Passenger cabin power bank thermal runaway active isolation control method and system for passenger aircraft

Abstract

The application provides a kind of passenger cabin power bank thermal runaway active isolation control method and system, the application is first to port voltage, current and temperature comprehensive judgment and identification power bank current working state, realize state precision monitoring;On this basis, when meeting the diagnostic safety constraints, short-time diagnostic excitation is applied through external port, further, individualized multi-time scale port fingerprint is constructed, and the above fingerprint is input into the state model trained by normal data, the model prediction deviation, initial reference deviation and long-term trend risk are weighted and fused into a comprehensive abnormal score, and the conformal calibration value is obtained by comparing with the normal sample calibration set, and the risk level is determined by combining the continuous abnormal criterion.The application realizes the change from passive dependence to active use of port electrical characteristics to identify the precursor, effectively improves the early warning lead time, reduces false positives and false negatives, and provides a basis for automated closed-loop disposal, thereby significantly enhancing the safety level of passenger cabin.

Description

Technical Field

[0001] This invention relates to the field of aircraft cabin safety system technology, and in particular to an active isolation control method and system for thermal runaway of a power bank in an aircraft cabin. Background Technology

[0002] With the widespread use of portable electronic devices, it is common for passengers to carry and use power banks during flights. Lithium-ion batteries may experience thermal runaway under conditions such as internal short circuits, overcharging, external heating, mechanical damage, or aging and deterioration. After thermal runaway occurs, the battery may undergo dangerous processes such as rapid temperature rise, release of flammable gases, flame ejection, and reignition. In the crowded, confined environment of the cabin, where emergency response conditions are limited, power bank malfunctions pose a significant safety risk.

[0003] Currently, existing technologies addressing the thermal runaway risk of portable lithium-ion battery devices such as power banks primarily focus on two aspects: detection and early warning, and physical intervention. For detection and early warning, temperature sensors are placed on the surface of the battery or device to issue an alert when the temperature exceeds a preset threshold. Voltage and current monitoring detects voltage drops or current surges at the detection ports to determine abnormal battery conditions. Regarding physical intervention, current solutions typically rely on passengers or crew members discovering smoke, odors, abnormally high temperatures, or open flames before implementing manual intervention.

[0004] However, the aforementioned existing technical solutions still have the following shortcomings in preventing thermal runaway risks from power banks in passenger cabins: First, alarms can only be triggered after thermal runaway has progressed to a certain point (e.g., a significant increase in temperature), making it difficult to provide an effective early warning. While techniques such as electrochemical impedance spectroscopy can provide earlier information to overcome this temperature hysteresis, their engineering applications are limited by the detection method: obtaining impedance information usually requires applying or utilizing current perturbations; if the perturbation method, amplitude, duration, and triggering timing are not properly controlled, it may cause problems such as port voltage fluctuations and charging protocol switching, affecting normal passenger use; furthermore, the cumulative energy injection resulting from an unconstrained detection process faces strict limitations in aviation safety certification. From the time the detection process identifies an anomaly to when passengers or crew discover the alarm, assess the situation, and then retrieve isolation containers for handling, the response time from risk identification to physical isolation is significantly prolonged.

[0005] Secondly, existing technologies use uniform absolute thresholds for voltage, current, or temperature to determine anomalies. Essentially, this uses static standards to measure dynamic objects, making it difficult to avoid false alarms or missed alarms. Because characteristic quantities of a power bank, such as impedance and recovery time constant, may drift slowly over a long period, the change between two adjacent measurements may be extremely small and not trigger any instantaneous thresholds. However, the long-term cumulative deviation may reach a level that is risky. Existing detection methods based on instantaneous thresholds or short-term change rates lack the ability to characterize and track long-term cumulative changes. Moreover, relying solely on fixed thresholds or short-term changes for judgment makes it difficult to identify small but continuous single-step changes as risk signals during long-term operation, leading to further degradation of detection capabilities.

[0006] In view of this, the inventor has specifically designed an active isolation control method and system for thermal runaway of power banks in passenger aircraft cabins, which leads to this invention. Summary of the Invention

[0007] To solve the above problems, the technical solution of the present invention is as follows: A method for active isolation and control of thermal runaway of power banks in passenger aircraft cabins includes the following steps: The power bank's external port voltage signal, port current signal, and temperature signal are collected at a preset sampling frequency. The power bank's current working status is identified based on the port voltage level, port current direction, power flow direction, and available interface protocol status. When the current working state meets the preset diagnostic safety constraints, a diagnostic excitation is applied through the external port of the power bank. During the application of the diagnostic excitation, the port voltage response and port current response are collected synchronously and the port equivalent dynamic impedance is determined. The recovery time constant, temperature rise rate and impedance drift are extracted. The port equivalent dynamic impedance, recovery time constant, temperature rise rate and impedance drift obtained from each effective diagnosis, as well as the port voltage, port current and temperature, are used to form a diagnostic feature vector. A multi-timescale port fingerprint is constructed as follows: When the power bank is first connected and the port voltage, port current, and temperature remain stable within a preset time, the current port voltage, port current, temperature, port equivalent dynamic impedance, and recovery time constant are recorded to form an initial safety benchmark corresponding to the current working state; the diagnostic feature vectors obtained from the most recent few diagnoses before the current diagnosis time are arranged in chronological order to form a short-term sliding window; and the long-term trend summary is constructed from the changing trends of the port equivalent dynamic impedance and recovery time constant within a preset time span. The multi-timescale port state fingerprint is input into a normal state model trained with normal state data. The normal state model includes a one-dimensional nonlinear function represented by truncated Fourier sine and cosine terms, used to output normal prediction features or normal reconstruction features. The prediction bias or reconstruction bias is obtained based on the difference between the current diagnostic feature vector and the normal prediction feature or normal reconstruction feature. A comprehensive anomaly score is obtained by weighting and combining the prediction bias or reconstruction bias, the degree of deviation of the current diagnostic feature vector from the initial security benchmark, and the degree of deviation of the long-term trend summary from the preset security boundary. Obtain a calibration score set consisting of comprehensive anomaly scores corresponding to sample data collected under normal operating conditions, corresponding to the current working state; calculate the proportion of samples in the calibration score set that are greater than or equal to the current comprehensive anomaly score to obtain a conformal calibration value; count the number of times the conformal calibration value is lower than a preset anomaly threshold in a specified number of consecutive diagnoses before the current time as the consecutive anomaly count; determine the risk level based on the conformal calibration value and the consecutive anomaly count. When the risk level reaches the preset isolation condition, an isolation control signal is output.

[0008] Preferably, the long-term trend summary is constructed in the following manner: the port equivalent dynamic impedance is smoothed by applying an exponential moving average to obtain the smoothed impedance characteristics; the impedance change slope is obtained based on the time change rate corresponding to the difference between the current port equivalent dynamic impedance and the Mth past diagnosis; the recovery time constant change slope is obtained based on the time change rate corresponding to the difference between the current recovery time constant and the Mth past diagnosis; and the long-term trend summary is composed of the smoothed impedance characteristics, the impedance change slope, and the recovery time constant change slope.

[0009] Preferably, the comprehensive anomaly score is determined as follows: the weighted distance between the current diagnostic feature vector and the prediction deviation or reconstruction deviation of the normal state model output is used as the model deviation term; the weighted distance between the current diagnostic feature vector and the initial safety benchmark is used as the initial benchmark deviation term; and the amount by which the slope of impedance change and the slope of recovery time constant change exceed their respective safety thresholds are used as the long-term trend risk term. The comprehensive anomaly score is obtained by weighted summing of the model deviation term, the initial benchmark deviation term, and the long-term trend risk term.

[0010] Preferably, the comprehensive anomaly score further includes a physical consistency risk item, which is determined as follows: when the impedance drift is greater than the impedance drift safety threshold and the temperature rise rate is greater than the temperature rise rate safety threshold, a first preset weight value is taken; when the impedance drift is greater than the impedance drift safety threshold and the recovery time constant change rate is greater than the recovery time constant change rate safety threshold, a second preset weight value is taken; the sum of the two is the physical consistency risk item; the comprehensive anomaly score is the weighted sum of the model deviation item, the initial baseline deviation item, the long-term trend risk item, and the physical consistency risk item.

[0011] Preferably, the risk level includes four levels: normal, suspicious, obvious risk, and high risk, determined according to the following rules: when the conformal calibration value is lower than the preset abnormality threshold and the number of consecutive abnormalities reaches a specified number, and the impedance drift and temperature rise rate both exceed their respective safety thresholds, it is judged as high risk; when the conformal calibration value is lower than the preset abnormality threshold and the number of consecutive abnormalities reaches a specified number, but the impedance drift and temperature rise rate do not both exceed their respective safety thresholds, it is judged as obvious risk; when the conformal calibration value is lower than the preset abnormality threshold and the number of consecutive abnormalities is less than a specified number, it is judged as suspicious; the rest are judged as normal.

[0012] Preferably, when the risk level is suspected or obvious risk, the parameter updates of the initial safety benchmark and the normal state model are frozen, that is, the current diagnostic features are not included in the initial safety benchmark, nor are the current diagnostic features used as training data for the normal state model. When the risk level is high risk, the isolation control signal is output.

[0013] Preferably, the port equivalent dynamic impedance is determined by performing a Fourier transform on the port voltage response and port current response during the diagnostic excitation period, extracting the voltage response amplitude and current response amplitude at the diagnostic excitation frequency, and using the ratio of the voltage response amplitude to the current response amplitude as the port equivalent dynamic impedance amplitude; and determining the complex form of the port equivalent dynamic impedance by combining the phase difference between the port voltage response and the port current response.

[0014] Preferably, the recovery time constant is extracted in the following way: after the diagnostic excitation stops, the voltage recovery curve of the port voltage recovering from the peak value to the stable value is collected, the voltage recovery curve is fitted with a first-order exponential decay function, and the decay time constant obtained by fitting is used as the recovery time constant; the temperature rise rate is the difference between the current temperature and the temperature at the time of the last diagnosis divided by the corresponding time interval.

[0015] Preferably, this solution also provides an active isolation control system for thermal runaway of power banks in passenger aircraft cabins, including: The signal acquisition module is used to acquire the port voltage signal, port current signal and temperature signal of the external port of the power bank, and identify the current working status of the power bank based on the port voltage level, port current direction, power flow direction and the available interface protocol status. The diagnostic excitation module is used to apply diagnostic excitation through the external port of the power bank when the preset diagnostic safety constraints are met. During the application of diagnostic excitation, the port voltage response and port current response are collected synchronously and the port equivalent dynamic impedance is determined. The recovery time constant, temperature rise rate and impedance drift are extracted. The port equivalent dynamic impedance, recovery time constant, temperature rise rate and impedance drift obtained from each effective diagnosis, as well as the port voltage, port current and temperature, are used to form a diagnostic feature vector. A control processor, electrically connected to the signal acquisition module and the diagnostic excitation module, is used to construct a multi-timescale port fingerprint. Specifically, when the power bank is first connected and the port voltage, port current, and temperature remain stable within a preset time, the processor records the current port voltage, port current, temperature, port equivalent dynamic impedance, and recovery time constant to form an initial safety benchmark corresponding to the current operating state. The processor arranges the diagnostic feature vectors obtained from the most recent diagnoses before the current diagnostic moment in chronological order to form a short-term sliding window. A long-term trend summary is constructed from the changing trends of the port equivalent dynamic impedance and recovery time constant within a preset time span. The multi-timescale port state fingerprint is input into a normal state model trained with normal state data. The normal state model includes a one-dimensional nonlinear function represented by truncated Fourier sine and cosine terms, used to output normal prediction features or normal reconstruction features. A prediction bias or reconstruction bias is obtained based on the difference between the current diagnostic feature vector and the normal prediction feature or normal reconstruction feature. A comprehensive anomaly score is obtained by weighting and combining the prediction bias or reconstruction bias, the degree of deviation of the current diagnostic feature vector from the initial safety benchmark, and the degree of deviation of the long-term trend summary from the preset safety boundary. The risk confirmation module acquires a set of calibration scores consisting of comprehensive anomaly scores from sample data collected during normal operation, corresponding to the current working state. It then calculates the proportion of samples in the calibration score set that are greater than or equal to the current comprehensive anomaly score to obtain a conformal calibration value. The module also calculates the number of consecutive anomaly counts before the current time where the conformal calibration value is lower than a preset anomaly threshold during a specified number of diagnoses. Finally, it determines the risk level based on the conformal calibration value and the consecutive anomaly count. The signal output module outputs an isolation control signal when the risk level is high.

[0016] Preferably, the isolation actuator is used to implement isolation measures in response to the isolation control signal. The isolation actuator includes a power source and a linkage mechanism. The linkage mechanism includes a main linear actuator, a first crank-rocker mechanism, and a second crank-rocker mechanism. The first crank-rocker mechanism is used to drive the detachable electrical connection component to disengage, and the second crank-rocker mechanism is used to open the lower flap door of the accommodating compartment. Both are driven by the same main linear actuator to form a preset mechanical sequence. The manual emergency trigger module includes a manual emergency trigger interface and a manual emergency trigger button. The manual emergency trigger interface is used to bypass the control processor and directly trigger the isolated actuator when it receives a manual emergency trigger signal generated by the manual emergency trigger button.

[0017] The technical solution provided by this invention has the following beneficial effects: This invention first identifies the current operating state of a power bank by comprehensively judging port voltage, current, and temperature, achieving precise monitoring of different states. Based on this, when diagnostic safety constraints are met, a short-term diagnostic excitation is applied to the external port, resulting in only instantaneous, small-amplitude, and constrained disturbances. This helps reduce the impact on normal charging and power supply states and protocol stability, and facilitates meeting the safety constraints on diagnostic disturbance energy in aviation environments. Based on this, early electrical characteristics such as port equivalent dynamic impedance, recovery time constant, temperature rise rate, and impedance drift are obtained, providing a basis for identifying precursors to thermal runaway before a significant temperature increase. Furthermore, this invention constructs an individualized multi-timescale port fingerprint. An initial safety benchmark establishes a dedicated baseline for each power bank, a short-term sliding window tracks recent rapid changes, and a long-term trend summary captures slow, cumulative drift. These fingerprints are input into a state model trained with normal data. The model prediction bias, initial benchmark deviation, and long-term trend risk are weighted and fused into a comprehensive anomaly score. This score is compared with a normal sample calibration set to obtain a conformal calibration value, and the risk level is determined by combining continuous anomaly criteria. This invention overcomes the shortcomings of fixed thresholds that cannot adapt to individual differences, and can identify minute but continuous single-step degradation, directly outputting isolation control signals when the risk reaches the isolation condition. This invention realizes a shift from passively relying on delayed alarms to actively using port electrical characteristics to identify precursors, helping to improve early warning lead time, reduce the risk of false alarms and missed alarms caused by individual differences, load fluctuations, and transient disturbances, and provide a basis for automated closed-loop handling, thereby enhancing cabin safety. Attached Figure Description

[0018] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, are illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention.

[0019] in: Figure 1 This is a schematic diagram of the risk identification and isolation control process based on port equivalent dynamic impedance fingerprint and normal state model in this invention.

[0020] Figure 2 This is a schematic diagram of the isolation actuator of the present invention in normal use, wherein the electrical connection remains connected and the lower flap door is closed.

[0021] Figure 3 This is a schematic diagram of the action state of the isolation actuator of the present invention after it is triggered, wherein the electrical connection has been disconnected and the lower flap door is in the open state.

[0022] Figure captions: 1. Passenger aircraft seat; 2. First crank rocker mechanism; 3. Second crank rocker mechanism; 4. Lower flap door; 5. Main linear push rod; 6. Reset elastic element; 7. Servo; 8. Rigid actuating element. Detailed Implementation

[0023] To make the technical problems, solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0024] Please refer to Figures 1-3 The present invention provides a method for active isolation and control of thermal runaway of a power bank in the passenger cabin of an aircraft, comprising the following steps: S1. Identify the current operating status of the power bank based on the port voltage level, port current direction, power flow direction, and available interface protocol status; The control processor identifies the current operating mode of the power bank based on the port voltage level, current direction, power flow direction, and available protocol status information. In the formula, Indicates the first The working mode corresponding to each diagnostic moment; This indicates a standby or detection state where the detection interface is connected but no power is continuously supplied to the outside. This indicates that the power bank outputs electrical energy to external devices; This indicates that the power bank accepts external power input; This indicates a state of simultaneous charging and discharging, where both input and output power exist.

[0025] The internal current path, power conversion control, and port response differ under different modes. Therefore, this invention employs a state-separated normal fingerprint model, or uses mode variables as conditional inputs:

[0026] For example, the port dynamic response measured by the same power bank in output power supply mode and the response measured in input charging mode cannot be directly compared with the same threshold. By identifying the status before performing diagnostics, false alarms caused by normal operating condition switching can be reduced.

[0027] This invention clearly distinguishes sensor sampling frequency Diagnostic excitation frequency and diagnostic update cycle : This indicates the number of times the sensor collects voltage and current waveforms per second; A single diagnostic window can contain a large number of sampling points. Therefore, continuous sampling is not equivalent to continuous perturbation. To ensure analytical excitation response, the following is preferably satisfied:

[0028] In practical engineering, Higher values ​​can be selected based on noise immunity and phase calculation accuracy. For example, when the candidate excitation frequency is... At that time, it is possible to use far higher The sampling frequency is used to obtain the complete waveform; the specific value should be determined by actual measurement, rather than fixing a certain frequency as the only implementation method in advance.

[0029] S2. When the preset diagnostic safety constraints are met in the current working state, a diagnostic excitation is applied through the external port of the power bank. During the application of the diagnostic excitation, the port voltage response and port current response are collected synchronously and the port equivalent dynamic impedance is determined. The recovery time constant, temperature rise rate and impedance drift are extracted. The port equivalent dynamic impedance, recovery time constant, temperature rise rate, impedance drift and port voltage, port current and temperature obtained from each effective diagnosis are combined to form a diagnostic feature vector. When the power bank is in a stable standby, stable charging, or stable discharging state, and the active detection meets the constraints of disturbance amplitude, duration, cumulative energy, and port stability, the control unit can apply a low-amplitude diagnostic current through a controllable detection load or power modulation module: The port voltage produces a corresponding response: In the formula, To diagnose the amplitude of the disturbance current, For the excitation frequency, The voltage response amplitude, In response to the phase difference, the dynamic impedance at the candidate frequency can be calculated based on the voltage and current responses:

[0030] For resource-constrained microprocessor implementations, it may be necessary to use only the impedance magnitude, the response change at a candidate frequency, or the equivalent resistance characteristics obtained from a controlled current step, without having to fully calculate the complex impedance spectrum.

[0031] Active perturbations must be implemented within safety boundaries. First, the amplitude of the diagnostic perturbation current should be constrained by the current stable operating current:

[0032] In the formula, This is the current basic operating current. This is the permissible perturbation ratio determined experimentally. Secondly, the duration of a single diagnosis should meet the following requirements:

[0033] Within the preset observation period, the cumulative diagnostic energy should meet the following requirements: If, during the detection period, the port voltage exceeds the allowable drop range, there is a sudden load change, the protocol is renegotiated, the power bank protection is activated, or the status mode changes, the control unit immediately stops active diagnosis and returns to passive continuous monitoring mode.

[0034] Each valid diagnostic event forms a diagnostic feature vector, instead of directly inputting all high-speed raw sampling points into the model. The feature vector obtained from the first effective diagnosis is defined as:

[0035] The features are defined as follows: Port equivalent dynamic impedance. The port equivalent dynamic impedance is determined as follows: Fourier transform is performed on the port voltage response and port current response during diagnostic excitation, and the amplitudes of the voltage response and current response at the diagnostic excitation frequency are extracted. The ratio of the voltage response amplitude to the current response amplitude is used as the amplitude of the port equivalent dynamic impedance. Furthermore, the complex form of the port equivalent dynamic impedance can be determined by combining the phase difference between the voltage response and the current response.

[0036] When using a small current change for lightweight estimation, the equivalent dynamic impedance of the port can be written as: For example, the current change within the diagnostic window is The voltage change caused by this change is Then the equivalent dynamic impedance of the port is: This value represents the overall response of the current port and is not claimed to be the actual internal resistance of a single internal cell.

[0037] Impedance drift relative to the initial safety reference Once the power bank is confirmed to be in an initial safe and stable state, its initial port equivalent dynamic impedance is recorded. The current relative drift is:

[0038] For example, if the initial value is The current value is ,but: That is, the current impedance is higher than the initial safe state. This feature can identify slow but continuously accumulating changes.

[0039] Temperature rise rate The rate of temperature rise is defined as: If the interval between the two diagnoses is Temperature from Rise to ,but: Voltage change rate The rate of change of voltage is defined as: This feature is used to describe whether the port voltage continues to drop under stable operating conditions, or whether there are abnormal changes that are inconsistent with the load.

[0040] Recovery time constant. The recovery time constant is extracted as follows: After the diagnostic excitation stops, the voltage recovery curve of the port voltage from the peak value to the stable value is collected, and the voltage recovery curve is fitted with a first-order exponential decay function. The decay time constant obtained by fitting the curve is used as the recovery time constant; the temperature rise rate is the difference between the current temperature and the temperature at the time of the last diagnosis divided by the corresponding time interval.

[0041] When a port is subjected to a controlled perturbation or a natural load change, the voltage typically recovers gradually to a new stable level. This recovery process can be approximated as follows:

[0042] In the formula, The initial voltage after the disturbance. To stabilize the voltage, This is the recovery time constant. Under the same operating conditions... A continuous increase indicates a slower port response recovery, a phenomenon that, along with impedance drift and temperature rise changes, forms a basis for risk assessment.

[0043] S3. Construct a multi-timescale port fingerprint, specifically: when the power bank is first connected and the port voltage, port current, and temperature remain stable within a preset time, record the current port voltage, port current, temperature, port equivalent dynamic impedance, and recovery time constant to form an initial safety benchmark corresponding to the current working state; arrange the diagnostic feature vectors obtained from the most recent few diagnoses before the current diagnostic time in chronological order to form a short-term sliding window; and construct a long-term trend summary based on the changing trends of the port equivalent dynamic impedance and recovery time constant within a preset time span. If a model uses only the most recent or a few recent data points to predict the next state, it may detect sudden anomalies but struggle to identify slow, gradual deterioration trends. Therefore, this invention employs a three-layer historical information structure: an initial safety baseline, a short-term sliding window, and a long-term trend summary.

[0044] Initial security baseline fingerprint When the power bank is first connected, and its voltage, current, temperature, and status mode stabilize within a preset time, an initial security baseline fingerprint is formed: The initial safety baseline is not updated immediately after each detection, but is saved as a long-term reference to prevent slow anomalies from being treated as new normal states by the model.

[0045] Short-term sliding window Let the short-term window length be... The current short-term window is represented as: This window is used to depict rapid changes over a recent period. If the... If the actual observation differs significantly from the normal response predicted based on the short-term window, a sudden anomaly can be identified.

[0046] Long-term trend summary For historical data over a longer period, it is not necessary to repeatedly input all the original data into the model; instead, compressed trend characteristics are preserved. An exponential moving average is used to obtain smooth impedance characteristics for the equivalent dynamic impedance of the ports. The exponential moving average of the port impedance can be defined as:

[0047] The impedance change slope is obtained by calculating the rate of change over time corresponding to the difference between the current and the Mth previous port equivalent dynamic impedance. The impedance change slope over a longer time span can be defined as: Similarly, based on the rate of change of the recovery time constant corresponding to the difference between the current recovery time constant and the Mth past diagnosis, the slope of the recovery time constant can be obtained, and the long-term trend of the recovery time constant and temperature can be defined: For example, the equivalent dynamic impedance of a power bank's port is from Gradually rise to Although each increment is small, and the model's short-term prediction error for the next point may be small, the cumulative drift relative to the initial baseline has reached: Therefore, long-term benchmark deviation and trend characteristics can complement the shortcomings of short-term forecasts in identifying slow-changing anomalies.

[0048] Baseline freeze protection during abnormal periods If the current state has been identified as suspicious or abnormal, then directly incorporating the current data into the normal baseline is prohibited. The baseline freeze rule can be expressed as:

[0049] In the formula, This represents the parameters of the normality model or online update statistics. This mechanism is used to prevent anomalous data from contaminating normal fingerprints.

[0050] S4. Input the multi-timescale port state fingerprint into the normal state model trained with normal state data. The normal state model includes a one-dimensional nonlinear function represented by truncated Fourier sine and cosine terms, which is used to output normal prediction features or normal reconstruction features. The prediction bias or reconstruction bias is obtained based on the difference between the current diagnostic feature vector and the normal prediction or reconstruction features. The comprehensive anomaly score is obtained by weighted combination based on the prediction bias or reconstruction bias, the degree of deviation of the current diagnostic feature vector from the initial safety benchmark, and the degree of deviation of the long-term trend summary from the preset safety boundary. The normal state model is a Kolmogorov-Arnold network model based on Fourier basis function expansion. The one-dimensional nonlinear function in the model is represented by a weighted combination of truncated sine and cosine terms, which is used to fit the smooth temporal evolution law of port features under the current working state and reduce the impact of isolated contact jitter and transient load spikes on normality judgment.

[0051] In a preferred embodiment, the normal state model is established based on confirmed safe operation data and is used to characterize the normal evolution relationship of port features under the corresponding working state. During operation, when the measured port features continuously deviate from the normal evolution relationship and the initial safety benchmark, the control processor increases the corresponding risk level.

[0052] The normal state model must have at least the following implementation characteristics: The training phase primarily relies on available normal data; Suitable for identifying novel anomalous evolution patterns that have not been fully enumerated; The model output can be interpreted in conjunction with physical characteristics such as port impedance, temperature rise, and recovery time; Suitable for deploying lightweight models on microprocessors or edge controllers.

[0053] In a preferred embodiment, the normal state model is a KAN model; further, the one-dimensional nonlinear function in the KAN model is represented by a truncated Fourier basis function to form a Fourier-KAN normal state model.

[0054] The KAN model expresses the effect of each input feature by setting a learnable one-dimensional nonlinear function. For input variables... Its corresponding learnable function is In this invention, truncated Fourier basis functions are preferably used to express one-dimensional functions:

[0055] In the formula, , and The parameters to be learned For the order of the Fourier expansion, The base frequency parameter is used. The significance of using truncated Fourier expressions is to prioritize learning the smooth evolution of normal port characteristics, thereby reducing the interference of isolated noise, contact jitter, or one-off load spikes on the normality judgment.

[0056] Model Input Structure Taking into account short-term window, long-term trend, initial benchmark deviation, and current operating mode, this invention constructs the model input: In the formula: For short-term sliding windows; Predictive normality model In the predictive implementation, the model predicts the characteristics that should appear under normal conditions at the next effective diagnostic moment based on current and historical information: Obtain the true diagnostic vector at the next time step Then, the predicted residuals can be calculated: In the formula, This represents the distance weighted according to the importance of each diagnostic variable. The predicted residuals are suitable for detecting sudden port response anomalies.

[0057] Reconstructed Normality Model In the reconfigurable implementation, the model recovers the current normal response based on multi-timescale inputs: Predictive and reconstructive approaches can be used individually or in combination to balance the identification of sudden anomalies and the evaluation of deviations from the current state.

[0058] This invention does not directly determine isolation actions based on a single model residual, but rather integrates different types of risk evidence to form a comprehensive anomaly score.

[0059] Predict or reconstruct the deviation term Prediction or reconstruction bias is used to detect whether the current response deviates from the normal short-term evolutionary pattern. Initial baseline deviation term The initial baseline deviation term is used to identify smooth, persistent slow drift anomalies: The weighted distance between the current diagnostic feature vector and the prediction or reconstruction deviation of the normal state model output is used as the model bias term, and the weighted distance between the current diagnostic feature vector and the initial safety benchmark is used as the initial benchmark deviation term. Even if the model can predict a slowly increasing impedance trend, as long as the current state has continuously deviated from the initial safety fingerprint, It will continue to increase.

[0060] Long-term trend risk items The long-term trend risk term can be represented as: In the formula, , and The long-term variation boundary is determined by safety samples or experiments, and the amount by which the slope of impedance change and the slope of recovery time constant change exceed their respective safety thresholds is taken as the long-term trend risk term.

[0061] The physical consistency risk item, which is also included in the overall anomaly score, is determined as follows: A first preset weight value is used when both the impedance drift and temperature rise rate exceed the impedance drift safety threshold; a second preset weight value is used when both the impedance drift and recovery time constant change rate exceed the recovery time constant change rate safety threshold. The sum of these two values ​​constitutes the physical consistency risk item. The overall anomaly score is a weighted sum of the model deviation item, the initial baseline deviation item, the long-term trend risk item, and the physical consistency risk item. A single jump in port impedance may originate from interface contact changes or load switching, rather than thermal runaway risk. Therefore, this invention uses the co-occurrence of multiple physical responses as evidence of aggravated risk.

[0062] In the formula, This is an indicator function. When impedance drift and accelerated temperature rise occur simultaneously, or when impedance drift and significantly slower recovery occur simultaneously, the physical consistency risk term increases.

[0063] Determination of Comprehensive Abnormal Score The final comprehensive abnormality score is: With the above structure, the present invention can simultaneously detect sudden anomalies, long-term slow degradation, and dangerous states corroborated by multiple physical signals.

[0064] S5. Obtain a calibration score set consisting of the comprehensive anomaly scores of sample data collected under normal operating conditions corresponding to the current working state. Calculate the proportion of samples in the calibration score set that are greater than or equal to the current comprehensive anomaly score to obtain the conformal calibration value. Calculate the number of times the conformal calibration value is lower than the preset anomaly threshold in a consecutive specified number of diagnoses before the current time as the consecutive anomaly count. Determine the risk level based on the conformal calibration value and the consecutive anomaly count. The fluctuation range of normal and abnormal scores may vary depending on the power bank and its operating mode. Using only a fixed score threshold can lead to an unstable false alarm rate. This invention utilizes normal, safe samples to form a calibration score set:

[0065] Among them, superscript This indicates that the calibration set corresponds to the current operating mode.

[0066] For the current comprehensive abnormal score Calculate the conformal calibration value: In the formula, This indicates how rare the current outlier score is in the normal reference distribution. The smaller the value, the less likely it is that such large outlier scores will appear in normal samples. It should be noted that... It is not directly equivalent to "normal probability of power banks", but rather a statistical calibration indicator used to assist in control and judgment.

[0067] For example, if the calibration set contains Only one safe sample score, only A normal score greater than or equal to the current score ,but: This indicates that the current situation is significantly incompatible with normal reference data, and the risk level should be increased.

[0068] To prevent isolation actions from being triggered directly by momentary contact fluctuations or one-time load switching, this invention includes a continuous anomaly detection mechanism: In the formula, To continuously determine the window length, For abnormal calibration threshold, This indicates the number of times significant anomalies have occurred within the recent window. When The control system will only enter a higher-level handling state when the preset requirements are met, or when the high-risk physical consistency conditions are met simultaneously.

[0069] Risk level and isolation control output: Risk levels include four categories: normal, suspicious, significant risk, and high risk, determined according to the following rules: High risk is defined as follows: when the conformal calibration value is lower than the preset abnormal threshold and the number of consecutive abnormalities reaches a specified number, and the impedance drift and temperature rise rate both exceed their respective safety thresholds; Significant risk is defined as follows: when the conformal calibration value is lower than the preset abnormal threshold and the number of consecutive abnormalities reaches a specified number, but the impedance drift and temperature rise rate do not both exceed their respective safety thresholds; Suspicious risk is defined as follows: when the conformal calibration value is lower than the preset abnormal threshold and the number of consecutive abnormalities is less than a specified number; all other cases are defined as normal.

[0070] This invention outputs discrete risk levels: The risk level and control meaning can be set as follows: Normal state: The current response is compatible with normal fingerprints and the long-term trend is stable, and the system maintains passive continuous monitoring.

[0071] Suspicious status: Occasional low conformal calibration values ​​or slight deviations occur. The system increases the monitoring or diagnostic frequency and maintains [the following information is missing from the original text]. Significant risk status: Abnormalities persist within a continuous window, or multiple diagnostic features deviate significantly, and the system output... High-risk condition: Conformal calibration values ​​are extremely low and impedance, temperature rise, or recovery characteristics exhibit consistent anomalies; system output... The control signal is defined as: In the formula, To control the output, This establishes a mapping relationship between risk level and current operating mode for selecting appropriate actions. For example, when a significant risk level is reached during charging, a stop charging signal can be prioritized; when a high risk level is reached during power supply, a control signal to cut off power and trigger isolation can be output.

[0072] S6. When the risk level reaches the preset isolation conditions, output an isolation control signal.

[0073] The present invention also provides an active isolation control system for thermal runaway of power banks in passenger aircraft cabins, which is installed on passenger aircraft seat 1 and includes: The present invention also provides an active isolation control system for thermal runaway of power banks in passenger aircraft cabins, comprising: The signal acquisition module is used to acquire the port voltage signal, port current signal and temperature signal of the external port of the power bank, and identify the current working status of the power bank based on the port voltage level, port current direction, power flow direction and the available interface protocol status. The diagnostic excitation module is used to apply diagnostic excitation through the external port of the power bank when the preset diagnostic safety constraints are met. During the application of diagnostic excitation, the port voltage response and port current response are collected synchronously and the port equivalent dynamic impedance is determined. The recovery time constant, temperature rise rate and impedance drift are extracted. The port equivalent dynamic impedance, recovery time constant, temperature rise rate and impedance drift obtained from each effective diagnosis, as well as the port voltage, port current and temperature, are used to form a diagnostic feature vector. A control processor, electrically connected to the signal acquisition module and the diagnostic excitation module, is used to construct a multi-timescale port fingerprint. Specifically, when the power bank is first connected and the port voltage, port current, and temperature remain stable within a preset time, the processor records the current port voltage, port current, temperature, port equivalent dynamic impedance, and recovery time constant to form an initial safety benchmark corresponding to the current operating state. The processor arranges the diagnostic feature vectors obtained from the most recent diagnoses before the current diagnostic moment in chronological order to form a short-term sliding window. A long-term trend summary is constructed from the changing trends of the port equivalent dynamic impedance and recovery time constant within a preset time span. The multi-timescale port state fingerprint is input into a normal state model trained with normal state data. The normal state model includes a one-dimensional nonlinear function represented by truncated Fourier sine and cosine terms, used to output normal prediction features or normal reconstruction features. A prediction bias or reconstruction bias is obtained based on the difference between the current diagnostic feature vector and the normal prediction feature or normal reconstruction feature. A comprehensive anomaly score is obtained by weighting and combining the prediction bias or reconstruction bias, the degree of deviation of the current diagnostic feature vector from the initial safety benchmark, and the degree of deviation of the long-term trend summary from the preset safety boundary. The risk confirmation module acquires a set of calibration scores consisting of comprehensive anomaly scores corresponding to sample data collected under normal operating conditions, corresponding to the current working state. It then calculates the proportion of samples in the calibration score set that are greater than or equal to the current comprehensive anomaly score to obtain a conformal calibration value. It also calculates the number of consecutive anomaly counts before the current time where the conformal calibration value is lower than a preset anomaly threshold during a specified number of diagnoses. Finally, it determines the risk level based on the conformal calibration value and the number of consecutive anomalies. The signal output module outputs an isolation control signal when the risk level is high. An isolation actuator, used to implement isolation measures in response to isolation control signals, includes a power source and a linkage mechanism. The linkage mechanism includes a main linear push rod 5, a first crank-rocker mechanism 2, and a second crank-rocker mechanism 3. The first crank-rocker mechanism 2 is used to drive the detachable electrical connection components to disengage, and the second crank-rocker mechanism 3 is used to open the lower flap door 4 of the storage compartment. Both are driven by the same main linear push rod 5 to form a preset mechanical sequence. The isolation actuator receives isolation control signals or manual emergency trigger signals output by the control processor to transfer the power bank from its normal power supply and storage position to a safe disposal position. The mechanism performs at least two actions: first, it separates the charging cable or detachable electrical connection components from the power bank to terminate the power supply link; then, or simultaneously, it opens the lower flap door 4 of the storage compartment, allowing the power bank to enter the safe transfer channel.

[0074] The power source is the servo motor 7, which uses an actuator with sufficient output torque and position feedback capability. Its output shaft can be connected to a lead screw pair or a gear rack pair via a coupling to convert rotary motion into reciprocating linear motion of the main linear push rod 5. The control unit drives the actuator to complete the preset stroke according to the isolation control signal.

[0075] Main linear push rod 5: serves as the common power input component for the linkage mechanism. A reset elastic element 6 can be installed on the push rod guide rail to return to the initial position under reset command or maintenance conditions.

[0076] First execution link: electrical connection separation. One hinge point of the main linear push rod 5 is connected to the first crank rocker mechanism 2, which drives the transverse rotating shaft to rotate. A rigid actuating element 8 is set on the transverse rotating shaft, which does not affect the placement of the power bank under normal conditions; after being triggered, the rigid actuating element 8 acts on the force-bearing position of the separable electrical connection component, so as to physically separate the power bank from the daily power supply link.

[0077] Second execution link: Physical transfer channel opening. The other hinge point of the main linear push rod 5 is connected to the second crank rocker mechanism 3. The mechanism drives the actuation fork to open the one-way flap door at the bottom of the receiving compartment, forming an opening to the safe transfer channel.

[0078] The first and second execution links are driven by the same main linear push rod 5. By setting the position of the toggle, the stroke of the flap door linkage, and the direction of the push rod movement, the electrical connection separation action can be made to occur before or no later than the flap door is fully opened, ensuring that the power bank is disconnected from the electrical connection before entering the safe transfer channel.

[0079] like Figure 2 As shown, under normal use, the flip door at the bottom of the compartment is in the closed position, and the charging cable or detachable interface remains connected; the power source, main linear push rod 5 and linkage are in the untriggered position, and the power bank can remain in the daily use area.

[0080] like Figure 3 As shown, when the control processor outputs an isolation control signal or the user triggers the manual emergency button, the power source drives the main linear push rod 5 to move: the upper linkage pushes the charging cable or detachable interface out of the connection position, and the lower linkage drives the flip door to open. At this time, the power bank is disconnected from the external power supply link and can fall into or be guided into the final safe disposal unit through the opened door and safety transfer channel.

[0081] It also includes a final safe disposal unit, comprising a housing structure, suppression and thermal protection modules, and exhaust and filtration structures; the final safe disposal unit is located at the end of the safe transfer channel to receive the isolated power banks. Its housing may employ a multi-layered composite protective structure, including an impact-resistant inner lining, a thermal insulation layer, and a load-bearing outer shell. The housing is equipped with a maintenance access port, which has a sealing and locking structure to facilitate ground maintenance, replacement of consumable parts, and removal of residues after an incident.

[0082] The interior of the enclosure can be equipped with heat suppression, cooling, or fire extinguishing medium release modules according to airworthiness and safety verification requirements. These modules can be triggered by internal temperature, pressure, or external control signals to reduce the spread of open flames, mitigate the thermal impact, and limit the risk of reignition.

[0083] The final safety disposal unit may be equipped with an exhaust port connected to the safety exhaust path, and heat-resistant and filtering components may be installed in the exhaust path to reduce the adverse effects of high-temperature flue gas and particulate matter on the cabin environment. The specific exhaust path and material configuration should be determined in conjunction with the aircraft structure, airworthiness requirements, and verification tests.

[0084] The workflow of this invention is as follows: Normal operation. The power bank is placed in the compartment on the seat back and forms a normal connection through the interface component, maintaining the isolation actuator. Figure 2 As shown, the lower flap door 4 is closed. The control unit identifies the current operating mode and continuously performs passive data acquisition of voltage, current, and temperature. If the operating state is stable and meets diagnostic safety constraints, the control unit performs short-term diagnostics and updates the port fingerprint characteristics according to a preset cycle.

[0085] Suspicious Status Handling. When the model output shows a slight anomaly, or when the conformal calibration results indicate that the current state is rarely seen in the normal reference data, the system enters a suspicious state. In this case, the passive sampling frequency can be increased, the diagnostic interval shortened, port characteristics recorded, and crew members alerted, but physical isolation is not immediately implemented.

[0086] Significant Risk Handling. When an abnormal state persists within a continuous window, or when impedance drift, slowed recovery, and accelerated temperature rise occur together in a series of abnormalities, the system enters a significant risk state. Based on the current energy flow, the control system can output control signals to stop charging, limit power supply, or enter an isolation preparation state.

[0087] High-risk isolation procedures. When the overall anomaly score reaches a high-risk level, or when direct danger signals such as temperature or smoke meet emergency trigger conditions, or when a passenger or crew member presses the manual emergency button, the control unit immediately outputs an isolation signal. The isolation execution mechanism is... Figure 2 The normal state shown is converted to Figure 3 As shown in the action state, the electrical connection is disconnected and the lower flip door 4 is opened first or simultaneously; the power bank then moves along... Figure 2 The safety transfer channel shown leads to the final safety disposal unit.

[0088] Event logging and maintenance. Before and after the incident is triggered, the control unit saves key port data, risk levels, control signals, and action feedback information for subsequent maintenance and incident analysis. After the incident, maintenance personnel check the final safety handling unit, actuators, and sensor control modules according to the operating procedures to restore the system to its operational state.

[0089] In summary, this invention first identifies the current operating state of the power bank by comprehensively judging port voltage, current, and temperature, achieving precise monitoring across different states. Based on this, when diagnostic safety constraints are met, a short-term diagnostic stimulus is applied to the external port, causing only a momentary, small-amplitude disturbance. This avoids affecting normal charging and protocol stability while also meeting the strict energy injection restrictions for aviation safety. Based on this, early electrical characteristics such as the port's equivalent dynamic impedance, recovery time constant, temperature rise rate, and impedance drift are obtained, capturing early signs of thermal runaway without relying on lagging temperature signals. Furthermore, this invention constructs an individualized multi-timescale port fingerprint. An initial safety benchmark establishes a dedicated baseline for each power bank, a short-term sliding window tracks recent rapid changes, and a long-term trend summary captures slow, cumulative drift. These fingerprints are input into a state model trained with normal data. The model prediction bias, initial benchmark deviation, and long-term trend risk are weighted and fused into a comprehensive anomaly score. This score is compared with a normal sample calibration set to obtain a conformal calibration value, and the risk level is determined by combining continuous anomaly criteria. This invention overcomes the shortcomings of fixed thresholds that cannot adapt to individual differences, and can identify minute but continuous single-step degradation, directly outputting isolation control signals when the risk reaches the isolation condition. This invention realizes a shift from passively relying on delayed alarms to actively using port electrical characteristics to identify precursors, effectively improving early warning lead time, reducing false alarms and missed alarms, and providing a basis for automated closed-loop handling, thereby significantly enhancing cabin safety.

[0090] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any non-substantial improvements made using the inventive concept and technical solution of the present invention, or the direct application of the inventive concept and technical solution of the present invention to other occasions without modification, are all within the protection scope of the present invention.

Claims

1. A method for active isolation and control of thermal runaway of a power bank in a passenger aircraft cabin, characterized in that, Includes the following steps: The power bank's external port voltage signal, port current signal, and temperature signal are collected at a preset sampling frequency. The power bank's current working status is identified based on the port voltage level, port current direction, power flow direction, and available interface protocol status. When the current working state meets the preset diagnostic safety constraints, a diagnostic excitation is applied through the external port of the power bank. During the application of the diagnostic excitation, the port voltage response and port current response are collected synchronously and the port equivalent dynamic impedance is determined. The recovery time constant, temperature rise rate and impedance drift are extracted. The port equivalent dynamic impedance, recovery time constant, temperature rise rate and impedance drift obtained from each effective diagnosis, as well as the port voltage, port current and temperature, are used to form a diagnostic feature vector. A multi-timescale port fingerprint is constructed as follows: When the power bank is first connected and the port voltage, port current, and temperature remain stable within a preset time, the current port voltage, port current, temperature, port equivalent dynamic impedance, and recovery time constant are recorded to form an initial safety benchmark corresponding to the current working state; the diagnostic feature vectors obtained from the most recent few diagnoses before the current diagnosis time are arranged in chronological order to form a short-term sliding window; and the long-term trend summary is constructed from the changing trends of the port equivalent dynamic impedance and recovery time constant within a preset time span. The multi-timescale port state fingerprint is input into a normal state model trained with normal state data. The normal state model includes a one-dimensional nonlinear function represented by truncated Fourier sine and cosine terms, used to output normal prediction features or normal reconstruction features. The prediction bias or reconstruction bias is obtained based on the difference between the current diagnostic feature vector and the normal prediction feature or normal reconstruction feature. A comprehensive anomaly score is obtained by weighting and combining the prediction bias or reconstruction bias, the degree of deviation of the current diagnostic feature vector from the initial security benchmark, and the degree of deviation of the long-term trend summary from the preset security boundary. Obtain a calibration score set consisting of comprehensive anomaly scores corresponding to sample data collected under normal operating conditions, corresponding to the current working state; calculate the proportion of samples in the calibration score set that are greater than or equal to the current comprehensive anomaly score to obtain a conformal calibration value; count the number of times the conformal calibration value is lower than a preset anomaly threshold in a specified number of consecutive diagnoses before the current time as the consecutive anomaly count; determine the risk level based on the conformal calibration value and the consecutive anomaly count. When the risk level reaches the preset isolation condition, an isolation control signal is output.

2. The active isolation and control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 1, characterized in that, The long-term trend summary is constructed as follows: the port equivalent dynamic impedance is smoothed by applying an exponential moving average to obtain the smoothed impedance characteristics; the impedance change slope is obtained based on the time change rate corresponding to the difference between the current port equivalent dynamic impedance and the Mth past diagnosis; the recovery time constant change slope is obtained based on the time change rate corresponding to the difference between the current recovery time constant and the Mth past diagnosis; and the long-term trend summary is composed of the smoothed impedance characteristics, the impedance change slope, and the recovery time constant change slope.

3. The active isolation control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 2, characterized in that, The comprehensive anomaly score is determined as follows: the weighted distance between the current diagnostic feature vector and the prediction deviation or reconstruction deviation of the normal state model output is used as the model deviation term; the weighted distance between the current diagnostic feature vector and the initial safety benchmark is used as the initial benchmark deviation term; and the amount by which the slope of impedance change and the slope of recovery time constant change exceed their respective safety thresholds are used as the long-term trend risk term. The comprehensive anomaly score is obtained by weighted summing of the model deviation term, the initial benchmark deviation term, and the long-term trend risk term.

4. The active isolation and control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 3, characterized in that, The comprehensive anomaly score also includes a physical consistency risk item, which is determined as follows: when the impedance drift is greater than the impedance drift safety threshold and the temperature rise rate is greater than the temperature rise rate safety threshold, a first preset weight value is taken; when the impedance drift is greater than the impedance drift safety threshold and the recovery time constant change rate is greater than the recovery time constant change rate safety threshold, a second preset weight value is taken; the sum of the two is the physical consistency risk item; the comprehensive anomaly score is the weighted sum of the model deviation item, the initial baseline deviation item, the long-term trend risk item, and the physical consistency risk item.

5. The active isolation control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 1, characterized in that, The risk levels are categorized into four levels: normal, suspicious, significant risk, and high risk, determined according to the following rules: High risk is defined as follows: when the conformal calibration value is below a preset anomaly threshold and the number of consecutive anomalies reaches a specified number, and both impedance drift and temperature rise rate exceed their respective safety thresholds; significant risk is defined as follows: when the conformal calibration value is below the preset anomaly threshold and the number of consecutive anomalies reaches a specified number, but impedance drift and temperature rise rate do not simultaneously exceed their respective safety thresholds; suspicious risk is defined as follows: when the conformal calibration value is below the preset anomaly threshold and the number of consecutive anomalies is less than a specified number; all other cases are defined as normal.

6. The active isolation and control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 5, characterized in that, When the risk level is suspected or obvious, the parameter updates of the initial safety benchmark and the normal state model are frozen, that is, the current diagnostic features are not included in the initial safety benchmark, nor are the current diagnostic features used as training data for the normal state model. When the risk level is high risk, the isolation control signal is output.

7. The active isolation and control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 1, characterized in that, The equivalent dynamic impedance of the port is determined by performing a Fourier transform on the port voltage response and port current response during the diagnostic excitation period, extracting the voltage response amplitude and current response amplitude at the diagnostic excitation frequency, and using the ratio of the voltage response amplitude to the current response amplitude as the amplitude of the port equivalent dynamic impedance. The complex form of the port equivalent dynamic impedance is determined by combining the phase difference between the port voltage response and the port current response.

8. The active isolation and control method for thermal runaway of a power bank in a passenger aircraft cabin according to claim 1, characterized in that, The recovery time constant is extracted in the following way: after the diagnostic excitation stops, the voltage recovery curve of the port voltage recovering from the peak value to the stable value is collected, the voltage recovery curve is fitted with a first-order exponential decay function, and the decay time constant obtained by fitting is used as the recovery time constant; the temperature rise rate is the difference between the current temperature and the temperature at the time of the last diagnosis divided by the corresponding time interval.

9. A thermal runaway active isolation control system for power banks in passenger aircraft cabins, characterized in that, include: The signal acquisition module is used to acquire the port voltage signal, port current signal and temperature signal of the external port of the power bank, and identify the current working status of the power bank based on the port voltage level, port current direction, power flow direction and the available interface protocol status. The diagnostic excitation module is used to apply diagnostic excitation through the external port of the power bank when the preset diagnostic safety constraints are met. During the application of diagnostic excitation, the port voltage response and port current response are collected synchronously and the port equivalent dynamic impedance is determined. The recovery time constant, temperature rise rate and impedance drift are extracted. The port equivalent dynamic impedance, recovery time constant, temperature rise rate and impedance drift obtained from each effective diagnosis, as well as the port voltage, port current and temperature, are used to form a diagnostic feature vector. A control processor, electrically connected to the signal acquisition module and the diagnostic excitation module, is used to construct a multi-timescale port fingerprint. Specifically, when the power bank is first connected and the port voltage, port current, and temperature remain stable within a preset time, the processor records the current port voltage, port current, temperature, port equivalent dynamic impedance, and recovery time constant to form an initial safety benchmark corresponding to the current operating state. The processor arranges the diagnostic feature vectors obtained from the most recent diagnoses before the current diagnostic moment in chronological order to form a short-term sliding window. A long-term trend summary is constructed from the changing trends of the port equivalent dynamic impedance and recovery time constant within a preset time span. The multi-timescale port state fingerprint is input into a normal state model trained with normal state data. The normal state model includes a one-dimensional nonlinear function represented by truncated Fourier sine and cosine terms, used to output normal prediction features or normal reconstruction features. A prediction bias or reconstruction bias is obtained based on the difference between the current diagnostic feature vector and the normal prediction feature or normal reconstruction feature. A comprehensive anomaly score is obtained by weighting and combining the prediction bias or reconstruction bias, the degree of deviation of the current diagnostic feature vector from the initial safety benchmark, and the degree of deviation of the long-term trend summary from the preset safety boundary. The risk confirmation module acquires a set of calibration scores consisting of comprehensive anomaly scores from sample data collected during normal operation, corresponding to the current working state. It then calculates the proportion of samples in the calibration score set that are greater than or equal to the current comprehensive anomaly score to obtain a conformal calibration value. The module also calculates the number of consecutive anomaly counts before the current time where the conformal calibration value is lower than a preset anomaly threshold during a specified number of diagnoses. Finally, it determines the risk level based on the conformal calibration value and the consecutive anomaly count. The signal output module outputs an isolation control signal when the risk level is high.

10. The active isolation control system for thermal runaway of a power bank in a passenger aircraft cabin according to claim 9, characterized in that, Also includes: An isolation actuator is used to implement isolation measures in response to the isolation control signal. The isolation actuator includes a power source and a linkage mechanism. The linkage mechanism includes a main linear actuator, a first crank-rocker mechanism, and a second crank-rocker mechanism. The first crank-rocker mechanism is used to drive the detachable electrical connection component to disengage, and the second crank-rocker mechanism is used to open the lower flap door of the accommodating compartment. Both mechanisms are driven by the same main linear actuator to form a preset mechanical sequence. The manual emergency trigger module includes a manual emergency trigger interface and a manual emergency trigger button. The manual emergency trigger interface is used to bypass the control processor and directly trigger the isolated actuator when it receives a manual emergency trigger signal generated by the manual emergency trigger button.

Images