verification step in the subsequent method of the present invention. They can significantly improve the accuracy of judging battery pressure relief valve opening events and arc generation events, effectively reduce false alarms caused by misjudgment of pure acoustic signals, and provide more comprehensive information support for more accurate and reliable thermal runaway risk level assessment.

[0042] Example 2

[0043] Example 2 of the present invention relates to a method for detecting thermal runaway of a battery module, referring to Figure 2 , which is used to determine the thermal runaway risk level of the battery module. Obviously, this detection method is implemented based on the thermal runaway detection system described in Example 1.

[0044] Among them, the battery module thermal runaway detection method mainly includes the following steps: using at least one sound sensor arranged inside or near the battery module to obtain a sound signal; based on the sound signal, identifying a first type of sound event related to the opening of the battery cell pressure relief valve in the battery module, and recording its occurrence time and location; based on the sound signal, identifying a second type of sound event related to the electric arc generated in the battery module, and recording its occurrence time and location; based on the relationship between the occurrence time and location of the first type of sound event and the second type of sound event, determining the thermal runaway risk level of the battery module.

[0045] Among them, in order to realize the identification of sound events, the step of identifying the first type of sound event related to the opening of the battery cell pressure relief valve includes: extracting the acoustic characteristics of the sound signal, the acoustic characteristics include at least one of frequency components, amplitude distribution and spectral shape, and comparing the extracted acoustic characteristics with the preset acoustic characteristic data of the first type of sound event. If the two match, it is identified as the first type of sound event; the step of identifying the second type of sound event related to the electric arc generated inside the battery cell includes: extracting the acoustic characteristics of the sound signal, the acoustic characteristics include at least one of frequency components, amplitude distribution and spectral shape, and comparing the extracted acoustic characteristics with the preset acoustic characteristic data of the second type of sound event. If the two match, it is identified as the second type of sound event.

[0046] In order to determine the thermal runaway risk level of the battery module, the step of determining the thermal runaway risk level of the battery module includes: judging whether they occur successively or simultaneously within a preset time window, and judging whether the occurrence location of the first type of sound event and the occurrence location of the second type of sound event are the same location or adjacent locations.

[0047] In addition, in order to improve the detection efficiency of the second type of sound event, after the first type of sound event is identified, the recognition sensitivity of the second type of sound event is improved within the preset time window for the location that is the same as or adjacent to the location where the first type of sound event occurs.

[0048] In addition, in order to accurately locate the battery cell that has undergone thermal runaway, when the at least one sound sensor includes multiple sound sensors, the step of determining the occurrence location of the first type of sound event and / or the second type of sound event includes: using the arrival time difference of the sound signals collected by the multiple sound sensors to locate the sound source.

[0049] Furthermore, in order to improve the accuracy of locating the battery cell that has undergone thermal runaway, before using the arrival time difference of multiple sound signals to locate the sound source, it also includes a step of locating the sound source to one of the multiple stacking areas pre-divided in the battery module by comparing the amplitudes of the sound signals collected by the multiple sound sensors; each of the stacking areas includes multiple battery cells.

[0050] In addition, in order to reduce the impact of noise on sound recognition, before or during the identification of the first type of sound events and the second type of sound events based on the sound signal, a step of performing noise filtering on the acquired sound signal is also included.

[0051] At the same time, in order to improve the accuracy of sound event judgment, when the at least one sound sensor is controlled by multiple control units respectively to obtain the sound signal, the step of identifying the first type of sound event and the second type of sound event based on the sound signal includes: when at least one control unit identifies the occurrence of the first type of sound event and / or the second type of sound event, the other control units synchronously perform sound recognition analysis, and judge whether the first type of sound event and / or the second type of sound event occurs based on the sound recognition results of the control units occupying a predetermined proportion.

[0052] In order to further improve the accuracy of sound event judgment, when the first type of sound event and / or the second type of sound event are identified, the temperature data and electrical parameter data of the battery module are obtained, and when the temperature data shows a temperature change related to the opening of the battery cell pressure relief valve, the occurrence of the first type of sound event is confirmed, and when the electrical parameter data shows a change related to the generation of an electric arc, the occurrence of the second type of sound event is confirmed.

[0053] Specifically, the battery module thermal runaway detection method involved in this embodiment can be implemented according to the following steps:

[0054] First, the sound signal acquisition and preprocessing steps are performed.

[0055] The control unit continuously collects sound signals from the battery module or its surrounding environment through the sound sensor array connected to it at a preset sampling frequency (for example, it can be set to 48kHz to cover a sufficiently wide acoustic frequency band) and bit depth (for example, it can be set to 16 bits or 24 bits to ensure the dynamic range and accuracy of the signal). The collected analog sound signal is converted into a digital sound signal sequence inside the sound sensor or through a high-precision analog-to-digital converter (ADC) integrated in the control unit. Subsequently, the control unit performs necessary preprocessing on the original digital sound signal sequence, mainly noise filtering. The purpose of this step is to filter out background noise that is irrelevant to the target sound event as much as possible, such as the steady-state mechanical noise generated by the operation of the battery module itself (such as the low-frequency hum of the cooling fan), the high-frequency electromagnetic interference noise generated when the power electronic converter is working, and the sudden or continuous interference sound from the external environment, thereby improving the signal-to-noise ratio (SNR) of subsequent target sound events (i.e., the first type of sound event and the second type of sound event). In this embodiment, a digital bandpass filter can be applied. The passband frequency range of this bandpass filter should be determined based on experimental test data and an analysis of the acoustic characteristics of the pressure relief valve opening sound (typically containing abundant mid- and high-frequency explosive components and sustained airflow) and the arcing sound (typically manifesting as a broadband "sizzling," "crackling," or popping sound, potentially containing significant high-frequency components). For example, the filter passband can be set within the range of 500 Hz to 20 kHz. The filter type can be either a finite impulse response (FIR) or infinite impulse response (IIR) filter. Specific design parameters (such as the filter order and window function type) should be carefully considered and optimized based on the noise characteristics of the actual application scenario, the tolerance for signal distortion, and the computational resource constraints of the control unit. When designing the filter, special attention should be paid to ensuring that the transient impulse characteristics (such as the explosive front of the pressure relief valve opening) or the persistent weak energy characteristics (such as the weak discharge sound of the early arcing) unique to the target sound event are preserved to the greatest extent possible while effectively suppressing noise.

[0056] Next, the step of recognizing the sound event is performed.

[0057] The control unit divides the pre-processed sound signal stream into a series of short-time analysis frames with a certain overlap (for example, the length of each frame can be set to 20 milliseconds to 40 milliseconds, and the shift between frames, i.e., the frame shift, can be set to 10 milliseconds to 20 milliseconds to ensure a good capture capability of the dynamic changes of the signal and a smooth transition between frames). Then, for each short-time analysis frame, the control unit extracts one or more groups of acoustic feature parameters that can effectively distinguish different types of sound events. The selection of these acoustic feature parameters is intended to comprehensively capture and quantify the energy distribution and persistence characteristics of the sound event in the time domain, as well as the key discriminant information such as the pitch and timbre composition in the frequency domain. In this embodiment, the extracted acoustic features may include but are not limited to the following categories:

[0058] In terms of time domain characteristics, for example, the short-time energy or root mean square (RMS) amplitude of each frame signal is calculated. This feature can directly reflect the intensity of the sound signal. The moment the pressure relief valve opens is usually accompanied by a higher energy peak; the zero crossing rate (ZCR) of each frame signal is calculated, that is, the number of times the signal sample sequence passes through the zero value. The zero crossing rate can reflect the level of the main frequency components of the signal to a certain extent; and by analyzing the energy change trend of multiple consecutive frames (for example, the number of consecutive frames whose energy values ​​continue to exceed a preset dynamic threshold), the start and end times of the detected sound event are determined, thereby estimating the duration of the sound event. The sound of the pressure relief valve opening is usually manifested as an instantaneous explosion followed by a relatively short period of airflow sound, while the arc sound may be manifested as a relatively long-lasting and irregular energy discharge sound.

[0059] In terms of frequency domain features, a fast Fourier transform (FFT) or short-time Fourier transform (STFT) is first performed on each short-time analysis frame to obtain a spectral representation of the frame signal, thereby analyzing the distribution of its energy at different frequency points or frequency bands. The main frequency components in the spectrum can then be extracted, that is, the frequencies and amplitudes corresponding to several energy peaks in the spectrogram are identified. The spectral centroid is calculated, which is the "weighted average" frequency of the spectral energy distribution and can measure the "brightness" or "sharpness" of the sound. The spectral spread is calculated, which measures the width of the spectral energy distribution around the spectral centroid and reflects the purity or complexity of the sound. Mel-frequency cepstral coefficients (MFCCs) are finally extracted. These are feature parameters widely used in speech recognition and sound event recognition. They are calculated based on the human ear's nonlinear perception of sound frequency (Mel-frequency scale) and can very effectively characterize the timbre characteristics of the sound. 12- to 20-dimensional MFCCs and their first- and second-order difference coefficients are usually extracted to include dynamic information.

[0060] To accurately identify target sound events, the control unit pre-builds and stores an acoustic feature database. The database construction process is as follows: First, in a controlled experimental environment, a large number of authentic and effective sound samples are recorded using high-fidelity recording equipment (such as professional acoustic sensors and data acquisition cards) for single cells or battery modules of different models, different states of health (SOH), and different states of charge (SOC), simulating various operating conditions that may cause the pressure relief valve to open (for example, through overcharging, over-discharging, puncture, local heating, etc.) or arcing (for example, simulating air breakdown caused by internal micro-shorts, loose external connectors, or damaged insulation materials). When constructing the feature library, special attention should be paid to ensuring the diversity and representativeness of the collected samples, covering as many target sound events as possible across different battery models, failure modes, aging levels, and ambient temperature and background noise conditions. This improves the robustness of the subsequent recognition model and its generalization ability under complex real-world operating conditions. Then, these carefully collected and labeled sound samples are subjected to the same preprocessing and acoustic feature extraction operations as in the real-time detection process to form standardized feature vectors. Finally, the feature vector set related to the "first type of sound event (i.e., the sound of the pressure relief valve opening)" (which should include the sound features of the pressure relief valve opening under various working conditions and intensities) is stored as the "first type of sound event feature library"; the feature vector set related to the "second type of sound event (i.e., the sound of arc generation)" (which should include the sound features of arcs of various types, intensities, and durations) is stored as the "second type of sound event feature library". These feature libraries can be solidified and stored in the non-volatile memory of the control unit, such as a Flash memory chip.

[0061] During real-time detection, the control unit matches and compares the acoustic feature vectors of each frame or a sound segment consisting of multiple consecutive frames extracted from the current sound signal with reference feature vectors in the "First Category Sound Event Feature Library" and "Second Category Sound Event Feature Library" preset in the feature database. This matching and comparison process can be implemented using a variety of algorithms. For example, traditional pattern recognition methods based on distance or similarity metrics can be used, such as calculating the Euclidean distance, Mahalanobis distance, or cosine similarity between the real-time feature vector and each reference feature vector in the feature library. When the minimum distance is less than a certain threshold or the maximum similarity is greater than a certain threshold, the corresponding event type is determined. For sequence features such as MFCCs, the dynamic time warping (DTW) algorithm can be used to compare the overall similarity between two sound event sequences of potentially different lengths. This algorithm can effectively handle the natural scaling of sound events on the time scale. Furthermore, advanced machine learning algorithms can be used to make more intelligent classification decisions. For example, shallow learning models such as support vector machines (SVM), K-nearest neighbor (KNN) algorithms, decision trees, and random forests, or deep learning models such as deep neural networks (DNN), convolutional neural networks (CNN), recurrent neural networks (RNN) and their variants (such as LSTM, GRU) can be used to build sound event classifiers. These machine learning models require sufficient offline training using a large amount of previously collected and carefully labeled sound sample data. After the training is completed and its performance is verified, the trained model parameters are deployed to the control unit for online real-time recognition. When the acoustic features extracted in real time are input into the classification model, if the discrimination probability (or confidence score) of a certain type of sound event output by the model exceeds a pre-set first discrimination threshold (for example, the recognition threshold T_event1 for the first type of sound event, and the recognition threshold T_event2 for the second type of sound event), it is preliminarily determined that the corresponding sound event (i.e., the first type of sound event or the second type of sound event) has occurred inside the battery module. The specific values ​​of these discrimination thresholds need to be carefully adjusted and optimized based on a large amount of experimental test data and the balance requirements of the false alarm rate (FAR) and missed detection rate (MDR) in actual application scenarios.

[0062] After identifying a first or second type of sound event, the sound event location step is then performed to determine the precise location of the sound event within the battery module. In this embodiment, to ensure both real-time positioning and accuracy, a two-step approach is employed: initial coarse regional positioning followed by precise point source location.

[0063] Preliminary area localization is triggered once the first type of sound event has been identified in the previous step. The localization principle exploits the fundamental physical property that sound energy attenuates with distance as it propagates through a medium. This means that sound sensors closer to the sound source typically receive sound signals of greater intensity (amplitude or energy). Specifically, this is accomplished by comparing the amplitudes of the sound signals received by each sensor in the sound sensor array during the time period of the recently identified sound event. For example, the peak amplitude or root mean square (RMS) energy of the signal received by each sensor during the sound event can be calculated. Typically, the sound sensor receiving the highest signal amplitude, or a few sound sensors with significantly higher signal amplitudes than the others, is preliminarily determined as the approximate area where the sound source event occurred, within a pre-divided logical stacking area or submodule within the battery module (for example, the entire battery module can be pre-divided into a number of three-dimensional grid units or logical sub-areas along its length, width, and height, with each area roughly corresponding to a number of battery cells). The role of this coarse positioning step is to quickly narrow the search range of the subsequent precise positioning algorithm, thereby effectively reducing the computational complexity of the precise positioning process and significantly improving the real-time performance of the overall positioning process.

[0064] After the approximate area of ​​the sound source has been determined through coarse localization, or when multiple acoustic sensors simultaneously detect a significant event signal, the precise localization process begins. Precise localization utilizes the slight differences in the time it takes for an acoustic signal to propagate from the source to sensors at different known locations in space, known as the Time Difference of Arrival (TDOA). The specific implementation method involves first accurately calculating the exact time difference between the arrival of the identified acoustic event (e.g., the initial shock wave front of a pressure relief valve opening, or a pulse segment with significant energy or unique waveform characteristics in arcing sound) at at least three different acoustic sensors surrounding the coarse localization area (or within the entire module). This typically requires highly precise time synchronization between the acoustic sensors in the array. For example, the system needs to ensure that the timestamps used by each acoustic sensor when collecting data have sub-millisecond or even microsecond synchronization accuracy. This high-precision time synchronization can be achieved through methods such as the Network Time Protocol (NTP), the Precision Time Protocol (PTP), or dedicated hardware synchronization signals (e.g., sharing a high-precision clock source or receiving a synchronization pulse signal). Time difference extraction can be achieved using advanced signal processing techniques, such as the Generalized Cross-Correlation (GCC) algorithm, particularly the Phase Transform (GCC-PHAT) algorithm, which is robust to noise and reverberation and can accurately estimate the delay between target signals received by different sensors. Then, combining the three-dimensional spatial coordinates of each sound sensor, which are precisely calibrated in the battery module coordinate system, a multi-point localization algorithm (also often called hyperbolic localization or multilateration localization) is used to calculate the three-dimensional coordinates (x, y, z) of the sound source. Commonly used TDOA localization algorithms include the Chan algorithm, the Chan-Ho algorithm, and the Taylor series expansion method. When using, for example, four or more sound sensors for localization calculations, multiple independent localization calculations can be performed by selecting different sensor subsets (for example, selecting M sensors from N available sensors to form a localization subarray, where M is typically 3 or 4). This may result in multiple slightly different candidate sound source locations due to measurement errors, noise interference, or model simplification. To obtain a final, more accurate, and reliable sound source location estimate from these candidate locations, the least squares method or other advanced optimization algorithms (such as weighted least squares and Kalman filtering) can be used to perform data fusion processing on these candidate locations. For example, the least squares method iteratively solves for the optimal sound source location by minimizing the mean square error between the actual measured TDOA value and the theoretical TDOA value calculated by reverse calculation based on the current estimated sound source location.After positioning is completed, the system will record in detail the type of sound event identified (first or second category), the precise timestamp of the event, and the three-dimensional coordinates of the location of the event inside the battery module calculated by the precise positioning algorithm.

[0065] The method then performs a step of determining the thermal runaway risk level.

[0066] The control unit comprehensively determines the current thermal runaway risk level of the battery module based on the type of sound event identified in the previous step and the determined time and location of the event. In this embodiment, in order to distinguish and manage different degrees of risks, the thermal runaway risk level can be divided into multiple different levels. For example, it can be set as: "Safety" (Level 0), "Attention / Pressure Relief Warning" (Level 1), "Arc Warning" (Level 2), and "Danger / High Risk" (Level 3). It should be emphasized that the division of these levels, the number and their corresponding specific names are only illustrative in this embodiment. In actual applications, the number, definition and degree of refinement of the risk levels can be flexibly adjusted according to the specific battery type, the safety requirements of the application scenario and the relevant industry standards or regulatory requirements.

[0067] In the initial state or during normal operation of the system, if no valid first-category sound event or second-category sound event is detected, the thermal runaway risk level of the battery module is judged to be "safe" (level 0).

[0068] When the system only detects the first type of sound event, that is, the pressure relief valve of one or more batteries is determined to be open, the control unit will raise the thermal runaway risk level of the battery module to "Attention / Pressure Relief Warning" (Level 1). At this time, the system should immediately record the time and location of the occurrence of the first type of sound event (the positioning result can be accurate to the specific single cell number or the physical space area where it is located). At the same time, the control unit should send a warning message to the battery management system (BMS) or a higher-level monitoring system through its communication interface. The information should at least include the event type (pressure relief valve opening), occurrence time, occurrence location and current risk level (Level 1). This warning is intended to remind operation and maintenance personnel or related systems that one or more cells inside the battery module may have activated pressure relief protection due to excessive internal pressure. There is a potential initial risk of electrolyte leakage or accumulation of flammable gas inside the module. It is recommended to pay close attention, further inspection or necessary maintenance operations.

[0069] When the system only detects the second type of sound event, that is, an arc is determined to have occurred inside the battery module, and no first type of sound event (i.e., pressure relief valve opening event) is detected within a reasonable time window before or in the vicinity of the arc, the control unit will raise the thermal runaway risk level of the battery module to "arc warning" (level 2). The system should also record the time and location of the occurrence of the second type of sound event in detail. The control unit should also send corresponding warning information to the BMS or upper-level monitoring system to indicate that there may be active electrical fault sources inside the battery module, such as internal micro-short circuits, poor connector contact, or damage to insulating materials, which constitute a potential ignition source risk and require vigilance and investigation.

[0070] The determination of a high-risk state is the most critical logical link in this method. Its purpose is to accurately identify whether the battery module is in a highly dangerous state where severe thermal runaway is about to occur (for example, about to cause open flame or explosion). The core basis for this determination is the close temporal and spatial coupling of the first type of sound event (pressure relief valve opening) and the second type of sound event (arcing). The specific judgment process is as follows: When the control unit detects the first type of sound event (i.e., any pressure relief valve is determined to be open), the system immediately enters a highly alert monitoring state. Specifically, the system will immediately start a preset "dangerous coupling judgment time window" based on the time of occurrence of the first type of sound event. The length of this time window needs to be scientifically set based on a variety of factors such as the specific battery type, the flammable and explosive properties of the electrolyte, the development speed of the thermal runaway process, etc., through a large amount of experimental test data and safety redundancy considerations. For example, it can be set to between 1 second and 10 seconds. In this embodiment, we take 5 seconds as an example. During the duration of this preset time window, the system will adaptively and specifically improve the recognition sensitivity of the second type of sound event (i.e., electric arc), especially for the location of the first type of sound event that has just occurred and its adjacent physical area (for example, it can be defined as a spherical space with a radius of 5 cm to 20 cm centered on the point where the first type of event occurred, or, based on the actual size of the battery cell and the physical layout structure inside the battery module, it can be defined as the area where the same single cell where the pressure release occurred or a few single cells directly adjacent to it are located). Specific technical means to improve the recognition sensitivity may include but are not limited to: temporarily and moderately lowering the recognition threshold used to determine the second type of sound event; or, if computing resources permit, calling a set of acoustic feature extraction algorithms or more sophisticated machine learning models with higher computational complexity but potentially better recognition performance to specifically analyze the sound signals from the key area; or, at the system level, increasing the sampling and analysis frequency of the sound sensor signals from the specific area, or dynamically allocating more processor computing resources to the second type of sound event recognition task for the area, so as to conduct more intensive key monitoring. If within the above-mentioned preset 5-second time window, and at the same physical location as the first type of sound event that has occurred or judged to be adjacent according to the preset proximity threshold (this judgment is made by comparing the three-dimensional coordinates of the occurrence locations of the two events recorded separately in the previous step to determine whether the spatial distance between them is less than the preset proximity threshold, or whether they clearly belong to the same single cell or a logical unit area composed of a few adjacent cells), the control unit detects the second type of sound event (that is, the arc is identified as being ongoing or recently occurring).If all the above conditions (within the time window, spatial proximity, and occurrence of the second type of event) are met, the system determines that the battery module already has both "combustible materials" (because the opening of the pressure relief valve usually means that flammable electrolyte vapor or combustible gases produced by battery thermal decomposition have leaked or are accumulating inside the module, forming the material basis for combustion or explosion) and "effective ignition source" (because the generation of the electric arc directly provides an energy source that can ignite these combustible materials) inside the battery module. These two most critical elements that cause combustion or explosion, and these two elements show a highly dangerous tight coupling relationship in time and space. At this point, it can be considered that the thermal runaway risk of the battery module has risen sharply to a critical state. Therefore, the system immediately determines the thermal runaway risk level of the battery module as the highest "dangerous / high risk" (level 3). By accurately judging the tight coupling relationship between the first and second types of sound events in time and space, this method can effectively screen out those extremely critical situations with urgent fire or explosion risks from the numerous possible battery abnormality indication signals, thereby avoiding the misjudgment that may result from relying solely on the monitoring of any one type of event (for example, when the pressure relief valve is only open but there is no arc, or when there is only a weak arc but the battery pack is sealed intact and no large amount of combustible materials have been leaked, the immediate risk of these situations may be relatively low), significantly reducing the system's false alarm rate and missed alarm rate, gaining valuable warning time for taking timely and effective emergency response measures, and greatly improving the accuracy of risk decision-making.

[0071] According to the final determined thermal runaway risk level, the system will execute the corresponding action response and conduct standardized information reporting. For "Attention / Pressure Relief Warning" (Level 1) and "Arc Warning" (Level 2), in addition to detailed event records (including event type, time, location, relevant acoustic characteristics, etc.), the system mainly sends warning information to the BMS or upper-level monitoring system through the communication interface, and may provide clear prompts on the local or remote user operation interface, suggesting that relevant operation and maintenance personnel pay close attention to the battery module, arrange subsequent detailed inspections, necessary maintenance operations or further fault diagnosis. For extreme situations judged to be "dangerous / high risk" (Level 3), the system should immediately trigger the highest level of alarm signal, for example, through the connected sound and light alarm device (such as sirens, warning lights) to issue an emergency alarm, and display the current serious alarm information in the most prominent way (such as red highlight, pop-up warning, etc.) on the main interface of the monitoring system. In terms of optional linkage control, when the level of danger / high risk is determined, the control unit can send linkage control instructions to other relevant safety execution units such as the BMS, the energy management system (EMS) of the energy storage system, and the fire control system through its communication interface according to the preset and strictly verified safety strategy, so as to automatically or semi-automatically execute a series of emergency safety measures. These emergency measures may include: requesting the BMS to immediately disconnect the electrical connection of the faulty battery module or the battery circuit in which it is located to prevent further expansion of the fault; activating the automatic fire extinguishing device at the battery module level or battery cluster level (for example, precisely spraying inert gas fire extinguishing agent or coolant); turning on or increasing the operating power of the forced ventilation and cooling system of the battery compartment or energy storage container to quickly disperse the possible accumulation of combustible gas and strive to cool the battery module, etc. Regardless of the level of warning triggered (especially for risk levels 1, 2, and 3 indicating abnormalities), the control unit will report a data packet containing detailed event information and the final risk assessment results in real time to the BMS, a higher-level supervisory control and data acquisition (SCADA) system, a cloud analysis platform, or a designated operation and maintenance management user terminal through its communication interface. The reported information should at least include: the type of sound event identified (type 1, type 2, or both), the exact timestamp of each relevant sound event, the three-dimensional coordinates of the location of each sound event inside the battery module calculated by the positioning algorithm, the thermal runaway risk level finally determined by the system, and optional relevant sensor raw data summaries or extracted key acoustic feature data. This comprehensive and timely information is of vital importance for subsequent accurate fault diagnosis, accident cause tracing, continuous optimization of safe operation strategies, and, when necessary, human intervention for higher-level emergency response.

[0072] In order to further improve the overall accuracy of the detection results and the reliability of the system operation, the method of the present invention may also selectively include the step of using non-acoustic data to perform multimodal information fusion verification, and the step of performing multi-control unit collaborative confirmation when adopting a distributed control system architecture.

[0073] The triggering timing of the multimodal verification step based on non-acoustic data can be after the preliminary identification of the first or second type of sound events in the previous step, or during or after the determination of the risk level, to start this verification process as a confirmation, evidence or auxiliary judgment method for the detection results based solely on acoustic signals. The control unit actively requests to obtain the real-time or historical temperature data and electrical parameter data of the battery module related to the physical location and time point (or a preset short time window before and after) of the acoustic event through the communication interface with the BMS. The verification logic for the first type of sound event (i.e., pressure relief valve opening) is: check the data reported by the temperature sensor of the single cell or its adjacent position where the pressure relief valve opening sound event occurs. If, in the time period surrounding the sound event, the temperature at the corresponding location exhibits a characteristic pattern of change associated with the opening of the pressure relief valve, such as a sustained rapid temperature rise due to the intensified internal chemical reaction at the onset of thermal runaway, or a brief, sharp drop in local temperature due to the Joule-Thomson effect when high-pressure gas is ejected from the pressure relief valve, followed by a rapid recovery due to the failure to alleviate the internal problem (or the formation of a significant, abnormal temperature gradient relative to surrounding cells that are not experiencing an abnormality), then this temperature change with clear physical significance can serve as strong evidence that the pressure relief valve has indeed opened, significantly increasing the control unit's confidence in the first-category sound event. The verification logic for the second-category sound event (i.e., arcing) involves checking the battery module's bus voltage and total current, or more specifically, checking the voltage of the cell (or cell string) inferred to be the source of the arc based on the sound source localization results, for violent, abnormal fluctuations or irregular high-frequency noise components associated with arcing. For example, arcing can often cause a transient drop in local circuit voltage, a sharp, pulsed increase in current, or unstable oscillations. If such abnormal characteristics of electrical parameters with clear indications are observed at the same time as or slightly earlier than the occurrence of the acoustic event, they can serve as evidence for the judgment that an arc has indeed occurred, thereby enhancing the confidence of the control unit in the judgment of the second type of sound event. If the result of the cross-validation is successful (that is, the non-acoustic data and the acoustic detection results support each other and are logically consistent), the validity of the sound event can be confirmed, and its weight or confidence level can be increased in the subsequent risk assessment model. On the contrary, if the results of the cross-validation are inconsistent, or the relevant non-acoustic data do not show obvious corresponding abnormalities, the confidence of the judgment of the acoustic event can be appropriately reduced, or it can be marked as a suspected interference event (for example, it may be caused by sudden external strong noise interference that is not related to battery failure), which helps to effectively avoid false alarms or unnecessary system responses caused by simple misjudgment of acoustic signals.This multimodal information fusion verification mechanism fully utilizes the complementarity between different physical quantity sensors in information sources and fault characterization, and can significantly enhance the overall accuracy and anti-interference ability of the entire detection system in judging battery abnormal events under complex actual working conditions.

[0074] The collaborative confirmation step based on multiple control units is mainly applicable to the specific situation where the detection system of the present invention adopts a distributed control system architecture. For example, in a large battery energy storage power station or an electric vehicle battery system composed of multiple independent battery packs, it may contain multiple battery clusters, and each battery cluster or several sub-modules under it may be equipped with its own independent battery management unit (BMU) or dedicated sound monitoring and processing sub-unit, which are responsible for the signal collection and preliminary event identification and positioning of a part of the sound sensors within their jurisdiction. Under this distributed architecture, the workflow of the collaborative confirmation mechanism is as follows: When at least one of the sub-control units (for example, we call it control unit A) preliminarily identifies a first-class or second-class sound event based on the sound signal within its monitoring range, it will immediately broadcast this preliminary discovery (which should include key information such as the type of event, the preliminary determined time of occurrence, the approximate physical location, the confidence level of the identification, etc.) to other relevant sub-control units in the system (for example, control unit B, control unit C, etc.) through the system's internal communication network (for example, a high-speed CAN bus or industrial Ethernet). Other sub-control units that receive the broadcast information will immediately or according to the preset response logic conduct targeted analysis of the sound signals in the areas they are responsible for monitoring (for example, checking whether similar acoustic signal characteristics appear within their own monitoring range at almost the same time point as the broadcast event), or retrieve their locally cached recent historical sound data for precise retrospective comparison to confirm whether signs of sound events that are consistent with or logically related to the broadcast event have been independently observed within their own monitoring range. The final decision logic is: only when more than a preset proportion (for example, two-thirds of the total number of control units participating in the collaboration in the system, or other specific proportions or numbers set according to the system topology and reliability redundancy design requirements) of control units, or multiple control units responsible for critical physical areas (for example, areas that are physically adjacent to the area where the event was initially reported or are closely related to the fault propagation path), jointly confirm or independently identify the (type of) sound event at a similar time (for example, the time difference between the event occurrence timestamps reported by each of them is within the preset system synchronization error allowable threshold) and a similar physical location (for example, the spatial distance between the event occurrence points located by each of them is within the preset collaborative positioning tolerance threshold, or can be logically attributed to the same potential fault source or fault impact area), the entire system will regard this event as a highly valid event that has been collaboratively confirmed by multiple parties, and use it for subsequent global thermal runaway risk level judgment and corresponding emergency response decisions.This decision-making mechanism, based on collaborative analysis across multiple control units and confirmation based on specific ratios (or specific logical rules), effectively improves the fault tolerance and overall robustness of the entire detection system to potential temporary failures of individual sound sensors, occasional strong transient electromagnetic or acoustic interference in a local area, or accidental misjudgments by the algorithms within a single sub-control unit. Through multi-point information collection, independent analysis from multiple angles, and cross-validation of results, it significantly enhances the overall stability of critical sound event identification and the accuracy and credibility of the final risk assessment results.

[0075] Embodiments of the present invention relate to a method for detecting thermal runaway in a battery module. This method utilizes at least one acoustic sensor positioned within or near the battery module to acquire acoustic signals. Based on these acoustic signals, the method then identifies a first-type acoustic event associated with the opening of a battery cell explosion-proof valve and a second-type acoustic event associated with an arc generated within the battery module. The time and location of each event are recorded. Ultimately, the method determines the thermal runaway risk level of the battery module based on the correlation between the time and location of the recorded first and second-type acoustic events. In the prior art, while detecting the sound of an explosion-proof valve opening alone can indicate an abnormal increase in internal pressure due to thermal runaway, this event alone is insufficient to fully assess the immediate fire risk. This is because an explosion-proof valve opening primarily indicates internal pressure buildup and potential leakage of electrolyte or combustible gas, which creates the catalyst for combustion. However, without an effective ignition source, the likelihood of a direct, large-scale combustion is relatively low. Similarly, detecting the sound of an arc generated within a battery cell alone can indicate the presence of an electrical short circuit or other form of abnormal discharge within the cell, potentially igniting an ignition source. However, if the sealing structure of the battery cell is intact at this time, the explosion-proof valve is not opened, and there is a lack of sufficient combustibles to contact the arc, then the risk of a serious thermal runaway event (such as fire or explosion) directly caused by the arc is relatively limited. After in-depth analysis, the inventor found that the risk of a serious thermal runaway of the battery module and causing combustion or explosion depends to a large extent on the simultaneous existence and interaction of combustibles (such as leaked electrolyte or internally generated combustible gas, whose appearance is often related to the opening of the explosion-proof valve) and effective ignition sources (such as electric arcs) in time and space. Therefore, when these two events show a close correlation in time and space, for example, when an arc is detected near or at the same location within a short period of time after the explosion-proof valve is opened, it strongly indicates that the combustion conditions are already in place and the risk of thermal runaway increases sharply. Based on the above-mentioned method steps, this technical solution more accurately assesses the actual fire risk, effectively distinguishes between operating conditions where there is only pressure anomaly, only electrical fault, and both together to form high-risk combustion conditions, and avoids the risk misjudgment (over- or under-assessment) that may be caused by separate detection; and, by capturing the combination of these two key events, it can identify impending serious thermal runaway events earlier than single event monitoring, buying valuable time for taking preventive and control measures; in addition, it improves the reliability of detection, and by correlating and analyzing the acoustic characterizations of two different physical processes (pressure release and arc discharge), it reduces false alarms and missed alarms caused by the randomness or interference of a single signal source. Among them, more importantly, this technical solution does not simply identify and distinguish the sound of the pressure relief valve opening and the sound of the arc generation separately, but organically combines the two, making them an inseparable and complete technical means in the process of determining the thermal runaway risk level of the battery module.Among them, the identification of the first type of sound event, namely the opening of the battery cell explosion-proof valve, establishes the prerequisite that there may be leakage or accumulation of combustible materials (such as electrolyte vapor or decomposition products) inside a battery cell. On this basis, the identification of the second type of sound event, namely the generation of an electric arc, closely couples and associates the signs of combustible materials with the signs of ignition sources in time and space, thus forming a basis for accurately judging the risk of thermal runaway. Without the effective identification of any event and the lack of temporal and spatial correlation between the two events, it is impossible to accurately judge the specific thermal runaway state of the current battery module. This synergistic effect overcomes the one-sidedness of single event detection, making risk detection no longer a simple superposition of isolated events, but a comprehensive judgment based on the logic of event development, thereby achieving a more comprehensive and in-depth detection of the safety status of the battery module. Overall, this detection method can effectively detect and judge the thermal runaway of the battery module, and can determine the thermal runaway risk of the battery module early, facilitating the determination of the necessary measures based on the actual situation.

[0076] The descriptions in the above specification and embodiments are intended to explain the scope of protection of the present method, but do not constitute a limitation on the scope of protection of the present method. Modifications, equivalent substitutions, or other improvements to the embodiments of the present method or portions of the technical features thereof that can be obtained by a person of ordinary skill in the art through logical analysis, reasoning, or limited experimentation based on the teachings of the present method or the above embodiments, combined with common knowledge, ordinary technical knowledge in the field, and / or existing technology, should be included in the scope of protection of the present method.

Claims

1. A battery module thermal runaway detection method, which is used to determine the thermal runaway risk level of a battery module, characterized in that: include: Acquiring a sound signal using at least one sound sensor disposed inside or near the battery module; Based on the sound signal, identifying a first type of sound event related to the opening of a cell pressure relief valve in the battery module, and recording the time and location of the occurrence of the first type of sound event; identifying, based on the sound signal, a second type of sound event associated with an arc generated in the battery module, and recording the time and location of occurrence of the second type of sound event; Based on the relationship between the occurrence time and the occurrence location of the first type of sound event and the second type of sound event, the thermal runaway risk level of the battery module is determined.

2. A battery module thermal runaway detection method according to claim 1, characterized in that: The step of identifying the first type of sound event related to the opening of the battery cell pressure relief valve includes: extracting the acoustic characteristics of the sound signal, the acoustic characteristics including at least one of frequency components, amplitude distribution and spectral shape, and comparing the extracted acoustic characteristics with the preset acoustic characteristic data of the first type of sound event. If the two match, it is identified as the first type of sound event; the step of identifying the second type of sound event related to the electric arc generated inside the battery cell includes: extracting the acoustic characteristics of the sound signal, the acoustic characteristics including at least one of frequency components, amplitude distribution and spectral shape, and comparing the extracted acoustic characteristics with the preset acoustic characteristic data of the second type of sound event. If the two match, it is identified as the second type of sound event.

3. A battery module thermal runaway detection method according to claim 1, characterized in that: The step of determining the thermal runaway risk level of the battery module includes: determining whether the first type of sound event occurs successively or simultaneously within a preset time window, and determining whether the occurrence location of the first type of sound event and the occurrence location of the second type of sound event are the same location or adjacent locations.

4. A battery module thermal runaway detection method as claimed in claim 3, characterized in that After the occurrence of the first type of sound event is identified, within the preset time window, the recognition sensitivity of the second type of sound event is increased for the location that is the same as or adjacent to the location where the first type of sound event occurs.

5. A battery module thermal runaway detection method according to claim 1, characterized in that: When the at least one sound sensor includes multiple sound sensors, the step of determining the occurrence location of the first type of sound event and / or the second type of sound event includes: performing sound source localization using the arrival time difference of the sound signals collected by the multiple sound sensors.

6. A battery module thermal runaway detection method according to claim 5, characterized in that: Before using the arrival time difference of multiple sound signals to locate the sound source, the method also includes a step of locating the sound source to one of the multiple stacking areas pre-divided in the battery module by comparing the amplitudes of the sound signals collected by the multiple sound sensors; each stacking area includes multiple battery cells.

7. A battery module thermal runaway detection method as claimed in claim 1, characterized in that Before or during identification of the first type of sound event and the second type of sound event based on the sound signal, the method further includes performing noise filtering processing on the acquired sound signal.

8. A battery module thermal runaway detection method according to claim 1, characterized in that: When the at least one sound sensor is controlled by multiple control units respectively to obtain the sound signal, the step of identifying the first type of sound event and the second type of sound event based on the sound signal includes: when at least one control unit identifies the occurrence of the first type of sound event and / or the second type of sound event, the other control units synchronously perform sound recognition analysis, and judge whether the first type of sound event and / or the second type of sound event occurs based on the sound recognition results of the control units occupying a predetermined proportion.

9. A battery module thermal runaway detection method as claimed in claim 1, characterized in that When the occurrence of the first type of sound event and / or the second type of sound event is identified, the temperature data and electrical parameter data of the battery module are obtained, and when the temperature data shows a temperature change related to the opening of the battery cell pressure relief valve, the occurrence of the first type of sound event is confirmed, and when the electrical parameter data shows a change related to the generation of an arc, the occurrence of the second type of sound event is confirmed.

10. A battery module thermal runaway detection system, characterized in that: include: At least one sound sensor is disposed inside or near the battery module to be inspected; and At least one control unit is communicatively connected to the sound sensor and uses the sound sensor to obtain a sound signal; the control unit is configured to identify a first type of sound event related to the opening of the battery cell pressure relief valve in the battery module and a second type of sound event related to the generated arc based on the sound signal, and record the respective occurrence time and location, and determine the thermal runaway risk level of the battery module based on the relationship between the occurrence time and location of the first type of sound event and the second type of sound event.

Images