A battery thermal runaway early warning method and system based on NV color center entropy flow pattern calibration

Abstract

The application belongs to the technical field of intelligent early warning, and particularly relates to a battery thermal runaway early warning method and system based on NV color center entropy production manifold calibration, which comprises an NV color center quantum sensing acquisition module, a battery state synchronous acquisition module, a pulse modulation module and an early warning output module; the method acquires a pulse ODMR high-dimensional quantum signal in the battery operation process, calculates the Gibbs free energy gradient, reaction enthalpy change and local irreversible entropy production rate, and embeds the same as a thermodynamic penalty term into the adjacent edge distance of a near neighbor graph, obtains a thermodynamic weighted geodesic distance from the shortest path of the near neighbor graph, and constructs a physically constrained low-dimensional phase space; the boundary distance is fed back to adjust the pulse quantity, pulse interval and scanning range, and the battery health state and early warning result are output. The application improves the physical consistency, early warning accuracy and self-adaptive detection capability of battery anomaly identification under complex working conditions.

Description

Technical Field

[0001] This application belongs to the field of intelligent early warning technology, specifically relating to a battery thermal runaway early warning method and system for NV color center entropy generation manifold calibration. Background Technology

[0002] Diamond NV centers are solid-state defect centers with quantum spin properties, enabling highly sensitive detection of physical quantities such as magnetic fields, temperature, stress, and electric fields using photodetector magnetic resonance (PDMR) technology. In recent years, NV center quantum sensing technology has been increasingly introduced into the field of battery detection to acquire signals related to local magnetic field distribution, temperature changes, and electrochemical reactions during battery operation. Compared to traditional macroscopic sensors, NV center sensors offer advantages such as high spatial resolution, high sensitivity, in-situ detection capability, and strong multi-physics response, providing a new technological approach for monitoring the internal state of batteries and providing early safety warnings.

[0003] However, existing battery detection methods based on NV centers mainly focus on the direct extraction and threshold determination of parameters such as ODMR spectral line shifts, linewidths, fluorescence intensity, temperature changes, or magnetic field changes. While these methods can acquire rich quantum sensing data, their diagnostic logic is mostly an open-loop processing mode of "signal acquisition—feature extraction—threshold determination" or "signal acquisition—model classification—state recognition." For battery systems under complex operating conditions, relying solely on changes in a single physical quantity or static spectral features is easily affected by environmental magnetic field disturbances, temperature drift, laser power fluctuations, material aging differences, and changes in charge and discharge conditions, resulting in insufficient stability of diagnostic results.

[0004] Therefore, there is an urgent need to propose a new closed-loop early warning method for battery thermal runaway. This method should be able to calibrate the manifold neighborhood relationship and low-dimensional phase space by introducing battery thermodynamic constraints based on the acquisition of high-dimensional ODMR quantum signals, and further establish the mapping relationship between local irreversible entropy yield and NV color center transverse relaxation time. Then, the pulse measurement parameters should be dynamically adjusted according to the distance from the current state to the thermal runaway safety boundary, thereby improving the reliability, physical interpretability and timely warning of early battery anomaly identification under complex operating conditions. Summary of the Invention

[0005] To address the problems of existing NV center battery detection methods, which mainly rely on direct extraction of ODMR spectral line parameters, lack of thermodynamic constraint manifold calibration mechanism, difficulty in establishing quantitative relationship between microscopic quantum decoherence characteristics and macroscopic irreversible battery decay process, and inability to dynamically adjust quantum measurement parameters according to battery safety status, this invention proposes a battery thermal runaway early warning method and system based on NV center entropy generation manifold calibration.

[0006] The purpose of this invention is to provide a battery diagnostic and early warning method that maintains high physical consistency and early warning reliability under complex charge-discharge conditions, environmental disturbances, and battery aging. This method does not simply perform conventional frequency shift, linewidth, or fluorescence intensity analysis on the NV center ODMR signal. Instead, it embeds thermodynamic state variables such as Gibbs free energy, local reaction enthalpy change, and local irreversible entropy yield during battery operation into the manifold modeling process of high-dimensional ODMR data. It uses thermodynamically weighted adjacent edge distances to correct candidate adjacency relationships between high-dimensional spectral state points and obtains thermodynamically weighted geodesic distances based on the shortest path in the physically constrained nearest neighbor graph, thereby obtaining a low-dimensional physical phase space that better reflects the true irreversible evolution of the battery. Furthermore, it establishes a cross-scale dynamic mapping relationship between the local irreversible entropy yield and the NV center transverse relaxation time, and adjusts the NV center pulse measurement parameters based on the distance feedback from the current state to the thermal runaway safety boundary manifold surface, achieving adaptive quantum measurement and early warning driven by the battery's safety state.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for early warning of battery thermal runaway based on NV color center entropy generation manifold calibration includes the following steps: S1. Obtain the high-dimensional quantum signal of NV color center pulse ODMR during the operation of the target battery to form a high-dimensional spectral state dataset. This includes at least the fitting characteristics of ODMR spectral line shift, fluorescence intensity, linewidth, coherent evolution time, and transverse relaxation time; S2. Simultaneously acquire the voltage, current, temperature, charge / discharge rate, and electrochemical impedance parameters of the target battery, and calculate the thermodynamic state quantities corresponding to the high-dimensional spectral state data. These thermodynamic state quantities include at least the Gibbs free energy. Gibbs free energy gradient Local reaction enthalpy change Local irreversible entropy production rate And the cumulative irreversible entropy output obtained by integrating the local irreversible entropy output rate. ; S3. Using the high-dimensional spectral state dataset as a node set, construct the thermodynamically weighted adjacency distance; S4, Based on the aforementioned thermodynamically weighted adjacent edge distance Construct a physically constrained nearest neighbor graph, calculate the shortest path between any two state points in the graph, and obtain the thermodynamically weighted geodesic distance. Based on the thermodynamically weighted geodesic distance, manifold dimensionality reduction is performed to generate low-dimensional physical phase space coordinates, so that the low-dimensional physical phase space coordinates simultaneously satisfy the ODMR spectral neighborhood preservation condition and the battery irreversible thermodynamic evolution consistency condition. S5. Establish local irreversible entropy production rate Lateral relaxation time of NV color center The cross-scale dynamic mapping relationship between them; S6. Using the cross-scale dynamic mapping relationship between the local irreversible entropy yield and the transverse relaxation time determined in step S5, perform physical consistency calibration on the normal operation samples and historical fault samples, so that the boundary samples simultaneously satisfy the decoherent evolution trend of the NV color center and the irreversible entropy production evolution trend of the battery; based on the physically consistent calibrated normal operation samples, historical fault samples, and thermodynamic stability constraints, determine the thermal runaway safety boundary manifold in the low-dimensional physical phase space, and calculate the current battery state coordinates. Boundary distance to the thermal runaway safety boundary manifold and based on the boundary distance Determine the battery safety level; S7. Based on the boundary distance In accordance with the battery safety level, the NV color center pulse measurement parameters are adjusted in a closed loop. The pulse measurement parameters include at least the number of pulses, the pulse interval, and the frequency scanning range, and the battery health status and warning results are output.

[0008] Preferably, the locally irreversible entropy yield in step S2 is obtained as follows: ; in, For the first Local irreversible entropy production rate corresponding to each state point; , and The locations within the area to be measured are respectively The volumetric heat generation rate of polarization heat, ohmic heat and side reaction heat at the point; This refers to the local absolute temperature. This is the area of ​​the battery to be tested. The cumulative irreversible entropy yield is obtained as follows: ; in, For the first The sampling time corresponding to each state point For integration time; The first The change in Gibbs free energy corresponding to each state point is obtained as follows: ; in, The number of electrons transferred in an electrochemical reaction. It is Faraday's constant. For the first The open-circuit voltage corresponding to each state point; The Gibbs free energy gradient is obtained as follows: ; The Gibbs free energy gradient of the first state point is determined by the forward difference between the first and second state points. The first The reaction entropy change and local reaction enthalpy change corresponding to each state point are as follows: ; ; in, For the first The absolute temperature corresponding to each state point For the first The reaction entropy change corresponding to each state point For the first The local enthalpy change of the reaction corresponding to each state point.

[0009] Preferably, the thermodynamic penalty term in step S3 It consists of a thermodynamic state difference term and a thermodynamic evolution direction penalty term, and its expression is: ; ; ; ; ; in, , , , and These are the non-negative weighting coefficients for the Gibbs free energy gradient difference term, the local reaction enthalpy difference term, the local irreversible entropy yield difference term, the cumulative irreversible entropy yield difference term, and the thermodynamic evolution direction penalty term, respectively, and they satisfy the following: ; , , and These are the scale normalization factors for the corresponding thermodynamic state quantities, and both are greater than zero; This is a penalty term for the direction of thermodynamic evolution.

[0010] Preferably, the thermodynamically weighted adjacent edge distance is: ; in, For the first The ODMR high-dimensional spectral state point and the ... Thermodynamic weighted adjacency distance between ODMR high-dimensional spectral state points; The normalized geometric distance between two ODMR high-dimensional spectral state points; The geometric distance weights are non-negative. The non-negative thermodynamic penalty weight is dynamically updated according to battery operating conditions; This is a thermodynamic penalty term; ; in, and The first The and the first A high-dimensional spectral state vector of ODMR , is the scale normalization factor for the ODMR high-dimensional spectral state dataset; Thermodynamic evolution direction penalty term Obtain it in the following way: ; in, For the first Thermodynamic state vectors at state points, and: ; For the first Thermodynamic state vectors at state points, and: ; ; in For the first Thermodynamic state vector of a state point; The thermodynamic driving direction vector is determined based on the thermodynamic state vectors of adjacent measurement periods; the local thermodynamics of the first state point. Sure; To prevent positive numbers with a denominator of zero; When the direction of state change corresponding to a candidate adjacent edge is opposite to the direction of local thermodynamic driving... This increases the edge weight of the candidate adjacent edge; when the state change direction corresponding to the candidate adjacent edge is aligned with the local thermodynamic driving direction... Or it may approach 0.

[0011] Preferably, the manifold dimensionality reduction in step S4 employs a thermodynamically constrained metric mapping algorithm, with the objective function being: , ; ; in, It is a set of low-dimensional physical phase space coordinates; For the first The low-dimensional physical phase space coordinates corresponding to each state point; The shortest path geodesic distance is calculated based on the physical constraint nearest neighbor graph. This is the normalized Gibbs free energy estimate obtained from reconstruction based on adjacent samples in the low-dimensional physical phase space; This is the normalized cumulative irreversible entropy yield estimate obtained from reconstruction based on adjacent samples in the low-dimensional physical phase space; and These are the normalized mappings of Gibbs free energy and cumulative irreversible entropy output, respectively; and These are the regularization coefficients for the Gibbs free energy consistency constraint and the cumulative irreversible entropy production consistency constraint, respectively.

[0012] Preferably, the cross-scale dynamic mapping relationship described in step S5 is as follows: ; in, These are cross-scale mapping coefficients related to the battery material system. , is a threshold activation function used to characterize the nonlinear enhancement of the decoherent response of the NV color center when the local irreversible entropy yield crosses a critical threshold; The threshold activation function is the Sigmoid threshold activation function, and its expression is: ; in, To activate the kurtosis factor, The critical entropy yield threshold; The discrete form of the cross-scale dynamic mapping relationship is: ; in, and The first The and the first Transverse relaxation time for each measurement cycle; For the first Local irreversible entropy production rate for each measurement cycle; To measure the update cycle, the measured transverse relaxation time at the first calibration time is used as the initial value of the discrete mapping. .

[0013] Preferably, the cross-scale mapping coefficients Activate the kurtosis factor and critical entropy yield threshold The calibration was obtained through joint calibration, which included the following steps: selecting at least five battery samples of the same type in different health states; performing spin echo attenuation measurement on each sample and extracting the transverse relaxation time. Simultaneously, the local irreversible entropy yield is calculated using electrochemical impedance data, current data, and temperature data. The lateral relaxation time With the aforementioned local irreversible entropy production rate Data pairs were formed, and the least squares regression method was used to determine the cross-scale mapping coefficients. Activate the kurtosis factor and critical entropy yield threshold Its target is: ; in, To determine the number of calibrated samples; The measured transverse relaxation time corresponding to the l-th calibration sample; To predict the lateral relaxation time.

[0014] Preferably, the thermal runaway safety boundary manifold surface in step S6 is jointly determined by historical fault samples, normal operation samples, and thermodynamic stability constraints that have undergone physical consistency calibration via the cross-scale dynamic mapping relationship described in step S5; the boundary distance is calculated using the directed projection distance of the local tangent plane of the manifold, and its expression is: ; in, This is the thermal runaway safety boundary discrimination function. These are the low-dimensional physical phase space coordinates of the current battery state. The gradient of the thermal runaway safety boundary discrimination function at the current battery state coordinates is given. To prevent positive numbers with a denominator of zero; The sign direction of the thermal runaway safety boundary discrimination function is uniformly defined as follows: This indicates that the state point is located on the safe side of the thermal runaway safety boundary. This indicates that the state point is located on the thermal runaway safety boundary manifold. This indicates that the state point has crossed the thermal runaway safety boundary and entered the danger side; The safety threshold, warning threshold, and danger threshold satisfy the following: ; when When this occurs, it is considered a normal state; when When this occurs, it is determined to be a state of attention; when When this occurs, it is determined to be a warning state; when When this occurs, it is determined to be a dangerous situation.

[0015] Preferably, the closed-loop regulation in step S7 adopts a pulse resource redistribution control law based on directed boundary distance error. The boundary distance error for each measurement cycle is: ; in, For the first The directed boundary distance for each measurement cycle; The amplitude-limited integral state of the boundary distance error is as follows: ; The pulse count is updated as follows: ; The pulse interval is updated as follows: ; The frequency scan range is updated as follows: ; in: ; This indicates that the number of pulses will be a positive integer; , and These are the number of reference pulses, the interval between reference pulses, and the range of reference frequency scanning, respectively. , , and These are non-negative control parameters; This represents the upper limit of the integration state. and These are the lower and upper limits for the number of pulses, respectively. and These are the lower and upper limits of the pulse interval, respectively. and These are the lower and upper limits of the frequency scan range, respectively. when At this time, both the boundary distance error and the integral state are zero, maintaining the low-power reference scan mode; when At that time, enter enhanced scanning mode; when At this time, it enters high-sensitivity scanning mode; when At that time, the emergency warning mode will be activated.

[0016] A battery thermal runaway early warning system with NV color center entropy generation manifold calibration includes an NV color center quantum sensing acquisition module, a battery state synchronous acquisition module, a pulse modulation module, and an early warning output module; The NV color center quantum sensing acquisition module is used to perform pulsed ODMR measurements on the target battery and acquire high-dimensional quantum signals. The battery status synchronous acquisition module is used to synchronously acquire the voltage, current, temperature, charge / discharge rate and electrochemical impedance parameters of the target battery. The processor module is used to complete the construction of thermodynamically weighted adjacent edge distances, the calculation of geodesic distances of the shortest path in the physically constrained nearest neighbor graph, the dimensionality reduction of the physically constrained manifold, the cross-scale mapping of entropy production rate and lateral relaxation time, the calculation of safety boundary distances, and closed-loop pulse modulation. The pulse modulation module is used to adjust the number of pulses, the pulse interval, and the frequency scanning range according to the closed-loop modulation command output by the processor module. The warning output module is used to output the target battery's health status, safety level, boundary distance, and warning information; When the NV color center quantum sensing acquisition module includes multiple NV color center sensing units, the processor module calculates the directed boundary distance of each sensing region and determines the minimum value among the directed boundary distances of each sensing region as the module-level boundary distance.

[0017] Compared with the prior art, the beneficial effects of this application are as follows:

[0018] First, this invention embeds the Gibbs free energy gradient, local reaction enthalpy change, local irreversible entropy yield, and cumulative irreversible entropy yield into the nearest neighbor graph construction process of high-dimensional ODMR signals by using thermodynamically weighted adjacent edge distances. The thermodynamically weighted geodesic distance is obtained by physically constraining the shortest path of the nearest neighbor graph, so that the low-dimensional embedding result not only maintains the geometric neighborhood structure of the spectral data, but also conforms to the direction of irreversible thermodynamic evolution of the battery, thereby reducing neighborhood misconnections and topological distortions caused by pure data manifold dimensionality reduction under complex operating conditions.

[0019] Second, the present invention penalizes candidate adjacent edges that are inconsistent with the current thermodynamic driving direction of the battery by using a thermodynamic evolution direction penalty term, so that the nearest neighbor graph construction process can exclude non-physical adjacent relationships caused by noise disturbance, temperature drift or transient abnormal sampling, thereby improving the physical consistency of the battery state phase space representation.

[0020] Third, this invention establishes the relationship between local irreversible entropy yield and NV color center transverse relaxation time. The cross-scale dynamic mapping relationship between them enables the macroscopic irreversible decay process of batteries to be characterized by microscopic quantum decoherence characteristics, which is beneficial for capturing early characteristics of local side reaction enhancement, lithium dendrite initiation, local short circuit precursors and early thermal runaway evolution.

[0021] Fourth, the present invention performs closed-loop adjustment of the number of pulses, pulse interval and frequency scanning range based on the boundary distance from the current battery state to the thermal runaway safety boundary manifold surface, so that the system maintains low power consumption reference scanning under normal conditions and automatically switches to enhanced scanning or high sensitivity measurement mode when approaching the warning boundary, thereby taking into account real-time performance, sensitivity and system energy consumption.

[0022] Fifth, this invention forms a process of "NV color center high-dimensional quantum signal acquisition—thermodynamically constrained manifold calibration—entropy production". The complete technical closed loop of "cross-scale mapping - safety boundary distance calculation - closed-loop pulse modulation - early warning output" can improve the reliability, interpretability and early warning capability of battery safety diagnosis under complex operating conditions compared with the scheme that only performs NV color center signal acquisition or static threshold judgment. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below; Figure 1 This diagram illustrates the data flow and feedback relationship between data acquisition, thermodynamic constraint calibration, core algorithm processing, early warning output, feedback control, and closed-loop adaptive pulse modulation in this invention. Figure 2 A logical diagram illustrating how a two-dimensional fuzzy controller is used to solve for thermodynamic penalty weights based on battery temperature change rate and charge / discharge rate, and how these thermodynamic penalty weights are fed back to the joint physical distance metric and low-dimensional physical phase space mapping process. Figure 3 This is a schematic diagram of the macroscopic irreversible entropy production path calculated based on electrochemical impedance parameters and thermal generation rate, the microscopic quantum dynamic path for extracting transverse relaxation time based on spin echo decay measurement, and the logic diagram of the cross-scale dynamic mapping relationship established by the two paths. Figure 4 This is a closed-loop control diagram that compares the current directed safety boundary distance with the safety threshold, and then generates control commands for the number of pulses, pulse interval, and frequency scan range through boundary distance error, proportional-integral adjustment, and amplitude limiting. Figure 5 This is a schematic diagram illustrating the principle of fitting the thermal runaway safety boundary using sample trajectories calibrated by thermodynamic constraints and cross-scale dynamic mapping, and completing the battery health state classification using the directed projection distance from the current state point to the local tangent plane of the boundary. Detailed Implementation

[0024] The technical solution of the present invention will be further described below with reference to specific embodiments. It should be understood that the following embodiments are only for illustrating the present invention and are not intended to limit the scope of protection of the present invention. Without departing from the core concept of the present invention, those skilled in the art can make adaptive adjustments to the specific parameters according to different battery types, sensor arrangements, pulse measurement parameters, and computing platforms.

[0025] The core of this invention lies in: acquiring high-dimensional quantum signals during battery operation through NV color center pulse ODMR measurement; embedding battery thermodynamic state variables into the manifold modeling process of the high-dimensional ODMR signal; constructing a thermodynamically weighted adjacent edge distance and a physically constrained nearest neighbor graph; obtaining the thermodynamically weighted geodesic distance from the shortest path in the nearest neighbor graph and generating a physically constrained low-dimensional phase space; further establishing a cross-scale mapping relationship between local irreversible entropy yield and NV color center transverse relaxation time; and then, based on the distance from the current battery state to the thermal runaway safety boundary manifold surface, adjusting the NV color center pulse measurement parameters in a closed loop and outputting the battery health status and early warning results.

[0026] This embodiment illustrates the specific implementation of the present invention in the diagnostic early warning of single-cell lithium-ion batteries. The target battery can be a lithium iron phosphate battery, a ternary lithium battery, a lithium titanate battery, or other rechargeable electrochemical energy storage devices. The NV color center sensing unit can be disposed on the outside of the battery casing, at the encapsulation observation window, in the lateral thermal / magnetic coupling area of ​​the cell, or at a location that forms an effective thermal and magnetic field coupling with the area of ​​the battery to be tested.

[0027] I. Acquisition of NV color center pulse ODMR signal:

[0028] In this embodiment, a diamond NV center quantum sensing unit is used to detect local multiphysics field changes during the operation of the target battery. The NV center quantum sensing unit includes a diamond NV center sensing sheet, a laser excitation unit, a microwave driving unit, a fluorescence collection unit, and a photoelectric detection unit.

[0029] During measurement, the NV centers are first optically initialized using a green laser. Then, the spin states of the NV centers are excited by microwave pulses, and fluorescence signals are acquired. The pulse sequence can be a Hahn-Echo spin echo sequence or a Carr-Purcell-Meiboom-Gill multi-pulse sequence. For each measurement cycle, the ODMR spectral line shift, fluorescence intensity, linewidth, microwave frequency, number of pulses, pulse interval, coherence evolution time, and reference temperature and reference magnetic field signals are recorded.

[0030] In one specific implementation, the normalized fluorescence signals obtained at different coherent evolution times are subjected to exponential decay fitting, and the characteristic decay time obtained from the fitting is used as the transverse relaxation time of the NV color center. The same fitting model, the same data preprocessing rules, and the same fitting termination conditions are used for different measurement periods to ensure that the transverse relaxation times obtained from each measurement period are comparable.

[0031] Through the above acquisition process, a high-dimensional ODMR spectral state dataset is generated: ; in, For the first The high-dimensional state vector corresponding to each measurement cycle can be represented as: ; in, For the reason A high-dimensional ODMR spectral state dataset formed from several measurement cycles; For the first High-dimensional ODMR spectral state vectors corresponding to each measurement cycle; This is due to the frequency shift of the ODMR spectral lines; The fluorescence intensity is corrected for reference light intensity, reference temperature, and reference magnetic field. The linewidth of the ODMR spectrum; For coherent evolution time; This represents the lateral relaxation time.

[0032] Microwave frequency, pulse count, pulse interval, reference temperature, and reference magnetic field are recorded synchronously as measurement metadata, environmental compensation parameters, or closed-loop control parameters, but are not directly used as state dimensions for low-dimensional manifold modeling. Before forming high-dimensional state vectors from different measurement cycles, each feature is processed according to the same reference correction method and scale normalization rule.

[0033] II. Calculation of battery thermodynamic state quantities:

[0034] While acquiring the NV color center pulse ODMR signal, the battery terminal voltage, current, surface or internal temperature, charge / discharge rate, and electrochemical impedance parameters are simultaneously acquired. Based on these parameters, the thermodynamic state variables corresponding to each ODMR high-dimensional state point are calculated.

[0035] The electrochemical impedance parameters include at least ohmic internal resistance and charge transfer resistance. The ohmic heat generation rate is determined based on the battery current and ohmic internal resistance at the same sampling time; the polarization heat generation rate is determined based on the battery current and charge transfer resistance at the same sampling time; and the side reaction heat generation rate is determined based on the calibration relationship of side reaction heat generated by the same type of battery at different temperatures and charge / discharge rates. The polarization heat generation rate, ohmic heat generation rate, and side reaction heat generation rate are all converted to heat generation rate per unit volume and then substituted into the formula for calculating the local irreversible entropy yield.

[0036] First, calculate the first based on the open-circuit voltage. The change in Gibbs free energy corresponding to each state point: ; in, The number of electrons transferred in an electrochemical reaction. It is Faraday's constant. For the first The open-circuit voltage corresponding to each state point.

[0037] The Gibbs free energy gradient is obtained as follows: .

[0038] The Gibbs free energy gradient at the first state point is: .

[0039] No. The reaction entropy corresponding to each state point becomes: .

[0040] The local enthalpy change is: ; in, For the first The absolute temperature corresponding to each state point.

[0041] The local irreversible entropy production rate is: .

[0042] The cumulative irreversible entropy output is: ; in, For integration time variable, For the first The sampling time corresponding to each state point.

[0043] III. Construction of Thermodynamically Weighted Adjacent Edge Distance: In traditional manifold learning, adjacency relationships between high-dimensional data points are typically established only based on Euclidean distance or local geometric distance. In this embodiment, to avoid non-physical adjacency relationships caused by noise, temperature drift, or transient anomalies under complex operating conditions, a thermodynamically weighted adjacency edge distance is introduced, and the thermodynamically weighted geodesic distance is obtained by physically constraining the shortest path in the nearest neighbor graph.

[0044] Thermodynamic penalty weight The values ​​are determined by a two-dimensional fuzzy controller. The two-dimensional fuzzy controller takes the absolute rate of change of battery temperature and the absolute value of charge / discharge rate of the current measurement cycle as input, performs fuzzy inference according to a pre-calibrated fuzzy rule matrix, and uses the centroid method to defuzzify, obtaining non-negative thermodynamic penalty weights. .

[0045] No. The state point and the th state point The thermodynamically weighted adjacent edge distance between each state point is: ; Wherein, the normalized geometric distance is: ; Thermodynamic penalty term is: ; in: ; ; ; ; in, , , , and All are non-negative weight coefficients, and satisfy the following: ; , , , and All are scale normalization factors greater than zero.

[0046] Figure 2 The "joint physical distance metric" shown is the thermodynamically weighted adjacent edge distance described in this embodiment. .

[0047] ; ; ; ; in, For the first Thermodynamic state vector of a state point; This is the local thermodynamic driving direction vector determined based on the thermodynamic state vectors of adjacent measurement cycles; the local thermodynamic driving direction vector of the first state point is adopted... Sure; To prevent positive numbers with a denominator of zero.

[0048] When the direction of state change corresponding to a candidate adjacent edge is opposite to the direction of local thermodynamic driving... This increases the edge weight of the candidate adjacent edge; when the direction of state change corresponding to the candidate adjacent edge is consistent with the direction of local thermodynamic driving, Or it may approach 0.

[0049] IV. Dimensionality Reduction of Physically Constrained Manifolds: Based on thermodynamic weighted adjacent edge distance Construct a physically constrained nearest neighbor graph. The nodes in the graph are high-dimensional ODMR state points, and the edge weights of candidate adjacent edges are... After the nearest neighbor graph is constructed, the shortest path between any two state points in the graph is calculated, and the length of the shortest path is recorded as the thermodynamically weighted geodesic distance. The low-dimensional physical phase space coordinate set is solved using a thermodynamically constrained metric mapping algorithm. , ; ; in, This is the estimated value of the normalized Gibbs free energy; This is the normalized cumulative irreversible entropy yield estimate; and These are the normalized mappings for the corresponding thermodynamic state quantities; and is the regularization coefficient.

[0050] V. Cross-scale mapping of entropy production rate and transverse relaxation time: After the low-dimensional physical phase space is constructed, a local irreversible entropy yield is established. Lateral relaxation time of NV color center The cross-scale dynamic mapping relationship between them: ; Where km is the cross-scale mapping coefficient. This is a non-linear threshold activation function.

[0051] In this embodiment, the nonlinear threshold activation function is the Sigmoid function: ; Where γ is the activation kurtosis factor. This is the critical entropy yield threshold.

[0052] The discrete form of this mapping relationship is: ; The measured transverse relaxation time at the first calibration moment is used as the initial value for the discrete mapping. .

[0053] When irreversible side reactions within the battery intensify, local heat generation increases, or microstructural distortion worsens, the local irreversible entropy yield rises, and magnetic field fluctuations, temperature fluctuations, or stress disturbances around the NV color center intensify, leading to a decrease in transverse relaxation time. Shortening. Through the above cross-scale mapping relationship, macroscopic thermodynamic decay processes can be linked to microscopic quantum decoherent responses.

[0054] VI. Safety Boundary Distance Calculation and Early Warning Output: Before being used to form the thermal runaway safety boundary manifold, the normal operation samples and historical failure samples are physically consistent with each other using the cross-scale dynamic mapping relationship between the local irreversible entropy production rate and the transverse relaxation time determined in step S5, so that the boundary samples simultaneously satisfy the decoherent evolution trend of the NV color center and the irreversible entropy production evolution trend of the battery.

[0055] In a low-dimensional physical phase space, the thermal runaway safety boundary manifold is determined based on normal samples, historical failure samples, and thermodynamic stability constraints. Let the thermal runaway safety boundary discrimination function be... The current state coordinates are Then the distance from the current state to the safety boundary is: .

[0056] The sign direction of the thermal runaway safety boundary discrimination function is defined as follows: This indicates that the state point is located on the safe side. This indicates that the state point lies on the boundary manifold. This indicates that the state point has crossed the thermal runaway safety boundary. The safety threshold, warning threshold, and danger threshold must satisfy: ; Based on boundary distance Based on the relationship with preset thresholds, the battery status is divided into normal state, attention state, warning state, and danger state.

[0057] Specifically: when When this occurs, it is considered a normal state; when When this occurs, it is determined to be a state of attention; when When this occurs, it is determined to be a warning state; when When this occurs, it is determined to be a dangerous situation.

[0058] The output includes battery health status, safety level, boundary distance, entropy yield change trend, lateral relaxation time change trend, and early warning information.

[0059] This embodiment illustrates how the present invention adaptively adjusts the NV color center pulse measurement parameters based on the safety boundary distance under high-rate charge / discharge or early-stage thermal runaway risk conditions.

[0060] Under high-rate charge and discharge conditions, the generation rates of polarization heat, ohmic heat, and side reaction heat within the battery may all increase, leading to an increase in local irreversible entropy production. Traditional fixed-parameter ODMR measurements are prone to two problems: first, prolonged use of high-sensitivity measurements during normal operation increases energy consumption and computational burden; second, using low-sensitivity scanning when approaching dangerous boundaries may fail to capture weak anomalies in a timely manner. Therefore, this embodiment adopts a boundary distance-driven pulse resource reallocation strategy.

[0061] I. Construction of safety boundary distance error: First, obtain the current battery state coordinates according to the method in Example 1. And calculate the current boundary distance. Then, the boundary distance error is defined as: ; in, For the first The directed boundary distance for each measurement cycle.

[0062] when At that time, the current state is within the normal range. The system maintains a low-power baseline scan; when hour, The system will enter enhanced scanning, high-sensitivity scanning, or emergency warning mode based on the distance to the boundary.

[0063] To avoid the infinite accumulation of historical errors leading to integral saturation, a limit integration state is defined: ; When the directed boundary distance exceeds the safety threshold again At that time, boundary distance error and integral state All are reset to zero.

[0064] II. Adaptive adjustment of pulse quantity: Pulse count This directly affects the coherent accumulation effect and the signal-to-noise ratio. In this embodiment, the number of pulses is updated as follows: ; in, This indicates that the number of pulses is a positive integer. The reference pulse number, This is the proportional control coefficient. The integral control coefficient, and These represent the lower and upper limits for the number of pulses, respectively.

[0065] When the boundary distance decreases rapidly As the distance increases, the number of pulses also increases to enhance the cumulative detection capability of weak abnormal signals; when the directed boundary distance recovers to... At that time, boundary distance error and limit integral state Simultaneously, the pulse count is reset to zero and restored to the reference pulse configuration. .

[0066] III. Adaptive adjustment of pulse interval: Pulse interval This affects the sensitivity of the NV color center to noise of different frequencies and local multiphysics fluctuations. In this embodiment, the pulse interval is updated as follows: ; in, This is the pulse interval adjustment coefficient. and These are the lower and upper limits of the pulse interval, respectively.

[0067] When the battery condition gradually approaches the warning boundary, the system shortens the pulse interval to improve its response to rapidly changing noise and local abnormal fluctuations; when the battery is in the safe range, it maintains a larger pulse interval to reduce the measurement burden.

[0068] IV. Adaptive adjustment of frequency scanning range: Frequency scanning range This is used to cover potential shifts in ODMR spectral lines due to temperature, magnetic field, and stress perturbations. In this embodiment, the frequency scan range is updated as follows: ; in, This is the frequency scan adjustment coefficient. and These are the lower and upper limits of the frequency scanning range, respectively.

[0069] When the battery is in the safe zone, the system uses a narrower scanning range to improve measurement efficiency; when the battery enters the warning zone, the scanning range is increased to avoid the loss of key features due to rapid shift of ODMR spectral lines.

[0070] V. Measurement Mode Switching: In this embodiment, based on boundary distance The size can be set to four measurement modes.

[0071] When the following conditions are met: ; The system enters a low-power reference scan mode, using a smaller number of pulses, a larger pulse interval, and a reference frequency scan range.

[0072] When the following conditions are met: ; The system enters enhanced scanning mode, appropriately increasing the number of pulses and expanding the scanning range.

[0073] When the following conditions are met: ; The system enters a high-sensitivity scanning mode, significantly increasing the number of pulses, shortening the pulse interval, and expanding the frequency scanning range.

[0074] When the following conditions are met: ;

[0075] The system enters emergency warning mode, outputs a danger status warning, and outputs the danger level and disposal suggestions such as current limiting, power reduction, load disconnection, or charging cessation to the external battery management system.

[0076] It should be noted that, Figure 4 The "alarm threshold distance" shown refers to the activation threshold of the closed-loop pulse parameter adjustment, which corresponds to the safety threshold in this embodiment. ; Figure 4 The "Danger Approaching" branch includes the warning state and the danger state described in this embodiment. The warning state corresponds to the high-sensitivity scanning mode, and the danger state corresponds to the emergency warning mode. Figure 4 In normal conditions, "wideband" refers to the bandwidth of the reference scan relative to single-frequency point monitoring, and does not mean that its frequency scanning range is greater than that of enhanced scan mode or high-sensitivity scan mode.

[0077] VI. Technical Effects of this Embodiment:

[0078] Through the closed-loop adaptive pulse modulation method in this embodiment, the NV color center measurement system no longer passively acquires signals with fixed parameters, but dynamically allocates measurement resources according to the distance from the current state of the battery to the thermal runaway safety boundary. This reduces system energy consumption and computational burden under normal conditions, and improves detection sensitivity and response speed under abnormal conditions, thereby enhancing the early warning capability of battery thermal runaway.

[0079] This embodiment illustrates a system-level implementation of the present invention in a battery module or battery pack. This system can be used in conjunction with a battery management system, or deployed as a standalone quantum sensing diagnostic unit in energy storage systems, new energy vehicle battery packs, or high-reliability power systems.

[0080] The system in this embodiment includes an NV color center quantum sensing acquisition module, a battery status synchronous acquisition module, a processor module, a pulse modulation module, and an early warning output module.

[0081] The NV color center quantum sensing acquisition module is used to perform pulsed ODMR measurements and obtain high-dimensional quantum signals; the battery state synchronous acquisition module is used to obtain battery terminal voltage, current, temperature, charge / discharge rate, and electrochemical impedance parameters; the processor module is used to perform thermodynamic state quantity calculations, thermodynamic weighted adjacent edge distance construction, physical constraint nearest neighbor graph shortest path calculation, manifold dimensionality reduction, cross-scale dynamic mapping, safety boundary distance calculation, and early warning determination; the pulse modulation module is used to adjust the number of pulses, pulse interval, and frequency scanning range according to the closed-loop modulation instructions output by the processor module; and the early warning output module is used to output battery health status, safety level, boundary distance, and early warning information.

[0082] Historical samples, calibration parameters, ODMR quantum signals, and operation records are managed by storage units configured in the processor module. These storage units are part of the processor module's data management function and are not independent, necessary system modules.

[0083] II. Multi-sensor module-level early warning: When the battery module is configured with multiple NV color center sensing units, each NV color center sensing unit corresponds to a local test area. The processor module obtains the corresponding directed boundary distance for each local test area using the same method described in Example 1.

[0084] Each local area to be measured uses the same feature scale normalization rule, boundary discrimination function sign direction, and boundary distance scale to ensure comparability between directed boundary distances of different sensing areas.

[0085] Let the first The directed boundary distance of each local region to be measured is The number of sensing areas is Then the module-level boundary distance is: ;

[0086] The processor module identifies the local test area with the smallest boundary distance as the risk-priority area and applies it according to the same set of safety thresholds. Warning threshold and danger threshold The module-level states are divided. Module-level diagnostics do not establish a second set of low-dimensional state vectors or a second set of safety boundary models.

[0087] When the module-level boundary distance is greater than the safety threshold, each sensing area maintains the baseline scanning mode; when the module-level boundary distance enters the attention state or warning state range, the measurement sensitivity of the risk priority area is increased first; when the module-level boundary distance is less than or equal to the danger threshold, the warning output module outputs the danger level and corresponding handling suggestions to the external battery management system.

[0088] III. Parameter Calibration and Offline Update:

[0089] In order to adapt to different battery material systems and different battery module structures, this embodiment performs offline calibration before the system is put into use.

[0090] First, samples of the same type of battery were selected, and pulsed ODMR signals, electrochemical impedance data, current data, and temperature data were collected under different health conditions. Then, the least squares regression method was used to determine the cross-scale mapping coefficients. Activate the kurtosis factor and critical entropy yield threshold Its target is: ; in, To determine the number of calibrated samples; The measured transverse relaxation time corresponds to the l-th calibration sample; To predict the lateral relaxation time.

[0091] New data recorded during system operation is used to form the subsequent offline review dataset. Updates to cross-scale mapping parameters, scale normalization factors, thermodynamic weighting coefficients, and safety boundary discrimination functions are written into the next running cycle after offline data review and model validation are completed. They are not directly modified within the current warning cycle, thereby avoiding the impact of unverified online parameter drift on the current warning results.

[0092] Through the system-level deployment method of this embodiment, the present invention can not only diagnose and warn of individual battery cells, but also identify local risk areas in battery modules. The system can adaptively adjust measurement resources according to the safety boundary distance of different sensing areas, prioritizing the allocation of high-sensitivity measurements to risk areas, thereby improving the early warning capability of module-level thermal runaway and reducing unnecessary measurement energy consumption in normal areas.

[0093] It should be noted that the above embodiments are only used to further illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although this specification has described the present invention in detail in conjunction with single-cell batteries, high-rate charge and discharge conditions, and battery module-level application scenarios, those skilled in the art should understand that, without departing from the core concept of the present invention, the pulse ODMR acquisition parameters, thermodynamic state quantity calculation methods, manifold dimensionality reduction algorithm parameters, safety boundary thresholds, and closed-loop modulation control parameters can be adaptively adjusted according to different battery material systems, different packaging structures, different NV color center sensor arrangements, and different computing platforms.

[0094] For example, the target battery is not limited to lithium iron phosphate batteries, ternary lithium batteries, or lithium titanate batteries, but can also be applied to sodium-ion batteries, solid-state batteries, lithium-sulfur batteries, or other electrochemical energy storage devices with thermal runaway risks; the NV color center sensing unit can be located inside the battery, or on the outside of the battery casing, in the encapsulation observation window, in the thermally sensitive area of ​​the module, or at a location that forms an effective thermal or magnetic field coupling with the area to be measured; in addition to the isometric mapping algorithm, the manifold dimensionality reduction method can also employ other nonlinear dimensionality reduction algorithms that can preserve the local topology and introduce physical constraint terms; in addition to the support vector machine boundary, the safety boundary manifold surface can also be determined based on the probability boundary model, kernel density estimation model, thermal stability discrimination model, or a combination thereof.

[0095] It should also be noted that the thermodynamically weighted adjacent edge distance, the thermodynamically weighted geodesic distance obtained from the shortest path of the physically constrained nearest neighbor graph, the local irreversible entropy yield, and the transverse relaxation time mentioned in this invention are all related to this invention. The cross-scale dynamic mapping relationship and the closed-loop pulse modulation strategy based on the directed boundary distance can be used for offline parameter calibration and verification updates according to the actual battery type, test environment and safety level requirements.

[0096] Therefore, any equivalent substitutions, simple modifications, combinations, adjustments, or adaptive improvements made to the relevant structures, steps, algorithm parameters, sensor arrangements, or data processing flows based on the technical concept disclosed in this invention should fall within the protection scope of this invention.

Claims

1. A method for early warning of battery thermal runaway based on NV color center entropy generation manifold calibration, characterized in that, Includes the following steps: S1. Obtain the high-dimensional quantum signal of NV color center pulse ODMR during the operation of the target battery to form a high-dimensional spectral state dataset. This includes at least the fitting characteristics of ODMR spectral line shift, fluorescence intensity, linewidth, coherent evolution time, and transverse relaxation time; S2. Simultaneously acquire the voltage, current, temperature, charge / discharge rate, and electrochemical impedance parameters of the target battery, and calculate the thermodynamic state quantities corresponding to the high-dimensional spectral state data. These thermodynamic state quantities include at least the Gibbs free energy. Gibbs free energy gradient Local reaction enthalpy change Local irreversible entropy production rate And the cumulative irreversible entropy output obtained by integrating the local irreversible entropy output rate. ; S3. Using the high-dimensional spectral state dataset as a node set, construct the thermodynamically weighted adjacency distance; S4, Based on thermodynamically weighted adjacent edge distance Construct a physically constrained nearest neighbor graph, calculate the shortest path between any two state points in the graph, and obtain the thermodynamically weighted geodesic distance. Based on the thermodynamically weighted geodesic distance, manifold dimensionality reduction is performed to generate low-dimensional physical phase space coordinates, so that the low-dimensional physical phase space coordinates simultaneously satisfy the ODMR spectral neighborhood preservation condition and the battery irreversible thermodynamic evolution consistency condition. S5. Establish local irreversible entropy production rate Lateral relaxation time of NV color center The cross-scale dynamic mapping relationship between them; S6. Using the cross-scale dynamic mapping relationship between the local irreversible entropy yield and the transverse relaxation time determined in step S5, perform physical consistency calibration on the normal operation samples and historical fault samples so that the boundary samples simultaneously satisfy the NV color center decoherence evolution trend and the battery irreversible entropy production evolution trend. Based on physically consistent calibrated normal operation samples, historical failure samples, and thermodynamic stability constraints, the thermal runaway safety boundary manifold is determined in a low-dimensional physical phase space, and the current battery state coordinates are calculated. Boundary distance to the thermal runaway safety boundary manifold And based on the boundary distance Determine the battery safety level; S7. Based on the boundary distance In accordance with battery safety level, the NV color center pulse measurement parameters are adjusted in a closed loop. The pulse measurement parameters include at least the number of pulses, pulse interval, and frequency scanning range, and the battery health status and warning results are output.

2. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 1, characterized in that, The local irreversible entropy yield in step S2 is obtained as follows: ； in, For the first Local irreversible entropy production rate corresponding to each state point; , and The locations within the area to be measured are respectively The volumetric heat generation rate of polarization heat, ohmic heat and side reaction heat at the point; This refers to the local absolute temperature. This is the area of ​​the battery to be tested. The cumulative irreversible entropy yield is obtained as follows: ； in, For the first The sampling time corresponding to each state point For integration time; No. The change in Gibbs free energy corresponding to each state point is obtained as follows: ； in, The number of electrons transferred in an electrochemical reaction. It is Faraday's constant. For the first The open-circuit voltage corresponding to each state point; The Gibbs free energy gradient is obtained as follows: ； The Gibbs free energy gradient of the first state point is determined by the forward difference between the first and second state points. No. The reaction entropy change and local reaction enthalpy change corresponding to each state point are as follows: ； ； in, For the first The absolute temperature corresponding to each state point For the first The reaction entropy change corresponding to each state point For the first The local enthalpy change of the reaction corresponding to each state point.

3. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 1, characterized in that, Thermodynamic penalty term in step S3 It consists of a thermodynamic state difference term and a thermodynamic evolution direction penalty term, and its expression is: ； ； ； ； ； in, , , , and These are the non-negative weighting coefficients for the Gibbs free energy gradient difference term, the local reaction enthalpy difference term, the local irreversible entropy yield difference term, the cumulative irreversible entropy yield difference term, and the thermodynamic evolution direction penalty term, respectively, and they satisfy the following: ； , , and These are the scale normalization factors for the corresponding thermodynamic state quantities, and both are greater than zero; This is a penalty term for the direction of thermodynamic evolution.

4. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 1, characterized in that, The thermodynamically weighted adjacent edge distance is: ； in, For the first The ODMR high-dimensional spectral state point and the ... Thermodynamic weighted adjacency distance between ODMR high-dimensional spectral state points; The normalized geometric distance between two ODMR high-dimensional spectral state points; The geometric distance weights are non-negative. The non-negative thermodynamic penalty weight is dynamically updated according to battery operating conditions; This is a thermodynamic penalty term; ； in, and The first The and the first A high-dimensional spectral state vector of ODMR , is the scale normalization factor for the ODMR high-dimensional spectral state dataset; Thermodynamic evolution direction penalty term Obtain it in the following way: ； in, For the first Thermodynamic state vectors at state points, and: ； For the first Thermodynamic state vectors at state points, and: ； ； in For the first Thermodynamic state vector of a state point; The thermodynamic driving direction vector is determined based on the thermodynamic state vectors of adjacent measurement periods; the local thermodynamics of the first state point. Sure; To prevent positive numbers with a denominator of zero; When the direction of state change corresponding to a candidate adjacent edge is opposite to the direction of local thermodynamic driving... This increases the edge weight of the candidate adjacent edge; when the state change direction corresponding to the candidate adjacent edge is aligned with the local thermodynamic driving direction... Or it may approach 0.

5. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 1, characterized in that, The manifold dimensionality reduction in step S4 employs a thermodynamically constrained metric mapping algorithm, with the objective function being: ， ； ； in, It is a set of low-dimensional physical phase space coordinates; For the first The low-dimensional physical phase space coordinates corresponding to each state point; The shortest path geodesic distance is calculated based on the physical constraint nearest neighbor graph. This is the normalized Gibbs free energy estimate obtained from reconstruction based on adjacent samples in the low-dimensional physical phase space; This is the normalized cumulative irreversible entropy yield estimate obtained from reconstruction based on adjacent samples in the low-dimensional physical phase space; and These are the normalized mappings of Gibbs free energy and cumulative irreversible entropy output, respectively; and These are the regularization coefficients for the Gibbs free energy consistency constraint and the cumulative irreversible entropy production consistency constraint, respectively.

6. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 1, characterized in that, The cross-scale dynamic mapping relationship in step S5 is as follows: ； in, These are cross-scale mapping coefficients related to the battery material system. , is a threshold activation function used to characterize the nonlinear enhancement of the decoherent response of the NV color center when the local irreversible entropy yield crosses a critical threshold; The threshold activation function is the Sigmoid threshold activation function, and its expression is: ； in, To activate the kurtosis factor, The critical entropy yield threshold; The discrete form of the cross-scale dynamic mapping relationship is: ； in, and The first The and the first Transverse relaxation time for each measurement cycle; For the first Local irreversible entropy production rate for each measurement cycle; To measure the update cycle, the measured transverse relaxation time at the first calibration time is used as the initial value of the discrete mapping. .

7. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 6, characterized in that, Cross-scale mapping coefficients Activate the kurtosis factor and critical entropy yield threshold The results were obtained through joint calibration, which included the following steps: selecting at least five battery samples of the same type in different health states; performing spin echo attenuation measurement on each sample and extracting the transverse relaxation time. Simultaneously, the local irreversible entropy yield is calculated using electrochemical impedance data, current data, and temperature data. ; lateral relaxation time With the aforementioned local irreversible entropy production rate Data pairs were formed, and the least squares regression method was used to determine the cross-scale mapping coefficients. Activate the kurtosis factor and critical entropy yield threshold Its target is: ； in, To determine the number of calibrated samples; The measured transverse relaxation time corresponding to the l-th calibration sample; To predict the lateral relaxation time.

8. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 1, characterized in that, The thermal runaway safety boundary manifold surface in step S6 is jointly determined by historical fault samples, normal operation samples, and thermodynamic stability constraints, which have undergone physical consistency calibration via the cross-scale dynamic mapping relationship in step S5; the boundary distance is calculated using the directed projection distance of the local tangent plane of the manifold, and its expression is: ； in, This is the thermal runaway safety boundary discrimination function. These are the low-dimensional physical phase space coordinates of the current battery state. This represents the gradient of the thermal runaway safety boundary discrimination function at the current battery state coordinates. To prevent positive numbers with a denominator of zero; The sign direction of the thermal runaway safety boundary discrimination function is uniformly defined as follows: This indicates that the state point is located on the safe side of the thermal runaway safety boundary. This indicates that the state point is located on the thermal runaway safety boundary manifold. This indicates that the state point has crossed the thermal runaway safety boundary and entered the danger side; Safety threshold Warning threshold and danger threshold satisfy: ； when When this occurs, it is considered a normal state; when When this occurs, it is determined to be a state of attention; when When this occurs, it is determined to be a warning state; when When this occurs, it is determined to be a dangerous situation.

9. The battery thermal runaway early warning method based on NV color center entropy generation manifold calibration according to claim 8, characterized in that, The closed-loop regulation in step S7 adopts a pulse resource reallocation control law based on directed boundary distance error. The boundary distance error for each measurement cycle is: ； in, For the first The directed boundary distance for each measurement cycle; The limit integral state of the boundary distance error is: ； The pulse count is updated as follows: ； The pulse interval is updated as follows: ； The frequency scan range is updated as follows: ； in: ； This indicates that the number of pulses will be a positive integer; , and These are the number of reference pulses, the interval between reference pulses, and the range of reference frequency scanning, respectively. , , and These are non-negative control parameters; This represents the upper limit of the integration state. and These are the lower and upper limits for the number of pulses, respectively. and These are the lower and upper limits of the pulse interval, respectively. and These are the lower and upper limits of the frequency scan range, respectively. when At this time, both the boundary distance error and the integral state are zero, maintaining the low-power reference scan mode; when At that time, enter enhanced scanning mode; when At this time, it enters high-sensitivity scanning mode; when At that time, the emergency warning mode will be activated.

10. A battery thermal runaway early warning system based on NV color center entropy generation manifold calibration, characterized in that, It includes an NV color center quantum sensing acquisition module, a battery status synchronous acquisition module, a pulse modulation module, and an early warning output module; The NV color center quantum sensing acquisition module is used to perform pulsed ODMR measurements on the target battery and acquire high-dimensional quantum signals. The battery status synchronous acquisition module is used to synchronously acquire the voltage, current, temperature, charge / discharge rate and electrochemical impedance parameters of the target battery. The processor module is used to complete the construction of thermodynamically weighted adjacent edge distances, the calculation of geodesic distances of the shortest path in the physically constrained nearest neighbor graph, the dimensionality reduction of the physically constrained manifold, the cross-scale mapping of entropy production rate and lateral relaxation time, the calculation of safety boundary distances, and closed-loop pulse modulation. The pulse modulation module is used to adjust the number of pulses, the pulse interval, and the frequency scanning range according to the closed-loop modulation command output by the processor module. The warning output module is used to output the target battery's health status, safety level, boundary distance, and warning information; When the NV color center quantum sensing acquisition module includes multiple NV color center sensing units, the processor module calculates the directed boundary distance of each sensing region and determines the minimum value among the directed boundary distances of each sensing region as the module-level boundary distance.

Images