Underground new energy vehicle fire intelligent identification and emergency disposal method and system

Abstract

The present application relates to the technical field of underground fire early warning, and discloses a kind of underground new energy vehicle fire intelligent identification and emergency disposal method and system, communication connection acquisition vehicle end BMS bottom layer data obtains internal view data, deployment field sensor network to collect field multi-modal perception data and then obtain external view data;Based on the internal view data and external view data, cross validation is carried out to determine the fire;The deep fusion and dynamic weighted evaluation of internal-external dual view data construct a robust comprehensive analysis mechanism, realize the prevention of false alarm and false alarm;Based on the cross validation of fire determination, the comprehensive analysis mechanism outputs trigger instructions to the executing mechanism when the critical alarm threshold is reached;The executing mechanism is driven by an energy storage unit, which releases the fire blanket to the target area by electromagnetic ejection release method, so that the fire blanket covers the whole vehicle, achieves the purpose of isolating the fire area from the air, suppresses the spread of fire and smoke, and prevents the ignition of adjacent vehicles.

Description

Technical Field

[0001] This invention relates to the field of underground fire early warning technology, and in particular, to a method and system for intelligent identification and emergency response to underground new energy vehicle fires. Background Technology

[0002] With the rapid development of the new energy vehicle industry, fires caused by thermal runaway of lithium-ion batteries occur frequently, especially in enclosed or semi-enclosed spaces such as underground garages and charging stations. Due to limited ventilation and relatively confined space, fires are characterized by rapid spread, easy reignition, and the generation of toxic and harmful gases, posing new and severe challenges to traditional fire rescue.

[0003] Currently, emergency response to fires in such scenarios mainly relies on two technological approaches: one is passive response based on traditional fire-fighting facilities, such as dry powder fire extinguishers and automatic sprinkler systems; the other is early warning based on conventional fire detectors, such as point-type smoke / heat detectors. However, when facing the special disaster of thermal runaway fires caused by power batteries, existing technologies have significant shortcomings in terms of early identification accuracy, response timeliness, and the targeted nature of emergency response measures. Specifically:

[0004] 1. Insufficient early warning sensitivity, high false alarm and missed alarm rates, leading to missed optimal intervention opportunities. Battery thermal runaway is a process that originates internally and gradually spreads outwards. Traditional fire detectors (such as smoke and temperature sensors) are typically deployed in the external space of the vehicle, sensing only the "consequence" characteristics such as smoke or high temperatures that have already overflowed in the later stages of thermal runaway, resulting in a response lag of minutes. By this time, irreversible chain reactions of heat may have already occurred inside the battery. More importantly, the complex environment of underground parking garages means that vehicle lights, engine waste heat, and human activity can all trigger false alarms from conventional detectors. This coexistence of "insensitivity" and "unreliability" prevents the system from providing high-confidence alarms within the "golden window" of the initial fire outbreak, missing the optimal opportunity to intervene before open flames appear.

[0005] 2. In confined spaces, existing technologies struggle to achieve rapid, accurate, and automated early emergency response. Underground parking garages typically have low ceilings, complex structures, and densely packed vehicles. Once a vehicle catches fire, heat and toxic fumes accumulate rapidly and spread quickly to nearby vehicles through heat radiation and debris, creating a chain reaction. However, physical space limitations hinder the rapid entry and deployment of large firefighting equipment. Relying on manual alarm response methods, from discovery and confirmation to personnel arriving with equipment, is too time-consuming and cannot meet the urgent need for "second-level response" in the initial stages of fire suppression.

[0006] In summary, existing technologies for fire prevention and control of new energy vehicles in underground spaces face problems such as inaccurate early detection, inappropriate emergency measures, and imprecise automatic response. The root cause lies in the failure to construct a closed-loop prevention and control system that deeply integrates internal battery status information with multi-dimensional external environmental perception information, enabling intelligent judgment, precise location, and rapid, accurate, and automatic response. Therefore, there is an urgent need for an intelligent fire-fighting technology solution capable of early, accurate, automatic, and precise suppression of thermal runaway fires caused by power batteries. Summary of the Invention

[0007] This invention provides a method and system for intelligent identification and emergency response to fires in underground new energy vehicles. By integrating vehicle BMS interior data and external multimodal perception data for cross-validation and dynamic weighted analysis, and based on the analysis results driving a servo-aiming electromagnetic catapult mechanism to quickly and accurately cover the fire source with a fire blanket, this invention solves the technical problems of traditional fire-fighting methods in underground spaces being insensitive to early identification of lithium battery fires, having a delayed response, and being prone to heat diffusion.

[0008] According to one aspect of the present invention, a method for intelligent identification and emergency response to underground new energy vehicle fires is provided, comprising the following steps: S100, acquiring internal view data by collecting vehicle-side BMS bottom-level data through communication connection, and acquiring external view data by deploying an external field sensor network to collect external field multimodal perception data; S200, performing cross-validation based on the internal-external dual-view data of internal and external view data, and then determining the fire; S300, constructing a robust comprehensive judgment mechanism through deep fusion and dynamic weighted evaluation of internal-external dual-view data to prevent missed alarms and false alarms; S400, based on the cross-validated fire determination and the comprehensive judgment mechanism, outputting a trigger command to the actuator when the critical alarm threshold is reached; S500, the actuator, driven by an energy storage unit, deploys a fire blanket to the target area via electromagnetic ejection, so that the fire blanket covers the entire vehicle, achieving the purpose of isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles.

[0009] Further, the acquisition of internal visual data in step S100 specifically involves: establishing a communication connection and interacting in real time during the charging process of the target new energy vehicle, and acquiring the underlying data of the vehicle-side BMS in real time, including at least one of the following: rapid voltage drop data of individual cells, abnormal voltage difference data between individual cells, and internal characteristic temperature change data; the acquisition of external visual data in step S100 specifically involves: deploying an external field sensor network in the underground garage and parking spaces, collecting external field multimodal perception data through the external field sensor network, including thermophysical field characteristics, characteristic gas concentration characteristics, and visual characteristics, and then constructing a triple external field perception matrix of temperature field-gas field-visual field, and achieving three-dimensional and highly robust identification of the initial stage of a fire through the coupled analysis of multiphysical field information.

[0010] Further, step S200 specifically includes: S201, connecting to and acquiring the target vehicle's Battery Management System (BMS) data through at least one of the following connection methods: charging pile communication interface, vehicle network gateway, wired connection, and wireless connection. The BMS data is used as an auxiliary criterion for early fire identification. The BMS data includes at least individual cell voltage anomalies, total voltage anomalies, individual cell voltage drop change ΔV, battery pack temperature, and battery pack temperature rise rate. S202. When the BMS data shows abnormal gradient, abnormal persistence, or abnormal combination characteristics, it is determined to be an internal thermal runaway potential risk, and the monitoring frequency of the external field sensor network is activated or increased, and the review and evaluation stage is entered; S203. In the review and evaluation stage, the BMS data and the external field multimodal sensing data are spatiotemporally matched. When the internal abnormal signal of the BMS data and the corresponding external field multimodal sensing data of the external field coordinates continuously correspond within a preset time window and point to the same target new energy vehicle and / or the same area in space, a fire diagnosis command is generated. Through the mechanism of internal abnormality - external thermal image review - spatiotemporal consistency diagnosis, false alarms are reduced and the early identification capability of battery thermal runaway is improved.

[0011] Further, step S300 specifically includes: False alarm prevention coordination: When the visible light sensor in the external view data captures a suspected smoke outline or the gas sensor shows a slight concentration fluctuation, the BMS data from the internal view data and the infrared hotspot data from the external view data are forcibly cross-verified. If the individual cell voltage in the BMS data is normal and the infrared matrix does not capture any local abnormal hotspots, it will be comprehensively determined as cold start exhaust gas or dust interference from a fuel vehicle, thereby actively shielding false fire alarms and reducing the false alarm rate; Early warning coordination: When the target new energy vehicle's power battery is in the very early stage of thermal runaway, during the safety valve venting phase, before open flames are generated, if the BMS data in the internal view data detects a characteristic voltage drop in a certain cell, the infrared thermal imaging and gas... The alarm sensitivity weight of the body sensor is increased. Once the external observation data is synchronously cross-verified to detect a trace amount of characteristic gas or a local abnormal temperature rise of tens of degrees, it will immediately be associated with and judge that an early thermal runaway has occurred, and trigger the pre-aiming state to drive the actuator to be positioned in advance. To prevent missed alarms, the backup coordination is based on the fact that a serious thermal runaway of the power battery of the target new energy vehicle can easily lead to damage to the BMS main control board and communication loss. When the internal observation data communication is suddenly lost, the alarm will not be lifted, and the environmental backup judgment logic will be automatically triggered. At this time, the decision-making power will be transferred to the multimodal matrix of the external observation data in sequence. If the infrared temperature gradient and characteristic gas concentration continue to rise exponentially, even if the BMS data signal verification is missing, it will be comprehensively judged as irreversible thermal runaway, and the actuator will be forcibly activated to perform physical intervention.

[0012] Furthermore, step S300 also includes constructing a multi-parameter weighted evaluation dynamic model architecture, specifically: embedding a dynamic multi-parameter weighted evaluation mathematical model into the edge computing decision node to calculate the comprehensive fire severity score R.

[0013] in These represent the outlier of the standardized BMS data, the infrared thermodynamic outlier, the characteristic gas concentration, and the confidence level of the visual features, respectively. The weighting coefficients are dynamically allocated by the system state machine; the collaborative calculation feature, namely the dynamic weight drift mechanism, means that the weighting coefficients α, β, γ, and δ are not static constants, but deeply correlated variables based on the coupling state of multi-source data. When the internal data detects a clear voltage drop in a single BMS cell, the system automatically increases the characteristic gas weight γ and the infrared thermal weight β in the algorithm logic to improve the sensitivity to early thermal runaway. Conversely, when a severe thermal runaway occurs and causes the BMS data communication to be interrupted, the system forces α to zero and starts an adaptive compensation mechanism to dynamically amplify the weighting coefficients of β, γ, and δ. Finally, the system is only allowed to output the final trigger command to the actuator when the dynamic weighted comprehensive score R exceeds the set critical alarm threshold.

[0014] Further, step S500 specifically includes: S501, the actuator receives the trigger command and immediately starts the predetermined emergency response procedure; S502, the energy storage unit releases an instantaneous pulse current to the electromagnetic coil, the electromagnetic coil generates a magnetic field and applies magnetic force to the armature of the ejection mechanism, the armature is connected to the traction structure of the folded fire blanket, thereby causing the folded fire blanket to be rapidly ejected and deployed to the target location; S503, the fire blanket is rapidly deployed to the fire and / or thermal runaway area of ​​the target new energy vehicle to cover the fire source and isolate oxygen, thereby suppressing the spread of the fire.

[0015] Furthermore, step S500 also includes a safety interlock mechanism, specifically: when the actuator receives the trigger command, it monitors the status of on-site personnel and equipment in real time to ensure that no one is staying in the target area and the system is operating normally before the fire blanket is deployed; when it is detected that on-site personnel accidentally enter the target area or the system malfunctions, the safety interlock mechanism will immediately intervene to suspend or stop the deployment of the fire blanket in order to avoid injury to on-site personnel and prevent equipment malfunction, thus ensuring the safety and reliability of the entire emergency response process.

[0016] According to another aspect of the present invention, an intelligent identification and emergency response system for underground new energy vehicle fires is also provided, characterized in that it includes: an internal-external dual-view data perception module, used to acquire and fuse battery status data inside the target new energy vehicle and multimodal perception data of the external environment of the underground garage to form a comprehensive data source for fire determination; a distributed intelligent execution module, communicatively connected to the internal-external dual-view data perception module, used to perform comprehensive assessment of fire risk and fire source location based on the comprehensive data source, generate a preliminary trigger command containing location information, and generate a final trigger command through deep fusion and dynamic weighted evaluation; and an electromagnetic catapult module, controlled and connected to the distributed intelligent execution module, used to drive the launch array to perform spatial alignment according to the location information, and under the control of the final trigger command, drive the fire extinguishing catapult through electromagnetic force to launch the fire extinguishing catapult to cover the target fire source area.

[0017] Furthermore, the internal-external dual-view data acquisition network module includes an internal-view perception unit, an external-field multimodal perception unit, and a spatiotemporal synchronous data preprocessing unit. The internal-view perception unit includes: a protocol communication interface for communicating with the charging pile at the parking space where the target new energy vehicle is located or the vehicle data gateway of the target new energy vehicle, and subscribing to or requesting BMS data; and a BMS data parser for real-time interception, parsing, and extraction of low-level time-series signals strongly correlated with early warning of thermal runaway from the massive data frames uploaded by the vehicle's BMS. The external-field multimodal perception unit includes: a thermophysical field sensing array for non-contact scanning of the external field and generating a two-dimensional temperature field distribution map, and extracting local abnormal hotspot areas. Spatial temperature gradient and expansion rate of high-temperature regions; a characteristic gas sensing array for monitoring the concentration of characteristic gases closely related to lithium battery thermal runaway in the air, and extracting the concentration values ​​and time-varying rates of carbon monoxide, hydrogen, volatile organic compounds, and electrolyte solvent decomposition products; a visual and optical sensing array for capturing visible light and optical features in the early stages of a fire, and extracting smoke, open flames, and abnormal brightness based on computer vision algorithms; and a spatiotemporal synchronized data preprocessing unit for assigning unified timestamps and spatial location labels to all heterogeneous data streams, associating them with specific location numbers, performing preliminary filtering, noise reduction, and format standardization preprocessing, thereby achieving effective cross-validation of internal and external dual-view data.

[0018] Furthermore, the distributed intelligent execution module adopts a distributed control architecture that coordinates upper and lower level computers, and separates the communication master control from the execution master control to ensure the unity of image processing computing power requirements and the real-time performance of underlying actions. The distributed intelligent execution module includes: an upper level computer, which serves as the image and decision-making hub, uses an edge computing core board to receive the high-frequency spatial temperature matrix from the infrared temperature measurement array, run temperature compensation algorithms and multimodal fusion judgment logic, determine the fire and calculate the coordinates, and then issue a trigger command to the lower level computer and associate it with the target spatial coordinates; and a lower level computer, which serves as the hard real-time execution hub, uses a microcontroller as the underlying hardware execution end. After stripping away the heavy computing tasks, the lower level computer focuses on millisecond or microsecond-level deterministic hardware timing control, accurately scheduling subsequent execution structures, including relay closing, supercapacitor charging, and MOSFET discharging.

[0019] Furthermore, the electromagnetic catapult module includes: a servo aiming subsystem, comprising a moving guide rail and a dual-axis gimbal, the dual-axis gimbal having a pitch axis and a yaw axis, the dual-axis gimbal being suspended and deployed on the moving guide rail at the top of the underground parking garage; an electromagnetic catapult execution subsystem, mounted on the dual-axis gimbal, comprising multiple non-magnetically insulated launch tubes arranged in parallel, drive coils wound around the outside of each launch tube, and fire extinguishing catapult bodies disposed inside the launch tubes, forming a multi-tube synchronous launch array; a host computer is connected to the servo aiming subsystem for controlling the movement of the dual-axis gimbal and the moving guide rail, so that the launch axis of the electromagnetic catapult execution subsystem is aligned with the center of the fire source; the fire extinguishing catapult body includes an electromagnetic armature serving as the traction end of the fire blanket, the drive coil being electrically connected to the execution unit of the lower-level computer, for energizing upon receiving a hardware drive signal to generate a pulsed magnetic field, driving the electromagnetic armature to accelerate and eject under the action of electromagnetic repulsion, traction of the fire blanket to cover the target area.

[0020] The present invention has the following beneficial effects: 1. An early perception system based on "dual-view and cross-validation" was constructed, significantly improving the reliability, accuracy (preventing missed alarms and false alarms), and foresight of fire identification, solving the problems of slow response and high false alarm rate of traditional single perception methods. In step S100, the underlying data of the vehicle-side BMS (internal view data) and the multimodal perception data of the field (external view data) are acquired simultaneously, establishing a synchronous monitoring dimension for the "internal state" and "external representation" of the same target; in steps S200 and S300, cross-validation is performed based on the internal-external dual-view data, and the two types of data are deeply fused and dynamically weighted for evaluation, constructing a highly robust comprehensive judgment mechanism; when internal view data (such as a sudden drop in single-unit voltage or a sharp rise in characteristic temperature) indicates an internal anomaly, and external view data (such as infrared thermal imager capturing the corresponding...) indicates an internal anomaly, the system can detect anomalies in the fire detection system. When surface hotspots at the location and gas sensors detect a rise in characteristic gas concentration simultaneously exhibit anomalies, the strong temporal and spatial correlation between the two can greatly improve the confidence of fire determination and achieve early warning. Conversely, if there is only a single-dimensional data anomaly (such as only external non-fire source heat radiation detected, or only temporary interference in BMS data), the dynamic weighting mechanism can reduce its weight, thereby effectively filtering false alarms. This "internal and external combination and mutual verification" perception-determination method is key to improving the system's anti-interference ability and early identification ability in complex underground parking garage environments.

[0021] 2. It has achieved a closed-loop automatic response of "intelligent judgment, precise triggering, and rapid execution", which has reduced the response time from several minutes in the traditional manual mode to seconds or even sub-seconds in the system automation, effectively seizing the golden intervention window in the early stage of a fire. The comprehensive judgment mechanism constructed in step S300 and the critical alarm threshold triggering in step S400 constitute an intelligent decision-making center, replacing the lengthy and uncertain response chain of traditional methods that rely on manual discovery, confirmation, alarm, and re-deployment. Once the internal and external dual-view data reach the preset critical alarm threshold representing an extremely high risk of thermal runaway after fusion and judgment, the system immediately outputs a trigger command to the actuator in step S400 without manual intervention. In step S500, the electromagnetic catapult deployment method driven by energy storage units (such as supercapacitors) has the core advantage of extremely rapid energy release, achieving millisecond-level acceleration response, thereby accurately delivering the fire blanket to the target area in a very short time. The automatic closed loop of the entire process from perception and judgment to execution coverage shortens the time from anomaly detection to intervention, making it possible to implement physical isolation before open flames or deflagration occur, thus solving the pain point of slow response in traditional methods from a time perspective.

[0022] 3. A targeted emergency response method of "non-contact, instantaneous, and physical coverage" was adopted, achieving efficient, safe, and non-hazardous suppression of lithium battery fires. The electromagnetic catapult deployment method used in step S500 is a non-contact remote delivery technology. The actuator does not need to approach the core area of ​​the high-temperature, toxic, and explosive dangerous fire source, improving the system's survivability and reliability in extreme fire environments. The launched fire blanket covers the entire vehicle, isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles. Isolating oxygen can cut off the combustion chain and block the chain reaction of thermal runaway.

[0023] 4. This method establishes a fully automated and intelligent operation mode encompassing the entire process of "perception-decision-execution," fundamentally mitigating the life safety risks associated with personnel entering high-risk enclosed spaces for reconnaissance and initial response. The overall process constitutes a complete autonomous operating system; from data collection, fusion analysis, and intelligent judgment to the final triggering of the execution mechanism and deployment of fire blankets, the entire process requires no human intervention, avoiding potential delays in response due to panic, misjudgment, or operational errors. It possesses extremely important safety value for high-risk enclosed spaces such as underground parking garages that suffer from poor ventilation, difficult evacuation, and a high accumulation of toxic fumes.

[0024] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a preferred embodiment of the intelligent identification and emergency response method for underground new energy vehicle fires of the present invention; Figure 2 This is a schematic diagram of the weighted evaluation and prevention decision-making logic of a preferred embodiment of the present invention; Figure 3 This is a schematic diagram of the four-corner traction and ejection deployment of a single fire blanket according to a preferred embodiment of the present invention.

[0026] Legend: 501. First catapult; 502. Second catapult; 503. Third catapult; 504. Fourth catapult; 45. Towing connector / towing rope; 60. Fire blanket; 61. Four corner towing points of the first fire blanket; 62. Four corner towing points of the second fire blanket; 63. Four corner towing points of the third fire blanket; 64. Four corner towing points of the fourth fire blanket; 80. Vehicle; 90. Parking area. Detailed Implementation

[0027] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, the present invention can be implemented in many different ways as defined and covered below.

[0028] like Figure 1As shown, the intelligent identification and emergency response method for underground new energy vehicle fires in this embodiment includes the following steps: S100, collecting the vehicle-side BMS bottom-level data through communication connection to obtain internal view data, and deploying an external field sensor network to collect external field multimodal perception data to obtain external view data; S200, performing cross-validation based on the internal-external dual-view data of internal and external view data to determine the fire; S300, constructing a robust comprehensive judgment mechanism through deep fusion and dynamic weighted evaluation of internal-external dual-view data to prevent missed alarms and false alarms; S400, based on the cross-validated fire determination and the comprehensive judgment mechanism, outputting a trigger command to the actuator when the critical alarm threshold is reached; S500, the actuator, driven by the energy storage unit, deploys the fire blanket to the target area through electromagnetic catapult deployment to cover the entire vehicle, achieving the purpose of isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles. This invention discloses an intelligent identification and emergency response method for underground new energy vehicle fires. In step S100, the underlying data of the vehicle-side BMS (internal view data) and external multimodal perception data (external view data) are acquired simultaneously, establishing a synchronous monitoring dimension for the "internal state" and "external representation" of the same target. In steps S200 and S300, cross-validation is performed based on the internal-external dual-view data, and the two types of data are deeply fused and dynamically weighted for evaluation, constructing a highly robust comprehensive judgment mechanism. When internal view data (such as a sudden drop in single-cell voltage or a sharp rise in characteristic temperature) indicates an internal anomaly, while external view data... When anomalies are simultaneously observed (such as an infrared thermal imager capturing a surface hot spot at a corresponding location and a gas sensor detecting a rise in the concentration of a characteristic gas), the strong temporal and spatial correlation between the two can greatly improve the confidence of fire detection and achieve early warning. Conversely, if there is only a single-dimensional data anomaly (such as only external non-fire source heat radiation detected, or only temporary interference in BMS data), the dynamic weighting mechanism can reduce its weight, thereby effectively filtering false alarms. This "internal and external combination and mutual verification" perception-judgment method is key to improving the system's anti-interference ability and early identification ability in complex underground garage environments.The comprehensive judgment mechanism constructed in step S300 and the critical alarm threshold triggering in step S400 constitute an intelligent decision-making center, replacing the lengthy and uncertain response chain of traditional methods that rely on manual discovery, confirmation, alarm, and re-deployment. Once the internal and external dual-view data reach the preset critical alarm threshold representing an extremely high risk of thermal runaway after fusion and judgment, the system immediately outputs a trigger command to the actuator in step S400 without manual intervention. In step S500, the electromagnetic catapult deployment method driven by energy storage units (such as supercapacitors) has the core advantage of extremely rapid energy release, achieving millisecond-level acceleration response, thereby accurately delivering the fire blanket to the target area in a very short time. The automatic closed loop of the entire process from perception and judgment to execution coverage shortens the time from anomaly detection to intervention, making it possible to implement physical isolation before open flames or deflagration occur, thus solving the pain point of slow response in traditional methods from a time perspective. The electromagnetic catapult deployment method used in step S500 is a non-contact, remote delivery technology. The actuator does not need to approach the core area of ​​a high-temperature, toxic, or explosive fire source, thus improving the system's survivability and reliability in extreme fire environments. The deployed fire blanket covers the entire vehicle, isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles. Isolating oxygen can break the combustion chain and block the chain reaction of thermal runaway. The overall process of this invention constitutes a complete autonomous operating system. From data acquisition, fusion analysis, and intelligent judgment to the final triggering of the actuator and deployment of the fire blanket, the entire process requires no human intervention, avoiding potential delays in response due to panic, misjudgment, or operational errors. It has extremely important safety value for high-risk enclosed spaces such as underground parking garages that are poorly ventilated, difficult to evacuate, and prone to the accumulation of toxic fumes. This invention presents an intelligent identification and emergency response method for underground new energy vehicle fires. Through cross-validation and dynamic weighted fusion analysis of internal and external dual-view data, a highly reliable early-stage intelligent fire identification mechanism is constructed. Automatic triggering based on critical thresholds and rapid, precise deployment via electromagnetic catapults achieve a rapid closed-loop response from identification to response. Finally, the physical coverage of fire blankets provides an efficient and safe means of suppressing lithium battery fires. Each step is interconnected and works synergistically to solve a series of technical problems in existing technologies, such as inadequate detection, false alarms, missed alarms, delayed response, and unsuitable methods for detecting new energy vehicle fires in underground spaces. This method achieves accurate early warning and rapid, effective, and automatic response to early fire risks, significantly improving the safety protection level of underground spaces.

[0029] In this embodiment, the acquisition of internal observation data in step S100 specifically involves: establishing a communication connection and interacting in real time during the charging process of the target new energy vehicle to acquire the underlying data of the vehicle-side BMS, including at least one of the following: rapid voltage drop data of individual cells, abnormal voltage difference data between individual cells, and internal characteristic temperature mutation data; the acquisition of external observation data in step S100 specifically involves: deploying an external field sensor network in the underground garage and parking spaces, and collecting external field multimodal perception data through the external field sensor network (including the extreme values ​​of surface hot spots and temperature gradients extracted by the infrared thermal imaging matrix, the rate of increase of characteristic exhaust concentrations such as CO / VOCs, and the confidence level of smoke / open flame features extracted by the visual algorithm, etc.). The external field multimodal perception data includes thermophysical field features, characteristic gas concentration features, and visual features, thereby constructing a triple external field perception matrix of temperature field, gas field, and visual field. Through the coupled analysis of multiphysical field information, a three-dimensional and highly robust identification of the initial stage of a fire is achieved. Thermal runaway of a power battery is a physicochemical process with a clear temporal characteristic. In step S100, a communication connection is established and real-time interaction is performed during the charging process to obtain the underlying data of the vehicle-side BMS, which is equivalent to directly "listening" to the "vital signs" of the battery system. A rapid drop in the voltage of a single cell usually points to a severe internal short circuit. The voltage difference between single cells often reflects the inconsistency of the battery pack, which may be a precursor to local overheating. Sudden changes in internal characteristic temperature are the most direct intrinsic signals for the initiation of a thermal runaway chain reaction. These intrinsic data are the "causal signals" and "early precursors" of thermal runaway, with the highest directness and foresight. The multimodal sensing data of the field collected by deploying the field sensing network in step S100 is the capture of the external "manifestation signals" of thermal runaway. The extreme values ​​of surface hot spots and temperature gradients (thermophysical field characteristics) extracted by the infrared thermal imaging matrix reflect the process of heat conduction and accumulation from the inside of the battery pack to the outer shell; the rate of increase of characteristic exhaust concentrations (such as CO / VOCs) (characteristic gas concentration characteristics) captures the characteristic gases released by electrolyte decomposition and pressure relief valve opening; the confidence level of smoke / open flame features extracted by the visual algorithm (visual features) confirms open flames and visible smoke. Constructing a triple external field perception matrix of temperature field, gas field, and visual field means observing the same phenomenon from three independent physical dimensions: thermal, chemical, and optical. The acquisition of internal visual data and external multimodal perception data constitutes a complete data chain of "internal electro-thermal state" and "external thermal-chemical-optical phenomena"; through the coupled analysis of multi-physics field information, the system can not only see external phenomena but also perceive changes in internal state, thereby achieving "three-dimensional and highly robust recognition" of the early stages of a fire, even before the appearance of open flames and dense smoke. While single-dimensional signals may be subject to interference or lag, the confidence level increases exponentially when multiple independent and physically related signals are coupled and corroborated in space and time. Traditional fire alarms rely on a single sensor reading exceeding a fixed threshold (such as temperature > 70℃ or smoke concentration > 5%obs / m).In complex underground parking environments, such thresholds are easily triggered by non-fire factors such as vehicle engine waste heat, human smoking, and dust, leading to false alarms. The data acquired in step S100 of this invention are all time-series data with high-dimensional characteristics, rather than isolated scalars. For example, the "rapid" characteristic of the rapid drop in voltage of a single battery cell, the rate of increase in characteristic gas concentration, and the spatial distribution and temporal rate of change of surface temperature gradient, including trends and spatial distribution; acquiring this native high-dimensional data makes it possible to perform cross-validation and dynamic weighted evaluation in subsequent steps S200 and S300. The system no longer needs to rely solely on whether the absolute value exceeds the limit at a certain instant, but can analyze whether the temporal pattern of voltage and temperature changes in the internal observation data matches the spatiotemporal pattern of hot spot diffusion and gas concentration increase in the external multimodal sensing data. For example, a slow and uniform temperature rise on the vehicle body surface (possibly due to sunlight) and the rapid drop in battery cell voltage inside the BMS cannot be coupled temporally and spatially, and the system can determine it as interference through coupling analysis of multiphysics information. This analytical capability based on multi-dimensional, high-dimensional feature temporal correlation is the core technological foundation for the system to achieve highly robust identification and effectively distinguish between real battery thermal runaway and various environmental disturbances, solving the problem of high false alarm rate in traditional systems. The perception system constructed in step S100 is a rich set of multi-physics field feature data with spatiotemporal labels describing the event evolution process, providing an information advantage for downstream processing modules (steps S200-S500). For example, the infrared thermal imaging matrix provides not only the highest temperature point but also the entire temperature field, enabling the system to not only determine "whether there is a fire" but also provide centimeter-level precision target area coordinates for electromagnetic catapult deployment in step S500 through the precise location and range of the hot spot. The location information of abnormal cells in the internal observation data can be cross-validated with the location of external infrared hot spots to achieve "three-dimensional location" of the fire source (location within the battery pack and vehicle location). The rate of change of characteristic gas concentration characteristics can serve as an important basis for judging the thermal runaway stage (such as before depressurization, during depressurization, and open flame stage), providing situation assessment for the decision-making system. Therefore, the data acquisition method defined in detail in step S100 ensures that subsequent functions such as precise positioning, stage judgment, and automatic aiming can be realized in the entire system.

[0030] Multimodal sensing data in the field includes, but is not limited to: infrared thermal imaging data, the temperature field distribution map of the target surface obtained after processing, and the extraction of local hot spot temperature extremes and spatial temperature gradients from it. and temperature change time series Gas sensor data is used to monitor the concentrations of carbon monoxide (CO), total volatile organic compounds (TVOC), and hydrogen (H2) in real time, and to calculate their instantaneous concentration values ​​(c) and concentration change rates. Visual analysis data, based on visual algorithms such as YOLO / Transformer, performs real-time analysis of video streams and outputs quantitative features such as smoke coverage area, optical density, pixel ratio of open flame area, and flame flicker frequency.

[0031] Multimodal sensing data acquisition in the field: Fire characteristic data of new energy vehicles in underground parking garages are acquired through multimodal fire sensing methods. The fire characteristic data includes at least several sources, such as infrared thermal imaging temperature information, visible light image information, environmental parameter information, and smoke concentration information. Infrared thermal imaging is used primarily to monitor abnormally high temperatures in the lithium battery area and extract features such as the highest temperature, temperature rise trend, and spatial location of high-temperature areas. Visible light video is used to capture the visual features of open flames or smoke as auxiliary verification of thermal anomalies. Environmental sensors provide parameters such as ambient temperature to reflect abnormal changes in the surrounding environment as verification information. Smoke sensors provide smoke concentration information to assist in locating the burning vehicle. The above multi-source fire characteristic data are fused and processed to form fire discrimination parameters, and a preliminary judgment of the fire state is made, i.e., whether the target vehicle shows signs of fire.

[0032] In this embodiment, step S200 specifically involves: S201, connecting to and acquiring the target vehicle's Battery Management System (BMS) data via at least one of the following connection methods: charging pile communication interface, vehicle network gateway, wired connection, and wireless connection. The BMS data is used as an auxiliary criterion for early fire identification. The BMS data includes at least individual cell voltage anomalies, total voltage anomalies, individual cell voltage drop change ΔV, battery pack temperature, and battery pack temperature rise rate. S202. When the BMS data shows abnormal gradient, abnormal persistence, or abnormal combination characteristics, it is determined to be an internal thermal runaway potential risk, and the monitoring frequency of the external field sensor network is activated or increased, and the review and evaluation stage is entered; S203. In the review and evaluation stage, the BMS data and the external field multimodal sensing data are spatiotemporally matched. When the internal abnormal signal of the BMS data and the corresponding external field multimodal sensing data of the external field coordinates continuously correspond within a preset time window and point to the same target new energy vehicle and / or the same area in space, a fire diagnosis command is generated. Through the mechanism of internal abnormality - external thermal image review - spatiotemporal consistency diagnosis, false alarms are reduced and the early identification capability of battery thermal runaway is improved. The target vehicle's Battery Management System (BMS) data acquired in step S201 covers comprehensive internal parameters from voltage and temperature to insulation status. Using this data as an auxiliary criterion for early fire identification is valuable because of its "leading" nature. The essence of thermal runaway in power batteries is thermal runaway caused by internal electrochemical imbalance. The appearance of signals such as abnormal voltage in individual cells and the rate of temperature rise in the battery pack often precedes externally observable heat, smoke, and fire phenomena. In step S202, when these BMS data exhibit abnormal gradients, abnormal persistence, or abnormal combinations, the system determines that there is a latent risk of internal thermal runaway. This determination initiates a review and assessment phase. The first step in this process is to activate or increase the monitoring frequency of the external field sensor network. This progressive architecture of "internal early warning - external verification" plays a crucial role: leveraging the high sensitivity of BMS data, it "sounds the alarm" at the earliest possible time, shifting the system from a routine monitoring state to a high-alert state, thus gaining a valuable time window for capturing fleeting early external features. It avoids hastily initiating final processing based solely on BMS data that may contain occasional interference, setting up the first "buffer" for system reliability. This phased and hierarchical processing logic is the core mechanism by which this method achieves the "early" goal while ensuring "accuracy." Step S203 performs spatiotemporal matching of the internal abnormal signals of the BMS data with the corresponding external field coordinates of the external field multimodal sensing data, which is a key leap in information fusion from "simple superposition" to "deep coupling." The correspondence between internal anomalies and external sensing data must be maintained within a preset time window. This means that an isolated, instantaneous BMS jump will not be diagnosed if it fails to trigger subsequent physical field changes that can be continuously observed by external sensors (such as a continuous rise in temperature or a continuous accumulation of gas concentration). This effectively filters out transient internal noise such as electrical interference.The system requires that internal and external signals spatially point to the same target new energy vehicle and / or the same area. For example, BMS data shows an abnormal temperature in a specific module, while an external infrared thermal imager accurately captures a continuously increasing surface hot spot at the corresponding location on the vehicle chassis; or a BMS fault code points to the vehicle, while a visual sensor identifies initial smoke around the vehicle. This spatial consistency is a powerful criterion for ruling out misjudgments caused by environmental heat sources (such as exhaust pipes of nearby vehicles) or fluctuations in the background concentration of ambient gases. Through the mechanism of internal anomaly - external thermal image verification - spatiotemporal consistency diagnosis, the system weaves the originally independent and potentially questionable internal visual data and external multimodal perception data into a chain of evidence that is closely related and mutually corroborating in time and space. Only when this chain of evidence is complete and robust will the system generate a fire diagnosis command. This mechanism fundamentally elevates fire judgment from a "possible" logic relying on a single sensor threshold to a "certain" logic based on the spatiotemporal correlation of multi-source information. It is a direct technical path to reduce false alarms and improve the early identification capability of battery thermal runaway. In step S202, the monitoring frequency of the field sensor network is activated or increased based on anomalies in BMS data, reflecting an intelligent resource scheduling strategy. Under normal, risk-free conditions, the field sensor network (especially high-power devices such as high-frequency scanning of infrared thermal imagers and high-frequency sampling of multi-component gas sensors) can be in a low-power routine inspection or dormant state, which helps reduce long-term energy consumption and equipment wear. Once the internal data indicates a risk, the system immediately and dynamically adjusts its strategy, concentrating limited monitoring and computing resources on the risk target. By increasing the sampling frequency and scanning density, more continuous and refined external feature data is obtained to support high-confidence spatiotemporal matching analysis. This "internal-driven, dynamically enhanced" monitoring mode enables the system to intelligently balance the contradictions between "all-weather coverage capability," "early warning sensitivity," and "system operation economy and lifespan," enhancing the feasibility and long-term operation and maintenance value of this technical solution in actual engineering deployments.

[0033] Based on the multimodal predictive early warning mechanism of internal and external dual-view, when the battery enters the initial stage of pressure relief and exhaust, this system uses the cross-fusion judgment of abnormal communication data (internal view data) of vehicle-side BMS and environmental multimodal sensors (infrared hot spots and characteristic gases, external view data) to drive the actuator to move above the target coordinates in advance; when thermal runaway exceeds the critical threshold, it automatically triggers ejection to suppress the fire before deflagration and effectively prevents the spread of heat radiation to the side vehicles.

[0034] In this embodiment, step S300 specifically includes: False alarm prevention coordination: When the visible light sensor in the external view data captures a suspected smoke outline or the gas sensor shows a slight concentration fluctuation, the BMS data of the internal view data and the infrared hotspot data of the external view data are forcibly correlated for cross-verification. If the individual cell voltage in the BMS data is normal and the infrared matrix does not capture any local abnormal hotspots, it will be comprehensively determined as cold start exhaust gas or dust interference from a fuel vehicle, thereby actively shielding false fire alarms and reducing the false alarm rate; Early warning coordination: When the target new energy vehicle's power battery is in the very early stage of thermal runaway, before open flames are generated, if the BMS data of the internal view data detects a characteristic drop in the voltage of a certain cell, the infrared thermal imaging and gas... The alarm sensitivity weight of the body sensor is increased. Once the external observation data is synchronously cross-verified to detect a trace amount of characteristic gas or a local abnormal temperature rise of tens of degrees, it will immediately be associated with and judge that an early thermal runaway has occurred, and trigger the pre-aiming state to drive the actuator to be positioned in advance. To prevent missed alarms, the backup coordination is based on the fact that a serious thermal runaway of the power battery of the target new energy vehicle can easily lead to damage to the BMS main control board and communication loss. When the internal observation data communication is suddenly lost, the alarm will not be lifted, and the environmental backup judgment logic will be automatically triggered. At this time, the decision-making power will be transferred to the multimodal matrix of the external observation data in sequence. If the infrared temperature gradient and characteristic gas concentration continue to rise exponentially, even if the BMS data signal verification is missing, it will be comprehensively judged as irreversible thermal runaway, and the actuator will be forcibly activated to perform physical intervention. Non-fire interference sources commonly found in underground parking environments (such as exhaust fumes from cold starts of fuel vehicles, dust, vehicle lights, etc.) can trigger some signals in the multimodal sensing data of the field, for example, causing the visible light sensor to capture a suspected smoke outline or the gas sensor to show slight fluctuations in concentration. If the system makes judgments based on the instantaneous signals of a single or a few sensors, it is very easy to generate false alarms. The false alarm prevention collaborative strategy of this invention solves this problem by introducing the logic of forced correlation cross-validation. Specifically, when the aforementioned easily interfered sensing signals are captured, the system does not directly trigger an alarm. Instead, it mandates that the internal observation data be collaboratively verified with another type of external observation data (referring to infrared hotspot data), which has different physical principles and stronger anti-interference capabilities. Since exhaust fumes or dust from fuel vehicles do not generate continuous, localized, high-power heat sources, the infrared matrix typically fails to capture localized abnormal hotspots. These interferences are unrelated to thermal runaway of new energy vehicle batteries, so the individual cell voltage in the BMS data should appear normal. By cross-verifying the suspicious signals from the visible light / gas sensor with the "normal state" of the BMS voltage data and the "no abnormal state" of the infrared hotspot data, the system can physically identify that the event does not possess the characteristics of battery thermal runaway (i.e., "smoke / gas but no heat, and the battery electrical state is normal"), thus comprehensively classifying it as interference and actively shielding false alarms. This proactive discrimination based on the logical association of multiple physical quantities can more intelligently and quickly suppress false alarms from the source compared to simple threshold filtering or delayed confirmation.In the very early stage of thermal runaway, during the safety valve venting phase, side reactions have already begun to occur inside the battery, generating characteristic gases. However, the temperature rise and gas release may not yet have reached the conventional alarm threshold. At this time, no open flame has been generated, and traditional smoke or flame detectors are completely ineffective. The early warning and coordination strategy of this invention utilizes the signal most sensitive to this stage: a characteristic voltage drop in a certain cell detected by the BMS in the internal data. Once this key precursor signal is captured, the system immediately enters a high-alert state. Its core action is to increase the alarm sensitivity weight of the infrared thermal imaging and gas sensor in the external data. This means that the system reduces the absolute requirements for the alarm thresholds of these external sensors, and... Instead, the system focuses on the presence and trend of the signal. As long as the external observation data is subsequently cross-verified to detect trace amounts of characteristic gas or local abnormal temperature rises of tens of degrees, the system considers that the internal electrical signal abnormality has triggered observable changes in the external physical field, thus meeting the early thermal runaway criteria. After the determination, the system does not wait for a more intense fire, but immediately triggers the pre-aiming state, driving the actuator to be positioned in advance. This linkage chain of "internal electrical signal triggering early warning - dynamically improving external monitoring sensitivity - weak external signal confirmation - pre-deployment of actuators" realizes the "prediction" and "pre-action" of the thermal runaway process, winning preparation time for subsequent instantaneous intervention at the deflagration critical point. Severe thermal runaway leads to damage to the BMS main control board and communication failure. At this point, the path relying on internal visual data for judgment completely fails. The anti-missing-report fallback collaborative strategy of this invention provides a redundant and reliable backup judgment channel. When internal visual data communication is suddenly lost, the system does not deactivate the alarm, but immediately switches to the environmental fallback judgment logic, sequentially transferring the judgment power to the multimodal matrix of external visual data. The "sequential" and "multimodal matrix" here reflect the rigor of the logic. It does not rely on a single external sensor, but requires that the two core physical quantities characterizing the fire intensity, infrared temperature gradient and characteristic gas concentration, both show a continuous and exponential upward trend. This trend is a typical characteristic of open flames or intense combustion in open spaces and has extremely high certainty. Even without BMS data signal verification, the exponential growth of the two physical quantities is sufficient to comprehensively determine irreversible thermal runaway. The ultimate guarantee of this strategy is reflected in the forced activation of the actuator for physical intervention, ensuring that under any circumstances, as long as the fire develops to the point of destroying the vehicle's own control system, the external system will definitely implement final intervention, forming the last and extremely robust line of defense for system reliability.

[0035] like Figure 2The diagram illustrates the weighted evaluation and prevention decision-making logic, constructing a hierarchical, progressive fire risk assessment and prevention decision-making logic system from multi-source input to intelligent decision-making. This system comprises three levels: a fire evaluation factor input layer, a weighted evaluation processing module, a fire risk assessment result layer, and a prevention decision-making logic layer. The fire evaluation factor input layer, as the data source, collects multi-dimensional raw data related to the fire state, specifically including temperature anomalies, temperature rise rates, fire duration, smoke characteristics, flame identification results, and scene risk factors. These factors cover multiple dimensions of fire characterization parameters, including thermal, optical, temporal, and environmental aspects. The weighted evaluation processing module receives the various evaluation factors from the input layer. Through the core processing logic of normalizing and weighting multiple evaluation factors, it transforms raw data of different dimensions and magnitudes into comparable standardized indicators. It then combines the weights of each factor for comprehensive evaluation, eliminating the bias of a single factor and reflecting the comprehensive impact of multiple factors on fire risk. The weighted data is output to the fire risk assessment result layer, generating a quantitative fire risk score, which is a comprehensive quantitative representation of the current fire situation. Ultimately, this score is passed as input to the prevention and control decision-making logic layer. Based on the fire risk score and preset thresholds / decision rules, prevention and control decision instructions are generated, realizing the mapping from risk assessment to specific prevention and control actions. Corresponding prevention and control measures are triggered according to the risk level. The overall logic embodies a closed-loop intelligent processing flow from data perception, feature extraction, comprehensive evaluation to decision execution, aiming to improve the accuracy of fire risk assessment and the pertinence of prevention and control decisions, achieving dynamic, hierarchical, and intelligent management of fire risks.

[0036] In this embodiment, step S300 further includes constructing a multi-parameter weighted evaluation dynamic model architecture, specifically: embedding a dynamic multi-parameter weighted evaluation mathematical model into the edge computing decision node to calculate the comprehensive fire urgency score R.

[0037] in These represent the outlier of the standardized BMS data, the infrared thermodynamic outlier, the characteristic gas concentration, and the confidence level of the visual features, respectively. The weighting coefficients are dynamically allocated by the system state machine; the collaborative calculation feature, namely the dynamic weight drift mechanism, means that the weighting coefficients α, β, γ, and δ are not static constants, but deeply correlated variables based on the coupling state of multi-source data. When the internal data detects a clear voltage drop in a single BMS cell, the system automatically increases the characteristic gas weight γ and the infrared thermal weight β in the algorithm logic to improve the sensitivity to early thermal runaway. Conversely, when a severe thermal runaway occurs and causes the BMS data communication to be interrupted, the system forces α to zero and starts an adaptive compensation mechanism to dynamically amplify the weighting coefficients of β, γ, and δ. Finally, the system is only allowed to output the final trigger command to the actuator when the dynamic weighted comprehensive score R exceeds the set critical alarm threshold.

[0038] Specific implementation of the weight dynamic shift mechanism (scenario calculation): To further clarify the dynamic weighted evaluation model in this invention ( The specific operating logic of the system is set to a critical threshold for triggering electromagnetic catapult launch. The system's preset initial baseline weights are: BMS weights. Infrared weight Gas weight Visual weight .

[0039] Example 1: Early pressure relief feature capture and adaptive sensitivity up-adjustment.

[0040] Scenario Input: An early internal short circuit occurs in the battery of a new energy vehicle. The vehicle-side BMS reports an abnormal voltage drop in a certain battery cell. The number surged to 90, but no open flame was observed at this time, and the infrared signature was not obvious. Only trace amounts of characteristic gases escaped ( ), visually smoke-free ( ).

[0041] Dynamic weight adjustment logic: The system master controller detects high confidence levels... The sudden change triggers the "early high-risk alert" state machine. The algorithm automatically increases the capture weights of gas and infrared light, adjusting them accordingly. .

[0042] Collaborative computing: .

[0043] Execution result: At this time If the system does not launch the vehicle but exceeds the pre-aiming threshold (e.g., 50 points), the system will automatically drive the electromagnetic catapult gimbal to move above the vehicle position and "lock the axle to stand by," thus achieving very early physical defense deployment.

[0044] Example 2: Adaptive fallback determination after BMS communication failure (preventing missed reports).

[0045] Scenario Input: The battery explodes and burns, instantly melting the BMS communication bus, resulting in the complete loss of BMS signals. It dropped to 0. Meanwhile, the infrared sensor detected extremely high temperatures (…). The gas sensor detected a high concentration of toxic gas. ), visually detected thick smoke ( ).

[0046] Dynamic weighting logic: If fixed weights are used, loss This will lead to a significantly lower total score. The algorithm in this invention, upon detecting a BMS disconnection, immediately triggers the "environmental fallback" state machine. This forces the... Return to zero ( ), and adaptively amplify the environmental perception weights proportionally, reallocating them to .

[0047] Collaborative computing: .

[0048] Execution result: At this time Under extremely harsh conditions where there was no support from the vehicle's internal data, the system successfully completed cross-validation by relying on weight drift, instantly triggering electromagnetic catapult and successfully achieving a last-ditch intervention.

[0049] Example 3: Cold start exhaust gas interference of fuel vehicles (preventing false alarms).

[0050] Scenario Input: An old gasoline-powered car starts cold in the target parking space, emitting a large amount of exhaust fumes. A visual sensor identifies the outline of dense smoke. The gas sensor captured part of the CO concentration ( However, since it is not a new energy vehicle, the BMS signal showed no abnormalities. Furthermore, the heat from the exhaust pipe did not reach the characteristic temperature gradient of battery thermal runaway. ).

[0051] Dynamic weighting logic: The system maintains the baseline anti-false alarm weights unchanged. ).

[0052] Collaborative computing: .

[0053] Execution result: At this time The system relied on the high weight of BMS and infrared to "veto" this environmental interference, and the electromagnetic catapult remained silent, completely avoiding serious false alarms caused by exhaust gas or dust.

[0054] In this embodiment, step S500 specifically includes: S501, the actuator receives a trigger command and immediately initiates a predetermined emergency response procedure; S502, the energy storage unit releases an instantaneous pulse current to the electromagnetic coil, the electromagnetic coil generates a magnetic field and applies magnetic force to the armature of the ejection mechanism, the armature is connected to the traction structure of the folded fire blanket, thereby causing the folded fire blanket to be rapidly ejected and deployed to the target location; S503, the fire blanket is rapidly deployed to the fire and / or thermal runaway area of ​​the target new energy vehicle to cover the fire source and isolate oxygen, thereby suppressing the spread of the fire. In step S501, the actuator immediately initiates the predetermined emergency response procedure upon receiving the trigger command, demonstrating the "hard real-time" characteristic of the system control. From the arrival of the command to the start of the program, there are no complex judgments or delays; it is a pre-fixed sequence of actions, ensuring that after the rigorous but potentially time-consuming intelligent analysis established in steps S200-S300, the execution link can start with the shortest possible additional delay. In step S502, the energy storage unit releases an instantaneous pulse current to the electromagnetic coil, utilizing the physical characteristic that energy storage components such as capacitors can instantly release ultra-large currents. The electromagnetic coil thereby generates a magnetic field and applies magnetic force to the armature of the catapult release mechanism. This process is based on the principle of electromagnetic induction. The energy conversion and motion generation process is almost inertia-free; from the release of current to the establishment of magnetic field and then to the generation of magnetic force, the entire process is completed within milliseconds or even microseconds. This direct and instantaneous "electric-magnetic-force" conversion mechanism fundamentally eliminates the mechanical transmission gaps, inertial delays, or pressure establishment times that are inevitable in traditional mechanical actuators (such as motor-driven gears, hydraulic cylinder pressure building, and spring compression and release). Therefore, the entire execution response (steps S501-S502) achieves an extremely fast and deterministic physical conversion from instruction to action, enabling the system's "decision speed" to be converted into "execution speed" without loss, which is the guarantee for ultimately achieving the goal of "second-level response and early suppression".The electromagnetic catapult launch mechanism is a non-contact drive method. The armature is accelerated and launched in a magnetic field, with no physical connection or friction between it and the drive coil. This avoids the need for traditional mechanical telescopic arms or mobile platforms to enter or approach the core area of ​​high-temperature, toxic, or explosive fire sources. It physically isolates the "fragile and delicate drive and control parts" from the "harsh fire zone," greatly improving the survivability and long-term reliability of the actuator under extreme conditions. The magnetic force generated by the electromagnetic coil can provide a large initial acceleration within a very short operating distance, causing the armature and its connected components to be launched. The folded fire blanket can be rapidly launched towards the target location at an extremely high initial velocity. This "explosive" delivery method allows it to quickly overcome air resistance over a distance of several meters and accurately reach the target, making it highly adaptable to the limited ceiling height and potential obstacles (such as pipelines) in underground parking garages. Because electromagnetic drive involves no combustion, no exhaust, and zero recoil, it does not impact the installation ceiling structure and does not introduce additional accelerants or explosion risks into the fire scene, ensuring safety and cleanliness. This execution method is tailor-made for the "limited space, high-risk environment, and requirement for speed and precision" of underground spaces. Step S503, by rapidly unfolding the folded fire blanket and covering the ignition and / or thermal runaway area of ​​the target new energy vehicle, instantly forms a dense physical barrier between the fire source and the air, cutting off the "accelerant" (oxygen) of the three elements of combustion, thereby rapidly isolating the heat source. This is particularly effective in suppressing battery fires that have already produced open flames.

[0055] The actuator employs a supercapacitor-driven "coilgun-style" electromagnetic catapult deployment: Upon receiving a trigger command (decision command), the actuator immediately initiates a predetermined emergency response procedure, rapidly deploying the fire blanket to the fire / thermal runaway area of ​​the target new energy vehicle via the electromagnetic catapult module. This covers the fire source, isolates oxygen, and thus suppresses the spread of the fire. To address the problems of slow response, high wear, and insufficient instantaneous power in traditional mechanical actuators such as springs / motors, this invention uses an electromagnetic catapult module driven by high-power-density energy storage devices. Each catapult unit is equipped with an energy storage device, preferably a supercapacitor bank, to provide instantaneous high-pulse discharge capability. The driving principle is as follows: Upon receiving a trigger signal, a power switching device (e.g., a relay, MOSFET, or IGBT) conducts, and the supercapacitor releases a pulse current to the electromagnetic coil. The coil generates a magnetic field that applies magnetic force to the armature connected to the fire blanket traction structure, causing the folded fire blanket to be deployed with a high initial velocity. This reduces mechanical wear, improves response speed and long-term standby reliability, and enhances consistency and safety across multiple operations.

[0056] In this embodiment, step S500 also includes a safety interlock mechanism, specifically: when the actuator receives the trigger command, it monitors the status of on-site personnel and equipment in real time to ensure that no one is staying in the target area and the system is operating normally before the fire blanket is deployed; when it detects that on-site personnel accidentally enter the target area or that the system malfunctions, the safety interlock mechanism will immediately intervene to suspend or stop the deployment of the fire blanket to avoid injury to on-site personnel and prevent equipment malfunction, ensuring the safety and reliability of the entire emergency response process. The emergency response execution in step S500 is an automated physical intervention process that pursues extreme speed. However, in real-world application scenarios such as underground parking garages, there is always the possibility of accidental movement of personnel (such as car owners and inspection personnel). The safety interlocking mechanism of this invention provides crucial safety assurance. Within the critical time window between the actuator receiving the trigger command and the actual ejection of the fire blanket, the system does not blindly and immediately execute the subsequent energy storage unit discharge and ejection actions. Instead, it initiates a new task in parallel: real-time monitoring of the status of on-site personnel and equipment. This mechanism expands the logic of "whether to execute" from a single "fire situation determination result" to the relationship between the fire situation determination result and the real-time safety status. Only when it is confirmed that no one is in the target area and the system is operating normally will subsequent physical intervention be permitted. While ensuring timely fire response, by adding a real-time, dynamic safety confirmation step, the possibility of accidental injury to on-site personnel due to the system's automated response is eliminated. This allows the highly automated system to be safely deployed in public or semi-public spaces where pedestrian traffic may exist, greatly expanding its application scope and acceptability. The synergistic effect of this mechanism is reflected in its independent monitoring and unified handling of two types of risks. On the one hand, regarding the risk of "people": monitoring (e.g., using visual sensors, microwave radar, etc. deployed on-site) for on-site personnel accidentally entering the target area. Fires in underground parking garages pose risks of toxic smoke and explosions, and personnel should ideally avoid the area. However, this interlock provides a final line of active protection as a precaution. On the other hand, regarding risks to equipment: abnormalities in the monitoring system's operation (such as actuator servo malfunctions, excessive aiming deviations, or abnormal energy storage unit status) are addressed. Whether the abnormality is related to personnel or equipment, the safety interlock mechanism intervenes immediately, suspending or halting the deployment of fire blankets. This reflects the "fail-safe" principle of safety design: when the system or its monitored objects experience any unexpected state that could affect safety, the system automatically adjusts to the safest state (suspending / aborting the action), preventing secondary accidents caused by equipment malfunctions (such as accidental damage to vehicles or equipment not on fire) and preventing injury to on-site personnel. By incorporating personnel safety and equipment safety into a single rapid decision-making loop, this mechanism significantly enhances the system's adaptability and reliability in complex and uncertain real-world environments.The judgment mechanism in steps S200-S300 mainly addresses the intelligent decision-making problem of whether a fire has occurred and its location, with the goal of accuracy. The main execution part of step S500 addresses the intelligent action problem of how to quickly respond to emergencies, with the goal of efficiency. The safety interlocking mechanism added in this invention solves the intelligent control problem of execution under the premise of ensuring absolute safety, with the goal of reliability. These three together constitute a complete intelligent system. The perception and judgment layer provides the basis for action, the rapid execution layer provides the action capability, and the purpose of the safety interlocking layer is to supervise the safety of action. The introduction of this mechanism makes the system not only an efficient emergency response automation tool, but also an intelligent safety entity with autonomous safety risk awareness, capable of dynamic risk assessment and intervention. It is not only an addition of functions, but also a sublimation of the system design concept. The entire technical solution, while pursuing performance indicators, strictly follows the highest principle of prioritizing the safety of personnel and property, and meets the stringent requirements of high-end fire-fighting equipment for system safety integrity.

[0057] During emergency response operations, the system also monitors the status of on-site personnel and equipment in real time. Safety interlock devices ensure that no one remains in the fire area and that the system is operating normally before the fire blanket is deployed. If personnel are detected accidentally entering the hazardous area or if the system malfunctions, the safety interlock mechanism will immediately intervene, suspending or halting the deployment of the fire blanket to prevent injury to personnel and equipment malfunction, ensuring the safety and reliability of the entire emergency response process. The safety interlock mechanism includes: The unmanned closed-loop automatic triggering module acts as the "brain" of the unmanned system, enabling automatic fire monitoring, judgment, and main switch triggering without manual intervention. It is suitable for unattended scenarios such as warehouses and computer rooms. The specific implementation is as follows: Hardware structure and connections: Thermal imaging temperature measurement unit: The MLX90640 infrared thermal imager is selected as the core of fire detection. Its resolution is 80×62 pixels and the field of view is 55°×35°. It communicates with the ESP32 master control unit through the SPI interface. The SPI clock frequency is configured to 1MHz. The SCK pin is connected to GPIO18 of ESP32, MOSI is connected to GPIO23, MISO is connected to GPIO19, and the CS chip select pin is connected to GPIO5. The module adopts 3.3V DC power supply and power consumption ≤100mW to meet the low power standby requirements of the device.

[0058] Temperature data processing chain: The ESP32 has a built-in dual-core processor, with Core0 dedicated to thermal imaging data acquisition and processing, and Core1 responsible for safety interlock verification and relay control. The dual cores process in parallel to avoid response delays caused by data blockage.

[0059] The logic of unmanned operation: After power-on, the ESP32 automatically enters unmanned monitoring mode and executes closed-loop control according to the following procedure: Initialization and self-test: Initialize the SPI interface and MLX90640 thermal imager, perform one full-frame temperature calibration, and if the thermal imager communication is abnormal, trigger a fault alarm and lock the device; Real-time temperature acquisition: Continuously acquires an 80×62 pixel temperature matrix at a sampling rate of 10Hz, acquiring 7680 temperature points in a single acquisition, with an acquisition period of ≤100ms; Data filtering and judgment: The collected temperature matrix is ​​filtered by moving average (window size 5) to remove environmental noise interference, and the highest temperature value of the whole field is calculated in real time; when the highest temperature is ≥150℃ and the duration is ≥200ms (to avoid false judgment of instantaneous high temperature, such as welding sparks), it is judged as a valid fire. Automatic trigger control: After confirming a valid fire and the personnel safety interlock signal is valid, the ESP32 outputs a high level for ≥500ms via GPIO14 to trigger the relay to engage. After the relay engages, the transmitter side automatically completes the entire process of "boost-charge-trigger-ejection". The ESP32 only acts as a "master switch" and does not participate in the timing and triggering details on the transmitter side, reducing program complexity and failure probability.

[0060] Quantitative results: Fire detection response time ≤100ms, false positive rate ≤0.1% (based on 1000 simulated open flame and interference tests), no manual operation is required in unattended scenarios, and the response speed is more than 90% faster than manual triggering; Optional embodiment: The thermal imaging temperature measurement unit can be replaced with MLX90641 (128×64 resolution), the sampling rate can be adjusted in the range of 5Hz-30Hz, and the fire judgment threshold can be adaptively set in the range of 100℃-200℃ according to the application scenario; the temperature data processing link can be replaced with an edge AI algorithm (such as a lightweight CNN) to realize fire type identification (such as open flame / smoldering) and further reduce the false judgment rate.

[0061] The safety interlock mechanism acts as the "highest safety gate," locking the device when personnel have not evacuated through hardware signal verification. This prevents accidental triggering and injury from both a programmatic and hardware perspective. The specific implementation is as follows: The hardware structure and connections employ a redundant "multi-signal AND logic" design, where the device is locked if any security signal is invalid. The hardware structure includes: Personnel detection input unit: A Jetson Orin Nano is used as the host computer to run the YOLO v8s model, detecting intrusion in the area in real time. If intrusion is detected, a high-level signal is output; otherwise, a low-level signal is output. If the ESP32 slave computer detects a low-level signal, it disconnects the relay and locks the device until no high-level signal is received for 500ms, at which point the locking stops. The output is connected to GPIO12 of the ESP32, configured as an input pin. Redundant emergency stop button unit: series normally closed industrial emergency stop button with IP65 protection rating, installed in an easily accessible location on the outside of the device; one end of the button is connected to a 3.3V DC power supply, and the other end is connected to GPIO13 of ESP32, which is configured as an input mode; Status indicator unit: GPIO12 (personnel detection) and GPIO13 (emergency stop) correspond to green LED1 and red LED2 indicator lights respectively. LED1 is lit to indicate that personnel have been evacuated, and LED2 is lit to indicate that the emergency stop has been triggered and the device is locked. The indicator lights are connected in series with a 1kΩ current-limiting resistor and connected to the corresponding GPIO pin and ground respectively.

[0062] The safety interlock and abnormal termination logic is fully embedded in the ESP32 program. It serves as the highest priority check before triggering and also supports abnormal termination after triggering. The specific logic is as follows: Interlock signal: Using "dual signal AND logic", the personnel safety interlock signal is considered valid only when GPIO12 (personnel detection) is low (personnel have been evacuated) and GPIO13 (emergency stop) is low (emergency stop not triggered); if either signal is high, the interlock signal is considered invalid. Forced verification before triggering: Before the relay is triggered by the output of GPIO14 high level, the ESP32 performs three consecutive samplings of the interlock signal (sampling interval 10ms). Only if all three sampling results are "valid" will the trigger action be performed. If the interlock signal is invalid, the ESP32 will always keep GPIO14 low level, the relay will be disconnected, and the transmitter side will not be powered on, thus achieving hardware-level safety lockout from the highest level. Triggered Abnormal Stop (Optional Extension): After the relay is energized and before the ejection action is completed, if the interlock signal becomes invalid (such as personnel entering the detection area or the emergency stop being pressed), the ESP32 can pull the GPIO14 level low within 10ms, disconnect the relay, cut off the 12V power supply on the transmitter side, and stop the subsequent boost, charging and ejection actions, realizing emergency stop in abnormal situations.

[0063] Quantitative results: Interlock signal response time ≤10ms, trigger success rate is 0 when personnel have not evacuated (verified by 1000 simulation tests), power outage time of device after emergency stop button trigger is ≤5ms, and safety protection level meets industrial safety standards; Optional embodiment: The personnel detection input unit can be replaced with a "personnel evacuation" command sent by a Bluetooth / Wi-Fi host computer (ESP32 receives the command via UART / Wi-Fi and parses it into an interlocking signal), the emergency stop button can be replaced with a safety light curtain (outputs a high level when an object blocks the detection area), and the interlocking logic can be expanded to "multi-sensor AND logic" (such as human infrared, safety door switch, host computer confirmation) to further improve safety redundancy.

[0064] This invention eliminates the reliance on manual on-site verification and operation. The system makes autonomous decisions through visual recognition and multimodal data algorithms; simultaneously, before physical intervention, it introduces a personnel prevention and accidental injury interlocking detection mechanism based on YOLO vision algorithms and microwave radar. From disaster early warning to precise fire blanket coverage, the entire process is unmanned, effectively ensuring the safety of on-site personnel and rescue forces.

[0065] The intelligent identification and emergency response system for underground new energy vehicle fires in this embodiment is characterized by comprising: an internal-external dual-view data perception module, used to acquire and fuse battery status data inside the target new energy vehicle and multimodal perception data of the external environment of the underground garage to form a comprehensive data source for fire determination; a distributed intelligent execution module, communicatively connected to the internal-external dual-view data perception module, used to perform comprehensive assessment of fire risk and fire source location based on the comprehensive data source, generate a preliminary trigger command containing location information, and generate a final trigger command through deep fusion and dynamic weighted evaluation; and an electromagnetic catapult module, controlled and connected to the distributed intelligent execution module, used to drive the launch array to perform spatial alignment according to the location information, and under the control of the final trigger command, drive the fire extinguishing catapult through electromagnetic force to launch the fire extinguishing catapult to cover the target fire source area.

[0066] In this embodiment, the internal-external dual-view data acquisition network module includes an internal view perception unit, an external field multimodal perception unit, and a spatiotemporal synchronous data preprocessing unit. The internal view perception unit includes a protocol communication interface for communicating with the charging pile at the parking space where the target new energy vehicle is located or the vehicle data gateway of the target new energy vehicle (parsing the charging pile communication protocol, such as GB / T 27930, ISO...). 15118, CCS, CHAdeMO, etc.), and subscribe to or request BMS data; BMS data parser, used to intercept, parse and extract low-level time-series signals strongly correlated with early warning of thermal runaway from massive data frames uploaded by vehicle BMS in real time (including: cell-level electrical characteristics, such as single cell voltage, voltage range, voltage difference, total voltage, current, etc.; thermal characteristics, such as temperature of key temperature measurement points in the battery pack, maximum temperature, minimum temperature, temperature rise rate, etc.; status and fault codes, such as insulation resistance, fault diagnosis codes, etc.); external field multimodal sensing unit includes: thermal physical field sensing array (such as infrared thermal imager array or high-density infrared temperature sensor array), used to scan the external field non-contactly and generate a two-dimensional temperature field distribution map, and extract local abnormal hot spot areas, spatial temperature gradient, and expansion rate of high temperature areas; characteristic gas sensing array (Such as electrochemical sensor arrays, photoionization detector arrays, or semiconductor sensor arrays), used to monitor the concentration of characteristic gases in the air closely related to the thermal runaway of lithium batteries, and extract the concentration values ​​and the rate of change over time of carbon monoxide, hydrogen, volatile organic compounds, and electrolyte solvent decomposition products; visual and optical sensing arrays (such as high-definition visible light camera arrays, smoke detector arrays, ultraviolet flame detector arrays, infrared flame detector arrays, etc.), used to capture visible light and optical features in the early stages of a fire, and extract smoke, open flames, and abnormal brightness based on computer vision algorithms; spatiotemporal synchronous data preprocessing unit is used to give all heterogeneous data streams a unified timestamp and spatial location label, and associate it with a specific location number, to perform preliminary filtering, noise reduction, and format standardization preprocessing, thereby achieving effective cross-validation of internal and external dual-view data.

[0067] In this embodiment, the distributed intelligent execution module adopts a distributed control architecture with collaboration between the upper and lower level computers, and separates the communication master control from the execution master control to ensure the unity of image processing computing power requirements and the real-time performance of underlying actions. The distributed intelligent execution module includes: an upper level computer, which serves as the image and decision-making hub, uses an edge computing core board to receive the high-frequency spatial temperature matrix from the infrared temperature measurement array, runs temperature compensation algorithms and multimodal fusion judgment logic, determines the fire and calculates the coordinates, and then issues a trigger command to the lower level computer and associates it with the target spatial coordinates; and a lower level computer, which serves as the hard real-time execution hub, uses a microcontroller as the underlying hardware execution end. After stripping away the heavy computing tasks, the lower level computer focuses on millisecond or microsecond-level deterministic hardware timing control, accurately scheduling subsequent execution structures, including relay closing, supercapacitor charging, and MOSFET discharging.

[0068] The distributed intelligent execution module includes a main control unit and relay execution units. These units form the core of the overall control and safety system, corresponding to the main control layer and the execution layer, respectively. They achieve electrical isolation between the low-voltage and high-voltage sides, normal hardware lock-up, and trigger-based on / off control. Their circuit structure and working principle are as follows: Main control core: ESP32 microcontroller is used, and GPIO14 is selected as the control output pin. This pin is configured in output mode and outputs a fixed low level under normal conditions to avoid false triggering caused by the uncertainty of the level at the moment of power-on.

[0069] Relay and isolation drive circuit: A 12V DC coil and a normally open high-power relay are used, with a rated contact voltage of 400V DC and a rated current of 40A, meeting the peak current requirements for synchronous operation of 4 ejection units; the relay coil is electrically isolated from the main control side through a PC817 linear optocoupler, with an isolation voltage ≥2500V AC. An S8050 NPN transistor is connected in series at the output of the optocoupler to drive the relay coil, and a 1N4007 freewheeling diode is connected in reverse parallel across the coil to absorb the reverse electromotive force at the moment the relay is disconnected, avoiding damage to the main control and drive devices by peak voltage.

[0070] Power on / off control logic: The normally open contact of the relay is connected in series in the +12V input main circuit of the high-voltage boost charging unit; in normal standby mode, the GPIO output is low, the relay coil is de-energized, the contact is open, and the subsequent high-voltage boost charging unit and ejection energy release unit are completely de-energized, with no high-voltage energy storage, achieving hardware-level safety lock-up; when triggered, the GPIO output is high for ≥500ms, the relay is energized, the +12V main circuit is turned on, and the subsequent circuit is synchronously powered on and started.

[0071] The distributed intelligent execution module includes a high-voltage boost charging unit. This unit is the core of the energy conversion process, corresponding to the energy layer, and converts a 12V low-voltage input to a 390V high-voltage DC output to charge the ejector energy storage capacitor. The specific circuit structure and working principle are as follows: Core control circuit: The UC3843 current-mode PWM controller is used as the boost control core. Its pin 8, VCC, is connected to a +12V input and controlled by a relay for switching on and off. A 150KΩ resistor and a 100nF capacitor are connected in series between pin 1 (COMP) and pin 2 (VFB) to form a loop compensation network. A 5.6KΩ resistor and a 4.7nF capacitor are connected externally to pin 3 (RT / CT) to set the PWM switching frequency to 65kHz. Pin 4 (GND) is grounded. A 0.1Ω / 2512 packaged current sampling resistor is connected in series to pin 5 (IS) to achieve primary-side cycle-by-cycle overcurrent protection. A 15Ω resistor is connected in series to pin 6 (OUT) to drive the subsequent switching transistor.

[0072] Flyback boost converter circuit: EE25 core flyback transformer T1 is used. Pin 5 of the primary winding is connected to the +12V input, and the other end is connected to the drain of a CSD18540Q5B N-channel MOSFET. The source of the MOSFET is connected to ground via a 0.1Ω sampling resistor, and the gate is connected to pin 6 of the UC3843 for output. The turns ratio of the transformer's secondary winding (pins 7-9) is 32:1. A US5M ultra-fast recovery diode is connected in series at the secondary output for rectification, outputting 390V high-voltage DC to charge the subsequent energy storage capacitor.

[0073] Voltage regulation feedback circuit: The high-voltage output terminal is divided by three 820KΩ / 1206 packaged resistors connected in series. The divided voltage signal is filtered by a 15KΩ resistor and a 10nF capacitor and then connected to pin 2 VFB of UC3843 to realize closed-loop voltage regulation of the output voltage. When the output voltage reaches the rated value of 390V, UC3843 automatically turns off the PWM output and stops charging to avoid overcharging and damage to the capacitor.

[0074] Redundancy protection circuit: Overvoltage protection: An overvoltage protection circuit is built using LM393 dual voltage comparators. The voltage divider sampling signal is connected to the non-inverting input of the comparator, and the reference voltage is connected to the inverting input. When the output voltage exceeds 400V, the comparator output flips, pulling down the COMP pin level of UC3843 and forcibly shutting down the PWM output. Over-temperature protection: A voltage divider network consisting of a 10KΩ NTC thermistor and a 10KΩ resistor is connected to the base of the J3YNPN transistor. When the ambient temperature or transformer temperature exceeds the threshold, the transistor turns on, pulling down the VCC power supply of the UC3843 and stopping the boost operation. Status indication: LED0 indicator in series in the +12V input circuit is used to indicate the on / off status of the relay; LED indicator in series in the output circuit is used to indicate the high voltage charging status.

[0075] The distributed intelligent execution module includes an electromagnetic catapult triggering and energy release unit. This unit is the core of the catapult execution, corresponding to the energy layer release branch of the solution, realizing high-voltage energy storage, automatic hardware triggering, and electromagnetic catapult energy release. The specific circuit structure and working principle are as follows: Energy storage unit: A 390V / 900μF electrolytic capacitor is used as the core of the ejection energy storage. The positive terminal of the capacitor is connected to the high-voltage output terminal of the high-voltage boost charging unit, and the negative terminal is connected to one end of the electromagnetic ejection coil. The rated energy storage is about 63.2J, which meets the energy requirements of the fire blanket ejection. A 4.7MΩ / 1206 packaged bleeder resistor is connected in parallel across the capacitor. After the device is powered off, the residual voltage of the capacitor can be discharged to below the safe voltage within 60 seconds. At the same time, a US1M diode and a charging indicator LED are connected in parallel. The LED lights up when the capacitor is charging and turns off when it is fully charged, which intuitively indicates the charging status.

[0076] Switching devices and drive circuit: An AUPS4070D1 IGBT is used as the main switch for ejector discharge. Its collector is connected to the positive terminal of the energy storage capacitor, and its emitter is connected to the other end of the electromagnetic ejector coil. The gate is driven by a hardware trigger circuit. The front stage of the trigger circuit uses a J3Y NPN transistor. Its base is connected to the PUT trigger signal input terminal through a 16KΩ resistor, and its collector is connected to the 21V drive power supply through a 2.2KΩ resistor. The emitter is grounded. The transistor's collector output is connected to the base of a Y2 PNP transistor through a 1KΩ resistor. The PNP transistor's emitter is connected to the 21V drive power supply, and its collector is connected to the IGBT gate, realizing level conversion and high-current drive to ensure rapid saturation and conduction of the IGBT.

[0077] RCD spike absorption circuit: An RCD absorption network is connected in parallel across the collector and emitter of the IGBT. It consists of two US5M ultra-fast recovery diodes connected in series and a non-inductive snubber coil L1. It can absorb the reverse spike voltage generated by the electromagnetic spring coil at the moment of IGBT turn-off, avoid voltage spike breakdown of IGBT, reduce electromagnetic interference, and improve device lifespan.

[0078] Hardware triggering logic: When the energy storage capacitor voltage is charged to 95% of the rated value of 390V, the PUT programmable unijunction transistor automatically outputs a trigger pulse, triggering the J3Y transistor to conduct, which in turn drives the PNP transistor to conduct. A 21V drive voltage is applied to the IGBT gate, and the IGBT quickly saturates and conducts. The energy storage capacitor discharges instantaneously with a large current through the electromagnetic catapult coil, converting electrical energy into mechanical energy and completing the fire blanket ejection action. The entire discharge process is ≤10ms, requiring no additional intervention from the main control unit, achieving hardware-level automatic triggering.

[0079] The system employs a distributed control architecture that combines upper and lower level control. The communication master control and execution master control are separated to ensure a balance between the computational demands of image processing and the extremely high real-time performance of underlying actions. The upper level computer (image and decision-making hub) uses an edge computing core board (such as Jetson Nano) specifically designed to receive the high-frequency spatial temperature matrix from the infrared thermometer array and run temperature compensation algorithms and multimodal fusion judgment logic. After determining the fire situation and calculating the coordinates, it issues simplified "trigger commands" and "azimuth parameters" to the lower level computer. The lower level computer (hard real-time execution hub) uses a microcontroller (such as the STC15 or ESP32 series) as the underlying hardware execution terminal. By removing the heavy computational tasks, the lower level computer focuses on millisecond / microsecond-level deterministic hardware timing control, precisely scheduling subsequent relay closing, supercapacitor charging, and MOSFET discharging.

[0080] In this embodiment, the electromagnetic catapult module includes: a servo aiming subsystem, comprising a moving guide rail and a dual-axis gimbal, the dual-axis gimbal having a pitch axis and a yaw axis, the dual-axis gimbal being suspended and deployed on the moving guide rail at the top of the underground parking garage; an electromagnetic catapult execution subsystem, mounted on the dual-axis gimbal, comprising multiple non-magnetically insulated launch tubes arranged in parallel, drive coils wound around the outside of each launch tube, and fire extinguishing catapults disposed inside the launch tubes, forming a multi-tube synchronous launch array; a host computer is connected to the servo aiming subsystem for controlling and driving the movement of the dual-axis gimbal and the moving guide rail, so that the launch axis of the electromagnetic catapult execution subsystem is aligned with the center of the fire source; the fire extinguishing catapult includes an electromagnetic armature serving as the traction end of the fire blanket, the drive coil being electrically connected to the execution unit of the lower computer, for energizing and generating a pulsed magnetic field when receiving a hardware drive signal, driving the electromagnetic armature to accelerate and eject under the action of electromagnetic repulsion, traction of the fire blanket to cover the target area, that is: traction of the fire blanket to cover the entire vehicle, achieving the purpose of isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles.

[0081] like Figure 3As shown, this top view illustrates the layout of an ejector-type fire blanket deployment system for new energy vehicles. The system includes a parking area 90, a fire blanket 60, and four ejectors. The parking area 90 is a rectangular frame structure used to define the working space. The vehicle 80 is parked in the center of the parking area (90) and serves as the target for emergency fire response. The fire blanket 60 is folded and positioned above the vehicle 80. Its four corners are equipped with a first fire blanket corner traction point 61, a second fire blanket corner traction point 62, a third fire blanket corner traction point 63, and a fourth fire blanket corner traction point 64, which are respectively connected to the corresponding traction connectors / traction ropes 45. The four ejectors are: the first ejector 50 1. Located on the outer side of the first corner of the parking space area 90, 2. Located on the outer side of the second corner, 3. Located on the outer side of the third corner, and 4. Located on the outer side of the fourth corner, 504 is connected to the four traction points of the fire blanket 60 (i.e., the four corner traction points 61, 62, 63, and 64 of the first, second, third, and fourth fire blankets) via their respective traction connectors / traction ropes 45. Upon receiving a trigger command, they can simultaneously release power to drive the folded fire blanket 60 to be rapidly ejected and deployed towards the fire or thermal runaway area of ​​the vehicle 80 to cover the fire source and isolate oxygen, thereby suppressing the spread of the fire.

[0082] The electromagnetic catapult module includes a multi-tube array synchronous expansion unit. This unit corresponds to a multi-tube array expansion design, achieving simultaneous launch of four tubes based on the aforementioned single-channel catapult unit. The specific implementation is as follows: Array topology: The four identical electromagnetic ejection triggering and energy release units are arranged in a rectangular four-corner symmetrical layout, corresponding to the four corners of the fire blanket; the +12V power input terminals of the four units are connected in parallel to the output terminal of the same relay, and the power supply of the four units is controlled by a single relay at the same time; the PUT trigger signal input terminals of the four units are connected in parallel to the same trigger signal source to achieve synchronous triggering.

[0083] Synchronization Guarantee Design: The four units adopt completely identical circuit design, component parameters, and PCB layout. The time deviation of the entire process of "power-on-boost-charging-trigger-ejection" for a single unit is ≤5ms. The relay is controlled by the same level signal output from a single GPIO pin of the main control unit, and the synchronization error of the trigger signal is ≤1μs. Each unit is equipped with an independent high-voltage boost charging circuit, with a charging voltage deviation of ≤2% and ejection energy consistency of ≥98%. From the hardware level, it is guaranteed that the four units eject synchronously, and the four corners of the fire blanket are subjected to force simultaneously without twisting or offset. The fire blanket can be fully deployed within ≤50ms.

[0084] Expansion capability: This topology can be directly expanded to 6-channel, 8-channel, 16-channel and other multi-channel arrays. Only the relays with the corresponding contact capacity need to be matched. There is no need to modify the main control hardware and control logic, which has a strong ability to adapt to different scenarios.

[0085] The system employs a mechanical deployment and a multi-tube array electromagnetic catapult mechanism. The guide rail and aiming system are integrated: the entire launch array and the host computer vision module are integrated and mounted on a dual-axis (pitch / yaw) gimbal, which is suspended on a moving guide rail mounted on the ceiling of the underground parking garage. After the host computer identifies an abnormal heat source, it drives the gimbal and guide rail motors to precisely align the physical axis of the launch array with the center of the fire source. The multi-tube launch array uses four parallel-arranged non-magnetic insulated launch tubes to form an array-type launch platform; each launch tube is independently wound with a high-frequency electromagnetic launch coil.

[0086] Projectile Launching Structure: The core of the projectile loaded inside the launch tube is a cylindrical metal electromagnetic armature (i.e., the traction end of the fire blanket). When the external coil is instantaneously energized to generate a pulsed alternating magnetic field, eddy currents are generated inside the metal cylinder, which is then subjected to a strong electromagnetic repulsion force (Lorentz force), causing it to accelerate instantaneously and be launched along the launch tube, precisely tractioning the fire blanket to cover the burning vehicle. More specifically, the fire blanket covers the entire burning vehicle, isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles.

[0087] This system effectively utilizes the space above the underground parking garage to lay miniature motion rails, avoiding traffic congestion and physical obstacles on the ground. Employing a supercapacitor-driven coil-gun electromagnetic catapult, it achieves precise load delivery with high kinetic energy and zero recoil within a compact mechanical structure, making it highly adaptable to low-headroom underground environments.

[0088] The servo aiming subsystem is a shared cruise and fixed-point delivery mechanism based on a top-mounted sliding rail, designed for multi-position navigation. Addressing the emergency response needs across parking spaces in large underground parking garages, this system overcomes the limitation of small coverage areas in traditional fixed sprinkler systems by designing a guide rail aiming system based on absolute position feedback and a visual servo closed-loop. Its underlying hardware and software control logic, relying on a master-slave architecture of a Jetson Nano host computer and an ESP32 slave computer, is as follows: Electrical topology for track drive and position feedback: Mechanical transmission unit: The guide rail system employs a linear transmission mechanism using a stepper motor in conjunction with an industrial-grade synchronous belt (or rack and pinion). The stepper motor's driver pulse (PUL) and direction (DIR) input terminals are connected to designated GPIO pins of the lower-level machine (ESP32 microcontroller). The ESP32, relying on its built-in hardware PWM peripheral (LEDC), can output high-frequency, extremely low-jitter, precise motion pulses without blocking, ensuring smooth gimbal movement.

[0089] Absolute position feedback unit: The system abandons the traditional open-loop stepper control and mounts a laser rangefinder sensor based on the TOF principle on the guide rail slider. The optical axis of this sensor is parallel to the guide rail, and it measures the absolute distance between the current gimbal and the physical origin of the guide rail in real time (with an accuracy of up to millimeters). The value is then reported to the host computer (Jetson Nano) via a high-frequency serial port, thus forming a fully closed-loop control system for the physical position.

[0090] Hierarchical control algorithms for "coordinate mapping preview" and "visual servo following": The system employs a hierarchical action logic, balancing the requirements for rapid response in cross-parking space dispatch with the precise aiming requirements for final firing. Rapid pre-aiming based on parking space coordinate mapping: When the system detects an internal warning signal from the BMS for a specific parking space, the host computer, Jetson Nano, immediately retrieves the "parking space number - guide rail absolute physical coordinates" mapping table stored in the memory. Jetson Nano calculates the physical difference between the target parking space coordinates and the current laser ranging feedback coordinates, and sends a long-distance movement command to the ESP32 via the high-speed serial bus. The ESP32 then drives the stepper motor to quickly slide the transmitting array directly above the target parking space, completing the initial "pre-aiming".

[0091] Two-dimensional hot spot feature extraction: When the system enters the infrared verification stage, Jetson Nano uses its GPU / CPU core to extract the pixel coordinates of the global highest temperature point in real time from the 32×24 temperature matrix transmitted back from the MLX90640 infrared array. ).

[0092] Visual servoing closed loop based on incremental PID: Jetson Nano compares the detected hot spot pixel coordinates with the preset ideal coordinates of the physical field of view center ( Perform a difference operation to obtain the current pixel offset. and .

[0093] (1) Offset After being calculated by the incremental PID algorithm inside the Jetson Nano, it is converted into the number of horizontal compensation pulses, which are sent to the ESP32 to control the guide rail stepper motor for fine-tuning of left and right translation; (2) Offset The angle of deflection of the gimbal pitch motor is calculated and driven by the PWM wave with the corresponding duty cycle output by the ESP32.

[0094] A multi-tube synchronous launch array is used to ensure the fire blanket deploys smoothly and avoids traction deviation and mid-air entanglement. This invention employs a multi-tube synchronous launch array, using a hardware synchronous triggering circuit to achieve microsecond-level synchronous conduction of four launch channels. This allows the metal blocks at the four corners of the fire blanket to be launched and pulled synchronously, ensuring uniform force on the fire blanket, preventing deviation and entanglement, and enabling it to deploy smoothly and completely cover the target area in one go. The synchronous triggering accuracy and launch parameters can be adaptively adjusted according to the parking space layout, target area, and safety interlock signals to improve the coverage success rate and reliability of underground space fire response.

[0095] Cross-sectional structure of single-tube launcher (non-magnetically insulated launch tube) and linkage deployment scheme of fire blanket: In order to achieve flat deployment of the fire blanket in the air and solve the problem of the static fixation and instantaneous release of the electromagnetic armature (projectile) in the launch tube, this system is designed with a dedicated throwing structure and linkage folding scheme. The specific static structure and connection logic are as follows: Static cross-sectional structure and fixing method of single-tube transmitter: (1) Layout of transmitting tube and coil: The transmitting tube is made of high-strength, insulating and non-magnetic composite material (such as epoxy resin fiberglass or special engineering plastic) to completely eliminate the electromagnetic eddy current loss of the tube wall itself. In terms of the winding method of the high-voltage driving electromagnetic coil, a multi-layer dense winding process is used to tightly wrap the enameled copper wire in the middle and lower section of the outer wall of the transmitting tube to form a high-density pulse excitation zone.

[0096] (2) Electromagnetic armature (projectile) structure: The cylindrical metal armature loaded inside the launch tube is preferably made of high magnetic permeability materials such as low carbon steel. The center of the tail of the armature is machined with a chamfered anti-cutting through hole (or a pre-embedded U-shaped metal pull ring) specifically for connecting the traction rope.

[0097] (3) Permanent magnet limiting base (core fixing method): To prevent the armature from slipping due to gravity during routine inspections or tilting aiming, a small high-strength permanent magnet (such as a neodymium iron boron magnet) is embedded in the bottom (rear end cover) of the launch tube. In normal standby mode, the tail end of the metal armature is firmly attracted and locked to the bottom of the launch tube by the permanent magnet; at the moment of launch, the huge electromagnetic pulse thrust of hundreds of Newtons generated by the electromagnetic coil can easily overcome the weak attraction of the permanent magnet, realizing frictionless and rapid unlocking and ejection of the armature.

[0098] Flexible traction connection and fire blanket compartment layout: (1) High-temperature resistant flexible traction link: The pull ring at the tail of the electromagnetic armature is connected to the edge of the fire blanket through a high-strength, high-temperature resistant flame-retardant flexible traction rope (preferably aramid fiber / Kevlar rope). The traction rope is led out through a reserved guide groove at the bottom of the launch tube to ensure that the rope is smooth and free from interference during launch.

[0099] (2) Central Blanket Storage Container and Four-Tube Diverging Array Layout: In terms of physical layout, a rectangular "central blanket storage container" is set in the center of the system, and four electromagnetic launch tubes are arranged on the four outer sides of the blanket storage container. In order to ensure that the fire blanket can be fully deployed in the air, the axes of the four launch tubes are not completely parallel, but are tilted outward relative to the central vertical line at a preset diffusion angle (for example, deflected outward by 5° to 15°).

[0100] The folding method and deployment logic of fire blankets in mid-air: (1) Z-shaped accordion folding method: In order to prevent the fire blanket from getting stuck or torn when it is pulled out at high speed, the fire blanket in the compartment adopts the "two-way Z-shaped accordion folding" process. First, the fire blanket is folded continuously in both directions along the length direction to form a long strip; then it is folded twice along the width direction, and finally compressed into a block and placed compactly in the central blanket storage compartment.

[0101] (2) Dynamic Deployment Process: When the four transmitters fire simultaneously, the four electromagnetic armatures fly out at high speed in four outward directions. With the enormous kinetic energy of the metal armatures and the outward flight trajectory, the fire blanket, which is in a folded state inside the storage compartment, is instantly pulled out through four aramid traction ropes of equal length. The fire blanket is subjected to centrifugal pull from the four corners in the air, and instantly unfolds completely from the center to the surrounding areas, finally covering the burning new energy vehicle precisely in a flat rectangular shape, achieving physical suffocation and heat source isolation.

[0102] Compared with existing technologies, the intelligent identification and emergency response method and system for underground new energy vehicle fires of the present invention has the following beneficial effects: (1) This invention improves the reliability of fire identification and reduces false alarms and missed alarms that may occur with a single sensing method by collecting and fusing multimodal information such as infrared, visible light and environmental parameters.

[0103] (2) The present invention introduces the fire discrimination logic of “internal-external dual view”, uses the vehicle BMS operation status as an early auxiliary criterion, and performs spatiotemporal consistency diagnosis with the external thermal imaging positioning results, thereby improving the ability to identify the early stage of lithium battery thermal runaway, while reducing false triggering caused by non-fire heat sources (such as brake disc heating).

[0104] (3) The present invention adopts a multi-parameter weighted risk assessment mechanism and combines time dimension (temperature rise rate, abnormal duration, etc.) to make graded decisions, thereby improving the accuracy and stability of prevention and control decisions and avoiding the limitations of traditional single-point threshold triggering.

[0105] (4) The present invention automatically generates prevention and control instructions and executes emergency response measures in the early stage of a fire, shortens the response time from identification to disposal, reduces the time spent on manual confirmation and disposal, helps to control the fire in the bud and reduce the probability of spread.

[0106] (5) The present invention adopts an electromagnetic catapult delivery mechanism driven by a supercapacitor, which provides higher instantaneous power and faster response capability, reduces mechanical wear and improves long-term standby reliability, enabling the fire blanket to cover the thermal runaway area more quickly and suppress the development of fire.

[0107] (6) The present invention adopts multi-tube array synchronous launch and microsecond-level trigger control to ensure that the traction force of the four corners of the fire blanket is uniform and the unfolding posture is regular, avoiding the blanket body tilting, curling or air entanglement interference during the ejection process, realizing rapid and complete sealing of the key areas of the whole vehicle, and greatly improving the coverage success rate and fire response reliability.

[0108] (7) The present invention introduces a personnel safety interlock and system status monitoring mechanism, which can automatically stop or adjust the handling action when personnel enter the danger zone or equipment malfunctions, thereby improving the system operation safety and engineering availability.

[0109] Matters not covered in this invention are common knowledge.

[0110] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0111] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

[0112] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent identification and emergency response to underground new energy vehicle fires, characterized in that, Includes the following steps: S100: Collects vehicle-side BMS underlying data through communication connection to obtain interior view data; and collects multimodal perception data in the field by deploying an external field sensor network to obtain external view data. S200: Based on the internal and external dual-view data, cross-validation is performed to determine the fire situation. S300 constructs a robust comprehensive judgment mechanism through deep fusion and dynamic weighted evaluation of internal and external dual-view data to prevent missed alarms and false alarms. S400, based on cross-validation fire detection and a comprehensive analysis mechanism, outputs a trigger command to the actuator when the critical alarm threshold is reached; The S500 actuator is driven by an energy storage unit and uses an electromagnetic catapult to deploy the fire blanket to the target area, so that the fire blanket covers the entire vehicle, isolating the fire area from the air, suppressing the spread of fire and smoke, and preventing the ignition of adjacent vehicles.

2. The method for intelligent identification and emergency response to underground new energy vehicle fires according to claim 1, characterized in that, The acquisition of internal view data in step S100 specifically involves: During the charging process of the target new energy vehicle, a communication connection is established and real-time interaction is performed to obtain the underlying data of the vehicle-side BMS in real time, including at least one of the following: data on the rapid voltage drop of individual cells, data on abnormal voltage differences between individual cells, and data on sudden temperature changes of internal characteristics. The acquisition of external view data in step S100 specifically involves: An outdoor sensor network is deployed in the underground garage and parking spaces to collect multimodal sensing data, including thermophysical field characteristics, characteristic gas concentration characteristics, and visual characteristics. This data is then used to construct a triple outdoor sensing matrix of temperature field, gas field, and visual field. Through the coupled analysis of multiphysical field information, a three-dimensional and highly robust identification of the initial stage of a fire can be achieved.

3. The method for intelligent identification and emergency response to underground new energy vehicle fires according to claim 2, characterized in that, Step S200 is as follows: S201. Connect to and acquire the target vehicle's Battery Management System (BMS) data via at least one of the following connection methods: charging pile communication interface, vehicle network gateway, wired connection, and wireless connection. Use the BMS data as an auxiliary criterion for early fire identification. The BMS data includes at least individual cell voltage anomalies, total voltage anomalies, and individual cell voltage drop changes. Battery pack temperature and battery pack temperature rise rate At least one of the following: charging / discharging status, abnormal current, abnormal insulation, or fault alarm signs; S202. When BMS data shows abnormal gradient, abnormal persistence, or abnormal combination characteristics, it is determined to be a potential risk of internal thermal runaway, and the monitoring frequency of the external field sensor network is activated or increased, and the process enters the review and evaluation stage. S203. During the review and evaluation phase, the BMS data and the external field multimodal sensing data are spatiotemporally matched. When the internal abnormal signal of the BMS data and the corresponding external field multimodal sensing data of the external field coordinates continuously correspond within a preset time window and point to the same target new energy vehicle and / or the same area in space, a fire diagnosis command is generated. Through the mechanism of internal abnormality - external thermal image review - spatiotemporal consistency diagnosis, false alarms are reduced and the early identification capability of battery thermal runaway is improved.

4. The intelligent identification and emergency response method for underground new energy vehicle fires according to claim 3, characterized in that, Step S300 specifically includes: To prevent false alarms, when the visible light sensor in the external view data captures a suspected smoke outline or the gas sensor shows a slight fluctuation in concentration, the BMS data in the internal view data and the infrared hot spot data in the external view data are forcibly linked for cross-verification. If the individual cell voltage in the BMS data is normal and the infrared matrix does not capture any abnormal hot spots, it will be comprehensively judged as exhaust gas or dust interference from a cold start of a fuel vehicle, thereby actively shielding false fire alarms and reducing the false alarm rate. Early warning coordination: When the power battery of the target new energy vehicle is in the safety valve venting stage of the very early stage of thermal runaway, before the open flame is generated, if the BMS data of the internal view data detects a characteristic drop in the voltage of a certain cell, the alarm sensitivity weight of the infrared thermal imaging and gas sensor of the external view data will be increased immediately. Once the external view data synchronously cross-verifies a trace amount of characteristic gas or a local abnormal temperature rise of tens of degrees, it will immediately be associated and determined that early thermal runaway has occurred, and trigger the pre-aiming state to drive the actuator to be positioned in advance. To prevent underreporting and ensure coordinated action, given that severe thermal runaway of the power battery of a target new energy vehicle can easily lead to damage to the BMS main control board and communication loss, when internal data communication is suddenly lost, the alarm is not deactivated, and the environmental fallback judgment logic is automatically triggered. At this time, the decision-making power is sequentially transferred to the multimodal matrix of external data. If the infrared temperature gradient and characteristic gas concentration continue to rise exponentially, even without BMS data signal verification, it will be comprehensively judged as irreversible thermal runaway, and the actuator will be forcibly activated to perform physical intervention.

5. The method for intelligent identification and emergency response to underground new energy vehicle fires according to claim 3, characterized in that, Step S300 also includes constructing a multi-parameter weighted evaluation dynamic model architecture, specifically: A dynamic multi-parameter weighted evaluation mathematical model is built into the edge computing decision node to calculate the comprehensive fire severity score R: in These represent the outlier of the standardized BMS data, the infrared thermodynamic outlier, the characteristic gas concentration, and the confidence level of the visual features, respectively. These are the weighting coefficients dynamically assigned by the system state machine; Collaborative computing features, namely the dynamic weight shifting mechanism: The weighting coefficients α, β, γ, and δ are not static constants, but rather deeply correlated variables based on the coupling state of multi-source data. When the internal data detects a clear voltage drop in a single BMS cell, the system automatically increases the characteristic gas weight γ and the infrared thermal weight β in the algorithm logic to improve the sensitivity to early thermal runaway. Conversely, when severe thermal runaway occurs and BMS data communication is interrupted, the system forces α to zero and activates an adaptive compensation mechanism to dynamically amplify the weighting coefficients of β, γ, and δ. Finally, the system is only allowed to output the final trigger command to the actuator when the dynamic weighted comprehensive score R exceeds the set critical alarm threshold.

6. The method for intelligent identification and emergency response to underground new energy vehicle fires according to any one of claims 1 to 5, characterized in that, Step S500 is as follows: S501. Upon receiving the trigger command, the executing agency shall immediately initiate the predetermined emergency response procedure; S502, the energy storage unit releases an instantaneous pulse current to the electromagnetic coil, the electromagnetic coil generates a magnetic field and applies magnetic force to the armature of the ejection mechanism, the armature is connected to the traction structure of the folded fire blanket, thereby causing the folded fire blanket to be ejected and deployed quickly to the target position. S503, fire blankets are rapidly deployed to the ignition and / or thermal runaway area of ​​the target new energy vehicle to cover the fire source and isolate oxygen, thereby suppressing the spread of the fire.

7. The method for intelligent identification and emergency response to underground new energy vehicle fires according to any one of claims 1 to 5, characterized in that, Step S500 also includes a safety interlock mechanism, specifically: When the executing agency receives the trigger command, it monitors the status of on-site personnel and equipment in real time to ensure that no one is in the target area and the system is operating normally before the fire blanket is deployed; When it is detected that on-site personnel have accidentally entered the target area or that the system is malfunctioning, the safety interlock mechanism will immediately intervene to suspend or stop the deployment of fire blankets in order to avoid injury to on-site personnel and prevent equipment malfunction, thus ensuring the safety and reliability of the entire emergency response process.

8. An intelligent identification and emergency response system for underground new energy vehicle fires, characterized in that, include: The internal-external dual-view data perception module is used to acquire and integrate battery status data inside the target new energy vehicle and multimodal perception data of the external environment of the underground garage to form a comprehensive data source for fire determination. The distributed intelligent execution module communicates with the internal-external dual-view data perception module to conduct comprehensive assessment of fire risk and fire source location based on comprehensive data sources, generate preliminary triggering instructions containing location information, and generate final triggering instructions through deep fusion and dynamic weighted evaluation. The electromagnetic catapult module, connected to the distributed intelligent execution module, is used to drive the launch array to perform spatial alignment based on position information. Under the control of the final trigger command, it uses electromagnetic force to drive the fire extinguishing projectile and launch it to cover the target fire source area.

9. The intelligent identification and emergency response system for underground new energy vehicle fires according to claim 8, characterized in that, The internal-external dual-view data acquisition network module includes an internal view perception unit, an external field multimodal perception unit, and a spatiotemporal synchronous data preprocessing unit; The introspective sensing unit includes: The protocol communication interface is used to establish a communication connection with the charging pile at the parking space where the target new energy vehicle is located or the vehicle data gateway of the target new energy vehicle, and to subscribe to or request BMS data. The BMS data parser is used to capture, parse, and extract low-level time-series signals that are strongly correlated with early warning of thermal runaway from the massive data frames uploaded by the vehicle BMS in real time. The external multimodal sensing unit includes: A thermophysical field sensing array is used to scan the external field in a non-contact manner and generate a two-dimensional temperature field distribution map, and to extract local anomalous hot spot regions, spatial temperature gradients, and the expansion rate of high-temperature regions. A characteristic gas sensing array is used to monitor the concentration of characteristic gases in the air that are closely related to the thermal runaway of lithium batteries, and to extract the concentration values ​​and the rate of change over time of carbon monoxide, hydrogen, volatile organic compounds, electrolyte solvent decomposition products, etc. A visual and optical sensing array is used to capture visible light and optical features in the early stages of a fire and extract smoke, open flames, and abnormal brightness based on computer vision algorithms. The spatiotemporal synchronization data preprocessing unit is used to assign a unified timestamp and spatial location label to all heterogeneous data streams and associate them with specific location numbers. It performs preliminary filtering, noise reduction, and format standardization preprocessing, thereby achieving effective cross-validation of internal and external dual-view data.

10. The intelligent identification and emergency response system for underground new energy vehicle fires according to claim 8, characterized in that, The distributed intelligent execution module adopts a distributed control architecture that coordinates upper and lower level machines, and uses an architecture that separates the communication master control and the execution master control to ensure the unity of image processing computing power requirements and the real-time performance of underlying actions. The distributed intelligent execution module includes: The host computer, as the image and decision-making center, adopts an edge computing core board to receive the high-frequency spatial temperature matrix of the infrared temperature measurement array, run the temperature compensation algorithm and multi-modal fusion judgment logic, determine the fire and calculate the coordinates, and then issue a trigger command to the lower computer and associate it with the target spatial coordinates. The lower-level machine, as the central hub for hard real-time execution, uses a microcontroller as the underlying hardware execution terminal. After stripping away the heavy computational tasks, the lower-level machine focuses on millisecond or microsecond-level deterministic hardware timing control, precisely scheduling subsequent execution structures, including relay closing, supercapacitor charging, and MOSFET discharging.

11. The intelligent identification and emergency response system for underground new energy vehicle fires according to claim 8, characterized in that, The electromagnetic catapult module includes: The servo aiming subsystem includes a moving rail and a dual-axis gimbal. The dual-axis gimbal has a pitch axis and a yaw axis. The dual-axis gimbal is suspended and deployed on the moving rail on the top of the underground parking garage. The electromagnetic catapult execution subsystem, installed on a dual-axis gimbal, includes multiple non-magnetic insulated launch tubes arranged in parallel, drive coils wound around the outside of each launch tube, and fire extinguishing catapults installed inside the launch tubes, forming a multi-tube synchronous launch array; The host computer is connected to the servo aiming subsystem for controlling and driving the movement of the dual-axis gimbal and the moving guide rail, so that the launch axis of the electromagnetic catapult execution subsystem is aligned with the center of the fire source; The fire extinguishing projectile includes an electromagnetic armature that serves as the traction end of the fire blanket. The drive coil is electrically connected to the execution unit of the lower-level machine. When a hardware drive signal is received, it is energized to generate a pulse magnetic field, which drives the electromagnetic armature to accelerate and eject under the action of electromagnetic repulsion, thereby traction of the fire blanket to cover the target area.

Images