Sodium-ion battery pack thermal runaway monitoring method based on multi-source parameter fusion

Abstract

The application discloses a sodium ion battery pack thermal runaway monitoring method based on multi-source parameter fusion, and relates to the technical field of battery pack safety monitoring, comprising the following steps: heterogeneous metal contacts are arranged on a parallel busbar conductor of a battery pack in the current direction at intervals and are connected in series to form a thermocouple array, a thermoelectric voltage sequence output by the thermocouple array is collected, and a branch level characteristic quantity corresponding to a parallel branch current distribution change is calculated according to the thermoelectric voltage sequence; on the basis of the branch level characteristic quantity, a battery management system is controlled to execute a preset pulse sequence, and pulse control operation is performed on a target monomer or a target branch. The application forms a thermocouple array on a parallel busbar and collects a thermoelectric voltage sequence, constructs a branch level characteristic quantity reflecting a parallel branch current distribution change, and obtains branch information and electrochemical evolution information of a finer granularity under the condition that a group level parallel average effect exists, thereby providing a multi-dimensional observation basis for early abnormal identification.

Description

Technical Field

[0001] This invention relates to the field of battery pack safety monitoring technology, specifically to a method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion. Background Technology

[0002] With the development of energy storage systems and electrification equipment, battery packs operate under conditions of high energy density, complex operating conditions and high integration. Safety monitoring and early warning have become an important part of battery management. For the risk of thermal runaway of battery packs, engineering usually requires identifying abnormal evolution trends as early as possible without significantly increasing system complexity, and implementing executable early warning output at the battery pack level. Existing thermal runaway monitoring solutions mostly revolve around electrical parameters such as temperature, voltage, and current, and combine them with information such as gas concentration and pressure for auxiliary judgment. At the same time, there are also methods that use pulse measurement to extract internal resistance or polarization-related features in order to obtain earlier abnormal clues. These methods have good applicability under common operating conditions, but under conditions such as parallel battery pack topology, non-co-located sensors, gas migration lag, and weak signal excitation, multi-source parameters may exhibit inconsistencies in time scale, spatial diffusion, and characteristic stages, thereby increasing the difficulty of early identification and localization at the pack level. To enhance thermal runaway monitoring capabilities at the battery pack level, the industry is exploring multi-parameter joint and structured modeling approaches. For example, this involves joint analysis using information such as busbar temperature difference, electrochemical response characteristics, and gas source strength estimation, incorporating electrical connections, thermal coupling, and gas migration relationships into a unified framework to better reflect the actual propagation paths and coupling mechanisms of battery packs. Against this backdrop, a monitoring method is needed that can collaboratively utilize electrical, thermal, and gas parameters at the battery pack scale, while also considering temporal accumulation and spatial correlation, to obtain more stable early warning criteria and clearer anomaly characterization. Therefore, this invention proposes a sodium-ion battery pack thermal runaway monitoring method based on multi-source parameter fusion. Summary of the Invention

[0003] The purpose of this invention is to provide a method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion, so as to solve the problems mentioned in the background art.

[0004] This invention can be achieved through the following technical solution: a method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion, comprising: Step 1: Set dissimilar metal contacts at intervals along the current direction on the parallel bus conductor of the battery pack and connect them in series to form a thermocouple array. Collect the temperature difference voltage sequence output by the thermocouple array and calculate the branch-level characteristic quantity corresponding to the change in the current distribution of the parallel branch based on the temperature difference voltage sequence. Step 2: Based on the branch-level characteristic quantities, control the battery management system to execute a preset pulse sequence to perform pulse control operation on the target cell or target branch, and simultaneously collect the cell voltage response, temperature difference voltage response and module temperature response. Analyze the transient changes in cell voltage and extract the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component and diffusion component. Step 3: Collect gas concentration data, temperature data and ventilation parameters at multiple points within the module, obtain the diffusion coefficient, convection parameters and adsorption parameters corresponding to the gas migration process, and perform partitioned calculations on the gas concentration data according to the diffusion, convection and adsorption changes of the gas inside the module to determine the unit-level gas generation intensity corresponding to each unit, and perform time matching between the unit-level gas generation intensity and the electrochemical response parameters. Step 4: Map the branch-level characteristic quantities, electrochemical response parameters, and cell-level gas generation intensity to the battery pack topology diagram. Calculate the correlation weights between nodes based on electrical connections, thermal coupling relationships, and gas migration processes to obtain the cell-level anomaly probability. Step 5: Accumulate the individual-level anomaly probability based on the first preset sampling period and the second preset sampling period to obtain the cumulative risk value. When the cumulative risk value exceeds the preset risk judgment threshold, output a thermal runaway early warning signal.

[0005] A further technical improvement of the present invention is that: the thermocouple array includes a first thermocouple array and a second thermocouple array respectively disposed on opposite sides of the parallel bus conductor, and the first thermocouple array and the second thermocouple array are divided into multiple one-to-one corresponding segments along the length direction of the bus. The difference between the first temperature difference voltage sequence and the second temperature difference voltage sequence of the corresponding segment is calculated to obtain the segment differential temperature difference voltage sequence. The temperature difference bias direction is determined according to the sign of the average value of the segment differential temperature difference voltage sequence in each segment. The gradient change is determined based on the first-order difference between adjacent segments of the differential temperature and voltage sequence, and the segment with the largest absolute value of the gradient change is identified as the extreme value segment. The segment differential temperature voltage value corresponding to the extreme value segment is used as a scalar weight, and the sign of the scalar weight is corrected according to the temperature difference bias direction. The sign-corrected scalar weight is multiplied with the temperature difference voltage sequence to obtain the corrected temperature difference voltage sequence, and the branch-level characteristic quantity is calculated based on the corrected temperature difference voltage sequence.

[0006] A further technical improvement of the present invention is that, in the process of determining the gradient change and the extreme value segment based on the first-order difference between adjacent segments of the segmented differential temperature-voltage sequence, the method includes: Centered on the extreme value segment, a preset number of adjacent segments are selected on both sides along the length of the busbar, and the gradient change corresponding to the adjacent segments is extracted. Calculate the attenuation trend of gradient change as segment distance increases, and determine the attenuation coefficient of gradient change in space; The attenuation coefficient is compared with the preset single-source diffusion attenuation threshold. When the attenuation coefficient meets the characteristics of rapid attenuation on one side, the extreme value segment is confirmed as an abnormal segment. When the attenuation coefficient shows the characteristics of gradual attenuation on both sides, the extreme value segment is determined to be an environmental disturbance segment. Only the segment differential temperature and voltage values ​​of the confirmed abnormal segments are extracted as scalar weights to participate in the correction calculation of the subsequent temperature and voltage sequence.

[0007] A further technical improvement of the present invention is that: the preset pulse sequence includes different combinations of time widths and is used to separate and extract the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component, including: The preset pulse sequence is constructed into a first pulse segment, a second pulse segment, and a third pulse segment arranged progressively according to the time width; During the execution of the first pulse segment, the second pulse segment, and the third pulse segment, the individual voltage response is time-aligned and the instantaneous change parameters corresponding to the first pulse segment, the establishment change parameters corresponding to the second pulse segment, and the recovery change parameters corresponding to the third pulse segment are extracted respectively. The instantaneous change parameters, the established change parameters, and the recovered change parameters are respectively mapped to the electrochemical response parameters corresponding to the ohmic internal resistance component, the charge transfer component, and the diffusion component according to the preset mapping relationship. The time width of the next cycle for the first, second, and third pulse segments is constrained and updated based on the variation amplitude of the module temperature response and thermoelectric voltage response within each pulse segment, so that different combinations of time widths can maintain the distinguishability of different electrochemical response parameters without changing the order of the pulse segments.

[0008] A further technical improvement of the present invention is that the time width of the first pulse segment is limited to make the initial voltage drop of the single-cell voltage response exhibit abrupt changes within a preset sampling resolution; The time width of the second pulse segment is limited to ensure that the individual cell voltage response exhibits a distinguishable gradual change during the pulse hold period; The time width of the third pulse segment is limited to ensure that the individual cell voltage response exhibits a discernible hysteresis characteristic during the recovery process after the pulse ends.

[0009] A further technical improvement of the present invention is that the temperature difference voltage response participates in the determination of electrochemical response parameters, including: While executing the preset pulse sequence, the individual voltage response and the temperature difference voltage response are simultaneously acquired, and the individual voltage response and the temperature difference voltage response are time-aligned according to the same sampling time scale. Calculate the amplitude of the change in the individual unit voltage response within the pulse segment and the amplitude of the change in the temperature difference voltage response within the same pulse segment, and calculate the consistency of the sign of the change direction of the individual unit voltage response and the change direction of the temperature difference voltage response. When the direction of change of the monomer voltage response is consistent with the direction of change of the thermoelectric voltage response, the proportionality coefficient of the two changes is calculated and compared with the preset proportional range. When the proportionality coefficient is within the preset proportional range, the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component and diffusion component extracted from the monomer voltage response are confirmed to be effective. When the proportional coefficient exceeds the preset proportional range or the direction of change is inconsistent, the electrochemical response parameters extracted from the monomer voltage response are corrected according to the change amplitude of the temperature difference voltage response.

[0010] A further technical improvement of the present invention is that: in step three, during the process of dividing and calculating the multi-point gas concentration data according to the diffusion, convection and adsorption change law of gas inside the module to determine the gas generation intensity at the monomer level, the module space is divided into multiple initial calculation units corresponding one-to-one with the monomer according to the physical location of the monomer. Centered on the physical location of each monomer, the diffusion distance of the gas within each time step is calculated based on the diffusion coefficient and time step, the main direction of gas migration is determined based on the convection parameters, and the effective concentration of the gas at the boundary of the corresponding calculation unit is corrected based on the adsorption parameters. Within a continuous time step, the spatial boundaries of each computational unit are updated based on the diffusion distance and the main migration direction. When the spatial boundaries of adjacent computing units overlap, the direction and magnitude of the gas concentration gradient within the overlapping region are calculated. The gradient component that is consistent with the direction of the unit location is allocated to the gas generation intensity calculation of the corresponding unit according to the gradient magnitude ratio, and the gradient component that is opposite to the direction of the unit location is deducted from the gas generation intensity calculation of the corresponding unit, thereby completing the update calculation of the gas generation intensity at the unit level.

[0011] A further technical improvement of the present invention is that: in step four, during the process of calculating the association weights between nodes based on electrical connection relationships, thermal coupling relationships, and gas migration processes, the association weights between nodes are divided into electrical propagation weights, thermal propagation weights, and gas propagation weights. For the electric propagation weight, the potential attenuation coefficient is calculated based on the equivalent resistance between adjacent nodes, and the reciprocal of the potential attenuation coefficient is used as the electric propagation weight. For the heat propagation weight, the heat flux transfer coefficient is calculated based on the thermal resistance between adjacent nodes, and the heat propagation weight is obtained by multiplying the heat flux transfer coefficient by the reciprocal of the square of the spatial distance between individual nodes. For the gas propagation weight, the direction correction coefficient is calculated based on the angle relationship between the distance between individual units and the gas migration direction, and the gas propagation weight is obtained by multiplying the direction correction coefficient by the distance exponential decay factor. The electrical propagation weight, thermal propagation weight, and gas propagation weight are synthesized according to a preset normalization rule to obtain the correlation weight between nodes.

[0012] A further technical improvement of the present invention is that step five, obtaining the cumulative risk value, includes: The difference between the cumulative result based on the first preset number of sampling periods and the cumulative result based on the second preset number of sampling periods is calculated respectively, and the rate of change of the difference is obtained by dividing the change of the difference within multiple consecutive sampling periods by the corresponding number of sampling periods. The rate of change of the difference is compared with a preset rate of change of the difference threshold. When the rate of change of the difference exceeds the preset rate of change of the difference threshold, the preset risk judgment threshold is reduced once by using the ratio of the rate of change of the difference to the preset rate of change of the difference threshold as a proportional coefficient. Meanwhile, the spatial clustering index is calculated by weighting and superimposing the individual-level anomaly probability according to the association weight between nodes. The spatial clustering index is then compared with the preset spatial clustering index threshold. When the spatial clustering index exceeds the preset spatial clustering index threshold, the preset risk judgment threshold is reduced and corrected a second time based on the ratio of the spatial clustering index to the preset spatial clustering index threshold as a proportional coefficient, after a first reduction correction. The modified preset risk assessment threshold is compared with the cumulative risk value. When the cumulative risk value exceeds the modified preset risk assessment threshold, a thermal runaway warning signal is output.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs branch-level characteristic quantities reflecting changes in the current distribution of parallel branches by forming a thermocouple array on a parallel bus and acquiring a temperature difference voltage sequence. Simultaneously, during the execution of a preset pulse sequence by the battery management system, the individual cell voltage response, temperature difference voltage response, and module temperature response are acquired synchronously. The transient changes in individual cell voltage are analyzed, and the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component are extracted. Thus, under the condition of the existence of group-level parallel averaging effect, more fine-grained branch information and electrochemical evolution information can be obtained, providing a multi-dimensional observation basis for early anomaly identification. Furthermore, this invention collects multi-point gas concentration data, temperature data, and ventilation parameters within the module. It then performs zoned calculations on the gas concentration data using diffusion coefficients, convection parameters, and adsorption parameters to determine the individual-level gas generation intensity. This intensity is then time-matched with electrochemical response parameters. Thus, even with the influence of gas spatial diffusion and retention, a gas intensity characterization corresponding to the abnormal evolution of the individual gaseous component can still be formed. This individual-level gas generation intensity, along with branch-level characteristic quantities and electrochemical response parameters, constitutes a multi-source parameter input, which helps improve the identifiability and consistency of gradual-type risks during the latency period. On the other hand, this invention maps branch-level characteristic quantities, electrochemical response parameters, and cell-level gas generation intensity to the battery pack topology diagram, and calculates the correlation weights between nodes based on electrical connections, thermal coupling relationships, and gas migration processes to obtain the cell-level anomaly probability. Subsequently, it performs cumulative calculations based on the first and second preset sampling periods to obtain the cumulative risk value, and completes the early warning output with a preset risk judgment threshold. This structured fusion and multi-period accumulation mechanism enables the early warning criteria to simultaneously reflect spatial propagation correlation and temporal evolution trends, thereby improving the stability, robustness, and interpretability of thermal runaway early warning, and making it more suitable for online monitoring in complex battery pack coupling scenarios. Attached Figure Description

[0014] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0017] Please see Figure 1 As shown, this invention provides a method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion, comprising: Step 1: Dissimilar metal contacts are spaced apart along the current direction on the parallel bus conductors of the battery pack and connected in series to form a thermocouple array. The temperature difference voltage sequence output by the thermocouple array is collected. Based on the temperature difference voltage sequence, branch-level characteristic quantities corresponding to changes in the current distribution of parallel branches are calculated. The temperature difference voltage sequence collected by the thermocouple array enables monitoring of temperature changes at the branch level of the battery pack, thus providing crucial data support for the early identification of current distribution and thermal runaway in individual battery cells. Calculating the temperature difference voltage sequence yields branch-level characteristic quantities related to current distribution, which is essential for battery state monitoring and fault diagnosis.

[0018] Specifically, dissimilar metal contacts are spaced apart along the current direction on the parallel bus conductors of the battery pack and connected in series to form a thermocouple array. The temperature difference voltage sequence output by the thermocouple array is collected, and the branch-level characteristic quantity corresponding to the change in the current distribution of the parallel branches is calculated based on the temperature difference voltage sequence.

[0019] Specifically, dissimilar metal contacts are arranged at equal intervals along the length of the busbar, with a spacing of 10 mm to 30 mm. Each adjacent contact is connected in series to form a continuous output terminal. The sampling frequency of the thermo-voltage sequence is set to 50 Hz to 500 Hz, and the sampling duration is set to 10 seconds to 60 seconds to obtain a stable thermo-voltage sequence.

[0020] Taking a battery pack with 16 parallel branches as an example, the temperature difference voltage sequence was collected under rated discharge conditions, and the amplitude of the temperature difference voltage sequence can fall within the range of 0.05 mV to 5 mV. When calculating the branch-level characteristic quantities corresponding to the changes in the current distribution of parallel branches based on the temperature difference voltage sequence, the branch-level characteristic quantities can be selected by combining the peak-to-peak value, root mean square value, and statistical quantities of the difference between adjacent sampling points of the temperature difference voltage sequence to obtain the branch-level characteristic quantities used to characterize the changes in the current distribution of parallel branches, thereby providing a data basis for the early identification of the current distribution and thermal runaway of battery cells.

[0021] The thermocouple array includes a first thermocouple array and a second thermocouple array respectively disposed on opposite sides of the conductor of the parallel bus, and the first thermocouple array and the second thermocouple array are divided into multiple corresponding segments along the length of the bus. The difference between the first and second temperature difference voltage sequences of the corresponding segments is calculated to obtain the segment differential temperature difference voltage sequence. The temperature difference bias direction is determined based on the sign of the average value of the segment differential temperature difference voltage sequence in each segment.

[0022] Specifically, the busbar length direction is divided into N segments, where N ranges from 8 to 32. Each segment contains several dissimilar metal contacts and forms the segment output terminal. After acquiring the first temperature difference voltage sequence and the second temperature difference voltage sequence, the difference between the first temperature difference voltage sequence and the second temperature difference voltage sequence of the same segment at the same time is calculated to obtain the segment differential temperature difference voltage sequence.

[0023] Taking N=16 as an example, if the average value of the differential temperature voltage sequence of the k-th segment within a 20-second window is +0.12 mV, then the temperature bias direction is positive; if the average value is -0.10 mV, then the temperature bias direction is negative. By determining the temperature bias direction, the subsequent sign correction of the scalar weights has a consistent basis, and misjudgments of the segment differential temperature voltage sequence are avoided due to common-mode components caused solely by ambient temperature rise or overall current changes.

[0024] The gradient change is determined based on the first-order difference between adjacent segments of the segment differential temperature and voltage sequence, and the segment with the largest absolute value of the gradient change is identified as the extreme value segment. The segment differential temperature and voltage value corresponding to the extreme value segment is used as a scalar weight, and the sign of the scalar weight is corrected according to the temperature difference bias direction. The sign-corrected scalar weight is multiplied with the temperature and voltage sequence to obtain the corrected temperature and voltage sequence, and the branch-level characteristic quantity is calculated based on the corrected temperature and voltage sequence.

[0025] Specifically, the first-order difference is calculated for the segment differential temperature-voltage sequence according to the segment index to obtain the gradient change between adjacent segments, and the segment with the largest absolute value of gradient change is taken as the extreme value segment; taking N as 16 as an example, if the absolute value of the gradient change between the 7th segment and the 8th segment is 0.35 mV and is the global maximum, then the extreme value segment is determined to be the 8th segment.

[0026] Subsequently, the average or peak value of the differential temperature voltage sequence within the extreme range is taken as the differential temperature voltage value of the range and used as a scalar weight. When the temperature difference bias direction is positive, the scalar weight retains its original sign; when the temperature difference bias direction is negative, the scalar weight takes the opposite sign. The sign-corrected scalar weight is multiplied point by point with the temperature voltage sequence to obtain the corrected temperature voltage sequence. Based on this, the branch-level characteristic quantity corresponding to the change in the current distribution of parallel branches is calculated, so that the contribution of the branch-level characteristic quantity is enhanced in the abnormal section while maintaining the consistency of sign.

[0027] The process of determining the gradient change and identifying the extreme value segment based on the first-order difference between adjacent segments of the differential temperature-voltage sequence includes: Centered on the extreme value segment, a preset number of adjacent segments are selected on both sides along the length of the busbar, and the gradient change corresponding to the adjacent segments is extracted. Calculate the attenuation trend of gradient change as segment distance increases, and determine the attenuation coefficient of gradient change in space; The attenuation coefficient is compared with the preset single-source diffusion attenuation threshold. When the attenuation coefficient meets the characteristics of rapid attenuation on one side, the extreme value segment is confirmed as an abnormal segment. When the attenuation coefficient shows the characteristics of gradual attenuation on both sides, the extreme value segment is determined to be an environmental disturbance segment. Only the segment differential temperature and voltage values ​​of the confirmed abnormal segments are extracted as scalar weights to participate in the correction calculation of the subsequent temperature and voltage sequence.

[0028] Specifically, the preset number is 2 to 5. Adjacent segments are selected on both sides of the extreme value segment to form a set of segments for calculating the attenuation trend. The attenuation trend is calculated based on the gradient change amount within each set of segments according to the segment distance. The attenuation coefficient can be determined by the ratio sequence of the absolute value of the gradient change amount with the distance or the logarithmic slope sequence.

[0029] Taking a preset quantity of 3 as an example, if the absolute value of the gradient change on the right side of the extreme value segment shows a decreasing sequence of 0.30, 0.16, and 0.08, while the left side shows a smooth sequence of 0.05, 0.04, and 0.03, then the attenuation coefficient exhibits a unilateral rapid attenuation characteristic. This attenuation coefficient is compared with a preset single-source diffusion attenuation threshold, which can be between 0.50 and 0.80. When the attenuation coefficient satisfies the unilateral rapid attenuation characteristic, the extreme value segment is confirmed as an abnormal segment. When the absolute values ​​of the gradient changes on both sides show a smooth attenuation or approximately symmetrical attenuation, the extreme value segment is determined to be an environmental disturbance segment and does not participate in scalar weight extraction. Only the segment differential temperature and voltage values ​​corresponding to the confirmed abnormal segments are extracted as scalar weights to participate in the subsequent correction calculation of the temperature and voltage sequence, thereby making the corrected temperature and voltage sequence more spatially consistent with the single-source diffusion characteristics and improving the stability and interpretability of branch-level features.

[0030] Step Two: Based on the branch-level characteristic quantities, the battery management system executes a preset pulse sequence to perform pulse control operations on the target cell or branch. Simultaneously, it collects the cell voltage response, temperature difference voltage response, and module temperature response. The transient changes in cell voltage are analyzed to extract the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component. To stimulate the electrochemical response inside the battery, the pulse duration and amplitude are controlled to effectively guide the cell or branch to generate a voltage response under specific conditions. The simultaneous collection of cell voltage response, temperature difference voltage response, and module temperature response combines the battery's voltage, temperature, and thermocouple signals to more accurately capture the dynamic changes inside and outside the battery in both time and space. Analysis of the transient changes in cell voltage allows for the extraction of the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component, further providing a reliable basis for battery state assessment, aging prediction, and fault early warning.

[0031] Specifically, based on the branch-level characteristic quantities, the battery management system is controlled to execute a preset pulse sequence to perform pulse control operations on the target cell or target branch, and the cell voltage response, temperature difference voltage response and module temperature response are collected simultaneously. The transient changes in cell voltage are analyzed, and the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component and diffusion component are extracted.

[0032] Specifically, a preset pulse sequence is inserted and executed during charging or discharging. Before execution, branch-level characteristic quantities are recorded to determine the target cell or target branch. The sampling frequency for synchronous acquisition is 200 Hz to 2000 Hz, with the sampling start time covering 0.2 seconds before the pulse rising edge and the sampling end time covering 2 seconds after the pulse recovery phase, to ensure complete analysis of transient changes in cell voltage. Taking a cell with a nominal capacity of 100 AH as an example, the pulse amplitude of the preset pulse sequence is 0.5 A to 5 A, and the pulse amplitude change adopts either a fixed amplitude or segmented amplitude method. In the cell voltage response, the instantaneous voltage drop can fall within the range of 1 mV to 30 mV, and the recovery process can last from 0.5 seconds to 10 seconds. By analyzing the transient changes in cell voltage, the instantaneous change parameters, established change parameters, and recovered change parameters required for subsequent separation and extraction are obtained. This provides the original data basis for subsequently mapping the instantaneous change parameters, established change parameters, and recovered change parameters to the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component, respectively.

[0033] The preset pulse sequence contains different combinations of time widths and is used to separate and extract the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component. This includes: constructing the preset pulse sequence into a first pulse segment, a second pulse segment, and a third pulse segment arranged progressively according to time width; during the execution of the first pulse segment, the second pulse segment, and the third pulse segment, the individual cell voltage response is time-aligned and the instantaneous change parameters corresponding to the first pulse segment, the establishment change parameters corresponding to the second pulse segment, and the recovery change parameters corresponding to the third pulse segment are extracted.

[0034] Specifically, the first, second, and third pulse segments are executed sequentially within the same preset pulse sequence, with intervals between them. The duration of these intervals ranges from 0.2 to 2 seconds to complete voltage baseline regression. The duration of the first pulse segment ranges from 5 to 50 milliseconds, the second from 0.1 to 1 second, and the third from 1 to 8 seconds. Taking a combination of durations of 10 milliseconds, 300 milliseconds, and 3 seconds as an example, the amplitude of the three pulse segments is consistent at 2 amps. At the moment the first pulse segment ends, the initial voltage drop value is extracted from the individual voltage response as the instantaneous change parameter. At the end of the second pulse segment, the voltage gradual increment and its rate of change are extracted as the establishment change parameter. After the third pulse segment ends, the half-recovery time and tail residual of the recovery curve are extracted as the recovery change parameter. When aligning the individual voltage response in time, the rising edge trigger point of each pulse segment is used as the time zero point. The individual voltage response is aligned to a unified time axis according to the sampling time scale, so that the instantaneous change parameters, establishment change parameters, and recovery change parameters extracted from different pulse segments are comparable and can be used for subsequent mapping.

[0035] The instantaneous change parameters, established change parameters, and recovered change parameters are mapped to the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component, respectively, according to a preset mapping relationship. The time width of the next cycle of the first, second, and third pulse segments is constrained and updated based on the change amplitude of the module temperature response and thermo-voltage response within each pulse segment, so that different time width combinations can maintain the distinguishability of different electrochemical response parameters without changing the order of the pulse segments.

[0036] Specifically, the preset mapping relationship is disclosed as follows: the instantaneous change parameter of the first pulse segment is mapped to the electrochemical response parameter corresponding to the ohmic internal resistance component; the establishment change parameter of the second pulse segment is mapped to the electrochemical response parameter corresponding to the charge transfer component; and the recovery change parameter of the third pulse segment is mapped to the electrochemical response parameter corresponding to the diffusion component. The electrochemical response parameter corresponding to the ohmic internal resistance component is obtained by the ratio of the instantaneous change parameter to the pulse amplitude; the electrochemical response parameter corresponding to the charge transfer component is obtained by the ratio of the voltage gradual increment to the pulse amplitude in the establishment change parameter; and the electrochemical response parameter corresponding to the diffusion component is obtained by the combination of the half-recovery time and the tail residual in the recovery change parameter.

[0037] The constraint update for the next cycle time width is as follows: when the change amplitude of the module temperature response within a certain pulse segment exceeds 0.2 degrees Celsius or the change amplitude of the thermoelectric voltage response within the same pulse segment exceeds 0.2 millivolts, the next cycle time width of the corresponding pulse segment will be reduced by 20% to 50% according to a preset ratio; when the change amplitude of the module temperature response within a certain pulse segment is less than 0.05 degrees Celsius and the change amplitude of the thermoelectric voltage response within the same pulse segment is less than 0.05 millivolts, the next cycle time width of the corresponding pulse segment will be extended by 10% to 30% according to a preset ratio; at the same time, the time widths of the first pulse segment, the second pulse segment, and the third pulse segment are always limited to satisfy the progressive relationship, so as to maintain the separation and extraction conditions of the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component on the time scale.

[0038] The duration of the first pulse segment is limited to ensure that the initial voltage drop of the individual cell voltage response exhibits abrupt changes within a preset sampling resolution; the duration of the second pulse segment is limited to ensure that the individual cell voltage response exhibits a recognizable gradual change during the pulse hold period; the duration of the third pulse segment is limited to ensure that the individual cell voltage response exhibits a recognizable hysteresis during the recovery process after the pulse ends; the thermoelectric voltage response is used to determine the electrochemical response parameters, including: simultaneously acquiring the individual cell voltage response and the thermoelectric voltage response while executing the preset pulse sequence, and aligning the individual cell voltage response and the thermoelectric voltage response according to the same sampling time scale; calculating the amplitude of the change in the individual cell voltage response within the pulse segment and the temperature difference voltage response, respectively. The amplitude of the differential voltage response within the same pulse segment is calculated, and the sign consistency between the direction of change of the individual voltage response and the direction of change of the thermoelectric voltage response is determined. When the direction of change of the individual voltage response and the direction of change of the thermoelectric voltage response are consistent, the proportionality coefficient of their amplitudes is calculated and compared with a preset proportional range. When the proportionality coefficient is within the preset proportional range, the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component, and diffusion component extracted from the individual voltage response are confirmed to be valid. When the proportionality coefficient exceeds the preset proportional range or the direction of change is inconsistent, the amplitude of the electrochemical response parameters extracted from the individual voltage response is corrected according to the amplitude of change of the thermoelectric voltage response.

[0039] Specifically, the preset sampling resolution is 1 to 5 millivolts. The first pulse segment is limited to ensuring that the initial voltage drop exceeds at least twice the preset sampling resolution. The second pulse segment is limited to ensuring that a continuous slowly varying sequence of no less than 5 sampling points is formed during the pulse holding period. The third pulse segment is limited to ensuring that a hysteresis tail sequence of no less than 20 sampling points is formed during the recovery process. The proportionality coefficient is taken as the ratio of the change amplitude of the individual voltage response within the pulse segment to the change amplitude of the thermoelectric voltage response within the same pulse segment, with a preset proportionality range of 5 to 200. When the proportionality coefficient is within the preset proportionality range and the signs of the change directions of the individual voltage response and the thermoelectric voltage response are consistent, the corresponding electrochemical response parameter is confirmed to be valid. When the proportionality coefficient exceeds the preset proportionality range or the signs are inconsistent, the electrochemical response parameter is amplitude corrected according to the ratio of the change amplitude of the thermoelectric voltage response to the preset thermoelectric voltage response reference amplitude, with the preset thermoelectric voltage response reference amplitude ranging from 0.1 mV to 0.5 mV. This ensures that the electrochemical response parameters remain comparable and can be used for subsequent fusion calculations even when there are changes in the current distribution of parallel branches and local heating disturbances.

[0040] Step 3: Collect multi-point gas concentration data, temperature data, and ventilation parameters within the module. Obtain the diffusion coefficient, convection parameters, and adsorption parameters corresponding to the gas migration process. Based on the diffusion, convection, and adsorption changes of gas within the module, perform zoned calculations on the multi-point gas concentration data to determine the cell-level gas generation intensity for each cell. Then, time-match the cell-level gas generation intensity with the electrochemical response parameters. During battery operation, gases may be released due to side reactions, leading to changes in gas concentration. These changes reflect the physical and chemical states inside the battery. By analyzing the diffusion, convection, and adsorption patterns of gas within the module, zoned calculations on the gas concentration data can be performed, and this data can be time-matched with the electrochemical response parameters. This process allows changes in gas generation intensity to be combined with the electrochemical reactions inside the battery, providing more refined anomaly monitoring, especially playing a crucial role in the early identification of thermal runaway.

[0041] Specifically, within the battery pack module space, based on the physical location of each cell, the module space is divided into multiple initial calculation units corresponding one-to-one with each cell. Each cell's physical location corresponds to one calculation unit, and the boundaries of these units are set according to the battery's physical dimensions and installation layout. Within each calculation unit, gas concentration, temperature, and ventilation data are collected at that location, providing spatial resolution and data support for subsequent gas migration analysis and calculations.

[0042] Within each computational unit, the diffusion distance of the gas within each time step is calculated based on the diffusion coefficient, convection parameters, and time step. The diffusion coefficient is set according to the gas type and temperature within the battery module and is obtained through experimental data or physical models. In this embodiment, the diffusion coefficient of hydrogen is assumed to be 0.61 cm² / s, and the gas diffusion time step is set to 1 to 5 seconds to accommodate dynamic changes in the gas within the battery system.

[0043] The main direction of gas migration is calculated based on the module's convection parameters. Assuming the airflow velocity within the module is 0.1 m / s, the main direction of gas migration is determined by considering the gas diffusion process. This step ensures the calculation of the dynamic changes of the gas within each time step.

[0044] Within each time step, the spatial boundaries of the computational cells are updated based on the diffusion distance and the main migration direction. The spatial boundaries of each cell are adjusted according to the dynamic changes in gas diffusion and convection migration to maintain a reasonable distribution of gas concentration. When the spatial boundaries of adjacent computational cells overlap, the gas concentration gradient within the overlapping region is calculated and corrected based on the direction and magnitude of the gradient. If the gas concentration within the overlapping region increases along the monomer direction, the gas concentration in that region is proportionally allocated to the monomer generation intensity calculation; conversely, a corresponding value is deducted from the monomer generation intensity.

[0045] Finally, the gas generation intensity of the monomer is calculated based on the updated gas concentration gradient. Specifically, the calculation of the gas generation intensity depends on the generation intensity of the previous time step and the amount of new generation. The updated formula for the gas generation intensity is: Gas generation intensity update calculation method: The gas generation intensity of each monomer is the generation intensity of the previous time step plus the newly generated amount. The generation amount is related to the amplitude and direction of the gas concentration gradient in the overlapping region.

[0046] Specifically, the updated generation intensity is calculated as follows: When the boundaries of adjacent computational cells overlap, the direction and magnitude of the gas concentration gradient in the overlapping region are calculated. The gradient component aligned with the cell's location direction is allocated to the corresponding cell's gas generation intensity calculation according to its gradient magnitude. The gradient component opposite to the cell's location direction is subtracted from the cell's gas generation intensity.

[0047] Gas generation intensity update method: Within each time step, the gas generation intensity corresponding to each monomer is updated as follows: First, the gas generation intensity of the monomer at the end of the previous time step is obtained as the baseline value for the calculation of the current time step.

[0048] Based on the gas concentration distribution inside and outside the corresponding computational cell for this monomer within this time step, an initial increase in gas generation intensity is calculated. This initial increase reflects the total amount of new gas generated within the computational cell due to factors such as electrochemical reactions.

[0049] The initial incremental amount is corrected to obtain the corrected effective incremental amount. The correction method is as follows: based on the gas concentration gradient in the overlapping region between the single-unit calculation cell and its adjacent calculation cells within the current time step, an adjustment coefficient between 0 and 1 is determined. When the gas concentration gradient direction in the overlapping region points towards the single-unit, the adjustment coefficient is set to a larger value (e.g., close to 1), making the effective incremental amount close to the initial incremental amount; when the gas concentration gradient direction in the overlapping region moves away from the single-unit, the adjustment coefficient is set to a smaller value (e.g., close to 0), making the effective incremental amount significantly reduced, or even reduced to a negative value, to reflect the diffusion of gas to other single-unit cells.

[0050] The sum of the baseline value and the effective incremental amount is the gas generation intensity of the monomer at the end of this time step.

[0051] Through the above iterative updates, the gas generation intensity of each monomer can dynamically follow the real-time changes in the gas concentration in its region.

[0052] Step 4: Map branch-level characteristic quantities, electrochemical response parameters, and cell-level gas generation intensity to the battery pack topology diagram. Calculate the correlation weights between nodes based on electrical connections, thermal coupling, and gas migration processes to obtain the cell-level anomaly probability. The battery pack topology diagram displays the connection relationships between battery cells, branches, busbars, and various sensing nodes. By calculating the correlation weights between nodes in the electrical connection, thermal coupling, and gas migration processes, a quantitative result regarding the anomaly probability of each cell can be obtained. This calculation process provides a precise spatial and functional mapping for the battery pack, making the location of anomaly sources more accurate. Real-time monitoring at the battery pack level is possible, and potential thermal runaway risks can be promptly warned. This step is the core link in achieving efficient monitoring and real-time early warning.

[0053] Specifically, first construct the battery pack topology diagram and establish node feature mapping, including: Establish a battery pack topology diagram. The battery pack topology diagram uses individual battery cells, branches, busbars, and each sensor node as graph nodes, and electrical connections, thermal coupling relationships, and gas migration processes as graph edges.

[0054] After constructing the topology graph, branch-level features, electrochemical response parameters, and cell-level gas generation intensity are mapped to corresponding cell nodes or branch nodes as node feature inputs. For example, in a module containing 12 cells: each cell node contains 3 electrochemical response parameters; each branch node contains 1 branch-level feature; and each cell node contains 1 cell-level gas generation intensity. After mapping, the correlation weight calculation stage begins.

[0055] The electrical propagation weight is calculated as follows: The potential decay coefficient is calculated based on the equivalent resistance between adjacent nodes.

[0056] The potential attenuation coefficient is calculated as follows: the equivalent resistance between adjacent nodes is divided by a preset reference resistance value, and then interval normalization is performed to ensure the potential attenuation coefficient falls between 0 and 1. The reference resistance value can be selected based on the module design; for example, 0.08 ohms can be used as the calibration reference. For example, if the equivalent resistance between adjacent nodes is 0.05 ohms, the ratio of the equivalent resistance to the reference resistance is 0.05 divided by 0.08, resulting in 0.625. After normalization, the potential attenuation coefficient is 0.6. Subsequently, the reciprocal of the potential attenuation coefficient is used as the electric propagation weight. Therefore, the electric propagation weight is 1 divided by 0.6, resulting in approximately 1.67.

[0057] The heat transfer weight is calculated using the following steps: The heat transfer coefficient is calculated based on the thermal resistance between adjacent nodes.

[0058] The heat transfer coefficient is defined as the reciprocal of the thermal resistance. For example, if the thermal resistance is 0.2 degrees Celsius per watt, then the heat transfer coefficient is 1 divided by 0.2, which gives 5 watts per degree Celsius.

[0059] Secondly, consider the effect of spatial distance. If the distance between adjacent nodes is 5 cm, then the square of the distance is 25 square centimeters. Taking the reciprocal of the square of the distance, i.e., 1 divided by 25, we get 0.04. The final heat transfer weight is: the heat flux transfer coefficient multiplied by the reciprocal of the square of the distance, i.e., 5 multiplied by 0.04, which gives 0.2.

[0060] The gas propagation weight is calculated as follows: The directional correction factor is calculated based on the angle between the distance between monomers and the direction of gas migration. The directional correction factor is the absolute value of the cosine of the angle. For example, if the angle is 30 degrees and the cosine is approximately 0.866, then the directional correction factor is 0.87.

[0061] Next, the distance exponential decay factor is calculated. The distance exponential decay factor is calculated according to the natural exponential decay law. The gas attenuation constant is set to 0.23 per centimeter. If the distance is 3 centimeters, the attenuation exponent is 0.23 multiplied by 3, resulting in 0.69, while the natural exponential decay result is approximately 0.5. Therefore, the gas propagation weight is: Multiplying the direction correction factor by the distance attenuation factor, i.e., 0.87 multiplied by 0.5, yields approximately 0.435.

[0062] After obtaining the electrical propagation weights, thermal propagation weights, and gas propagation weights, normalization synthesis is performed. The normalization steps are as follows: The first step is to normalize the maximum value of each of the three types of weights, so that they are mapped to the interval between 0 and 1.

[0063] The second step is to set the preset scaling factors. For example: the scaling factor for electric propagation is 0.4; the scaling factor for thermal propagation is 0.35; and the scaling factor for gas propagation is 0.25.

[0064] The third step is to multiply the three normalized weights by their respective proportional coefficients and then sum them. For example, if the normalized results are: electric propagation weight 0.8, thermal propagation weight 0.6, and gas propagation weight 0.4; then the final association weight is: 0.8 multiplied by 0.4, plus 0.6 multiplied by 0.35, plus 0.4 multiplied by 0.25, which equals 0.32 plus 0.21 plus 0.10, resulting in 0.63. This final value is the association weight between the nodes.

[0065] The fourth step is to calculate the anomaly probability at the individual node level: This involves weighted propagation of the association weights between nodes and the node feature values. For each individual node, its anomaly probability is calculated as follows: First, multiply the node's own feature by its own node weight; second, multiply the features of adjacent nodes by their corresponding association weights; finally, sum and normalize all weighted results to obtain the individual-level anomaly probability.

[0066] For example, if a single node has a weight of 1 and an outlier of 0.6, its neighboring node A has a weight of 0.63 and an outlier of 0.5, and its neighboring node B has a weight of 0.42 and an outlier of 0.4, then the weighted result is: 0.6 + 0.63 multiplied by 0.5 + 0.42 multiplied by 0.4, which equals 0.6 + 0.315 + 0.168, or 1.083. After normalization, the single-node outlier probability is approximately 0.72.

[0067] Step 5: Accumulate the anomaly probability at the individual cell level based on the first and second preset sampling periods to obtain the cumulative risk value. When the cumulative risk value exceeds the preset risk judgment threshold, a thermal runaway warning signal is output. Through the cumulative calculation of short and long time windows, the risk change trend of the battery pack can be quantified according to the anomaly probability at different time scales within the battery pack. When the cumulative risk value exceeds the preset risk judgment threshold, the system will output a thermal runaway warning signal. This mechanism ensures that the thermal runaway warning system can not only respond quickly to sudden problems, but also cope with relatively gradual and long-term battery performance degradation problems, thereby improving the safety and reliability of the battery pack.

[0068] Specifically, in step five, the cumulative risk value is calculated based on the first and second preset sampling periods for the individual-level anomaly probability. When the cumulative risk value exceeds the preset risk judgment threshold, a thermal runaway warning signal is output. In this embodiment, the sampling period is defined as the time interval for completing one update calculation of the individual-level anomaly probability, and the sampling period is 1 second. The first preset sampling period is 30, and the second preset sampling period is 300, corresponding to the cumulative intervals of 30 seconds and 300 seconds, respectively. Taking the individual-level anomaly probability sequence obtained by a certain individual within a continuous sampling period as input, the cumulative result based on the first preset sampling period is defined as the arithmetic mean of the individual-level anomaly probability within the most recent 30 sampling periods, and the cumulative result based on the second preset sampling period is defined as the arithmetic mean of the individual-level anomaly probability within the most recent 300 sampling periods. The cumulative risk value is defined as the weighted sum of the above two cumulative results, where the weight of the cumulative result corresponding to the first preset sampling period is 0.6, and the weight of the cumulative result corresponding to the second preset sampling period is 0.4. The preset risk judgment threshold is 0.75, which is used for subsequent comparison with the cumulative risk value to trigger the output of the thermal runaway warning signal.

[0069] In the process of calculating the difference between the cumulative result based on the first preset number of sampling periods and the cumulative result based on the second preset number of sampling periods, and obtaining the rate of change of the difference by dividing the change in the difference over multiple consecutive sampling periods by the corresponding number of sampling periods, the difference is defined as "the cumulative result based on the first preset number of sampling periods minus the cumulative result based on the second preset number of sampling periods"; the multiple consecutive sampling periods are taken as 10 sampling periods; the change in the difference is defined as "the difference in the current sampling period minus the difference before the 10th sampling period"; the rate of change of the difference is defined as "the change in the difference divided by 10"; for example, if the cumulative result based on the first preset number of sampling periods is 0.72 and the cumulative result based on the second preset number of sampling periods is 0.60 at a certain moment, then the difference is 0.12; if the difference was 0.06 10 sampling periods ago, then the change in the difference is 0.06, and the rate of change of the difference is 0.06 divided by 10 to get 0.006.

[0070] When comparing the rate of change of the difference with a preset threshold, if the rate of change of the difference exceeds the preset threshold, the preset risk assessment threshold is reduced once using the ratio of the rate of change of the difference to the preset threshold as a proportional coefficient. Simultaneously, a spatial clustering index is calculated by weighting and superimposing the individual-level anomaly probabilities based on the association weights between nodes. This spatial clustering index is then compared with a preset threshold. If the spatial clustering index exceeds the preset threshold, the preset risk assessment threshold is reduced a second time using the ratio of the spatial clustering index to the preset threshold as a proportional coefficient, building upon the initial reduction. During this process, the preset rate of change of the difference threshold is set to 0.005. When the rate of change of the difference exceeds 0.005, the proportional coefficient is defined as "the rate of change of the difference divided by the preset threshold," and the initial reduction is defined as "the preset risk assessment threshold divided by the proportional coefficient." Continuing with the example above, if the rate of change of the difference is 0.006, then the proportional coefficient is 0.0. 0.06 divided by 0.005 yields 1.2. After the first reduction correction, the preset risk judgment threshold is adjusted from 0.75 to 0.75 divided by 1.2, resulting in 0.625. The spatial clustering index is defined as "the sum of all individual-level anomaly probabilities weighted according to the association weights between nodes," where the contribution of each pair of adjacent nodes is "the individual-level anomaly probability of the adjacent node multiplied by the association weight of the adjacent node to the current node," and the summation is obtained over all nodes to obtain the spatial clustering index. The preset spatial clustering index threshold is 1.8. When the spatial clustering index exceeds 1.8, the proportional coefficient for the second reduction correction is defined as "the spatial clustering index divided by the preset spatial clustering index threshold," and the second reduction correction is defined as "dividing the preset risk judgment threshold after the first reduction correction by this proportional coefficient." For example, if the spatial clustering index is 2.16, then the proportional coefficient for the second reduction correction is 2.16 divided by 1.8, resulting in 1.2. After the second reduction correction, the preset risk judgment threshold is adjusted from 0.625 to 0.625 divided by 1.2, resulting in approximately 0.521.

[0071] Finally, in the process of comparing the corrected preset risk judgment threshold with the cumulative risk value, and outputting a thermal runaway warning signal when the cumulative risk value exceeds the corrected preset risk judgment threshold, the cumulative risk value is calculated by multiplying the cumulative result corresponding to the first preset sampling period by 0.6 and then adding the cumulative result corresponding to the second preset sampling period by 0.4. Continuing the above example, if the cumulative result corresponding to the first preset sampling period is 0.72 and the cumulative result corresponding to the second preset sampling period is 0.60, then the cumulative risk value is 0.72 multiplied by 0.6 plus 0.60 multiplied by 0.4, which equals 0.672. The cumulative risk value of 0.672 is compared with the corrected preset risk judgment threshold of 0.521. Since 0.672 exceeds 0.521, a thermal runaway warning signal is output. Conversely, if the cumulative risk value does not exceed the corrected preset risk judgment threshold, no thermal runaway warning signal is output, and the process of cumulative calculation and threshold correction based on the first and second preset sampling periods is repeated in the next sampling period.

[0072] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion, characterized in that, include: Step 1: Set dissimilar metal contacts at intervals along the current direction on the parallel bus conductor of the battery pack and connect them in series to form a thermocouple array. Collect the temperature difference voltage sequence output by the thermocouple array and calculate the branch-level characteristic quantity corresponding to the change in the current distribution of the parallel branch based on the temperature difference voltage sequence. Step 2: Based on the branch-level characteristic quantities, control the battery management system to execute a preset pulse sequence to perform pulse control operation on the target cell or target branch, and simultaneously collect the cell voltage response, temperature difference voltage response and module temperature response. Analyze the transient changes in cell voltage and extract the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component and diffusion component. Step 3: Collect gas concentration data, temperature data and ventilation parameters at multiple points within the module, obtain the diffusion coefficient, convection parameters and adsorption parameters corresponding to the gas migration process, and perform partitioned calculations on the gas concentration data according to the diffusion, convection and adsorption changes of the gas inside the module to determine the unit-level gas generation intensity corresponding to each unit, and perform time matching between the unit-level gas generation intensity and the electrochemical response parameters. Step 4: Map the branch-level characteristic quantities, electrochemical response parameters, and cell-level gas generation intensity to the battery pack topology diagram. Calculate the correlation weights between nodes based on electrical connections, thermal coupling relationships, and gas migration processes to obtain the cell-level anomaly probability. Step 5: Accumulate the individual-level anomaly probability based on the first preset sampling period and the second preset sampling period to obtain the cumulative risk value. When the cumulative risk value exceeds the preset risk judgment threshold, output a thermal runaway early warning signal.

2. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 1, characterized in that, The thermocouple array includes a first thermocouple array and a second thermocouple array respectively disposed on opposite sides of the conductor of the parallel bus, and the first thermocouple array and the second thermocouple array are divided into multiple corresponding segments along the length of the bus. The difference between the first temperature difference voltage sequence and the second temperature difference voltage sequence of the corresponding segment is calculated to obtain the segment differential temperature difference voltage sequence. The temperature difference bias direction is determined according to the sign of the average value of the segment differential temperature difference voltage sequence in each segment. The gradient change is determined based on the first-order difference between adjacent segments of the differential temperature and voltage sequence, and the segment with the largest absolute value of the gradient change is identified as the extreme value segment. The segment differential temperature voltage value corresponding to the extreme value segment is used as a scalar weight, and the sign of the scalar weight is corrected according to the temperature difference bias direction. The sign-corrected scalar weight is multiplied with the temperature difference voltage sequence to obtain the corrected temperature difference voltage sequence, and the branch-level characteristic quantity is calculated based on the corrected temperature difference voltage sequence.

3. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 2, characterized in that, The process of determining the gradient change and identifying the extreme value segment based on the first-order difference between adjacent segments of the differential temperature-voltage sequence includes: Centered on the extreme value segment, a preset number of adjacent segments are selected on both sides along the length of the busbar, and the gradient change corresponding to the adjacent segments is extracted. Calculate the attenuation trend of gradient change as segment distance increases, and determine the attenuation coefficient of gradient change in space; The attenuation coefficient is compared with the preset single-source diffusion attenuation threshold. When the attenuation coefficient meets the characteristics of rapid attenuation on one side, the extreme value segment is confirmed as an abnormal segment. When the attenuation coefficient shows the characteristics of gradual attenuation on both sides, the extreme value segment is determined to be an environmental disturbance segment. Only the segment differential temperature and voltage values ​​of the confirmed abnormal segments are extracted as scalar weights to participate in the correction calculation of the subsequent temperature and voltage sequence.

4. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 1, characterized in that, The preset pulse sequence contains different combinations of time widths and is used to separate and extract the electrochemical response parameters corresponding to the ohmic resistance component, charge transfer component, and diffusion component, including: The preset pulse sequence is constructed into a first pulse segment, a second pulse segment, and a third pulse segment arranged progressively according to the time width; During the execution of the first pulse segment, the second pulse segment, and the third pulse segment, the individual voltage response is time-aligned and the instantaneous change parameters corresponding to the first pulse segment, the establishment change parameters corresponding to the second pulse segment, and the recovery change parameters corresponding to the third pulse segment are extracted respectively. The instantaneous change parameters, the established change parameters, and the recovered change parameters are respectively mapped to the electrochemical response parameters corresponding to the ohmic internal resistance component, the charge transfer component, and the diffusion component according to the preset mapping relationship. The time width of the next cycle for the first, second, and third pulse segments is constrained and updated based on the variation amplitude of the module temperature response and thermoelectric voltage response within each pulse segment, so that different combinations of time widths can maintain the distinguishability of different electrochemical response parameters without changing the order of the pulse segments.

5. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 4, characterized in that, The time width of the first pulse segment is limited to ensure that the initial voltage drop of the single-unit voltage response exhibits abrupt changes within a preset sampling resolution; The time width of the second pulse segment is limited to ensure that the individual cell voltage response exhibits a distinguishable gradual change during the pulse hold period; The time width of the third pulse segment is limited to ensure that the individual cell voltage response exhibits a discernible hysteresis characteristic during the recovery process after the pulse ends.

6. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 5, characterized in that, The temperature-voltage response is involved in the determination of electrochemical response parameters, including: While executing the preset pulse sequence, the individual voltage response and the temperature difference voltage response are simultaneously acquired, and the individual voltage response and the temperature difference voltage response are time-aligned according to the same sampling time scale. Calculate the amplitude of the change in the individual unit voltage response within the pulse segment and the amplitude of the change in the temperature difference voltage response within the same pulse segment, and calculate the consistency of the sign of the change direction of the individual unit voltage response and the change direction of the temperature difference voltage response. When the direction of change of the monomer voltage response is consistent with the direction of change of the thermoelectric voltage response, the proportionality coefficient of the two changes is calculated and compared with the preset proportional range. When the proportionality coefficient is within the preset proportional range, the electrochemical response parameters corresponding to the ohmic internal resistance component, charge transfer component and diffusion component extracted from the monomer voltage response are confirmed to be effective. When the proportional coefficient exceeds the preset proportional range or the direction of change is inconsistent, the electrochemical response parameters extracted from the monomer voltage response are corrected according to the change amplitude of the temperature difference voltage response.

7. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 1, characterized in that, Step 3 involves partitioning and calculating the gas concentration data at multiple points based on the diffusion, convection, and adsorption patterns of the gas within the module to determine the gas generation intensity at the monomer level. In this process, the module space is divided into multiple initial calculation units corresponding to each monomer according to the physical location of the monomer. Centered on the physical location of each monomer, the diffusion distance of the gas within each time step is calculated based on the diffusion coefficient and time step, the main direction of gas migration is determined based on the convection parameters, and the effective concentration of the gas at the boundary of the corresponding calculation unit is corrected based on the adsorption parameters. Within a continuous time step, the spatial boundaries of each computational unit are updated based on the diffusion distance and the main migration direction. When the spatial boundaries of adjacent computing units overlap, the direction and magnitude of the gas concentration gradient within the overlapping region are calculated. The gradient component that is consistent with the direction of the unit location is allocated to the gas generation intensity calculation of the corresponding unit according to the gradient magnitude ratio, and the gradient component that is opposite to the direction of the unit location is deducted from the gas generation intensity calculation of the corresponding unit, thereby completing the update calculation of the gas generation intensity at the unit level.

8. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 1, characterized in that, In step four, during the calculation of the association weights between nodes based on electrical connections, thermal coupling, and gas migration processes, the association weights between nodes are divided into electrical propagation weights, thermal propagation weights, and gas propagation weights. For the electric propagation weight, the potential attenuation coefficient is calculated based on the equivalent resistance between adjacent nodes, and the reciprocal of the potential attenuation coefficient is used as the electric propagation weight. For the heat propagation weight, the heat flux transfer coefficient is calculated based on the thermal resistance between adjacent nodes, and the heat propagation weight is obtained by multiplying the heat flux transfer coefficient by the reciprocal of the square of the spatial distance between individual nodes. For the gas propagation weight, the direction correction coefficient is calculated based on the angle relationship between the distance between individual units and the gas migration direction, and the gas propagation weight is obtained by multiplying the direction correction coefficient by the distance exponential decay factor. The electrical propagation weight, thermal propagation weight, and gas propagation weight are synthesized according to a preset normalization rule to obtain the correlation weight between nodes.

9. The method for monitoring thermal runaway of sodium-ion battery packs based on multi-source parameter fusion according to claim 1, characterized in that, Step five, obtaining the cumulative risk value, includes: The difference between the cumulative result based on the first preset number of sampling periods and the cumulative result based on the second preset number of sampling periods is calculated respectively, and the rate of change of the difference is obtained by dividing the change of the difference within multiple consecutive sampling periods by the corresponding number of sampling periods. The rate of change of the difference is compared with a preset rate of change of the difference threshold. When the rate of change of the difference exceeds the preset rate of change of the difference threshold, the preset risk judgment threshold is reduced once by using the ratio of the rate of change of the difference to the preset rate of change of the difference threshold as a proportional coefficient. Meanwhile, the spatial clustering index is calculated by weighting and superimposing the individual-level anomaly probability according to the association weight between nodes. The spatial clustering index is then compared with the preset spatial clustering index threshold. When the spatial clustering index exceeds the preset spatial clustering index threshold, the preset risk judgment threshold is reduced and corrected a second time based on the ratio of the spatial clustering index to the preset spatial clustering index threshold as a proportional coefficient, after a first reduction correction. The modified preset risk assessment threshold is compared with the cumulative risk value. When the cumulative risk value exceeds the modified preset risk assessment threshold, a thermal runaway warning signal is output.

Images