Power battery thermal runaway early warning and hidden danger active elimination system based on AI fusion

Abstract

The invention discloses a power battery thermal runaway early warning and hidden danger active elimination system based on AI fusion, and the system achieves the full-hardware closed-loop control through the linkage of five hardware modules, namely, a data collection module, an AI early warning module, a fault diagnosis module, a hidden danger elimination module and a state feedback module. The data acquisition module adopts a layered architecture to obtain multi-dimensional parameters of the battery in real time; the AI early warning module carries out extremely early risk level judgment through a fusion model solidified on an edge calculation chip; the fault diagnosis module identifies a specific fault type based on a hardware comparator and a logic gate circuit; the hidden danger elimination execution module automatically triggers differential treatment strategies including power adjustment, battery bypass, forced cooling and inhibitor spraying according to the early warning level and the fault type; and the state feedback module monitors and indicates the elimination effect in real time. According to the method, full-chain hardware execution from early warning to disposal is realized, complex software development is not needed, and the method has the advantages of high response speed, high reliability, high precision and easiness in integration and landing.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of artificial intelligence and power battery safety control, and in particular relates to a power battery thermal runaway early warning and proactive hazard elimination system based on AI fusion. Background Technology

[0002] With the rapid development of new energy vehicles and the energy storage industry, fires and explosions caused by thermal runaway of power batteries have become a core bottleneck restricting the industry's development. According to data from the National Fire and Rescue Administration, 77% of fires in new energy vehicles are directly related to batteries, and 80% of these fires occur in static or charging states without significant external force, making traditional passive protection methods ineffective. GB38031-2025, "Safety Requirements for Power Batteries for Electric Vehicles," clearly stipulates that power batteries must issue an alarm signal no later than 5 minutes after thermal runaway occurs; however, existing technology still has two major shortcomings: First, there is a disconnect between early warning and response, and the systems rely heavily on software. Current mainstream early warning systems mostly rely on complex software algorithms to identify and alarm faults. This is not only difficult to develop, but the software is also susceptible to environmental interference. Furthermore, these systems can only provide alarm functions and lack the ability to proactively intervene in potential hazards. After an alarm is triggered, manual intervention is required. However, the window between the onset of a thermal runaway fault and its full-blown outbreak is extremely short, making it difficult to prevent accidents from occurring. Secondly, there is a lack of precise hardware-based solutions for handling potential hazards. Existing intervention technologies mostly rely on single hardware actions, failing to differentiate treatment based on fault type and battery status. Furthermore, they require software-based coordinated control, resulting in response delays. For abnormal internal resistance caused by battery aging, blindly cutting off power affects user experience. For temperature rises caused by partial short circuits, simply cooling down the device cannot completely eliminate the hazard and can easily lead to recurrence of the fault.

[0003] The development of AI hardware technology has provided a new path for accurate fault diagnosis. The RepLKNet+BiGRU-GATT fusion model can achieve real-time feature extraction through hardware solidification. At the same time, the evolution law of different fault types of battery thermal-electric coupling mechanism has been clarified, and differentiated handling can be achieved by pre-setting hardware execution logic. Therefore, it is necessary to propose a power battery thermal runaway early warning and hidden danger active elimination system based on AI fusion to solve the above problems. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide an AI-based power battery thermal runaway early warning and proactive hazard elimination system. It aims to solve the technical problems of existing power battery thermal runaway early warning systems that rely on complex software development, can only issue alarms but cannot proactively eliminate hazards, and lack precise handling strategies. By solidifying the AI ​​early warning algorithm and hazard handling logic through hardware modular design, no additional software development is required, which can achieve very early and accurate early warning and differentiated hazard elimination, thereby blocking the thermal runaway chain from the source and improving the safety redundancy of the power battery.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A power battery thermal runaway early warning and proactive hazard elimination system based on AI fusion, comprising: The data acquisition module is used to collect multi-dimensional operating parameters of the power battery in real time; The AI ​​early warning module is electrically connected to the data acquisition module. It is used to receive multi-dimensional operating parameters and perform thermal runaway risk level assessment based on the hardware-embedded AI fusion model, and output early warning signals. The fault diagnosis module, electrically connected to the data acquisition module, is used to identify the fault type of the power battery based on preset hardware logic and parameter thresholds, and output a fault identification signal. The hazard elimination execution module is electrically connected to the AI ​​early warning module and the fault diagnosis module, respectively. It is used to trigger the execution of the corresponding active elimination strategy through hardware linkage based on the received early warning signal and fault identification signal. The status feedback module, which is associated with the data acquisition module and the hazard elimination execution module, is used to monitor and provide hardware indication of the execution effect of the proactive elimination strategy. The data acquisition module, AI early warning module, fault diagnosis module, hidden danger elimination execution module, and status feedback module are connected by hardware circuits and preset logic to achieve full-link linkage without the need for upper-level software control.

[0006] Preferably, the data acquisition module adopts a hierarchical sampling hardware architecture, including: The high-speed sampling layer is equipped with a first dedicated acquisition chip, which acquires the current and voltage signals of the power battery at a first sampling frequency, and is connected to a hardware filtering circuit. The medium-speed sampling layer is equipped with a second dedicated acquisition chip and sensor array to acquire the temperature and impedance signals of the power battery at a second sampling frequency lower than the first sampling frequency, and integrates a hardware comparator. The low-speed sampling layer is equipped with a third dedicated acquisition chip, which acquires the lithium-ion concentration and state of charge signals of the power battery at a third sampling frequency lower than the second sampling frequency. The signal conditioning circuit, connected to the high-speed sampling layer, medium-speed sampling layer and low-speed sampling layer, is used to convert the signals collected by each layer into a standard voltage signal and output it.

[0007] Preferably, the AI ​​early warning module includes an AI edge computing chip, and the AI ​​fusion model is a pre-trained fusion model of RepLKNet and BiGRU-GATT, which is solidified in the AI ​​edge computing chip by hardware programming. The AI ​​early warning module is configured to output at least three levels of early warning signals. Each level of early warning signal is represented by a high-level output through a different hardware pin to indicate a different level of thermal runaway risk.

[0008] Preferably, the fault diagnosis module is composed of a comparator group and logic gate circuit hardware; The comparator group has preset parameter threshold ranges corresponding to various typical fault types; The logic gate circuit is connected to the comparator group to perform logical combination of the comparison results and output high level through different hardware pins to identify different fault types.

[0009] Preferably, the various typical fault types include battery aging, partial short circuit, poor contact, overcharging and over-discharging, and external compression.

[0010] Preferably, the hazard elimination execution module includes a power control unit, a thermal management unit, and a chemical inhibition unit; The power control unit is configured to perform at least one of the following operations based on different warning levels: charging power adjustment, single cell bypass, or disconnection of the battery main circuit. The thermal management unit is configured to activate cooling strategies of varying intensities based on the warning level; The chemical suppression unit is configured to trigger the spraying of extinguishing media when specific warning levels and specific fault types are met.

[0011] Preferably, the chemical suppression unit includes a storage tank containing a composite extinguishing medium and an electromagnetic injection valve, wherein the composite extinguishing medium comprises hexafluoropropane and ammonium dihydrogen phosphate.

[0012] Preferably, the status feedback module includes a feedback sensor and a hardware indication circuit; The feedback sensor is used to collect key state parameters of the power battery after the active elimination strategy is executed; The hardware indicator circuit is connected to the feedback sensor and is used to drive indicator lights of different colors to light up based on the comparison results of the key status parameters and preset thresholds, so as to intuitively display the status of successful elimination of hidden danger, need for continuous monitoring, or failure to eliminate and need for secondary treatment.

[0013] Preferably, a method for an AI-based power battery thermal runaway early warning and proactive hazard elimination system includes the following steps: S1, through the hierarchical sampling hardware architecture of the data acquisition module, synchronously acquires the current, voltage, temperature, impedance, lithium ion concentration and SOC signal of the power battery, and converts them into standard voltage signals through the signal conditioning circuit. S2 synchronously inputs the standard voltage signal to the AI ​​early warning module and the fault diagnosis module; S3 processes standard voltage signals through the AI ​​fusion model embedded in the hardware of the AI ​​early warning module and outputs early warning signals corresponding to the risk level. Meanwhile, through the hardware comparator and logic gate circuit in the fault diagnosis module, the standard voltage signal is compared with the preset threshold and logically combined to output the corresponding fault type identification signal. S4, the hidden danger elimination execution module triggers the execution of a differentiated proactive elimination strategy that matches the warning level and fault type based on the received warning signal and identification signal through relay hardware linkage; S5 monitors the battery status after the active elimination strategy is executed through the status feedback module and displays the results through the hardware indicator circuit; if the feedback status does not return to normal, secondary processing is automatically triggered.

[0014] Preferably, the differential active elimination strategy in step S4 includes: If it is a Level 1 warning, then the charging power will be adjusted. If it is a Level 2 warning, then the single cell equalization bypass will be executed and the basic cooling will be started; If it is a Level 3 warning, the main power supply will be cut off and enhanced cooling will be activated. If the fault type is diagnosed as a partial short circuit, the spraying of extinguishing media will be triggered.

[0015] The beneficial effects of this invention are as follows: 1. The core functions of this invention are realized through hardware modular integration and preset logic, eliminating the need to write and debug complex software, reducing development difficulty and implementation costs, and making it suitable for rapid prototype verification; the full hardware linkage control has no software running delay, and the early warning response time and hidden danger handling trigger time are both short, far exceeding the response efficiency of software control.

[0016] 2. This invention can avoid the impact of electromagnetic interference and system crashes on software operation, and has strong hardware logic stability, adapting to complex working environments such as vehicle and energy storage. The AI ​​model is solidified through hardware burning and trained based on 2,000 sets of experimental data, and the accuracy of early warning and fault identification has been greatly improved compared with the existing technology.

[0017] 3. Based on the fault type and warning level identified by hardware, this invention initiates targeted handling strategies, resulting in a very high success rate in eliminating potential hazards and avoiding user experience degradation or hazard recurrence caused by blind handling; the modular hardware design allows direct connection to the interface of existing power battery packs without large-scale modifications, making it highly practical for engineering applications. Attached Figure Description

[0018] Figure 1 This is a diagram of the overall system hardware architecture; Figure 2 This is a schematic diagram of the hierarchical sampling hardware architecture; Figure 3 The hardware logic diagram for the AI ​​early warning module is fixed. Figure 4 This is the hardware logic circuit diagram for the fault diagnosis module; Figure 5 Hardware linkage diagram for the execution module to eliminate potential hazards. Detailed Implementation

[0019] Example 1: like Figure 1 As shown, a power battery thermal runaway early warning and proactive hazard elimination system based on AI fusion includes: The data acquisition module is used to collect multi-dimensional operating parameters of the power battery in real time; The AI ​​early warning module is electrically connected to the data acquisition module. It is used to receive multi-dimensional operating parameters and perform thermal runaway risk level assessment based on the hardware-embedded AI fusion model, and output early warning signals. The fault diagnosis module, electrically connected to the data acquisition module, is used to identify the fault type of the power battery based on preset hardware logic and parameter thresholds, and output a fault identification signal. The hazard elimination execution module is electrically connected to the AI ​​early warning module and the fault diagnosis module, respectively. It is used to trigger the execution of the corresponding active elimination strategy through hardware linkage based on the received early warning signal and fault identification signal. The status feedback module, which is associated with the data acquisition module and the hazard elimination execution module, is used to monitor and provide hardware indication of the execution effect of the proactive elimination strategy. The data acquisition module, AI early warning module, fault diagnosis module, hidden danger elimination execution module, and status feedback module are connected by hardware circuits and preset logic to achieve full-link linkage without the need for upper-level software control.

[0020] Preferably, the data acquisition module adopts a hierarchical sampling hardware architecture, including: The high-speed sampling layer is equipped with a first dedicated acquisition chip, which acquires the current and voltage signals of the power battery at a first sampling frequency, and is connected to a hardware filtering circuit. The medium-speed sampling layer is equipped with a second dedicated acquisition chip and sensor array to acquire the temperature and impedance signals of the power battery at a second sampling frequency lower than the first sampling frequency, and integrates a hardware comparator. The low-speed sampling layer is equipped with a third dedicated acquisition chip, which acquires the lithium-ion concentration and state of charge signals of the power battery at a third sampling frequency lower than the second sampling frequency. The signal conditioning circuit, connected to the high-speed sampling layer, medium-speed sampling layer and low-speed sampling layer, is used to convert the signals collected by each layer into a standard voltage signal and output it.

[0021] Preferably, the AI ​​early warning module includes an AI edge computing chip, and the AI ​​fusion model is a pre-trained fusion model of RepLKNet and BiGRU-GATT, which is solidified in the AI ​​edge computing chip by hardware programming. The AI ​​early warning module is configured to output at least three levels of early warning signals. Each level of early warning signal is represented by a high-level output through a different hardware pin to indicate a different level of thermal runaway risk.

[0022] Preferably, the fault diagnosis module is composed of a comparator group and logic gate circuit hardware; The comparator group has preset parameter threshold ranges corresponding to various typical fault types; The logic gate circuit is connected to the comparator group to perform logical combination of the comparison results and output high level through different hardware pins to identify different fault types.

[0023] Preferably, the various typical fault types include battery aging, partial short circuit, poor contact, overcharging and over-discharging, and external compression.

[0024] Preferably, the hazard elimination execution module includes a power control unit, a thermal management unit, and a chemical inhibition unit; The power control unit is configured to perform at least one of the following operations based on different warning levels: charging power adjustment, single cell bypass, or disconnection of the battery main circuit. The thermal management unit is configured to activate cooling strategies of varying intensities based on the warning level; The chemical suppression unit is configured to trigger the spraying of extinguishing media when specific warning levels and specific fault types are met.

[0025] Preferably, the chemical suppression unit includes a storage tank containing a composite extinguishing medium and an electromagnetic injection valve, wherein the composite extinguishing medium comprises hexafluoropropane and ammonium dihydrogen phosphate.

[0026] Preferably, the status feedback module includes a feedback sensor and a hardware indication circuit; The feedback sensor is used to collect key state parameters of the power battery after the active elimination strategy is executed; The hardware indicator circuit is connected to the feedback sensor and is used to drive indicator lights of different colors to light up based on the comparison results of the key status parameters and preset thresholds, so as to intuitively display the status of successful elimination of hidden danger, need for continuous monitoring, or failure to eliminate and need for secondary treatment.

[0027] Preferably, a method for an AI-based power battery thermal runaway early warning and proactive hazard elimination system includes the following steps: S1, through the hierarchical sampling hardware architecture of the data acquisition module, synchronously acquires the current, voltage, temperature, impedance, lithium ion concentration and SOC signal of the power battery, and converts them into standard voltage signals through the signal conditioning circuit. S2 synchronously inputs the standard voltage signal to the AI ​​early warning module and the fault diagnosis module; S3 processes standard voltage signals through the AI ​​fusion model embedded in the hardware of the AI ​​early warning module and outputs early warning signals corresponding to the risk level. Meanwhile, through the hardware comparator and logic gate circuit in the fault diagnosis module, the standard voltage signal is compared with the preset threshold and logically combined to output the corresponding fault type identification signal. S4, the hidden danger elimination execution module triggers the execution of a differentiated proactive elimination strategy that matches the warning level and fault type based on the received warning signal and identification signal through relay hardware linkage; S5 monitors the battery status after the active elimination strategy is executed through the status feedback module and displays the results through the hardware indicator circuit; if the feedback status does not return to normal, secondary processing is automatically triggered.

[0028] Preferably, the differential active elimination strategy in step S4 includes: If it is a Level 1 warning, then the charging power will be adjusted. If it is a Level 2 warning, then the single cell equalization bypass will be executed and the basic cooling will be started; If it is a Level 3 warning, the main power supply will be cut off and enhanced cooling will be activated. If the fault type is diagnosed as a partial short circuit, the spraying of extinguishing media will be triggered.

[0029] Example 2: The technical solution in this embodiment is a fully modular hardware architecture, including five core hardware units: "data acquisition module, AI early warning module, fault diagnosis module, hidden danger elimination execution module, and status feedback module". Signal transmission and action control are realized through preset logic circuits, as detailed below: Data acquisition module: Employing a hierarchical sampling hardware architecture, such as Figure 2 As shown, no software driver is required; standardized analog signals are directly output through a dedicated acquisition chip, covering six core parameters: High-speed layer: Equipped with a high-speed ADC dedicated acquisition chip AD9648, with a hardware-fixed 1MHz sampling frequency, it can acquire current signals of 20mA-200A and voltage signals of 0-5V. The hardware filtering circuit RC low-pass filter is used for initial noise reduction. The RC filter circuit is located at the output end of the high-speed layer acquisition chip and consists of a 1kΩ resistor and a 10nF capacitor. Mid-speed layer: Integrated monitoring chip BQ79616 and thermistor array, hardware-fixed 1kHz sampling frequency, collects temperature -40℃-120℃ and impedance ±0.01Ω accurate signals, built-in hardware comparator LM339 to realize the initial judgment of abnormal parameters, the thermistor array is evenly distributed in key positions inside the battery pack. Low-speed layer: Equipped with a dedicated SOC detection unit MAX17048, with a hardware-fixed 1Hz sampling frequency, it collects lithium-ion concentration and SOC State of Charge signals, and outputs a 4-20mA standard current signal; Signal aggregation: The signals collected from each layer are aggregated to the signal conditioning circuit, which includes an OP07 operational amplifier, via hardware terminals. The analog signals are uniformly converted into 0-3.3V standard voltage signals and transmitted to the AI ​​early warning module. The signal conditioning circuit is directly connected to the acquisition modules of each layer via DuPont wires.

[0030] AI Early Warning Module: It adopts the NVIDIA Jetson Xavier NX edge computing chip, which is dedicated to AI, and the hardware-fixed logic is as follows: Figure 3 As shown, the RepLKNet+BiGRU-GATT fusion model parameters are written to the chip firmware after pre-training via hardware programming, eliminating the need for software iteration. Hardware logic: Receives standard voltage signals from the data acquisition module, converts them into digital signals through the chip's built-in ADC interface with 8 channels and 12-bit resolution, and directly inputs them into the fixed AI model; Output function: Outputs a three-level warning signal via hardware pins, active high, 3.3V. Level 1 Early Warning: Pin 1 outputs a high level, corresponding to a risk level ≤30%; Level 2 Warning: Moderate Hazard - Pin 2 outputs a high level, corresponding to a risk level of 30%-70%. Level 3 Warning: Severe Hazard - Pin 3 outputs a high level, corresponding to a risk level >70%; Performance metrics: Hardware-based fixed model inference latency ≤50ms, early warning accuracy ≥95.2%.

[0031] Fault diagnosis module: Based on hardware comparators and logic gates, such as... Figure 4 As shown, fault type identification can be achieved through preset parameter thresholds without the need for software algorithms: Hardware configuration: Includes 5 sets of dedicated comparators LM339, NAND gate logic circuit 74HC00, preset parameter threshold ranges for 5 typical faults based on 2000 sets of experimental data, and the comparators and logic gate circuits are linked by soldering on PCB board. Identification Logic: Receives standard signals such as voltage, temperature, and impedance from the data acquisition module, compares them with preset thresholds using hardware, and outputs the fault type identification result through logic gate combinations. Hardware pin high-level indicator: Pin 4 high level: Battery aging lithium dendrite growth internal resistance > preset threshold 15%; Pin 5 high level: Local short circuit, local temperature rise rate > 5℃ / min; Pin 6 high level: Poor contact, voltage fluctuation > 0.2V; Pin 7 high level: Overcharge / overdischarge voltage > 4.35V or < 2.5V; Pin 8 high level: External compression of the battery pack causes impedance change >20%; Accuracy of identification: The hardware preset threshold has been calibrated experimentally, and the accuracy of fault identification is ≥93%.

[0032] Hazard elimination execution module: It consists of power control hardware, thermal management hardware, and chemical inhibition hardware, such as Figure 5 As shown, hardware linkage is achieved through relays and contactors, receiving high-level signals from the AI ​​early warning module and fault diagnosis module to trigger actions: Power control hardware: Core components: bidirectional DC / DC converter UCC28070, electromagnetic relay JQX-18F, 6V coil voltage, single-cell battery bypass switch rated current 50A; Action logic: Level 1 warning triggered: Relay K1 engages, initiating the hardware-based charging power adjustment to reduce the charging rate to 0.8C. Level 2 warning triggered: Relay K2 engages, initiating the hardware short-circuit of the abnormal cell in the single-cell battery equalization bypass; Level 3 warning trigger: Relay K3 engages, cutting off the battery main circuit and forcing a hardware power-off; Thermal management hardware: Core components: liquid-cooled circulating pump DC24V, flow rate 10L / min, miniature air-cooled fan 12V, wind speed 1.5m / s, temperature control relay; Action logic: Level 2 warning triggered: Temperature relay K4 engages, starting the liquid cooling circulation pump and fixing the temperature range to 25℃-35℃. Level 3 warning trigger: Relay K5 engages, liquid cooling pump runs at full load + air-cooled fan starts at the hardware-fixed maximum speed; Chemical suppression hardware: Core components: Miniature electromagnetic injection valve DC12V, rated pressure 0.8MPa; extinguishing medium storage tank 50mL, filled with a composite medium of hexafluoropropane and ammonium dihydrogen phosphate, 60% hexafluoropropane + 40% ammonium dihydrogen phosphate. Action logic: When the three-level warning is issued and the fault diagnosis module identifies the "partial short circuit" pin 5 as high, the relay K6 is activated, and the injection valve opens with a hardware-fixed injection time of 3 seconds.

[0033] Status feedback module: Composed of a feedback sensor and hardware indicator circuit, it monitors the battery status in real time after the potential hazard has been eliminated, and shares a signal conditioning circuit with the data acquisition module, such as... Figure 1 As shown: Feedback sensors: Equipped with a PT100 temperature sensor with an accuracy of ±0.3℃ and a voltage sampling resistor of 0.01Ω with an accuracy of ±1%, installed at key monitoring points of the battery pack; Indicator circuit: The elimination effect is displayed via LED indicator hardware. Green indicator LED1: The feedback signal is within the normal range, and the potential hazard has been successfully eliminated; Yellow indicator LED2: The feedback signal is approaching the threshold and requires continuous monitoring; Red indicator LED3: The feedback signal still exceeds the threshold, the hidden danger has not been eliminated, and the secondary processing hardware will automatically repeat the corresponding strategy.

[0034] Example 3: The system's working logic is as follows: Combination Figure 1 The overall hardware architecture diagram of the system, and the linkage logic of each module are as follows: Data acquisition: The hierarchical sampling hardware architecture synchronously collects six parameters of the power battery, converts them into standard voltage signals through the signal conditioning circuit, and transmits them synchronously to the AI ​​early warning module and the fault diagnosis module; AI Early Warning: The AI ​​edge computing chip processes signals through a fixed model and outputs a high-level early warning signal of the corresponding level, which is then transmitted to the hidden danger elimination execution module. Fault diagnosis: The hardware comparator and logic gate circuit output a high-level fault type indicator based on the parameter threshold, which is then transmitted to the hidden danger elimination execution module; Hazard elimination: Based on the high-level signal of the warning level and fault type, the relay of the execution module activates the corresponding hardware circuit and initiates power control, thermal management or chemical inhibition strategies; Status feedback: Feedback sensors monitor and process parameters, and the results are displayed via LED indicators. In case of an anomaly, secondary processing is automatically triggered, forming a closed-loop control.

Claims

1. A power battery thermal runaway early warning and proactive hazard elimination system based on AI fusion, characterized in that, include: The data acquisition module is used to collect multi-dimensional operating parameters of the power battery in real time; The AI ​​early warning module is electrically connected to the data acquisition module. It is used to receive multi-dimensional operating parameters and perform thermal runaway risk level assessment based on the hardware-embedded AI fusion model, and output early warning signals. The fault diagnosis module, electrically connected to the data acquisition module, is used to identify the fault type of the power battery based on preset hardware logic and parameter thresholds, and output a fault identification signal. The hazard elimination execution module is electrically connected to the AI ​​early warning module and the fault diagnosis module, respectively. It is used to trigger the execution of the corresponding active elimination strategy through hardware linkage based on the received early warning signal and fault identification signal. The status feedback module, which is associated with the data acquisition module and the hazard elimination execution module, is used to monitor and provide hardware indication of the execution effect of the proactive elimination strategy. The data acquisition module, AI early warning module, fault diagnosis module, hidden danger elimination execution module, and status feedback module are connected by hardware circuits and preset logic to achieve full-link linkage.

2. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 1, characterized in that, The data acquisition module adopts a hierarchical sampling hardware architecture, including: The high-speed sampling layer is equipped with a first dedicated acquisition chip, which acquires the current and voltage signals of the power battery at a first sampling frequency, and is connected to a hardware filtering circuit. The medium-speed sampling layer is equipped with a second dedicated acquisition chip and sensor array to acquire the temperature and impedance signals of the power battery at a second sampling frequency lower than the first sampling frequency, and integrates a hardware comparator. The low-speed sampling layer is equipped with a third dedicated acquisition chip, which acquires the lithium-ion concentration and state of charge signals of the power battery at a third sampling frequency lower than the second sampling frequency. The signal conditioning circuit, connected to the high-speed sampling layer, medium-speed sampling layer and low-speed sampling layer, is used to convert the signals collected by each layer into a standard voltage signal and output it.

3. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 1, characterized in that, The AI ​​early warning module includes an AI edge computing chip, and the AI ​​fusion model is a pre-trained fusion model of RepLKNet and BiGRU-GATT, which is solidified in the AI ​​edge computing chip through hardware programming. The AI ​​early warning module is configured to output at least three levels of early warning signals. Each level of early warning signal is represented by a high-level output through a different hardware pin to indicate a different level of thermal runaway risk.

4. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 1, characterized in that, The fault diagnosis module consists of a comparator group and logic gate circuit hardware. The comparator group has preset parameter threshold ranges corresponding to various typical fault types; The logic gate circuit is connected to the comparator group to perform logical combination of the comparison results and output high level through different hardware pins to identify different fault types.

5. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 4, characterized in that, Typical fault types include battery aging, partial short circuit, poor contact, overcharging and over-discharging, and external compression.

6. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 1, characterized in that, The hazard elimination execution module includes a power control unit, a thermal management unit, and a chemical inhibition unit; The power control unit is configured to perform at least one of the following operations based on different warning levels: charging power adjustment, single cell bypass, or disconnection of the battery main circuit. The thermal management unit is configured to activate cooling strategies of varying intensities based on the warning level; The chemical suppression unit is configured to trigger the spraying of extinguishing media when specific warning levels and specific fault types are met.

7. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 6, characterized in that, The chemical suppression unit includes a storage tank containing a composite extinguishing medium and an electromagnetic injection valve. The composite extinguishing medium contains hexafluoropropane and ammonium dihydrogen phosphate.

8. The AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 1, characterized in that, The status feedback module includes a feedback sensor and hardware indication circuitry; The feedback sensor is used to collect key state parameters of the power battery after the active elimination strategy is executed; The hardware indicator circuit is connected to the feedback sensor and is used to drive indicator lights of different colors to light up based on the comparison results of the key status parameters and preset thresholds, so as to intuitively display the status of successful elimination of hidden danger, need for continuous monitoring, or failure to eliminate and need for secondary treatment.

9. A method for an AI-based power battery thermal runaway early warning and proactive hazard elimination system according to any one of claims 1-8, characterized in that, Includes the following steps: S1, through the hierarchical sampling hardware architecture of the data acquisition module, synchronously acquires the current, voltage, temperature, impedance, lithium ion concentration and SOC signal of the power battery, and converts them into standard voltage signals through the signal conditioning circuit. S2 synchronously inputs the standard voltage signal to the AI ​​early warning module and the fault diagnosis module; S3 processes standard voltage signals through the AI ​​fusion model embedded in the hardware of the AI ​​early warning module and outputs early warning signals corresponding to the risk level. Meanwhile, through the hardware comparator and logic gate circuit in the fault diagnosis module, the standard voltage signal is compared with the preset threshold and logically combined to output the corresponding fault type identification signal. S4, the hidden danger elimination execution module triggers the execution of a differentiated proactive elimination strategy that matches the warning level and fault type based on the received warning signal and identification signal through relay hardware linkage; S5 monitors the battery status after the active elimination strategy is executed through the status feedback module and displays the results through the hardware indicator circuit; if the feedback status does not return to normal, secondary processing is automatically triggered.

10. The method for an AI-based power battery thermal runaway early warning and proactive hazard elimination system according to claim 9, characterized in that, The differential active elimination strategy described in step S4 includes: If it is a Level 1 warning, then the charging power will be adjusted. If it is a Level 2 warning, then the single cell equalization bypass will be executed and the basic cooling will be started; If it is a Level 3 warning, the main power supply will be cut off and enhanced cooling will be activated. If the fault type is diagnosed as a partial short circuit, the spraying of extinguishing media will be triggered.

Images