Solid-state battery performance test method and system based on data analysis

Abstract

The invention relates to the technical field of data analysis, in particular to a solid-state battery performance testing method and system based on data analysis, and the method comprises the steps: obtaining voltage time sequence data of a solid-state battery, and removing a DC component to obtain a fluctuation voltage sequence; and on the basis of the ratio of the second-order difference to the first-order difference of the fluctuation voltage sequence and the signal-to-noise ratio, calculating a local transient response index at each moment for representing the local steep degree of the voltage fluctuation. Aiming at the problem that micro-short-circuit waveforms of the solid-state battery are diversified and are easy to be covered by noise, the steep degree of voltage fluctuation is quantified in real time by constructing a local transient response index, a deformation wavelet basis function highly matched with a current signal form is dynamically generated accordingly, and a characteristic value calculation method of introducing a form proximity penalty term is combined to calculate the micro-short-circuit waveform of the solid-state battery. Accurate matching and enhancement of weak short-circuit signals are achieved, non-fault interference is effectively eliminated, and the signal-to-noise ratio and accuracy of detection are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology. More specifically, this invention relates to a data analysis-based method and system for testing the performance of solid-state batteries. Background Technology

[0002] Solid-state batteries, by using solid electrolytes instead of liquid electrolytes, have the potential for high energy density and high safety. However, in short-circuit testing scenarios, especially in the initial stages of forced internal short circuits or nail penetration tests, the solid electrolyte is highly susceptible to micro-short circuits caused by lithium dendrite growth. These micro-short circuits manifest as extremely weak voltage drop signals, often masked by environmental noise, and are precursors to thermal runaway.

[0003] Currently, for the detection of such weak signals, existing technologies often employ singularity detection algorithms based on wavelet transform. This technique utilizes the multi-scale focusing characteristics of wavelet transform to attempt to capture abrupt changes in voltage signals. However, existing technologies have significant limitations: they typically pre-select a single wavelet basis function (such as the Haar or Daubechies system) for full-time analysis. In solid-state batteries, due to differences in solid electrolyte materials (oxides, sulfides, etc.) and dendrite growth morphologies (grain boundary penetration or bulk cracking), the voltage drop waveform characteristics (such as fall edge steepness and rebound damping oscillation morphology) generated during micro-short circuits exhibit significant uncertainty and diversity. A fixed mother wavelet cannot achieve optimal matching with the diverse fault signal morphology, resulting in a low signal-to-noise ratio for feature extraction. This makes it highly susceptible to missed detection of micro-short circuit signals when waveform mismatches occur, failing to meet the testing requirements for high safety levels. Summary of the Invention

[0004] This invention provides a data analysis-based method and system for testing the performance of solid-state batteries. It aims to solve the problem in related technologies where fixed mother wavelets cannot achieve optimal matching with various types of fault signals, resulting in low signal-to-noise ratio of feature extraction and easy failure to detect micro-short-circuit signals when waveforms are mismatched, thus failing to meet the requirements of high-safety-level testing.

[0005] In a first aspect, the present invention provides a data analysis-based method for testing the performance of solid-state batteries, comprising: acquiring voltage time-series data of a solid-state battery and removing the DC component to obtain a fluctuating voltage sequence; calculating a local transient response index at each time step based on the ratio of the second-order difference to the first-order difference and the signal-to-noise ratio of the fluctuating voltage sequence, used to characterize the local steepness of voltage fluctuations; generating a deformable wavelet basis function, wherein the shape of the deformable wavelet basis function is determined by a morphological control parameter, the morphological control parameter being proportional to the local transient response index; performing a wavelet transform on the fluctuating voltage sequence based on the deformable wavelet basis function to obtain corresponding wavelet coefficients; calculating a morphologically matched micro-short-circuit feature value, the calculation of the morphologically matched micro-short-circuit feature value being based on the wavelet coefficients and the difference between the morphological control parameter and a standard morphological control parameter; comparing the morphologically matched micro-short-circuit feature value with a dynamic threshold to determine whether a micro-short circuit has occurred. By calculating the local transient response index characterizing the steepness of voltage fluctuations, a morphologically variable deformable wavelet basis function is dynamically generated, and the micro-short-circuit feature value is calculated in conjunction with the morphological matching degree. This method can adaptively adjust the wavelet basis shape according to the real-time physical characteristics of the signal (such as sharpness), achieving the best match with the fault signal, significantly improving the ability to capture weak short-circuit signals and the detection accuracy, and effectively solving the problem of missed detection caused by waveform mismatch.

[0006] Furthermore, the method for calculating the local transient response exponent includes: multiplying the ratio of the second-order difference to the first-order difference of the fluctuating voltage sequence by an exponential term based on the signal-to-noise ratio of the fluctuating voltage, and performing a hyperbolic tangent operation on the product. By combining the ratio of the second-order difference to the first-order difference (characterizing curvature and acceleration) with an exponential term based on the signal-to-noise ratio for hyperbolic tangent operation, this calculation method not only captures the geometric steepness of the waveform but also amplifies the exponent when the signal has significant structural features using the signal-to-noise ratio term. This allows the local transient response exponent to sensitively reflect the microscopic physical nature of voltage fluctuations (such as the instantaneous nonlinear changes during dendrite piercing), providing precise control parameters for the subsequent generation of high-matching wavelet basis functions.

[0007] Furthermore, the formula for calculating the local transient response exponent is as follows: In the formula, Indicating the fluctuating voltage sequence The local transient response exponent at time t; and These represent the first and second differences of the fluctuating voltage sequence, respectively. To prevent the minimum value where the denominator is zero, for Voltage fluctuations at any given moment The standard deviation of ambient background noise. Represents the hyperbolic tangent function. It is an exponential function with the natural constant e as its base. This specific calculation formula can quantify the local sharpness and abrupt nonlinearity of the signal, effectively distinguish between gentle impedance changes and sharp dendrite piercing characteristics, and ensure that the parameter adjustment of the adaptive wavelet transform has clear physical meaning and mathematical robustness.

[0008] Furthermore, the standard deviation of the ambient background noise is obtained by extracting a segment of fluctuating voltage sequence at the initial moment of the test or during the battery's resting phase, and calculating the standard deviation of this fluctuating voltage sequence as the standard deviation of the ambient background noise. This step provides a clean reference for subsequent signal-to-noise ratio calculations and threshold settings, avoiding contamination of the noise reference by load changes or interference signals, thereby improving the reliability of micro-short circuit detection.

[0009] Furthermore, the method for calculating the morphological matching micro-short-circuit eigenvalue includes multiplying the wavelet coefficients by a penalty term characterizing the closeness between the current morphological control parameter and the standard morphological control parameter. A penalty term is introduced when calculating the micro-short-circuit eigenvalue, taking into account the closeness between the current waveform morphological parameters and the standard morphological parameters. If the signal energy is high but the morphology does not conform to typical short-circuit characteristics, the eigenvalue will be reduced by the penalty term. This mechanism effectively eliminates non-short-circuit interference signals (such as mechanical vibrations) and significantly reduces the false alarm rate.

[0010] Furthermore, the standard morphological control parameters are obtained by averaging multiple morphological control parameters corresponding to historical micro-short circuit events. By statistically analyzing historical data, the system can establish the most typical fault waveform model for a specific battery system, thereby providing a more representative reference benchmark in morphological matching calculations and further improving the accuracy of identifying real short circuit events.

[0011] Furthermore, the dynamic threshold is calculated based on the mean and standard deviation of multiple morphologically matched micro-short circuit feature values ​​within a preset time period prior to the current moment.

[0012] Furthermore, when the duration for which the morphological matching micro-short circuit feature value is greater than the dynamic threshold is between 3 and 100 sampling points, a micro-short circuit is determined to have occurred.

[0013] Furthermore, acquiring voltage timing data of the solid-state battery includes: sampling the voltage of the solid-state battery at a sampling rate of not less than 10 kHz.

[0014] In a second aspect, the present invention also provides a solid-state battery performance testing system based on data analysis, comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the solid-state battery performance testing method based on data analysis described in any of the above embodiments.

[0015] Beneficial effects: To address the issue of diverse waveforms in solid-state batteries' micro-short circuits that are easily masked by noise, a local transient response index is constructed to quantify the steepness of voltage fluctuations in real time. Based on this, a deformable wavelet basis function that closely matches the current signal shape is dynamically generated. Combined with an eigenvalue calculation method that introduces a shape proximity penalty term, accurate matching and enhancement of weak short-circuit signals are achieved, effectively eliminating non-fault interference and significantly improving the signal-to-noise ratio and accuracy of detection. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating a short-circuit test flowchart for a solid-state battery according to an embodiment of the present invention. Detailed Implementation

[0017] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] like Figure 1 As shown, S101: High-frequency voltage data acquisition and preprocessing.

[0019] During the short-circuit test of the solid-state battery, the battery's voltage timing data was acquired in real time using a high-precision data acquisition card at a high sampling rate of 10kHz. The raw voltage data was then detrended to remove DC component drift caused by battery discharge or rest, resulting in a fluctuating voltage containing only high-frequency dynamic characteristics.

[0020] S102: Construct the local transient response exponent at each time step.

[0021] To address the issue of fixed wavelet bases being unable to adapt to varying waveforms, it is first necessary to analyze the physical nature of voltage fluctuations at the current moment from a data perspective. In solid-state batteries, dendrite piercing the electrolyte is a microscopic process of conduction-melting-reconduction, which manifests as a specific decay rate and curvature change in the voltage signal.

[0022] A local transient response exponent is constructed to characterize the steepness of the microscopic morphology of voltage fluctuations. Since dendrite penetration causes a dramatic step change in local resistance, resulting in a nonlinear exponential voltage change rather than a simple linear decrease, this local transient response exponent is constructed by analyzing the ratio of the second-order difference to the first-order difference of the voltage fluctuation. In the formula, Indicating the fluctuating voltage sequence The local transient response index at time t is a quantification of the sharpness and abrupt nonlinearity of the signal at a given time. and Let these represent the first and second differences of the fluctuating voltage sequence, respectively. The fluctuating voltage sequence is: The sequence of voltage fluctuations at time 1 and the previous N time points reflects the rate and acceleration of voltage change. In this embodiment, the value of N ranges from 5 to 20, and preferably, the value of N is 10. To prevent the minimum value where the denominator is zero, The value is ; for Voltage fluctuations at any given moment; To calculate the standard deviation of the ambient background noise, specifically, a segment of the fluctuating voltage sequence is extracted at the initial moment of the test or during the battery's resting phase, for example, the first 1000 sampling points, and the standard deviation of this fluctuating voltage sequence is obtained. This value represents the inherent noise level of the measurement system and environment, serving as a benchmark for determining whether subsequent voltage surges are significant. This represents the hyperbolic tangent function.

[0023] The above formula first calculates the ratio of acceleration to velocity of the voltage change to capture the curvature characteristics of the signal waveform, i.e., whether the waveform declines gently or drops sharply. Further, an exponential term with the signal-to-noise ratio as the independent variable is introduced to amplify the exponent when the signal amplitude is weak but has significant structural characteristics. When A larger value indicates The voltage fluctuation at any given moment exhibits extremely high transient nonlinearity, corresponding to sharp dendrite piercing characteristics; when the value is small, it corresponds to a gradual impedance change.

[0024] S103: Generate the deformation wavelet basis functions at each time step.

[0025] Based on the local transient response exponents obtained in the above steps, a non-standard wavelet basis function with variable parameters is constructed. Existing techniques use fixed wavelet bases, which cannot take into account the signal characteristics under different dendrite growth modes. This step constructs a generalized Gaussian derivative wavelet with a shape factor, and dynamically adjusts the wavelet shape using the local transient response exponents at each time step.

[0026] The morphological control parameters at each time step are constructed, and the deformation wavelet basis functions at each time step are generated accordingly. The formulas for calculating the morphological control parameters at each time step are as follows: In the formula, According to The morphological control parameters are calculated from the signal characteristics at a given time. Let be the maximum fundamental morphological constant, for example, 8; the formula for calculating the deformation wavelet basis function at each time step is: ; for The wavelet basis functions generated at each time step, For normalized time variables; This is a normalization constant to ensure that the energy of the generated wavelet basis function is always 1, thus guaranteeing the comparability of wavelet transform coefficients at different times. for The morphological control parameters are calculated from the signal characteristics at a given time.

[0027] From the two formulas above, it can be seen that when an extremely steep voltage drop is detected, that is... When it increases in size, it indicates that the dendrites penetrate at an extremely high speed, producing very sharp singularities. As the value of increases, the value of the wavelet basis function decreases, causing its peaks near the center point to narrow and its decay to accelerate, thus generating a sharp wavelet basis that can achieve maximum cross-correlation with the fast-piercing signal; conversely, when When the size is small, a wider wavelet basis is generated to match the slow contact short circuit.

[0028] S104: Weighted wavelet transform based on morphological matching.

[0029] The volatile voltage sequence is transformed using the deformation wavelet basis functions generated in step S103, and the shape matching confidence score is further introduced to calculate the final anomaly index. A large wavelet coefficient alone does not necessarily indicate a short circuit; the determination is based on the wavelet coefficient being large under the optimal matching shape.

[0030] First, the wavelet coefficients at each scale and displacement are calculated using the following formula: In the formula, In order to scale and displacement Below, the wavelet coefficients calculated using the wavelet basis functions at the corresponding time points are scaled. The value is ; Let t be the time axis of the voltage fluctuation sequence at time t. The displacement factor is the current time point in the analysis. The voltage fluctuation sequence at time t; Then, the final shape-matching micro-short circuit characteristic value under each displacement is calculated, and the calculation formula is as follows: In the formula, displacement The final morphological matching micro-short-circuit feature value; In order to scale and displacement The wavelet coefficients are calculated using the wavelet basis functions at the corresponding time points. Displacement The following are the morphological control parameters; The standard morphological control parameters corresponding to historical short-circuit waveforms can be the average value of the morphological control parameters corresponding to each historical short-circuit waveform. This is a penalty factor used to adjust the sensitivity to morphological deviations. The value of is 2.

[0031] First, the adaptive wavelet coefficients are calculated. Then, when calculating the final morphological matching micro-short-circuit eigenvalues, instead of simply taking the modulus maxima of the coefficients, a correction term is introduced, which characterizes the displacement. The degree of closeness between the lower morphological control parameters and the standard morphological control parameters corresponding to the historical short-circuit waveform.

[0032] The logical relationship of the above formula is as follows: If the algorithm automatically generates an unusual morphological control parameter to fit the current signal, it indicates that although the signal has high energy, its shape is likely due to mechanical vibration interference rather than a dendrite short circuit. Therefore, the correction term will reduce the final morphological matching micro-short circuit characteristic value. Conversely, if the adaptive morphology closely matches the typical fault morphology and the wavelet coefficients are large, the final morphological matching micro-short circuit characteristic value will increase significantly.

[0033] S105: Micro-short circuit detection and early warning.

[0034] Specifically, the numerical change of the final morphological matching micro-short-circuit feature value at the current moment is obtained, and a dynamic threshold is set. ,in and These represent the mean and standard deviation of the final morphological matching micro-short circuit characteristic values ​​for the M time steps prior to the current time step, where M ranges from 100 to 1000. The final form of the time-limited micro-short-circuit eigenvalues If the duration is between 3 and 100 sampling points, a micro short circuit is determined to have occurred inside the solid-state battery, and the safety protection mechanism of the test system is immediately triggered and the fault time is recorded. If the duration exceeds 100 sampling points, it is identified as continuous environmental oscillation or operating condition switching interference, and a short circuit alarm is not triggered.

[0035] The present invention also provides a solid-state battery performance testing system based on data analysis. The system includes a processor and a memory, the memory storing computer program instructions. When the computer program instructions are executed by the processor, they implement the solid-state battery performance testing method based on data analysis according to the first aspect of the present invention.

[0036] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0037] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0038] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for testing performance of a solid-state battery based on data analysis, characterized by, The method comprises: obtaining voltage time series data of a solid-state battery, and removing a direct current component to obtain a fluctuation voltage sequence; calculating a local transient response index at each time point based on a ratio of a second-order difference to a first-order difference of the fluctuation voltage sequence and a signal-to-noise ratio, the local transient response index being used to represent a local steepness of voltage fluctuation; generating a morphing wavelet basis function, wherein a shape of the morphing wavelet basis function is determined by a morphological control parameter, and the morphological control parameter is in proportional relationship with the local transient response index; performing wavelet transform on the fluctuation voltage sequence based on the morphing wavelet basis function to obtain corresponding wavelet coefficients; calculating a morphologically matched micro-short circuit feature value, the calculation of the morphologically matched micro-short circuit feature value being based on a difference between the wavelet coefficients and the morphological control parameter and a standard morphological control parameter; and comparing the morphologically matched micro-short circuit feature value with a dynamic threshold to determine whether a micro-short circuit occurs.

2. The data analysis based solid state battery performance test method of claim 1, wherein, The method for calculating the local transient response index comprises: multiplying a ratio of a second-order difference to a first-order difference of the fluctuation voltage sequence by an exponential term based on a signal-to-noise ratio of the fluctuation voltage, and performing a hyperbolic tangent operation on a product result.

3. The data analysis based solid state battery performance test method of claim 1, wherein, The calculation formula of the local transient response index is: where, denotes the local transient response index at time t, and denote the first and second order difference of the fluctuating voltage sequence, to prevent the denominator from being zero, is the fluctuating voltage at time is the standard deviation of the ambient background noise, denotes the hyperbolic tangent function, is the exponential function with base the natural constant e.

4. The data analysis based solid state battery performance test method of claim 3, wherein, The method for obtaining a standard deviation of environmental background noise comprises: extracting a fluctuation voltage sequence at a test initial time or a battery resting stage, and calculating a standard deviation of the fluctuation voltage sequence as the standard deviation of the environmental background noise.

5. The data analysis based solid state battery performance test method of claim 1, wherein, The method for calculating the morphologically matched micro-short circuit feature value comprises: multiplying the wavelet coefficients by a penalty term representing a closeness between the morphological control parameter at the current time point and the standard morphological control parameter.

6. The data analysis based solid state battery performance test method of claim 1, wherein, The standard morphological control parameter is obtained according to an average of a plurality of morphological control parameters corresponding to historical micro-short circuit events.

7. The data analysis based solid state battery performance test method of claim 1, wherein, The dynamic threshold is calculated according to a mean value and a standard deviation of a plurality of morphologically matched micro-short circuit feature values within a preset time period before the current time point.

8. The data analysis based solid state battery performance testing method of claim 7, wherein, When the morphologically matched micro-short circuit feature value is greater than the dynamic threshold for a duration of 3 sampling points to 100 sampling points, it is determined that a micro-short circuit occurs.

9. The data analysis based solid state battery performance test method of claim 1, wherein, The method for obtaining voltage time series data of a solid-state battery comprises: collecting the voltage of the solid-state battery at a sampling rate of no less than 10 kHz.

10. A data analytics based solid state battery performance testing system comprising a processor and a memory, wherein, The memory stores a computer program, and the processor executes the computer program to implement the data analysis-based solid-state battery performance test method according to any one of claims 1-9.

Images