A method for actively preventing thermal runaway in new energy vehicles based on vehicle-charging station-parking station coordination

Abstract

This invention discloses a method for actively preventing thermal runaway in new energy vehicles based on vehicle-charging pile-parking station coordination, comprising the following steps: upon detecting a charging start signal, controlling the ground active protection unit to perform reverse purging and forward suction operations of the air circuit system, constructing a clean air sampling boundary layer between the vehicle chassis and the ground; synchronously collecting multi-source raw data from the charging pile and the ground active protection unit in real time using a unified clock reference; performing feature decoupling and extraction on the collected data to obtain multi-source feature data; inputting the multi-source feature data into a multi-dimensional risk assessment model, using an improved D-S evidence theory for decision-level fusion, calculating the confidence level of each risk state and determining the current risk state; and implementing three-level control measures based on the risk assessment results to achieve early detection and active prevention of thermal runaway in new energy vehicles. This invention can significantly improve charging safety, reduce the risk of thermal runaway accidents, and has rapid response and precise intervention capabilities.

Description

Technical Field

[0001] This invention belongs to the field of new energy vehicle charging safety control technology, specifically relating to a method for actively preventing thermal runaway in new energy vehicles based on vehicle-charging pile-parking space coordination. Background Technology

[0002] With the global energy structure transformation and the rapid development of the new energy vehicle industry, the installed capacity of power batteries continues to expand, and high-voltage, high-power fast charging technology is widely used. Under high-rate charging conditions, lithium-ion power batteries are prone to thermal runaway under abnormal conditions such as overcharging, internal micro-short circuits within the cell, lithium plating due to aging, or external mechanical damage. Thermal runaway is characterized by its suddenness, rapid temperature rise rate, obvious chain propagation, and susceptibility to reignition. If it cannot be effectively identified and intervened in the early stages, it may develop from the failure of a single cell to the fire or even explosion of the entire vehicle in a very short time, seriously threatening the safety of people and public facilities.

[0003] Existing technologies mainly include vehicle-side battery management system (BMS) monitoring, electrical protection of charging piles, and fixed fire-fighting facilities in parking lots, but there are still significant shortcomings.

[0004] First, vehicle-side BMS primarily relies on voltage, current, and temperature parameters for threshold judgment, focusing on electrical anomaly monitoring and lacking the ability to perceive external environmental information such as gas release, acoustic anomalies, and chassis thermal runaway. During severe thermal runaway, information may be lost due to wiring erosion or communication interruption, affecting external systems' ability to obtain information about the risk status.

[0005] Secondly, in scenarios such as underground parking garages, top-mounted smoke or heat detectors are commonly used. Their triggering conditions depend on the smoke or high temperature rising to the detection height, and the response has a significant lag, making it difficult to cover the early stages such as battery safety valve exhaust or abnormal local chassis temperature rise.

[0006] Furthermore, in existing charging scenarios, vehicles, charging piles, and parking space fire-fighting facilities are independent of each other and lack a unified data interaction and collaborative analysis mechanism. This makes it impossible to achieve multimodal correlation analysis between current micro-perturbations, abnormal acoustic signals, gas concentration changes, and thermal image diffusion characteristics, resulting in both false alarms and false negatives, making it difficult to form highly reliable early identification results.

[0007] In addition, most existing automatic fire extinguishing devices use top spraying, which makes it difficult to accurately target the battery pack area at the bottom of the vehicle. They also lack a graded active intervention strategy linked to risk assessment models, making it difficult to implement precise blocking in the early stages of thermal runaway.

[0008] Regarding related improvements, Chinese patent application CN121417435A discloses a quantum-sensory driven active protection method and system for thermal runaway in charging piles. This solution collects quantum thermal runaway data in real time, constructs a thermal runaway spatiotemporal matrix, and utilizes a risk principal component extraction model and a trend prediction model to achieve early judgment of thermal runaway risk. Then, it implements graded protection responses and closed-loop optimization based on safety thresholds. This technology introduces a data modeling and prediction mechanism at the charging pile side, improving risk prediction capabilities to a certain extent.

[0009] However, this existing technology primarily analyzes thermal risk data from the charging pile side, with its sensing sources concentrated within a single system. It lacks deep collaboration with vehicle-side electrical status and parking space environmental sensing data, and has not established a unified risk identification framework across multiple nodes (vehicle-charging pile-parking space). Furthermore, this solution focuses on risk trend modeling and threshold judgment, failing to perform multi-source evidence-level fusion of gas anomalies, current micro-short circuits, acoustic rupture signals, and thermal diffusion characteristics, making it difficult to effectively address multi-sensor conflicts and false alarms. In addition, its protective response mechanism does not incorporate near-field directional intervention structures for coordinated control, making it difficult to achieve precise blocking at critical locations such as the bottom of the battery pack.

[0010] Therefore, existing technologies still lack a collaborative defense method that can connect multiple data channels between vehicles, charging piles, and parking spaces, integrate multi-dimensional electrical, gaseous, acoustic, and thermal features, achieve highly reliable risk identification through an evidence-level decision fusion model, and simultaneously coordinate tiered proactive intervention measures. This invention, by constructing a vehicle-charging pile-parking space collaborative architecture, decoupling and extracting multi-source features, and employing an improved DS evidence theory fusion mechanism, effectively overcomes the problems of single perception, fragmented data, and delayed intervention in existing technologies, achieving early and accurate identification and proactive prevention of thermal runaway risks in new energy vehicles. Summary of the Invention

[0011] The purpose of this invention is to overcome the defects of the prior art and provide a method for actively preventing thermal runaway of new energy vehicles based on vehicle-pile-parking space coordination.

[0012] The objective of this invention can be achieved through the following technical solutions:

[0013] This invention provides a method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination, comprising the following steps:

[0014] After detecting the charging start signal, the ground active protection unit is controlled to perform the sequential actions of reverse purging and forward suction of the air circuit system to build a clean air sampling boundary layer between the vehicle chassis and the ground.

[0015] Through a unified clock reference, multi-source raw data from charging piles and ground active protection units are collected in real time and synchronously.

[0016] Feature decoupling and extraction are performed on the collected multi-source raw data to obtain multi-source feature data;

[0017] The extracted multi-source feature data is input into a multi-dimensional risk assessment model, and decision-level fusion is performed through improved DS evidence theory to determine the current risk status.

[0018] Based on the assessment of the current risk status, three levels of control measures are implemented sequentially to actively prevent thermal runaway in new energy vehicles.

[0019] Furthermore, the control ground active protection unit executes the sequential actions of reverse purging and forward suction of the airway system to construct a clean air sampling boundary layer between the vehicle chassis and the ground, specifically including:

[0020] Step A1: After detecting the charging start signal, send an activation command to the ground active protection unit;

[0021] Step A2: After the ground active protection unit receives the activation command;

[0022] Step A3: Control the built-in air pump to rotate in reverse, spraying positive pressure airflow into the ground air intake micro-holes, and continue for a preset time to blow away the dust, water and debris deposited under the vehicle;

[0023] Step A4: Switch the air pump to forward rotation mode to generate a negative pressure flow field, and actively draw air from the vicinity of the vehicle battery pack into the sensor cavity through suction.

[0024] Step A5: Repeat steps A3 and A4, periodically performing reverse purging and forward aspiration operations until the gas sample collected by forward aspiration is stable and meets the preset cleanliness standard, then stop the cycle to obtain a clean air sampling boundary layer.

[0025] Furthermore, the multi-source raw data includes:

[0026] Charging current from the charging station;

[0027] Multi-component gas concentration data from the ground active protection unit is collected in real time by gas sensors on the vehicle chassis, including the concentration of each component of the collected gas.

[0028] The acoustic signal from the ground active protection unit is obtained by real-time acquisition of the acoustic signal during the charging process through a wideband microphone;

[0029] The infrared thermal image matrix from the ground active protection unit is obtained by monitoring the vehicle chassis and battery pack through an infrared thermal imager.

[0030] The spectral signals from the ground-based active protection unit are obtained by real-time acquisition of the spectral signals from the vehicle chassis through optical sensors.

[0031] Furthermore, the step of decoupling and extracting features from the collected multi-source raw data to obtain multi-source feature data specifically includes:

[0032] Statistical analysis is performed on the charging current in the multi-source raw data to calculate the deviation and variance between the real-time charging current and the reference current, and to obtain the electrical micro-short circuit characteristics.

[0033] Wavelet packet decomposition was performed on the acoustic signals in the multi-source raw data to extract the energy of different frequency bands, and the frequency band energy corresponding to the gas leakage and jet characteristics and the micro-fracture characteristics of battery materials was used as acoustic features.

[0034] Spatiotemporal gradient analysis is performed on the infrared thermal image matrix in multi-source raw data to calculate the temperature rise rate and thermal diffusion intensity, and thermal characteristics are generated by weighting.

[0035] Multi-component gas concentration data, electrical micro-short circuit characteristics, acoustic characteristics, and thermal characteristics are combined as multi-source feature data.

[0036] Furthermore, the statistical analysis of the charging current in the multi-source raw data, calculating the deviation and variance between the real-time charging current and the reference current, and obtaining the electrical micro-short circuit characteristics specifically includes:

[0037] Calculate the current deviation based on the charging current and the reference current. ,in, , These are the vehicle charging current and the reference current for normal vehicle charging, respectively, collected in real time by the ground active protection unit.

[0038] Within the preset time window Internal, for current deviation Integrate to obtain the cumulative deviation. :

[0039]

[0040] Calculate the current deviation within the preset time window variance within :

[0041]

[0042] in, This represents the total number of sampling points; For the first Current deviation at each sampling point; The mean of the deviation;

[0043] Based on cumulative deviation Sum of deviations and variances Determine the characteristics of electrical micro-short circuits:

[0044] when or At that time, electrical micro-short circuit characteristics are set. Otherwise ;in, and These are preset thresholds, obtained through statistical analysis of historical vehicle charging data, and preset according to battery type and charging environment.

[0045] Furthermore, the step of performing wavelet packet decomposition on the acoustic signals in the multi-source raw data to extract the energy of different frequency bands, and using the frequency band energy corresponding to the gas leakage and ejection characteristics and the microscopic fracture characteristics of the battery material as acoustic features, specifically includes:

[0046] The acoustic signals collected in real time by the ground active protection unit conduct Layer wavelet packet decomposition yields the node coefficients of each frequency band. The calculation formula is:

[0047]

[0048] in, Indicates the first Layer The frequency band node at the th frequency band node Frequency band node coefficients of each sampling point; These represent the filter coefficients of the selected orthogonal wavelet basis functions; For discrete-time indexing, The number of decomposition levels; This is the index of the current layer frequency band node. The original sound wave signal;

[0049] Based on the acoustic frequency range of the characteristics of micro-short circuits and gas leaks in vehicle batteries, a set of characteristic frequency band nodes is defined. This includes the frequency bands characteristic of battery rupture sound and gas leakage sound;

[0050] Based on the frequency band node coefficients obtained from wavelet packet decomposition, the values ​​in the characteristic frequency band node set are calculated. Characteristic frequency band energy within :

[0051]

[0052] Based on characteristic frequency band energy The ratio of the total energy across the entire frequency band to the normalized acoustic characteristics is calculated using the following formula:

[0053]

[0054] in, For acoustic characteristics, it represents the proportion of characteristic frequency band energy in the total energy of the entire frequency band, and is used to identify abnormal signal strengths of microscopic cracks or gas leaks in battery materials.

[0055] Furthermore, the process of performing spatiotemporal gradient analysis on the infrared thermal image matrix from the multi-source raw data, calculating the temperature rise rate and thermal diffusion intensity, and generating weighted thermal features specifically includes:

[0056] Set the infrared thermal image matrix acquired by the ground active protection unit to ,in, Represents the spatial pixel coordinates in an infrared image. Indicates the time of data collection. Indicates at time Time pixel Temperature value at that location;

[0057] At adjacent sampling times and Calculate the time gradient of temperature. The formula is:

[0058]

[0059] in, The sampling time interval between two adjacent infrared images. Represents pixels The rate of temperature rise;

[0060] At the same time The infrared thermal imaging matrix below Spatial gradient calculations are performed to obtain the thermal diffusivity. The formula is:

[0061]

[0062] in, It represents the magnitude of the temperature field gradient in space and is used to characterize the intensity of heat diffusion;

[0063] According to the preset temperature threshold Extracting regions of interest at high temperatures The high-temperature region of interest To meet A set of pixels;

[0064] Calculate the region of interest at high temperature The maximum temperature rise rate and average thermal diffusion intensity within the interior were determined, and thermal characteristics were constructed. The formula is:

[0065]

[0066] in, Indicates the region of interest for high temperature. The maximum rate of temperature rise of an internal pixel; Indicates the region of interest for high temperature. The average value of internal thermal diffusion intensity; , These are preset weighting coefficients.

[0067] Furthermore, the step of inputting the extracted multi-source feature data into a multi-dimensional risk assessment model and performing decision-level fusion through an improved DS evidence theory to determine the current risk status specifically includes:

[0068] Building a risk identification framework ,in, Indicates normal or sensor interference status. This indicates a thermal runaway warning state. Indicates the confirmed status of a fire;

[0069] Each feature extracted from the multi-source feature data is used as an independent source of evidence for the multi-component gas concentration data in the multi-source feature data. Electrical micro-short circuit characteristics Acoustic characteristics and thermal characteristics Calculate the basic probability assignment function separately, and use the sigmoid function to map each feature value to the support for different risk states, thus constructing the corresponding basic probability assignment function values. and :

[0070]

[0071]

[0072]

[0073]

[0074] in, The sensitivity coefficients for gaseous, electrical, acoustic, and thermal evidence were determined through historical data analysis and sensor performance evaluation. The preset threshold for the corresponding feature is set based on experimental data and actual application scenarios; and These represent the degree of support that gaseous, electrical, acoustic, and thermal characteristics provide for their respective risk states;

[0075] Calculate the conflict coefficient between each source of evidence based on the basic probability allocation function value of each independent source of evidence. The formula is:

[0076]

[0077] in, Risk identification framework any subset in Assign function values ​​to the basic probability of mutually exclusive propositions from different sources of evidence;

[0078] When the conflict coefficient Greater than the preset conflict threshold At that time, a weighting coefficient based on the historical reliability of the sensor is introduced. The basic probability allocation function values ​​for each source of evidence are corrected:

[0079]

[0080] in, These are weighting coefficients pre-calibrated based on the sensor's historical false alarm rate and stability.

[0081] For the revised sources of evidence The Dempster orthogonal sum rule is used for synthesis, and the joint fundamental probability assignment function value is calculated. :

[0082]

[0083] in, Representing different risk states; It is a subset obtained from the intersection of the basic probability allocation function values ​​of various evidence sources, representing the information synthesis of multiple evidence sources in the risk assessment process;

[0084] Based on the joint basic probability assignment function value Calculate the confidence level for each risk state. And determine the current risk status based on the level of trust, including;

[0085] when At that time, it was determined to be a confirmed fire situation; when and When, it is determined to be a thermal runaway warning state; when When this occurs, it is determined to be either a normal state or a sensor interference state, whereby... and The preset risk assessment threshold is determined based on historical data and experimental verification.

[0086] Furthermore, the allocation function value based on the joint basic probability... Calculate the confidence level for each risk state. The formula is:

[0087]

[0088] in, Trust level.

[0089] Furthermore, the aforementioned three-level control measures specifically include:

[0090] When the current risk status is normal or sensor interference status, maintain the normal charging status of the vehicle and continuously collect and update the features of multi-source raw data in real time; at the same time, perform consistency verification on sensor data. When a single sensor is detected as abnormal and does not meet the multi-source consistency conditions, it is marked as a sensor interference status and status record information is sent to the background monitoring platform, but charging interruption control is not triggered.

[0091] When the current risk status is a thermal runaway warning status, implement Level 1 active intervention measures, including:

[0092] The system controls the charging pile to reduce the output current to a preset safe current value and limits the rate of current rise; it sends a charging power limit command to the vehicle, causing the vehicle's battery management system to enter thermal safety monitoring mode; it activates the enhanced gas extraction mode of the ground active protection unit to increase the gas sampling flow rate; it activates the directional cooling or suppression device to pre-cool the vehicle chassis and battery pack area; and it sends early warning information to the background monitoring platform and records the current multi-source characteristic data and risk confidence level.

[0093] When the current risk status is confirmed as a fire, implement combined Level 2 and Level 3 containment measures, including:

[0094] Immediately send an emergency power-off command to the charging station to cut off the charging circuit; control the vehicle's high-voltage system to shut down, isolating the power battery from the external circuit; activate the fire extinguishing or inerting system of the ground active protection unit to spray fire extinguishing medium or inert gas onto the vehicle chassis area; simultaneously trigger the audible and visual alarm device and send fire confirmation signals and real-time data to the remote monitoring platform; after completing the power-off and suppression actions, continuously monitor changes in gas concentration, temperature gradient, and acoustic characteristics until the multi-source characteristic data recovers to within the safe threshold range.

[0095] Compared with the prior art, the present invention has the following advantages:

[0096] (1) In the existing technology, the vehicle-side BMS mainly relies on single electrical parameters such as voltage, current and temperature for monitoring. In the early stage of thermal runaway (such as safety valve venting, the initial stage of micro short circuit and material micro cracking), there are often no obvious electrical anomalies. In extreme cases, key status information may be "silent" due to communication interruption or sensor failure, making it difficult to identify the dangerous evolution process inside the battery in time. This invention synchronously collects multi-source raw data from charging piles and ground active protection units through a unified clock reference. It performs deviation and variance statistical analysis on charging current, extracts acoustic features by wavelet packet decomposition of acoustic signals, and generates thermal features by spatiotemporal gradient analysis of infrared thermal imaging matrix. At the same time, it combines multi-component gas concentration data for feature decoupling and extraction, realizing the transformation from single electrical monitoring to multi-dimensional collaborative sensing of "electricity-gas-sound-heat-light". This enables the detection of early warning signals such as micro short circuit, electrolyte venting and material rupture in the early stage when electrical anomalies are not obvious, significantly improving the sensitivity and reliability of early thermal runaway identification.

[0097] (2) In the prior art, top-mounted smoke or heat detectors are usually far from the vehicle chassis and are easily affected by airflow organization, spatial obstruction and smoke diffusion path. They can only trigger an alarm after a large amount of smoke or high temperature has diffused, resulting in a significant time lag and missing the best intervention window in the early stage of thermal runaway. This invention, after detecting the charging start signal, controls the ground active protection unit to perform the reverse purging and forward absorption of the gas path system in sequence, constructing a clean air sampling boundary layer between the vehicle chassis and the ground, and collecting gas samples from the battery pack area in real time through a gas sensor. This breaks the traditional passive waiting diffusion perception mode and changes the response to gas anomalies from passive diffusion detection to active near-field sampling detection, effectively shortening the detection response time and improving the sampling purity, thereby enabling rapid early warning during the safety valve opening and jetting stage.

[0098] (3) In the existing technology, the data between the "vehicle-charging pile-parking space" system are isolated from each other. The vehicle electrical data, charging pile operation data and environmental perception data lack a unified time sequence and collaborative analysis mechanism, making it difficult to cross-verify multimodal anomalies. It is easy to misjudge the exhaust gas of neighboring vehicles, normal relay sounds or high ambient temperature as dangerous signals, resulting in a high false alarm rate. This invention constructs a vehicle-charging pile-parking space collaborative architecture and uses a unified clock reference to synchronously collect multi-source data. Furthermore, it inputs multi-component gas concentration, electrical micro-short circuit characteristics, acoustic characteristics and thermal characteristics into a multi-dimensional risk assessment model, uses the improved DS evidence theory for decision-level fusion, and uses the conflict coefficient to perform decision-level fusion. K By correcting conflicting evidence with a weighting coefficient based on historical reliability, the credibility redistribution and consistency judgment among multi-source evidence can be achieved, thereby effectively suppressing false alarms from a single sensor and improving the stability and accuracy of risk identification results.

[0099] (4) Existing technologies lack a multi-stage, refined identification mechanism for the thermal runaway evolution process. They typically rely on simple thresholds or single indicators for judgment, making it difficult to distinguish between static high-temperature components (such as exhaust pipe waste heat) and thermal runaway heat sources with a diffusion trend, which can easily lead to misjudgments or omissions. This invention performs spatiotemporal joint analysis of the temporal and spatial gradients of the infrared thermal image matrix to extract the temperature rise rate and thermal diffusion intensity. Based on the high-temperature region of interest, it constructs weighted thermal features, enabling risk assessment to consider not only the absolute temperature value but also the growth trend and diffusion characteristics of the heat source. This allows for effective differentiation between stable heat sources and dynamic thermal runaway heat sources, achieving more accurate early risk identification.

[0100] (5) Most existing fire sprinkler systems are top-covering sprayers, which are difficult to penetrate the vehicle structure to reach the core fire area such as the bottom of the battery pack, resulting in low cooling and suppression efficiency and difficulty in timely blocking the spread of thermal runaway. This invention sets up a ground active protection unit and activates a near-field directional fire extinguishing or inerting system under high-risk conditions, so that the fire extinguishing medium or inert gas acts on the battery pack area from under the vehicle at close range, shortening the action path and improving the efficiency of medium utilization, thereby achieving precise suppression and rapid cooling of the core area of ​​thermal runaway, and significantly improving the initial fire extinguishing success rate.

[0101] (6) Existing technologies for charging safety protection often adopt a "one-size-fits-all" power-off or top spray strategy, lacking a graded and gradual active intervention mechanism. This may affect the normal charging experience due to premature power-off, or may fail to suppress the development of thermal runaway due to delayed intervention. Based on the risk status obtained by the fusion of improved DS evidence theory, this invention constructs a three-level control measure of "normal monitoring - early warning intervention - fire blocking". In the early warning stage, it reduces the charging current, limits the power and starts enhanced gas extraction and directional pre-cooling. In the fire confirmation stage, it implements emergency power-off, high-voltage system power-off and fire extinguishing / inerting linkage control, realizing a tiered response from flexible regulation to forced blocking. This reduces system interference caused by false triggering and enables rapid and effective intervention when the risk escalates. Attached Figure Description

[0102] Figure 1 This is a flowchart of the active prevention method for thermal runaway in new energy vehicles according to an embodiment of the present invention;

[0103] Figure 2 This is a model diagram of an active thermal runaway prevention system for new energy vehicles according to an embodiment of the present invention. Detailed Implementation

[0104] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0105] Example 1:

[0106] like Figure 1 As shown, this embodiment discloses a method for actively preventing thermal runaway in new energy vehicles based on vehicle-charging station-ground coordination. This method relies on a coordinated defense system composed of a new energy vehicle (vehicle), an intelligent charging terminal (charging pile), and a ground-based active protection unit (ground), and includes the following steps:

[0107] Step S1: After detecting the charging start signal, control the ground active protection unit to perform the sequential actions of reverse purging and forward suction of the air circuit system to build a clean air sampling boundary layer between the vehicle chassis and the ground.

[0108] In this embodiment, the ground-based active protection unit is embedded in a predetermined position in the central area of ​​the charging parking space or directly below the battery pack, and is connected to the charging pile via a wired communication bus or industrial Ethernet. The ground-based active protection unit includes an air path control module, an air pump assembly, a micro-hole air intake array, a gas sensor cavity, a wideband microphone array, an infrared thermal imaging acquisition assembly, an optical sensor assembly, and a local edge processing module. The air pump assembly uses a reversible brushless DC air pump, and the micro-hole air intake array is distributed in a ring on the upper surface of the unit to form a uniform airflow field.

[0109] After the charging gun establishes a charging handshake signal with the vehicle, the control module first drives the air pump to rotate in the opposite direction, spraying positive pressure airflow through the micro-hole air intake array to continuously blow back the area under the vehicle for 3 to 8 seconds, removing dust, moisture and suspended particles attached to the chassis. Then, it switches to forward rotation mode to form a negative pressure flow field, actively drawing air near the battery pack into the sensor cavity for sampling. The reverse blowing and forward suction cycle according to a preset duty cycle until the gas concentration fluctuation is less than the stable threshold and the particulate matter concentration is lower than the set cleanliness standard, thereby forming a stable and low-disturbance air sampling boundary layer between the vehicle chassis and the ground. This process can significantly improve the purity and response speed of near-field gas and spectral sampling.

[0110] Step S2: Real-time synchronous collection of multi-source raw data from charging piles and ground active protection units using a unified clock reference;

[0111] In this embodiment, the system is equipped with a high-precision clock synchronization module, which is integrated into the edge processing module and communicates with the charging pile main control unit and the ground active protection unit via a wired bus, using the system clock of the charging pile main control unit as the global time reference. Each acquisition node automatically adds a high-resolution timestamp during data acquisition, and cross-device data synchronization calibration is achieved through timestamp alignment and delay compensation algorithms, ensuring that data from different sources maintain a strict correspondence under the same time coordinate system, thereby improving the temporal consistency of multi-source information fusion.

[0112] In actual operation, multi-source raw data are continuously acquired and transmitted to the edge processing module according to a preset sampling period. The sampling frequency is preferably 10–50 Hz to balance real-time performance and data stability. The edge processing module has a built-in cache unit and a preprocessing unit. It sequentially performs sliding window caching, outlier removal, filtering and noise reduction, and normalization on the acquired data. This helps to reduce the random disturbances caused by environmental airflow fluctuations, electromagnetic interference, and ground vibrations on the detection results, thereby improving the reliability of subsequent thermal runaway detection.

[0113] Specifically, multi-source raw data includes at least the following types:

[0114] (1) Charging current data and charging voltage data from the charging pile are acquired in real time through the electrical parameter acquisition circuit inside the charging pile and are used to characterize the charging load change status.

[0115] (2) Multi-component gas concentration data from the ground active protection unit are collected in real time in the vehicle chassis area by a gas sensor array set in the gas sampling chamber. The gas sensor array includes a carbon monoxide sensor, a hydrogen sensor, a volatile organic compound sensor and an electrolyte characteristic gas sensor, which are used to reflect the abnormal gas evolution characteristics of the battery.

[0116] (3) Acoustic signals from the ground active protection unit are collected by a wideband microphone to collect acoustic emission information of the chassis area during the charging process. The acoustic signals can reflect the characteristic acoustic changes caused by abnormal internal structure of the battery or minor thermal instability.

[0117] (4) The infrared thermal image matrix from the ground active protection unit performs non-contact temperature field monitoring on the vehicle chassis and battery pack area through the area array infrared thermal imager, forming continuous two-dimensional temperature distribution data.

[0118] (5) The spectral signal from the ground active protection unit is acquired in real time by an optical sensor set in the chassis monitoring area, and is used to identify the characteristic spectral changes caused by early electrolyte volatilization or thermal decomposition.

[0119] By using the unified clock synchronization acquisition and preprocessing of the above multi-source data, gas, thermal imaging, acoustic and electrical parameter information can form a one-to-one corresponding data stream in the time dimension, which can more accurately reflect the early abnormal evolution process of the vehicle chassis area during charging, thereby providing highly consistent basic data support for the accurate identification of subsequent thermal runaway risks.

[0120] Step S3: Perform feature decoupling and extraction on the collected multi-source raw data to obtain multi-source feature data, specifically including:

[0121] Step S301: Perform statistical analysis on the charging current in the multi-source raw data, calculate the deviation and variance between the real-time charging current and the reference current, and obtain the electrical micro-short circuit characteristics.

[0122] In this embodiment, the edge processing module has a built-in electrical feature analysis unit for performing sliding window statistical calculations on the synchronized charging current data. The reference current is obtained by matching a historical normal charging curve library, which is categorized and stored according to battery type, rated capacity, SOC range, and ambient temperature. After charging begins, the reference current curve matching the current vehicle is automatically called as a benchmark, thereby improving the accuracy of deviation judgment.

[0123] Calculate the current deviation based on the charging current and the reference current. ,in, , These are the vehicle charging current and the reference current for normal vehicle charging, respectively, collected in real time by the ground active protection unit.

[0124] Within the preset time window Internal, for current deviation Integrate to obtain the cumulative deviation. :

[0125]

[0126] Calculate the current deviation within the preset time window variance within :

[0127]

[0128] in, This represents the total number of sampling points; For the first Current deviation at each sampling point; The mean of the deviation;

[0129] Based on cumulative deviation Sum of deviations and variances Determine the characteristics of electrical micro-short circuits:

[0130] when or At that time, electrical micro-short circuit characteristics are set. Otherwise ;in, and These are preset thresholds, obtained through statistical analysis of historical vehicle charging data, and preset according to battery type and charging environment.

[0131] When the cumulative deviation or deviation variance exceeds a preset threshold, the system determines that there is an abnormal current fluctuation trend. This trend can reflect the hidden electrical anomalies caused by micro short circuits or local insulation degradation inside the battery, thereby achieving sensitive identification of early electrical instability.

[0132] Step S302: Perform wavelet packet decomposition on the acoustic signals in the multi-source raw data, extract the energy of different frequency bands, and use the frequency band energy corresponding to the gas leakage and ejection characteristics and the micro-fracture characteristics of the battery material as acoustic features, specifically including:

[0133] The acoustic signals collected in real time by the ground active protection unit conduct Layer wavelet packet decomposition yields the node coefficients of each frequency band. The calculation formula is:

[0134]

[0135] in, Indicates the first Layer The frequency band node at the th frequency band node Frequency band node coefficients of each sampling point; These represent the filter coefficients of the selected orthogonal wavelet basis functions; For discrete-time indexing, The number of decomposition levels; This is the index of the current layer frequency band node. The original sound wave signal;

[0136] Based on the acoustic frequency range of the characteristics of micro-short circuits and gas leaks in vehicle batteries, a set of characteristic frequency band nodes is defined. This includes the frequency bands characteristic of battery rupture sound and gas leakage sound;

[0137] Based on the frequency band node coefficients obtained from wavelet packet decomposition, the values ​​in the characteristic frequency band node set are calculated. Characteristic frequency band energy within :

[0138]

[0139] Based on characteristic frequency band energy The ratio of the total energy across the entire frequency band to the normalized acoustic characteristics is calculated using the following formula:

[0140]

[0141] in, This acoustic feature represents the proportion of energy in a specific frequency band within the total energy across the entire frequency range. It is used to identify abnormal signal strengths caused by microscopic cracks or gas leaks in battery materials. This acoustic feature effectively characterizes the proportion of abnormal sound sources in the overall sound field, enabling the system to have a higher resolution for weak leakage and micro-crack sounds, and maintaining stable identification performance even in complex charging environments.

[0142] Step S303: Perform spatiotemporal gradient analysis on the infrared thermal image matrix in the multi-source raw data, calculate the temperature rise rate and thermal diffusion intensity, and generate weighted thermal features, specifically including:

[0143] Set the infrared thermal image matrix acquired by the ground active protection unit to ,in, Represents the spatial pixel coordinates in an infrared image. Indicates the time of data collection. Indicates at time Time pixel Temperature value at that location;

[0144] At adjacent sampling times and Calculate the time gradient of temperature. The formula is:

[0145]

[0146] in, The sampling time interval between two adjacent infrared images. Represents pixels The rate of temperature rise;

[0147] At the same time The infrared thermal imaging matrix below Spatial gradient calculations are performed to obtain the thermal diffusivity. The formula is:

[0148]

[0149] in, It represents the magnitude of the temperature field gradient in space and is used to characterize the intensity of heat diffusion;

[0150] According to the preset temperature threshold Extracting regions of interest at high temperatures High-temperature region of interest To meet A set of pixels;

[0151] Calculate the region of interest at high temperature The maximum temperature rise rate and average thermal diffusion intensity within the interior were determined, and thermal characteristics were constructed. The formula is:

[0152]

[0153] in, Indicates the region of interest for high temperature. The maximum rate of temperature rise of an internal pixel; Indicates the region of interest for high temperature. The average value of internal thermal diffusion intensity; , The preset weighting coefficients are used. This thermal characteristic can simultaneously reflect local rapid heating phenomena and heat diffusion trends, making the identification of the early stages of potential thermal runaway inside the battery more sensitive and stable.

[0154] Step S304: Combine multi-component gas concentration data, electrical micro-short circuit characteristics, acoustic characteristics and thermal characteristics as multi-source characteristic data.

[0155] Step S4: Input the extracted multi-source feature data into the multidimensional risk assessment model, and perform decision-level fusion using the improved DS evidence theory to determine the current risk status. Specifically, this includes:

[0156] Building a risk identification framework ,in, Indicates normal or sensor interference status. This indicates a thermal runaway warning state. Indicates the confirmed status of a fire;

[0157] Each feature extracted from the multi-source feature data is used as an independent source of evidence for the multi-component gas concentration data in the multi-source feature data. Electrical micro-short circuit characteristics Acoustic characteristics and thermal characteristics Calculate the basic probability assignment function separately, and use the sigmoid function to map each feature value to the support for different risk states, thus constructing the corresponding basic probability assignment function values. and :

[0158]

[0159]

[0160]

[0161]

[0162] in, The sensitivity coefficients for gaseous, electrical, acoustic, and thermal evidence were determined through historical data analysis and sensor performance evaluation. The preset threshold for the corresponding feature is set based on experimental data and actual application scenarios; and These represent the degree of support that gaseous, electrical, acoustic, and thermal characteristics provide for their respective risk states;

[0163] Calculate the conflict coefficient between each source of evidence based on the basic probability allocation function value of each independent source of evidence. The formula is:

[0164]

[0165] in, Risk identification framework any subset in Assign function values ​​to the basic probability of mutually exclusive propositions from different sources of evidence;

[0166] When the conflict coefficient Greater than the preset conflict threshold At that time, a weighting coefficient based on the historical reliability of the sensor is introduced. The basic probability allocation function values ​​for each source of evidence are corrected:

[0167]

[0168] in, These are weighting coefficients pre-calibrated based on the sensor's historical false alarm rate and stability.

[0169] For the revised sources of evidence The Dempster orthogonal sum rule is used for synthesis, and the joint fundamental probability assignment function value is calculated. :

[0170]

[0171] in, Representing different risk states; It is a subset obtained from the intersection of the basic probability allocation function values ​​of various evidence sources, representing the information synthesis of multiple evidence sources in the risk assessment process;

[0172] Based on the joint basic probability assignment function value Calculate the confidence level for each risk state. And determine the current risk status based on the level of trust, including;

[0173] when At that time, it was determined to be a confirmed fire situation; when and When, it is determined to be a thermal runaway warning state; when When this occurs, it is determined to be either a normal state or a sensor interference state, whereby... and The preset risk assessment threshold is determined based on historical data and experimental verification. It is established through statistical analysis of historical operational data and calibration using multi-scenario thermal runaway experiments to ensure clear distinctions between different risk levels, thereby enabling precise, tiered identification of thermal runaway in new energy vehicles, from early anomalies to fire development.

[0174] Based on the joint basic probability assignment function value Calculate the confidence level for each risk state. The formula is:

[0175]

[0176] in, Trust level.

[0177] Step S5: Based on the assessed current risk status, implement three levels of control measures sequentially to actively prevent thermal runaway in new energy vehicles;

[0178] In this embodiment, the three-level control measures specifically include:

[0179] When the current risk status is normal or sensor interference, the system maintains the normal charging process of the vehicle and continuously collects, caches, and updates features of multi-source raw data at a set sampling frequency. At the same time, it performs multi-source consistency verification on gas concentration data, current data, acoustic features, and thermal features, and determines whether there is a single sensor abnormal drift by calculating the correlation coefficient and stability index between each evidence source. When only a single evidence source shows a sudden change and does not meet the conditions for coordinated change of multi-source features, the state is judged as a sensor interference state, the abnormal sensor data is downweighted and marked with a fault label. The system uploads status record information, sensor health status, and timestamp data to the background monitoring platform, but does not trigger current reduction or power failure control, thereby avoiding charging interruption due to false alarms and improving system operation stability.

[0180] When the current risk status is a thermal runaway warning state, the system executes Level 1 active intervention measures, specifically: sending a current limiting control command to the charging pile main control module through the communication interface, gradually reducing the output current to a preset safe current value in a slope-controlled manner, and limiting the current rise rate to not exceed a set threshold to suppress the internal heating rate of the battery; simultaneously sending power limiting and safety monitoring commands to the vehicle through the charging communication protocol, causing the vehicle battery management system to enter an enhanced thermal safety monitoring mode, increasing the monitoring frequency of battery temperature, voltage anomalies, and insulation status; controlling the ground active protection unit to switch from the conventional sampling mode to the enhanced gas extraction mode, increasing the gas extraction flow rate and update frequency of the gas circuit system, and accelerating the replacement and detection of gas under the vehicle; activating the directional cooling or suppression device to implement low-intensity continuous pre-cooling of the area under the vehicle chassis and power battery pack, reducing the local temperature rise rate and delaying the evolution of thermal runaway; at the same time, the edge processing module uploads the current multi-source feature data, joint basic probability allocation function value, and various risk trust levels to the background monitoring platform in real time, generating warning logs and triggering remote visual monitoring to achieve traceable management of early risks.

[0181] When the current risk status is confirmed as a fire, the system executes a combined Level 2 and Level 3 blocking measure, forming a rapid physical isolation and near-field suppression coordinated control mechanism. Specifically: it immediately sends an emergency power-off control command to the charging pile, cutting off the charging circuit output and locking the charging interface; it sends a high-voltage power-off command to the vehicle through the vehicle-to-pile communication link, causing the high-voltage relay of the power battery to disconnect, achieving electrical isolation between the power battery and the external circuit, and preventing continuous energy input from exacerbating thermal runaway; simultaneously, it controls the ground active protection unit to activate the fire extinguishing or inerting system, driving the telescopic spray actuator to aim at the battery pack area under the vehicle chassis, directionally spraying fire extinguishing medium or inert gas, creating a local suppression environment at the bottom of the battery pack, achieving rapid... Rapid cooling and chemical suppression are implemented; simultaneously, a high-intensity audible and visual alarm signal is emitted to guide on-site personnel to evacuate in a timely manner, and fire confirmation signals, real-time gas concentrations, temperature gradients, and acoustic anomaly data are uploaded to the remote monitoring platform and fire linkage system; after power outage and near-field suppression are completed, the system maintains enhanced monitoring mode, continuously tracking changes in multi-component gas concentrations, infrared thermal image temperature rise rates, and acoustic characteristic energy. When the multi-source characteristic data recovers to within the safe threshold range for several consecutive monitoring cycles and the confidence level drops below the preset reset threshold, the emergency suppression state is lifted and the system enters the subsequent safety monitoring stage, thereby achieving closed-loop active blocking and safety control of the entire process of thermal runaway of new energy vehicles.

[0182] Example 2:

[0183] This embodiment provides an active thermal runaway prevention system for new energy vehicles based on vehicle-pile-parking space coordination, such as... Figure 2As shown, it includes a vehicle-side unit, a charging pile unit, a ground active protection unit, an edge processing module, and a remote monitoring platform. The units interact and coordinate control with each other through a wired communication bus and a wireless communication network.

[0184] Among them, the vehicle-side unit is used to provide vehicle operating status data and receive safety control commands. The vehicle-side unit and the charging pile unit establish a data interaction connection through the charging communication protocol, which is used to upload the status information of the battery management system and receive power limiting commands and high voltage power-off commands.

[0185] The charging pile unit is used to realize charging control and electrical data acquisition, including a charging main control module, a current and voltage acquisition module, a communication interface module and an execution control module. The current and voltage acquisition module is used to collect charging current and charging voltage data in real time and generate electrical raw data with timestamps. The charging main control module uses a local high-precision clock as the system reference clock and sends a synchronization clock signal to the edge processing module. At the same time, it executes current reduction control, power limiting control and emergency power cut-off control according to the control instructions issued by the edge processing module.

[0186] The ground-based active protection unit is installed on the parking space floor and located in the area corresponding to the vehicle chassis. It is used for near-field environmental perception and active intervention, and includes an air sampling module, a gas sensing module, an acoustic acquisition module, an infrared thermal imaging acquisition module, an optical acquisition module, an air path control module, and a near-field suppression execution module. Among these:

[0187] The air sampling module includes an air intake channel, an air extraction fan, and a reverse purging channel. It is used to build an air sampling boundary layer between the vehicle chassis and the ground after charging starts, and to realize the timing control of reverse purging and forward absorption.

[0188] The gas sensing module is used to detect the concentration data of multi-component gases in the undercarriage area in real time, including the concentrations of CO, H2, VOC and electrolyte characteristic gases.

[0189] The acoustic acquisition module uses a wideband microphone array to acquire sound wave signals during the charging process and output raw acoustic data.

[0190] The infrared thermal imaging acquisition module is used to acquire infrared thermal image matrix data of the vehicle chassis and battery pack area;

[0191] The optical acquisition module uses optical sensors to collect spectral signals or abnormal radiation signals in the chassis area in real time to help identify early thermal anomalies or weak signs of combustion.

[0192] The gas path control module is used to adjust the gas extraction flow rate and purging intensity;

[0193] The near-field suppression execution module includes a telescopic spray mechanism, a fire extinguishing medium storage tank, and an electromagnetic control valve, used to directionally spray fire extinguishing medium or inert gas into the bottom area of ​​the battery pack under high-risk conditions;

[0194] The edge processing module is electrically or communicatively connected to both the charging pile unit and the ground-based active protection unit. It is used for synchronous acquisition, caching, feature decoupling extraction, and risk assessment of multi-source data. Internally, the edge processing module includes a clock synchronization module, a data preprocessing module, a feature extraction module, a multi-dimensional risk assessment module, and a decision control module.

[0195] The clock synchronization module uses the charging pile's main control clock as a unified reference to align the timestamps of the data collected by each sensor.

[0196] The data preprocessing module is used to denoise, filter, and remove outliers from multi-source raw data;

[0197] The feature extraction module is used to extract electrical micro-short circuit features, gas features, acoustic features, and thermal features;

[0198] The multidimensional risk assessment module performs decision-level fusion of multi-source feature data based on the improved DS evidence theory and outputs the current risk status.

[0199] The decision control module generates hierarchical control commands based on the risk status and sends them to the charging pile unit, vehicle-end unit and ground active protection unit respectively, so as to realize the coordinated control of vehicle-charging pile-parking space.

[0200] The remote monitoring platform communicates with the edge processing module via the network to receive multi-source feature data, risk assessment results, and equipment operating status information uploaded by the system. It also performs visual monitoring, historical data storage, alarm information push, and system operation and maintenance management, thus forming a collaborative protection system that integrates near-field perception, multi-source fusion judgment, and active blocking control.

[0201] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0202] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for actively preventing thermal runaway in new energy vehicles based on vehicle-charging station-parking station coordination, characterized in that, Includes the following steps: After detecting the charging start signal, the ground active protection unit is controlled to perform the sequential actions of reverse purging and forward suction of the air circuit system to build a clean air sampling boundary layer between the vehicle chassis and the ground. Through a unified clock reference, multi-source raw data from charging piles and ground active protection units are collected in real time and synchronously. Feature decoupling and extraction are performed on the collected multi-source raw data to obtain multi-source feature data; The extracted multi-source feature data is input into a multi-dimensional risk assessment model, and decision-level fusion is performed through improved DS evidence theory to determine the current risk status. Based on the assessment of the current risk status, three levels of control measures are implemented sequentially to actively prevent thermal runaway in new energy vehicles. The process of decoupling and extracting features from the collected multi-source raw data to obtain multi-source feature data specifically includes: Statistical analysis is performed on the charging current in the multi-source raw data to calculate the deviation and variance between the real-time charging current and the reference current, and to obtain the electrical micro-short circuit characteristics. Wavelet packet decomposition was performed on the acoustic signals in the multi-source raw data to extract the energy of different frequency bands, and the frequency band energy corresponding to the gas leakage and jet characteristics and the micro-fracture characteristics of battery materials was used as acoustic features. Spatiotemporal gradient analysis is performed on the infrared thermal image matrix in multi-source raw data to calculate the temperature rise rate and thermal diffusion intensity, and thermal characteristics are generated by weighting. Multi-component gas concentration data, electrical micro-short circuit characteristics, acoustic characteristics and thermal characteristics are combined as multi-source feature data; The statistical analysis of charging current in multi-source raw data, calculating the deviation and variance between real-time charging current and reference current, and obtaining electrical micro-short circuit characteristics specifically includes: Calculate the current deviation based on the charging current and the reference current. ,in, , These are the vehicle charging current and the reference current for normal vehicle charging, respectively, collected in real time by the ground active protection unit. Within the preset time window Internal, for current deviation Integrate to obtain the cumulative deviation. : Calculate the current deviation within the preset time window variance within : in, This represents the total number of sampling points; For the first Current deviation at each sampling point; The mean of the deviation; Based on cumulative deviation Sum of deviations and variances Determine the characteristics of electrical micro-short circuits: when or At that time, electrical micro-short circuit characteristics are set. Otherwise ;in, and These are preset thresholds, obtained through statistical analysis of historical vehicle charging data, and preset according to battery type and charging environment.

2. The method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination according to claim 1, characterized in that, The control ground active protection unit executes the sequential actions of reverse purging and forward suction of the airway system to construct a clean air sampling boundary layer between the vehicle chassis and the ground, specifically including: Step A1: After detecting the charging start signal, send an activation command to the ground active protection unit; Step A2: After the ground active protection unit receives the activation command; Step A3: Control the built-in air pump to rotate in reverse, spraying positive pressure airflow into the ground air intake micro-holes, and continue for a preset time to blow away the dust, water and debris deposited under the vehicle; Step A4: Switch the air pump to forward rotation mode to generate a negative pressure flow field, and actively draw air from the vicinity of the vehicle battery pack into the sensor cavity through suction. Step A5: Repeat steps A3 and A4, periodically performing reverse purging and forward aspiration operations until the gas sample collected by forward aspiration is stable and meets the preset cleanliness standard, then stop the cycle to obtain a clean air sampling boundary layer.

3. The method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination according to claim 1, characterized in that, The multi-source raw data includes: Charging current from the charging station; Multi-component gas concentration data from the ground active protection unit is collected in real time by gas sensors on the vehicle chassis, including the concentration of each component of the collected gas. The acoustic signal from the ground active protection unit is obtained by real-time acquisition of the acoustic signal during the charging process through a wideband microphone; The infrared thermal image matrix from the ground active protection unit is obtained by monitoring the vehicle chassis and battery pack through an infrared thermal imager. The spectral signals from the ground-based active protection unit are obtained by real-time acquisition of the spectral signals from the vehicle chassis through optical sensors.

4. The active prevention method for thermal runaway of new energy vehicles based on vehicle-pile-parking space coordination according to claim 1, characterized in that, The process of performing wavelet packet decomposition on the acoustic signals in the multi-source raw data to extract the energy of different frequency bands, and using the frequency band energy corresponding to the gas leakage and ejection characteristics and the microscopic fracture characteristics of battery materials as acoustic features, specifically includes: The acoustic signals collected in real time by the ground active protection unit conduct Layer wavelet packet decomposition yields the node coefficients of each frequency band. The calculation formula is: in, Indicates the first Layer The frequency band node at the th frequency band node Frequency band node coefficients of each sampling point; These represent the filter coefficients of the selected orthogonal wavelet basis functions; For discrete-time indexing, The number of decomposition levels; This is the index of the current layer frequency band node. The original sound wave signal; Based on the acoustic frequency range of the characteristics of micro-short circuits and gas leaks in vehicle batteries, a set of characteristic frequency band nodes is defined. This includes the frequency bands characteristic of battery rupture sound and gas leakage sound; Based on the frequency band node coefficients obtained from wavelet packet decomposition, the values ​​in the characteristic frequency band node set are calculated. Characteristic frequency band energy within : Based on characteristic frequency band energy The ratio of the total energy across the entire frequency band to the normalized acoustic characteristics is calculated using the following formula: in, For acoustic characteristics, it represents the proportion of characteristic frequency band energy in the total energy of the entire frequency band, and is used to identify abnormal signal strengths of microscopic cracks or gas leaks in battery materials.

5. The method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination according to claim 1, characterized in that, The process of performing spatiotemporal gradient analysis on the infrared thermal image matrix in the multi-source raw data, calculating the temperature rise rate and thermal diffusion intensity, and generating weighted thermal features specifically includes: Set the infrared thermal image matrix acquired by the ground active protection unit to ,in, Represents the spatial pixel coordinates in an infrared image. Indicates the time of data collection. Indicates at time Time pixel Temperature value at; At adjacent sampling times and Calculate the time gradient of temperature. The formula is: in, The sampling time interval between two adjacent infrared images. Represents pixels The rate of temperature rise; At the same time The infrared thermal imaging matrix below Spatial gradient calculations are performed to obtain the thermal diffusivity. The formula is: in, It represents the magnitude of the temperature field gradient in space and is used to characterize the intensity of heat diffusion; According to the preset temperature threshold Extracting regions of interest at high temperatures The high-temperature region of interest To meet A set of pixels; Calculate the region of interest at high temperature The maximum temperature rise rate and average thermal diffusion intensity within the interior were determined, and thermal characteristics were constructed. The formula is: in, Indicates the region of interest for high temperature. The maximum rate of temperature rise of an internal pixel; Indicates the region of interest for high temperature. The average value of internal thermal diffusion intensity; , These are preset weighting coefficients.

6. The method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination according to claim 1, characterized in that, The step of inputting the extracted multi-source feature data into a multi-dimensional risk assessment model and performing decision-level fusion through an improved DS evidence theory to determine the current risk status specifically includes: Building a risk identification framework ,in, Indicates normal or sensor interference status. This indicates a thermal runaway warning state. Indicates the confirmed status of a fire; Each feature extracted from the multi-source feature data is used as an independent source of evidence for the multi-component gas concentration data in the multi-source feature data. Electrical micro-short circuit characteristics Acoustic characteristics and thermal characteristics Calculate the basic probability assignment function separately, and use the sigmoid function to map each feature value to the support for different risk states, thus constructing the corresponding basic probability assignment function values. and : in, The sensitivity coefficients for gaseous, electrical, acoustic, and thermal evidence were determined through historical data analysis and sensor performance evaluation. The preset threshold for the corresponding feature is set based on experimental data and actual application scenarios; and These represent the degree of support that gaseous, electrical, acoustic, and thermal characteristics provide for their respective risk states; Calculate the conflict coefficient between each source of evidence based on the basic probability allocation function value of each independent source of evidence. The formula is: in, Risk identification framework any subset in Assign function values ​​to the basic probability of mutually exclusive propositions from different sources of evidence; When the conflict coefficient Greater than the preset conflict threshold At that time, a weighting coefficient based on the historical reliability of the sensor is introduced. The basic probability allocation function values ​​for each source of evidence are corrected: in, These are weighting coefficients pre-calibrated based on the sensor's historical false alarm rate and stability. For the revised sources of evidence The Dempster orthogonal sum rule is used for synthesis, and the joint fundamental probability assignment function value is calculated. : in, Representing different risk states; It is a subset obtained from the intersection of the basic probability allocation function values ​​of various evidence sources, representing the information synthesis of multiple evidence sources in the risk assessment process; Based on the joint basic probability assignment function value Calculate the confidence level for each risk state. And determine the current risk status based on the level of trust, including; when When, it is determined to be a confirmed fire state; when and When, it is determined to be a thermal runaway warning state; when When this occurs, it is determined to be either a normal state or a sensor interference state, whereby... and The preset risk assessment threshold is determined based on historical data and experimental verification.

7. A method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination according to claim 6, characterized in that, The value of the joint basic probability allocation function Calculate the confidence level for each risk state. The formula is: in, Trust level.

8. The method for actively preventing thermal runaway in new energy vehicles based on vehicle-pile-parking space coordination according to claim 1, characterized in that, The three-level control measures specifically include: When the current risk status is normal or sensor interference status, maintain the normal charging status of the vehicle and continuously collect and update the features of multi-source raw data in real time; at the same time, perform consistency verification on sensor data. When a single sensor is detected as abnormal and does not meet the multi-source consistency conditions, it is marked as a sensor interference status and status record information is sent to the background monitoring platform, but charging interruption control is not triggered. When the current risk status is a thermal runaway warning status, implement Level 1 active intervention measures, including: The system controls the charging pile to reduce the output current to a preset safe current value and limits the rate of current rise; it sends a charging power limit command to the vehicle, causing the vehicle's battery management system to enter thermal safety monitoring mode; it activates the enhanced gas extraction mode of the ground active protection unit to increase the gas sampling flow rate; it activates the directional cooling or suppression device to pre-cool the vehicle chassis and battery pack area; and it sends early warning information to the background monitoring platform and records the current multi-source characteristic data and risk confidence level. When the current risk status is confirmed as a fire, implement combined Level 2 and Level 3 containment measures, including: Immediately send an emergency power-off command to the charging station to cut off the charging circuit; control the vehicle's high-voltage system to shut down, isolating the power battery from the external circuit; activate the fire extinguishing or inerting system of the ground active protection unit to spray fire extinguishing medium or inert gas onto the vehicle chassis area; simultaneously trigger the audible and visual alarm device and send fire confirmation signals and real-time data to the remote monitoring platform; after completing the power-off and suppression actions, continuously monitor changes in gas concentration, temperature gradient, and acoustic characteristics until the multi-source characteristic data recovers to within the safe threshold range.

Images