A battery pack thermal runaway early warning and collaborative disposal system and method

Abstract

This invention discloses a battery pack thermal runaway early warning and collaborative handling system and method. The system includes a multi-dimensional sensor array and an on-board edge computing unit. The multi-dimensional sensor array is used to monitor multi-dimensional data of internal battery changes, and its output is connected to the on-board edge computing unit. The on-board edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk and issues early warning information based on the identified signs. Alternatively, after identifying signs of thermal runaway risk, the on-board edge computing unit uploads the multi-dimensional data to a cloud big data platform, which analyzes and identifies the thermal runaway risk and outputs early warning information. This invention achieves early warning data acquisition through multi-parameter sensing monitoring, thereby enabling accurate early warning of the battery pack. Simultaneously, the multi-dimensional sensor parameters can detect abnormal thermal runaway information early, achieving the goal of early detection of thermal runaway risks.

Description

Technical Field

[0001] This invention relates to the field of remote monitoring, and in particular to an early warning and collaborative handling system and method for thermal runaway of electric vehicle battery packs based on multi-sensor fusion and cloud big data. Background Technology

[0002] Existing thermal runaway technologies monitor and alarm based on preset fixed thresholds. In terms of handling after a dangerous situation occurs, existing technologies rely on passive protection measures, with the core objective of controlling the consequences as much as possible after thermal runaway occurs.

[0003] However, the aforementioned existing technologies have significant drawbacks and shortcomings. Their early warning mechanisms are severely lagging, relying on signals such as sudden temperature rises or open flames, meaning that by the time the system intervenes, the battery has often already entered an irreversible thermal runaway phase. This makes it impossible to provide preventative warnings when the earliest signs, such as lithium plating or internal short circuits, appear. Furthermore, the sensing dimensions are limited, relying solely on finite parameters like voltage and temperature, which easily leads to false alarms (e.g., misjudging normal temperature increases caused by aggressive driving as a malfunction) and missed alarms (insensitivity to internal short circuits without significant heat generation), resulting in insufficient system reliability. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a battery pack thermal runaway early warning and collaborative handling system and method, which realizes multi-parameter sensing monitoring to acquire early warning data, thereby achieving accurate early warning of the battery pack.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A battery pack thermal runaway early warning and collaborative handling system includes a multi-dimensional sensor array and an in-vehicle edge computing unit; the multi-dimensional sensor array is used to monitor multi-dimensional data of changes inside the battery, and its output is connected to the in-vehicle edge computing unit.

[0007] The vehicle-mounted edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk, and issues early warning information based on the identified signs of thermal runaway risk. Alternatively, after identifying signs of thermal runaway risk, the vehicle-mounted edge computing unit uploads the multi-dimensional data to the cloud big data platform, which then analyzes and identifies the thermal runaway risk and outputs early warning information.

[0008] The multidimensional sensor array includes a voltage sensor, a temperature sensor, an electrochemical gas sensor, a high-frequency ultrasonic sensor, a micro-pressure sensor, and / or an acoustic vibration monitoring sensor; used to monitor voltage, temperature, trace gas characteristics inside the battery, changes in internal battery pressure, and abnormal acoustic signals, respectively.

[0009] The electrochemical gas sensor is installed in the gap between battery modules or in the top gas chamber to collect the concentrations of carbon monoxide and hydrogen in the air between battery modules or in the top gas chamber.

[0010] The high-frequency ultrasonic sensor is located on both sides of the battery module and is used to detect trace amounts of gas generated by lithium plating or electrolyte decomposition based on the minute changes in the propagation characteristics of ultrasonic waves in the electrolyte and electrode materials.

[0011] The micro-pressure sensor is installed on the sealed housing of the battery pack and is used to detect changes in the overall pressure inside the battery pack housing.

[0012] The acoustic vibration monitoring sensor is an acoustic vibration monitoring system based on fiber optic sensors. It lays thin optical fibers on the surface of the module in a set path and characterizes abnormal data of battery thermal runaway by analyzing the phase changes of the optical signal caused by internal micro short-circuit arcs or mechanical deformation.

[0013] The vehicle-mounted edge computing unit connects to the cloud-based big data platform via the vehicle's built-in wireless communication module. After identifying signs of thermal runaway risk, it uploads multi-dimensional data to the cloud-based big data platform. The cloud-based big data platform has a built-in battery fault prediction model library. It analyzes the uploaded data based on the battery fault prediction model library, obtains thermal runaway prediction results, and outputs thermal runaway handling instructions based on the prediction results.

[0014] The cloud-based big data platform connects to the vehicle domain controller via an in-vehicle wireless communication module, and is used to send processing instructions to the vehicle domain controller for execution.

[0015] A control method for a battery pack thermal runaway early warning and collaborative handling system includes using a multi-dimensional sensor array to monitor multi-dimensional data of internal battery changes; an on-board edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk, and issues early warning information based on the identified signs of thermal runaway risk; or, after identifying signs of thermal runaway risk, the on-board edge computing unit uploads the multi-dimensional data to a cloud big data platform, which then analyzes and identifies the thermal runaway risk and outputs early warning information.

[0016] The vehicle-mounted edge computing unit connects to the cloud-based big data platform via the vehicle's built-in wireless communication module. After identifying signs of thermal runaway risk, it uploads multi-dimensional data to the cloud-based big data platform. The cloud-based big data platform has a built-in battery fault prediction model library. It analyzes the uploaded data based on the battery fault prediction model library, obtains thermal runaway prediction results, and outputs thermal runaway handling instructions based on the prediction results.

[0017] The advantages of this invention are as follows: First, it achieves extremely early warning of battery safety by capturing weak chemical and physical signals at the forefront of thermal runaway chain reactions, providing a valuable time window for intervention and potentially preventing serious accidents such as fires. Second, the dual mechanism of multi-sensor fusion and cloud-based big data cross-validation greatly improves the accuracy and reliability of warnings, effectively reducing false alarms caused by interference and missed alarms caused by insufficient perception, thus enhancing user trust in the system. Third, this invention breaks through the limitations of single-vehicle intelligence, endowing the system with new characteristics of collaborative intelligence. Through information interaction between vehicles, roads, and the cloud, risk management is upgraded from single vehicle behavior to systematic optimal decision-making and resource scheduling, significantly improving overall safety performance. Finally, this system not only directly protects the lives and property of drivers and passengers but also uses accumulated real data to feed back into the optimization of battery design and management systems, forming a positive cycle that has profound significance for improving the safety level and user confidence of the entire electric vehicle industry. Attached Figure Description

[0018] The following is a brief explanation of the contents of each of the accompanying drawings and the markings in the drawings:

[0019] Figure 1 This is a schematic diagram of the early warning and collaborative response system of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and the description of the preferred embodiments.

[0021] The technical problem this invention aims to solve is to overcome the inherent defects of existing battery safety technologies, such as delayed early warning, single-sensor operation, and lack of coordination. Existing solutions cannot effectively intervene in the irreversible stage before thermal runaway occurs, and their reliance on single sensor threshold judgments results in a high false alarm rate. Furthermore, the isolated handling mode for individual vehicles is difficult to minimize risk in complex real-world traffic environments. Therefore, this invention aims to construct a comprehensive solution capable of achieving very early warning, accurate judgment, and initiating coordinated internal and external vehicle-mounted responses, thereby transforming battery safety from a passive "post-disaster response" to a proactive "pre-disaster prevention."

[0022] This embodiment of a battery pack thermal runaway early warning and collaborative handling system includes a multi-dimensional sensor array, an in-vehicle edge computing unit, and a cloud big data platform; the multi-dimensional sensor array is used to monitor multi-dimensional data of changes inside the battery, and its output is connected to the in-vehicle edge computing unit.

[0023] The vehicle-mounted edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk and issues warning information based on the identified signs of thermal runaway risk. Alternatively, after identifying signs of thermal runaway risk, the vehicle-mounted edge computing unit uploads the multi-dimensional data to the cloud big data platform, which then analyzes and identifies the thermal runaway risk and outputs warning information.

[0024] The multi-dimensional sensor array includes voltage sensors, temperature sensors, electrochemical gas sensors, high-frequency ultrasonic sensors, micro-pressure sensors, and / or acoustic vibration monitoring sensors; these are used to monitor voltage, temperature, trace gas characteristics inside the battery, changes in internal battery pressure, and abnormal acoustic signals, respectively. Multiple sensors in the multi-dimensional sensor array monitor thermal runaway data from temperature and voltage after thermal runaway, and also monitor changes in electrochemical properties in the early stages of thermal runaway, thus achieving more comprehensive monitoring of thermal runaway anomalies. Furthermore, by observing early-stage or pre-existing electrochemical properties, it can identify thermal runaway anomalies earlier, providing more time for early warning and subsequent rescue and protection strategies, reducing the occurrence of battery pack thermal runaway, and even minimizing the safety hazards caused by thermal runaway.

[0025] In this embodiment, the sensor array monitors battery pack thermal runaway-related parameters from different angles and time points. Temperature and voltage sensors monitor battery-related data in the later stages of thermal runaway, after it has already occurred. Electrochemical gas sensors, high-frequency ultrasonic sensors, micro-pressure sensors, and acoustic vibration monitoring sensors monitor battery pack thermal runaway-related data in the early or pre-stages of thermal runaway. These sensors collectively form a multi-source data set for monitoring, predicting, and identifying thermal runaway. The specific settings and acquired data for each sensor include the following:

[0026] Temperature sensors are placed at monitoring points inside, outside, and between modules of the battery pack to collect temperature data of the battery pack, battery modules, or the battery pack itself at the monitoring points. Voltage data is used to collect the voltage inside the battery pack, such as individual cell voltage and module voltage, to monitor the voltage data of the corresponding thermal runaway anomaly, which facilitates subsequent monitoring, early warning, and protection strategies for thermal runaway.

[0027] Electrochemical gas sensors are placed in the gaps or top gas chambers between battery modules to collect the concentrations of carbon monoxide and hydrogen in the air between battery modules or in the top gas chamber. Carbon monoxide and hydrogen are products of the decomposition of battery materials in the early stages of overheating, and their concentration changes much earlier than obvious temperature changes. Therefore, they are introduced into multi-source data as one of the parameters of thermal runaway, which facilitates the early identification of the risk of thermal runaway of the battery pack.

[0028] High-frequency ultrasonic sensors are positioned on both sides of the battery module to detect trace gases generated by lithium plating or electrolyte decomposition based on minute changes in the propagation characteristics of ultrasonic waves in the electrolyte and electrode materials. Pairs of high-frequency ultrasonic sensors are arranged on both sides of the module to detect these trace gases by analyzing minute changes in the propagation characteristics of ultrasonic waves in the electrolyte and electrode materials. Since thermal runaway may occur before or in its early stages due to lithium plating or electrolyte decomposition, the occurrence of lithium plating or electrolyte decomposition before thermal runaway is used as a characterizing parameter for the early stage of thermal runaway. Based on the parameters acquired by ultrasonic waves, the monitoring of lithium plating or electrolyte decomposition before thermal runaway is achieved, using this as one of the monitoring parameters for thermal runaway. This facilitates subsequent identification and analysis of thermal runaway, enabling earlier or more timely identification of thermal runaway in the battery pack.

[0029] A micro-pressure sensor is mounted on the sealed housing of the battery pack to detect changes in the overall pressure inside the battery pack housing. The micro-pressure sensor can be positioned on the inner or outer side of the sealed housing to identify minute changes in the internal pressure of the battery pack. A highly sensitive micro-pressure sensor integrated into the sealed housing of the battery pack is used to capture subtle increases in the overall internal pressure. Because the battery pack is sealed, minute pressure changes occur within it before or in the early stages of thermal runaway. Monitoring these changes can serve as parameters for early or pre-existing thermal runaway, thereby identifying its occurrence.

[0030] The acoustic vibration monitoring sensor is a fiber optic sensor-based acoustic vibration monitoring system. It lays thin optical fibers along a predetermined path on the module surface and characterizes abnormal data related to battery thermal runaway by analyzing phase changes in the optical signal caused by internal micro-short-circuit arcs or mechanical deformation. This fiber optic sensing-based acoustic vibration monitoring system is cleverly deployed inside the battery pack. The thin optical fibers are laid along a specific path on the module surface, and by analyzing the phase changes in the optical signal caused by internal micro-short-circuit arcs or mechanical deformation, it achieves precise location and identification of abnormal events. When no micro-short-circuit arc or mechanical deformation occurs inside the battery pack, the optical fiber passing through the pack is in a normal state, and its output optical signal remains unchanged. When a micro-short-circuit or mechanical deformation occurs, the optical signal changes. The more detectable optical signal can monitor the micro-short-circuit or mechanical deformation within the battery pack, providing early warning of thermal runaway.

[0031] The various data collected by the multi-dimensional sensor array are first transmitted to the vehicle-mounted edge computing unit for processing. The edge computing unit preprocesses the data to identify risks before sending it to the cloud-based big data platform for further processing and identification. Otherwise, if no risk is identified, the data is not uploaded, thus reducing the amount of data uploaded to the big data platform and reducing bandwidth and traffic consumption. The data collected by the multi-dimensional sensor array is massive. This massive amount of raw data generated by these sensors is processed in real time by the edge computing unit deployed on the vehicle. This unit is typically supported by a high-performance automotive-grade system-on-a-chip (SoC) and runs optimized data fusion algorithms. The data fusion processing algorithms running inside the vehicle-mounted edge computing unit include: cleaning, aligning, and standardizing multi-source heterogeneous data to eliminate noise interference. Subsequently, the core function is to perform multi-dimensional anomaly pattern recognition based on machine learning. This algorithm does not judge whether a single sensor exceeds the threshold in isolation, but continuously searches for correlated anomalies across physical quantities (electrical, chemical, acoustic, and mechanical). For example, when the algorithm detects a slight voltage fluctuation in a battery cell, a nearby fiber optic sensor captures an acoustic emission signal in a specific spectrum, and a gas sensor reading shows a slow, trending increase, this multi-signal coupling anomaly pattern will be identified as a high-risk event. Once the confidence level of such a pattern exceeds a set threshold, the edge computing unit immediately packages and encrypts the relevant time-series feature data, transmitting it to the cloud platform via the vehicle's built-in wireless communication module. This achieves initial extraction from "raw data" to "valuable information," significantly reducing network transmission burden. The multi-dimensional anomaly pattern recognition algorithm built into the edge computing unit uses machine learning to fuse and identify collected multi-source data, obtaining anomaly state data of multi-signal coupling between multiple sources. When an anomaly is detected and identified as a high-risk event, the data is packaged and uploaded to the cloud big data platform. The packaged data is uploaded to the cloud big data platform in a time-series format for further identification and processing, thus providing data for subsequent identification. This solution does not upload all data to the cloud big data platform; instead, it preprocesses the data in the edge computing unit before uploading, reducing network transmission burden.

[0032] The vehicle-mounted edge computing unit connects to a cloud-based big data platform via the vehicle's built-in wireless communication module. This allows it to upload multi-dimensional data to the cloud platform after identifying signs of thermal runaway risk. The cloud-based big data platform, with its built-in battery fault prediction model library, analyzes the uploaded data to obtain thermal runaway prediction results and outputs thermal runaway handling commands based on these results. The cloud-based big data platform also connects to the vehicle domain controller via the vehicle's wireless communication module to issue these commands for execution. Further analysis by the cloud-based big data platform identifies the risk of thermal runaway. If it is not considered a risk, no warning or safety strategy is required. Otherwise, if the cloud-based big data platform identifies a risk, it issues a battery pack thermal runaway safety warning and executes safety strategies simultaneously, thus achieving simultaneous warning and protection to reduce the safety risks and losses caused by thermal runaway.

[0033] The cloud-based big data platform serves as the system's decision-making hub, built upon a scalable cloud computing infrastructure. Its core asset is a continuously optimized battery fault prediction model library, built by constantly learning from massive amounts of fleet data. Upon receiving a warning from a vehicle, the platform initiates a complex cross-validation and decision-making process. This process includes: First, horizontal comparison: The platform retrieves data from other vehicles of the same model and batch as the alarming vehicle, operating under similar conditions, to analyze whether the anomaly is unique to an individual or a group phenomenon, thus ruling out false alarms caused by common environmental interference or batch defects. Second, vertical analysis: The platform deeply mines the vehicle's historical operating data to assess whether the anomaly is a sudden, accidental occurrence or an acceleration of a long-term deterioration trend. Finally, based on the prediction results from the fault model library, combined with horizontal and vertical analysis, the platform generates tiered handling instructions. For low-confidence risks, a "suggested inspection" prompt may be sent to the vehicle; for confirmed medium risks, the vehicle is instructed to enter "on-the-go safety mode"; for high-confidence emergency risks, the highest-level handling plan is triggered. The cloud-based big data platform calculates the confidence level of the current thermal runaway fault risk based on the fault model library and horizontal and vertical analysis. It calculates a confidence level value by weighting the prediction results of the fault model library, the horizontal analysis results, and the vertical analysis results, and outputs corresponding handling instructions based on the confidence level data.

[0034] The final handling phase demonstrates the system's collaborative intelligence. When the vehicle receives instructions from the cloud, its response is no longer isolated. The vehicle's domain controller coordinates and executes the entire vehicle's handling process. In "On-the-Go Safety Mode," the vehicle issues clear warnings to the driver via the cockpit display and voice system, while the navigation system automatically replans the route, guiding the vehicle to the nearest open space or safe parking lot. During this process, the battery management system gently but effectively limits the battery's charging and discharging power and initiates maximum cooling to suppress thermal runaway and buy time for a safe stop. Crucially, the vehicle domain controller initiates inter-vehicle collaboration via V2X communication technology. Using V2V communication, the vehicle can broadcast its status and location to other vehicles within a 100-meter radius, reminding them to maintain a safe distance or change lanes in advance. Through V2I communication, the vehicle can interact with intelligent traffic lights or roadside units to attempt to obtain "green wave" priority, allowing for a quick exit from high-risk areas such as tunnels and bridges. At the same time, the vehicle automatically sends precise location, risk level, battery type and other key information to the designated rescue service center, enabling rescue forces to prepare in advance and dispatch accurately.

[0035] In a preferred embodiment, when the vehicle is in motion, the sensor array and edge computing unit are in real-time operation. After the vehicle is turned off, the edge computing unit and sensor array are periodically woken up to prevent thermal runaway after parking. Some thermal runaways occur during driving but slowly develop until thermal runaway occurs after parking and turning off. To avoid monitoring and warning of thermal runaway after the vehicle is powered off, this solution uses a timed wake-up method for thermal runaway monitoring. The RTC wake-up module periodically wakes up the sensor array and edge computing unit within a set time range after the vehicle is powered off to monitor the data. In this embodiment, the RTC wake-up time setting includes an initial wake-up time. Each time the vehicle is powered down, the RTC wake-up module wakes up according to the initial wake-up time. Upon reaching the initial wake-up time, the RTC wake-up module activates the edge computing unit, which then enters a silent monitoring mode. In silent monitoring mode, the edge computing unit acquires and analyzes data through a multi-dimensional sensor array. Upon detecting signs of thermal runaway risk, it uploads the multi-dimensional data to the cloud big data platform, which analyzes and identifies the thermal runaway risk and outputs a warning message. Otherwise, after the wake-up working time reaches the set working time, silent monitoring is completed, and the module enters hibernation mode, waiting for the next time the RTC wake-up module will periodically activate the edge computing unit. During each timed wake-up after a power outage, the RTC wake-up module's timed wake-up time is longer than the previous timed wake-up time, gradually increasing until the vehicle is powered on again and restarted, entering the next timed start-up after a power outage. This timed start strategy involves gradually increasing the time interval for each wake-up call. Specifically, the second wake-up time for the RTC module is longer than the first wake-up time. This dynamic setting of the timed start time better adapts to actual battery malfunctions. In the immediate period after power-off, the battery temperature may be high due to the vehicle's recent use, and the probability of thermal runaway due to collisions or scrapes during driving is significantly increased. Therefore, after power-off, the time interval gradually lengthens. This results in more frequent RTC wake-up calls immediately after power-off, while the interval increases after a period of inactivity. This achieves the goal of frequent monitoring after power-off and reduced timed start frequency when the vehicle is stationary for extended periods.This strategy has at least two effects: First, it monitors the vehicle's electric safety status by frequently monitoring it during the initial period after power-off, which is high-risk, and extending the timed monitoring cycle after prolonged periods of inactivity. The timing is dynamically adjusted to monitor safety conditions such as battery thermal runaway, effectively improving monitoring results and enabling timely detection of battery anomalies. Second, because the timing of the RTC wake-up module is variable, the longer the parking time, the longer the wake-up time. This reduces the power consumption of frequent starts, improves vehicle safety when stationary, reduces energy consumption, increases vehicle range, and reduces energy waste.

[0036] In another preferred embodiment of this example, after power-off, the RTC wake-up module periodically starts waking up the vehicle edge computing unit according to the initial wake-up time. The initial wake-up time is dynamically set based on the vehicle's usage time. Because the longer the vehicle travels, the higher its battery temperature becomes, and the more likely it is to experience collisions, scrapes, or other situations that pose a safety hazard to the battery, the initial wake-up time can be set based on the vehicle's travel time. The longer the vehicle's driving time, the shorter the initial wake-up time should be. A pre-calibrated lookup table between the initial wake-up time and the vehicle's driving time is stored in the controller. Before each power-down, the corresponding initial wake-up time is obtained based on the vehicle's driving time and set in the RTC wake-up module. The initial wake-up time can have a pre-defined range, with a pre-set maximum and minimum value to prevent excessive fluctuations and keep it within a reasonable range. The initial wake-up time is calculated based on the vehicle's driving time before each power-down, and then compared to the maximum and minimum values. If the calculated initial wake-up time is between the maximum and minimum values, it is stored as the final wake-up time in the vehicle controller. If the calculated initial wake-up time is greater than the maximum value, the maximum value is used as the initial wake-up time; if the calculated initial wake-up time is less than the minimum value, the minimum value is used as the initial wake-up time. Calculating the initial wake-up time by associating it with driving time effectively and accurately obtains the base time for RTC timed wake-up, meeting the requirements for subsequent timed startup and monitoring.

[0037] This embodiment also provides a control method for a battery pack thermal runaway early warning and collaborative handling system. The method includes using a multi-dimensional sensor array to monitor multi-dimensional data of internal battery changes; an on-board edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk, and issues a warning message based on the identified signs of thermal runaway risk; or, after identifying signs of thermal runaway risk, the on-board edge computing unit uploads the multi-dimensional data to a cloud big data platform, which then analyzes and identifies the thermal runaway risk and outputs a warning message.

[0038] The vehicle-mounted edge computing unit connects to a cloud-based big data platform via the vehicle's built-in wireless communication module. This allows it to upload multi-dimensional data to the cloud platform after identifying signs of thermal runaway risk. The cloud-based big data platform, with its built-in battery fault prediction model library, analyzes the uploaded data based on this library to obtain thermal runaway prediction results and output thermal runaway handling instructions. Multi-source data undergoes preliminary fusion analysis at the vehicle-mounted edge computing unit to identify early risk signs formed by combinations of various weak abnormal signals. This encrypted data is then uploaded to the cloud-based big data platform. Instead of judging individual alarms in isolation, the platform cross-validates with its massive fleet operation data and fault mode library. By comparing data from similar vehicles horizontally and analyzing historical trends longitudinally, it distinguishes between common phenomena and individual risks, thereby making tiered decisions. Ultimately, the system executes precise responses through the vehicle-to-vehicle cooperative network: for confirmed risks, the vehicle will activate "on-the-go safety mode," automatically navigate to a safe area, and limit battery power; at the same time, it will use V2V communication to warn surrounding vehicles, coordinate traffic signals through V2I communication, and automatically send detailed risk information to the rescue center, forming a comprehensive and collaborative safety response from inside the vehicle to outside.

[0039] The core of this embodiment's technical solution is to construct a multi-layered intelligent system composed of an in-vehicle terminal, a cloud-based big data platform, and a vehicle-to-everything (V2X) network. The in-vehicle terminal, by introducing a multi-dimensional sensor array including ultrasonic sensors, gas sensors, micro-pressure sensors, and distributed acoustic sensors, transcends traditional methods that only monitor voltage and temperature, enabling the acquisition of trace characteristic gases, pressure changes, and abnormal acoustic signals within the battery. This multi-source data undergoes preliminary fusion analysis in the in-vehicle edge computing unit to identify early risk signs formed by combinations of various weak abnormal signals, and this encrypted data is then uploaded to the cloud-based big data platform. This platform does not judge individual alarms in isolation but cross-validates them using its massive stored fleet operation data and fault mode library. By comparing data from similar vehicles horizontally and analyzing historical trends of the vehicle vertically, it distinguishes between common phenomena and individual risks, thereby making tiered decisions. Ultimately, the system executes precise responses through the vehicle-to-vehicle cooperative network: for confirmed risks, the vehicle will activate "on-the-go safety mode," automatically navigate to a safe area, and limit battery power; at the same time, it will use V2V communication to warn surrounding vehicles, coordinate traffic signals through V2I communication, and automatically send detailed risk information to the rescue center, forming a comprehensive and collaborative safety response from inside the vehicle to outside.

[0040] like Figure 1As shown, this embodiment can be achieved by constructing a comprehensive system integrating advanced sensing technology, edge computing, cloud computing, and vehicle-to-everything (V2X) communication. The core of this system lies in the organic integration of various resources scattered across vehicle terminals, cloud platforms, and the road environment to form an intelligent safety network with predictive, decision-making, and collaborative execution capabilities. The following will elaborate on the technology selection, implementation methods, and workflow of the key components of the system.

[0041] At the vehicle terminal level, the system's hardware foundation is an enhanced battery pack assembly. This battery pack, based on a traditional structure, integrates multiple high-performance sensors, forming a three-dimensional sensing network for the battery's internal state. In addition to high-precision voltage sensors for monitoring basic electrical parameters and temperature sensors distributed at key hot spots, the system specifically incorporates detection units sensitive to early chemical changes. For example, electrochemical gas sensors optimized for carbon monoxide and hydrogen are installed in the gaps between battery modules or in the top gas chamber. These gases are products of the decomposition of battery materials in the early stages of overheating, and their concentration changes much earlier than significant temperature rises. Simultaneously, high-frequency ultrasonic sensors are arranged in pairs on both sides of the modules to detect trace amounts of gas generated by lithium plating or electrolyte decomposition by analyzing minute changes in the propagation characteristics of ultrasound in the electrolyte and electrode materials. To monitor the physical effects caused by these gas-generating behaviors, a highly sensitive micro-pressure sensor is integrated into the battery pack's sealed housing to capture subtle increases in overall internal pressure. In addition, an acoustic vibration monitoring system based on fiber optic sensing has been cleverly deployed, with thin optical fibers laid on the surface of the module in a specific path. By analyzing the phase changes of the optical signal caused by internal micro short-circuit arcs or mechanical deformation, the system can accurately locate and identify abnormal events.

[0042] The massive amounts of raw data generated by these sensors are processed in real time by edge computing units deployed on the vehicle. These units are typically supported by automotive-grade system-on-a-chips with powerful computing capabilities, running optimized data fusion algorithms. The primary task of these algorithms is to clean, align, and standardize the multi-source heterogeneous data, eliminating noise interference. Subsequently, the core function is to perform multi-dimensional anomaly pattern recognition based on machine learning. This algorithm does not judge whether a single sensor exceeds a threshold in isolation, but continuously searches for correlated anomalies across physical quantities (electrical, chemical, acoustic, and mechanical). For example, when the algorithm detects a slight fluctuation in the voltage of a battery cell, a nearby fiber optic sensor captures an acoustic emission signal of a specific spectrum, and a gas sensor reading shows a slow, trending increase, this multi-signal coupling anomaly pattern will be identified as a high-risk event. Once the confidence level of such a pattern exceeds a set threshold, the edge computing unit immediately packages and encrypts the relevant time-series feature data, transmitting it to the cloud platform via the vehicle's built-in wireless communication module. This achieves the initial extraction from "raw data" to "valuable information," significantly reducing the network transmission burden.

[0043] The cloud-based big data platform serves as the system's decision-making hub, built upon a scalable cloud computing infrastructure. Its core asset is a continuously optimized battery fault prediction model library, built by constantly learning from massive amounts of fleet data. Upon receiving a warning from a vehicle, the platform initiates a complex cross-validation and decision-making process. First, it performs a horizontal comparison: the platform retrieves data from other vehicles of the same model and batch as the alarming vehicle, operating under similar conditions, to analyze whether the anomaly is unique to an individual or a group phenomenon, thus ruling out false alarms caused by common environmental interference or batch defects. Second, it performs a vertical analysis: the platform deeply mines the vehicle's historical operating data to assess whether the anomaly is a sudden, accidental occurrence or an acceleration of a long-term deterioration trend. Finally, based on the prediction results from the fault model library, combined with the horizontal and vertical analyses, the platform generates tiered handling instructions. For example, for low-confidence risks, it may only send a "suggested check" prompt to the vehicle; for confirmed medium risks, it instructs the vehicle to enter "on-the-go safety mode"; and for high-confidence emergency risks, it triggers the highest-level handling plan.

[0044] The final handling phase demonstrates the system's collaborative intelligence. When the vehicle receives instructions from the cloud, its response is no longer isolated. For example, in "On-the-Go Safety Mode," the vehicle issues clear warnings to the driver via the cockpit display and voice system, while the navigation system automatically replans the route, guiding the vehicle to the nearest open space or safe parking lot. During this process, the battery management system gently but effectively limits the battery's charging and discharging power and initiates maximum cooling to suppress thermal runaway and buy time for a safe stop. Crucially, the system initiates vehicle-to-everything (V2X) communication. Using V2V communication, the vehicle can broadcast its status and location to other vehicles within a 100-meter radius, reminding them to maintain a safe distance or change lanes in advance. Through V2I communication, the vehicle can interact with intelligent traffic lights or roadside units to attempt to obtain "green wave" priority for quickly leaving high-risk areas such as tunnels and bridges. Simultaneously, the vehicle automatically sends precise location, risk level, battery type, and other critical information to the designated rescue service center, enabling rescue forces to prepare in advance and respond accurately.

[0045] The entire system operates within a closed loop, from microscopic perception to macroscopic coordination. It begins with the most minute chemical and physical changes within the battery, undergoes initial identification by vehicle-side intelligence, followed by in-depth verification and decision-making in the cloud, and finally, precise handling through network-connected coordination. This elevates battery safety from passive, delayed protection to proactive, early-stage, and systematic early warning and prevention. This implementation method fully integrates existing mature technologies and features innovative integration at the system architecture and algorithm levels, demonstrating high feasibility and practicality.

[0046] The technical solution of this invention offers significant benefits on multiple levels. The most fundamental effect is the realization of extremely early battery safety warnings. By capturing weak chemical and physical signals at the forefront of thermal runaway chain reactions, it provides a valuable time window for intervention, potentially preventing serious accidents such as fires. Secondly, the dual mechanism of multi-sensor fusion and cloud-based big data cross-validation greatly improves the accuracy and reliability of warnings, effectively reducing false alarms caused by interference and missed alarms due to insufficient perception, thus enhancing user trust in the system. Thirdly, this invention breaks through the limitations of single-vehicle intelligence, endowing the system with new characteristics of collaborative intelligence. Through information interaction between vehicles, roads, and the cloud, risk management is upgraded from single vehicle behavior to systematic optimal decision-making and resource scheduling, significantly improving overall safety performance. Ultimately, this system not only directly protects the lives and property of passengers but also uses accumulated real-world data to optimize battery design and management systems, forming a positive cycle that has profound implications for improving the safety level and user confidence of the entire electric vehicle industry.

[0047] Obviously, the specific implementation of this invention is not limited to the above-described methods. Any non-substantial improvements made using the inventive concept and technical solution of this invention are within the protection scope of this invention.

Claims

1. A battery pack thermal runaway early warning and collaborative handling system, characterized in that: It includes a multi-dimensional sensor array and an in-vehicle edge computing unit; the multi-dimensional sensor array is used to monitor multi-dimensional data of changes inside the battery, and its output is connected to the in-vehicle edge computing unit. The vehicle-mounted edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk, and issues early warning information based on the identified signs of thermal runaway risk. Alternatively, after identifying signs of thermal runaway risk, the vehicle-mounted edge computing unit uploads the multi-dimensional data to the cloud big data platform, which then analyzes and identifies the thermal runaway risk and outputs early warning information.

2. The battery pack thermal runaway early warning and collaborative handling system as described in claim 1, characterized in that: The multidimensional sensor array includes a voltage sensor, a temperature sensor, an electrochemical gas sensor, a high-frequency ultrasonic sensor, a micro-pressure sensor, and / or an acoustic vibration monitoring sensor; used to monitor voltage, temperature, trace gas characteristics inside the battery, changes in internal battery pressure, and abnormal acoustic signals, respectively.

3. The battery pack thermal runaway early warning and collaborative handling system as described in claim 2, characterized in that: The electrochemical gas sensor is installed in the gap between battery modules or in the top gas chamber to collect the concentrations of carbon monoxide and hydrogen in the air between battery modules or in the top gas chamber.

4. The battery pack thermal runaway early warning and collaborative handling system as described in claim 2, characterized in that: The high-frequency ultrasonic sensor is located on both sides of the battery module and is used to detect trace amounts of gas generated by lithium plating or electrolyte decomposition based on the minute changes in the propagation characteristics of ultrasonic waves in the electrolyte and electrode materials.

5. The battery pack thermal runaway early warning and collaborative handling system as described in claim 2, characterized in that: The micro-pressure sensor is installed on the sealed housing of the battery pack and is used to detect changes in the overall pressure inside the battery pack housing.

6. The battery pack thermal runaway early warning and collaborative handling system as described in claim 2, characterized in that: The acoustic vibration monitoring sensor is an acoustic vibration monitoring system based on fiber optic sensors. It lays thin optical fibers on the surface of the module in a set path and characterizes abnormal data of battery thermal runaway by analyzing the phase changes of the optical signal caused by internal micro short-circuit arcs or mechanical deformation.

7. A battery pack thermal runaway early warning and collaborative handling system as described in any one of claims 1-6, characterized in that: The vehicle-mounted edge computing unit connects to the cloud-based big data platform via the vehicle's built-in wireless communication module, and is used to upload multi-dimensional data to the cloud-based big data platform after identifying signs of thermal runaway risk. The cloud-based big data platform has a built-in battery fault prediction model library. It analyzes the uploaded data based on the battery fault prediction model library, obtains thermal runaway prediction results, and outputs thermal runaway handling instructions based on the prediction results.

8. The battery pack thermal runaway early warning and collaborative handling system as described in claim 7, characterized in that: The cloud-based big data platform connects to the vehicle domain controller via an in-vehicle wireless communication module, and is used to send processing instructions to the vehicle domain controller for execution.

9. A control method for a battery pack thermal runaway early warning and collaborative handling system as described in any one of claims 1-8, characterized in that: The method includes using a multi-dimensional sensor array to monitor multi-dimensional data on changes inside the battery. The vehicle-mounted edge computing unit performs fusion analysis on the monitored multi-dimensional data to identify signs of thermal runaway risk and issues warning information based on the identified signs of thermal runaway risk. Alternatively, after identifying signs of thermal runaway risk, the vehicle-mounted edge computing unit uploads the multi-dimensional data to the cloud big data platform, which then analyzes and identifies the thermal runaway risk and outputs warning information.

10. The control method of the battery pack thermal runaway early warning and collaborative handling system as described in claim 9, characterized in that: The vehicle-mounted edge computing unit connects to the cloud big data platform through the vehicle's built-in wireless communication module, and is used to upload multi-dimensional data to the cloud big data platform after identifying signs of thermal runaway risk. The cloud-based big data platform has a built-in battery fault prediction model library. It analyzes the uploaded data based on the battery fault prediction model library, obtains thermal runaway prediction results, and outputs thermal runaway handling instructions based on the prediction results.

Images