Battery short circuit test device and detection method

Abstract

The invention relates to the technical field of batteries, and provides a battery short-circuit testing device and a battery short-circuit detection method, and the method comprises the steps: obtaining a current signal and a voltage signal in a battery operation process, carrying out the de-noising processing, obtaining a de-noised current signal and a de-noised voltage signal, and constructing a current and voltage dynamic characteristic spectrum; generating a short-time current and voltage characteristic mapping matrix, and performing hierarchical clustering analysis on the short-time current and voltage characteristic mapping matrix to obtain a short-circuit evaluation result; and comparing the short-circuit assessment result with a preset risk degree, and if the short-circuit trend of the short-circuit assessment result exceeds a preset safety range of the risk degree, generating a short-circuit alarm signal. The current signal and the voltage signal are processed and analyzed, the detection threshold value is dynamically adjusted to adapt to different use environments, the detection sensitivity and accuracy are enhanced, and the problems that the threshold value adaptability is poor, and the false alarm or missing alarm phenomenon is likely to be caused when the battery is in different use environments and battery types are solved.

Description

Technical Field

[0001] This application relates to the technical field of batteries, and in particular to a battery short - circuit test device and a detection method. Background Art

[0002] With the rapid development of the new energy field, lithium - ion batteries have been widely used in fields such as electric vehicles, energy storage systems, and mobile electronic devices. However, during long - term operation, batteries may be affected by factors such as external mechanical damage and internal molecular structure degradation, resulting in short - circuit phenomena. This can not only lead to a decline in device performance, but more seriously, may cause safety problems such as fires or explosions. Therefore, battery short - circuit detection technology has become an important research direction in the industry.

[0003] In related technical means, identification is carried out through current and voltage data monitoring and anomaly detection algorithms. For example, when the battery is operating, the changes in current and voltage are monitored, and by setting fixed thresholds, it is judged whether there are abnormal situations. When an anomaly is detected, an alarm signal is triggered to alert the operator. This method can achieve a rapid response to some short - circuit risks and reduce the probability of safety accidents caused by short - circuits.

[0004] For the above - mentioned technical solution, although the fixed - threshold detection method can achieve a rapid judgment of short - circuits to a certain extent, when facing different usage environments and battery types, there is a problem of poor threshold adaptability, which is prone to false alarms or missed alarms. Summary of the Invention

[0005] In order to improve the problem of poor threshold adaptability and the tendency to cause false alarms or missed alarms when facing different usage environments and battery types, this application provides a battery short - circuit test device and a detection method.

[0006] The present invention provides a method for detecting battery short circuit, including: obtaining current signals and voltage signals during the operation of the battery, and performing denoising processing on the current signals and the voltage signals to obtain denoised current signals and denoised voltage signals; constructing a current-voltage dynamic feature map based on the denoised current signals and the denoised voltage signals, generating a short-time current-voltage feature mapping matrix based on the current-voltage dynamic feature map, performing hierarchical clustering analysis on the short-time current-voltage feature mapping matrix to obtain a short-circuit feature classification data set and a short-circuit anomaly factor; using the short-circuit anomaly factor to perform dynamic threshold adjustment on the denoised current signals and the denoised voltage signals to obtain an adaptive short-circuit detection threshold, and performing real-time matching on the short-circuit feature classification data set according to the adaptive short-circuit detection threshold to obtain a short-circuit anomaly determination result; performing multi-scale time series decomposition on the short-circuit anomaly determination result to obtain a short-term trend component, a long-term trend component, and an abnormal impact component, calculating the short-term trend component, the long-term trend component, and the abnormal impact component to obtain a short-circuit evaluation result; comparing the short-circuit evaluation result with a preset risk level, and if the short-circuit trend of the short-circuit evaluation result exceeds the safe range preset by the risk level, generating a short-circuit warning signal.

[0007] As a preferred solution, the step of obtaining current signals and voltage signals during the operation of the battery, and performing denoising processing on the current signals and the voltage signals to obtain denoised current signals and denoised voltage signals includes: collecting current signals and voltage signals of the battery in an operating state through sensors, constructing a current-voltage correlation data set based on the current signals and the voltage signals, performing signal decomposition on the current-voltage correlation data set to obtain a fluctuation component and a pulse component; calculating a short-time fluctuation trend according to the fluctuation component, analyzing current fluctuation characteristics based on the short-time fluctuation trend, and constructing a current transient feature mapping using the pulse component; performing noise reduction filtering on the current signals and the voltage signals based on the current fluctuation characteristics and the current transient feature mapping to obtain denoised current signals and denoised voltage signals.

[0008] As a preferred solution, the step of performing noise reduction filtering on the current signal and the voltage signal based on the current fluctuation characteristics and the current transient characteristic mapping to obtain a denoised current signal and a denoised voltage signal includes: performing frequency domain decomposition on the current fluctuation characteristics by using wavelet transform to obtain a low-frequency steady component and a high-frequency fluctuation component, performing envelope analysis on the current transient characteristic mapping by using Hilbert transform to obtain transient characteristic parameters; calculating steady-state trend parameters based on the low-frequency steady component, calculating short-time fluctuation parameters based on the high-frequency fluctuation component, and calculating pulse timing parameters based on the transient characteristic parameters; constructing a short-time current change characteristic set through the steady-state trend parameters and the short-time fluctuation parameters, performing time series analysis on the short-time current change characteristic set to obtain a short-time steady component and a short-time change component, and calculating a current denoising filter matrix based on the short-time steady component and the short-time change component; calculating a transient influence factor according to the short-time current change characteristic set and the pulse timing parameters, calculating a transient influence correction parameter through the transient influence factor, and adjusting the characteristic weight of the current denoising filter matrix by using the transient influence correction parameter to obtain a corrected current denoising filter matrix; performing wavelet decomposition on the voltage signal to obtain a low-frequency voltage component and a high-frequency voltage component, and calculating current-voltage joint denoising parameters by using the low-frequency voltage component, the high-frequency voltage component, and the current denoising filter matrix; performing joint filtering on the low-frequency voltage component and the high-frequency voltage component based on the current-voltage joint denoising parameters to obtain a filtered voltage component, and calculating a filtered current signal by using the filtered voltage component and the corrected current denoising filter matrix; comparing the characteristics of the filtered voltage component and the filtered current signal to obtain a denoised current signal and a denoised voltage signal.

[0009] As a preferred solution, the steps of constructing a current-voltage dynamic feature map based on the denoised current signal and the denoised voltage signal, generating a short-time current-voltage feature mapping matrix based on the current-voltage dynamic feature map, and performing hierarchical clustering analysis on the short-time current-voltage feature mapping matrix to obtain a short-circuit feature classification data set and a short-circuit anomaly factor include: calculating the instantaneous current change rate using the denoised current signal, calculating the instantaneous power change rate using the denoised voltage signal, and generating a short-time current-voltage change pair based on the instantaneous current change rate and the instantaneous power change rate; constructing a short-time correlation matrix based on the short-time current-voltage change pair, performing principal component analysis on the short-time correlation matrix, and generating a current-voltage time-series feature matrix based on the analysis results; performing feature aggregation on the current-voltage time-series feature matrix to obtain a short-time current change map and a short-time voltage change map, performing map mapping on the short-time current change map and the short-time voltage change map to obtain a short-time current-voltage feature mapping matrix; using the short-time current-voltage feature mapping matrix for feature screening, generating a short-circuit candidate feature set based on the screening results, performing clustering analysis on the short-circuit candidate feature set, and calculating the classification confidence by combining the analysis results with historical short-circuit data, and generating a short-circuit feature classification data set and a short-circuit anomaly factor according to the classification confidence; wherein, the historical short-circuit data is the current, voltage, temperature, and time of the battery when a short-circuit event occurs in the historical working state.

[0010] As a preferred solution, the steps of constructing a short-time correlation matrix based on the short-time current-voltage change pair, performing principal component analysis on the short-time correlation matrix, and generating a current-voltage time-series feature matrix based on the analysis results include: partitioning the short-time current-voltage change pair through a time sliding window to obtain an initial feature set, and extracting the feature clustering center of the initial feature set based on the neighborhood density estimation algorithm; calculating the current change gradient and the voltage change gradient of the area corresponding to the feature clustering center, calculating the current-voltage correlation parameter using the current change gradient and the voltage change gradient, and constructing a short-time correlation matrix based on the current-voltage correlation parameter; performing principal component analysis on the short-time correlation matrix to obtain the principal component feature vector, calculating the current change principal component and the voltage change principal component based on the principal component feature vector, constructing a current change mapping matrix using the current change principal component, and constructing a voltage change mapping matrix using the voltage change principal component; calculating the current time-series feature through the current change mapping matrix, calculating the voltage time-series feature through the voltage change mapping matrix, and performing feature matching on the current time-series feature and the voltage time-series feature to obtain a current-voltage time-series feature matrix.

[0011] As a preferred solution, the step of using the short - circuit anomaly factor to dynamically adjust the thresholds of the denoised current signal and the denoised voltage signal to obtain an adaptive short - circuit detection threshold, and performing real - time matching on the short - circuit feature classification data set according to the adaptive short - circuit detection threshold to obtain a short - circuit anomaly determination result includes: obtaining the historical distribution of the short - circuit anomaly factor, calculating the anomaly factor drift rate of the short - circuit anomaly factor based on the historical distribution, generating an anomaly factor time - evolution model according to the anomaly factor drift rate, using the anomaly factor time - evolution model to calculate the anomaly sensitivity of the short - circuit feature classification data set, and performing hierarchical screening based on the anomaly sensitivity to obtain a short - circuit feature screening set and a short - circuit anomaly feature set; performing statistical analysis on the short - circuit feature screening set to obtain a feature stability parameter, calculating the feature anomaly intensity using the feature stability parameter and the short - circuit anomaly feature set, dynamically correcting the weights in the short - time current - voltage feature mapping matrix according to the feature anomaly intensity to obtain a short - circuit feature correction matrix; adjusting a preset short - circuit detection threshold based on the short - circuit feature correction matrix to obtain an adaptive short - circuit detection threshold, using the adaptive short - circuit detection threshold to perform real - time matching on the short - circuit feature classification data set to obtain a short - circuit matching error matrix; performing an anomaly trend analysis on the short - circuit matching error matrix, and calculating an anomaly determination value based on the analysis result in combination with the short - circuit anomaly factor, and generating a short - circuit anomaly determination result according to the anomaly determination value.

[0012] As a preferred solution, the step of performing multi - scale time - series decomposition on the short - circuit anomaly determination result to obtain a short - term trend component, a long - term trend component, and an anomaly shock component, and calculating a short - circuit evaluation result by calculating the short - term trend component, the long - term trend component, and the anomaly shock component includes: constructing a short - circuit time series according to the short - circuit anomaly determination result, calculating the frequency and intensity of the occurrence of short - circuit events based on the short - circuit time series, and generating a short - circuit time - evolution feature according to the frequency and intensity of the occurrence of short - circuit events; calculating the short - circuit influence range using the short - circuit time - evolution feature, performing short - circuit trend correction on the short - circuit feature correction matrix according to the short - circuit influence range to obtain short - circuit correction trend data, and performing multi - scale time - series decomposition on the short - circuit correction trend data to obtain a short - term trend component, a long - term trend component, and an anomaly shock component; calculating a short - term change rate according to the short - term trend component, calculating an accumulated offset according to the long - term trend component, and generating a short - circuit trend change parameter based on the short - term change rate and the accumulated offset; calculating the short - circuit future change probability using the short - circuit trend change parameter and the anomaly shock component, performing trend fusion on the short - circuit future change probability and the short - circuit feature classification data set to obtain short - circuit trend prediction data, calculating a short - circuit safety boundary using the short - circuit trend prediction data, and performing risk matching on the short - circuit safety boundary to obtain a short - circuit evaluation result.

[0013] The present application also provides a battery short-circuit test device, including: an acquisition module, configured to acquire current signals and voltage signals during the operation of the battery, and perform denoising processing on the current signals and the voltage signals to obtain denoised current signals and denoised voltage signals; a calculation module, configured to construct a current-voltage dynamic characteristic map based on the denoised current signals and the denoised voltage signals, generate a short-time current-voltage characteristic mapping matrix based on the current-voltage dynamic characteristic map, and perform hierarchical clustering analysis on the short-time current-voltage characteristic mapping matrix to obtain a short-circuit characteristic classification data set and a short-circuit anomaly factor; an adjustment module, configured to perform dynamic threshold adjustment on the denoised current signals and the denoised voltage signals by using the short-circuit anomaly factor to obtain an adaptive short-circuit detection threshold, and perform real-time matching on the short-circuit characteristic classification data set according to the adaptive short-circuit detection threshold to obtain a short-circuit anomaly determination result; a decomposition module, configured to perform multi-scale time series decomposition on the short-circuit anomaly determination result to obtain a short-term trend component, a long-term trend component, and an anomaly impact component, and calculate the short-term trend component, the long-term trend component, and the anomaly impact component to obtain a short-circuit evaluation result; a generation module, configured to compare the short-circuit evaluation result with a preset risk level, and if the short-circuit trend of the short-circuit evaluation result exceeds the safe range preset by the risk level, generate a short-circuit warning signal.

[0014] Compared with the prior art, the present application has the following beneficial effects: strong adaptability. By combining technologies such as denoising processing of current signals and voltage signals, construction and calculation of current-voltage dynamic characteristic maps, hierarchical clustering analysis, adaptive dynamic threshold adjustment, and multi-scale time series decomposition, the short-circuit characteristics of the battery can be analyzed from multiple perspectives, the detection threshold can be dynamically adjusted to adapt to different usage environments, the sensitivity and accuracy of detection can be enhanced, and potential short-circuit risks can be effectively prevented through the precise comparison of short-circuit evaluation results and real-time warning functions, improving the safety and reliability during the use of the battery, and solving the problems of poor threshold adaptability, easy false alarms or missed alarms when facing different usage environments and battery types. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] The structures, proportions, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the conditions for the implementation of the present invention. Therefore, they do not have substantial technical significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.

[0017] Figure 1 It is a schematic flowchart of the battery short - circuit detection method provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of the battery short - circuit test device provided by an embodiment of the present invention.

[0018] Explanation of reference numerals: 10. Battery short - circuit test device; 11. Acquisition module; 12. Calculation module; 13. Adjustment module; 14. Decomposition module; 15. Generation module. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] The flowcharts shown in the drawings are only illustrative examples, and do not necessarily include all contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can also be decomposed, combined, or partially merged. Therefore, the actual execution order may change according to the actual situation.

[0021] It should also be understood that the terms used in this specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification of the present application and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0022] It should be further understood that the term " / and" used in this specification of the present application and the appended claims refers to any combination and all possible combinations of one or more of the related listed items, and includes these combinations.

[0023] Next, the technical solutions of the present invention will be further described in conjunction with the drawings and through specific implementation manners.

[0024] Embodiment 1: As shown Figure 1 in the figure, the present application provides a battery short - circuit detection method, including steps S100 to S500.

[0025] Step S100: Obtain the current signal and voltage signal during the operation of the battery, and perform denoising processing on the current signal and voltage signal to obtain a denoised current signal and a denoised voltage signal.

[0026] In this step, a high - precision sensor is used to collect the current signal and voltage signal during the operation of the battery in real - time, and the noise in the original signal is eliminated through a denoising algorithm. Specifically, the wavelet transform method is used to perform multi - scale decomposition on the current signal and voltage signal, soft - threshold filtering is performed on the high - frequency components of each decomposition layer, and then inverse transformation is used for reconstruction to obtain the denoised current signal and the denoised voltage signal.

[0027] For example, when monitoring the operation process of a certain type of battery, the original current signal fluctuates greatly. After wavelet transform denoising processing, the signal becomes smooth and the features are more obvious, which is convenient for subsequent analysis.

[0028] Step S200: Construct a current - voltage dynamic feature map based on the denoised current signal and the denoised voltage signal, generate a short - time current - voltage feature mapping matrix based on the current - voltage dynamic feature map, and perform hierarchical clustering analysis on the short - time current - voltage feature mapping matrix to obtain a short - circuit feature classification data set and a short - circuit anomaly factor.

[0029] In this step, the denoised current signal and the denoised voltage signal are used to construct a current - voltage dynamic feature map reflecting time - series characteristics, and key parameters in the dynamic feature map are calculated through statistical analysis. Specifically, the sliding window technique is used to segment the time period of the map, generate a short - time current - voltage feature mapping matrix, and perform hierarchical clustering analysis on the eigenvalues in the matrix, extract relevant features to form a short - circuit feature classification data set, and at the same time calculate the short - circuit anomaly factor.

[0030] For example, in an experiment, the dynamic feature map is segmented into multiple time periods through the sliding window technique. In the short - time current - voltage feature mapping matrix generated for each segment, the part with significant anomaly factors corresponds to the early - stage short - circuit trend characteristics, indicating that the clustering analysis effect is obvious.

[0031] Step S300: Use the short - circuit anomaly factor to perform dynamic threshold adjustment on the denoised current signal and the denoised voltage signal to obtain an adaptive short - circuit detection threshold, and perform real - time matching on the short - circuit feature classification data set according to the adaptive short - circuit detection threshold to obtain a short - circuit anomaly determination result.

[0032] In this step, based on the short - circuit anomaly factor, the denoised current signal and the detection threshold of the denoised current signal are dynamically adjusted through an algorithm model to be adaptively optimized according to the environmental conditions and signal characteristics. Specifically, through the joint analysis of historical data and real - time data, an adaptive short - circuit detection threshold is generated, and real - time comparison is made with the short - circuit feature classification data set to identify the short - circuit anomaly determination result.

[0033] For example, under different temperature conditions, historical data shows that a fixed threshold is difficult to adapt to changes, while the dynamically adjusted adaptive short - circuit detection threshold significantly improves the detection accuracy of short - circuit anomalies.

[0034] Step S400: Perform multi - scale time - series decomposition on the short - circuit anomaly determination result to obtain a short - term trend component, a long - term trend component, and an abnormal shock component, and calculate the short - term trend component, the long - term trend component, and the abnormal shock component to obtain a short - circuit evaluation result.

[0035] In this step, multi - scale time - series decomposition technology is used to refine the analysis of the short - circuit anomaly determination result, which is decomposed into a short - term trend component, a long - term trend component, and an abnormal shock component, and quantitative calculations are performed on each component. Finally, a comprehensive short - circuit evaluation result is obtained. Specifically, based on the eigenvalue of each decomposed component, the severity of the short - circuit trend is evaluated.

[0036] For example, through the decomposition analysis of a specific set of data, the short - term trend component reflects the characteristics of sudden changes, the long - term trend component indicates the gradually accumulating short - circuit risk, and the abnormal shock component is directly related to sudden abnormal events.

[0037] Step S500: Compare the short - circuit evaluation result with the preset risk level. If the short - circuit trend of the short - circuit evaluation result exceeds the safe range preset by the risk level, a short - circuit warning signal is generated.

[0038] In this step, according to the risk assessment model, the short - circuit evaluation result is compared item by item with the safe range of the preset risk level to determine whether its short - circuit trend reaches a dangerous level. Specifically, the out - of - range situation is converted into a warning signal output through a logic determination module, and relevant protection devices are linked to respond.

[0039] For example, in an application, the short - circuit evaluation result shows that the short - circuit trend significantly exceeds the safe range, and the system quickly generates a short - circuit warning signal and triggers an automatic power - off measure, avoiding potential accidents.

[0040] In this embodiment, by acquiring the current signal and voltage signal during the operation of the battery, and performing denoising processing on the current signal and voltage signal, a denoised current signal and a denoised voltage signal are generated. Then, based on the denoised current signal and the denoised voltage signal, a current-voltage dynamic feature map is constructed, and calculations are performed on the current-voltage dynamic feature map to generate a short-term current-voltage feature mapping matrix. Next, hierarchical clustering analysis is performed on the short-term current-voltage feature mapping matrix to obtain a short-circuit feature classification data set and a short-circuit anomaly factor. The denoised current signal is dynamically threshold-adjusted using the short-circuit anomaly factor to generate an adaptive short-circuit detection threshold, and the short-circuit feature classification data set is compared in real time according to the adaptive short-circuit detection threshold to obtain a short-circuit anomaly determination result. Finally, by performing multi-scale time series decomposition on the short-circuit anomaly determination result, a short-term trend component, a long-term trend component, and an anomaly impact component are obtained, and a short-circuit evaluation result is generated in combination with the calculation results. The short-circuit evaluation result is compared with a preset risk level. If it is found that the short-circuit trend exceeds the preset safety range, a short-circuit warning signal is generated.

[0041] By performing denoising processing on the current signal and voltage signal, more accurate denoised current signal and denoised voltage signal can be obtained, thereby improving the detection accuracy. At the same time, based on the analysis of the current-voltage dynamic feature map and the short-term current-voltage feature mapping matrix, the dynamic change characteristics of the battery short circuit can be fully captured. Through hierarchical clustering analysis and real-time adjustment of the adaptive short-circuit detection threshold, the sensitivity and adaptability of short-circuit detection can be significantly improved. In addition, the multi-scale time series decomposition method can not only identify short-term anomalies, but also analyze long-term trends and anomaly impacts, effectively avoiding false alarms or missed alarms, and improving the problem of poor threshold adaptability and easy occurrence of false alarms or missed alarms in the face of different usage environments and battery types.

[0042] Embodiment 2: In step S100, a current signal and a voltage signal of the battery in the operating state are collected by sensors, and a current-voltage correlation data set is constructed based on the current signal and the voltage signal. The current-voltage correlation data set is signal-decomposed to obtain a fluctuation component and a pulse component.

[0043] The current signal and the voltage signal are collected in real time during the actual operation of the battery by high-precision current sensors and voltage sensors, and the collected original signals are input into a data preprocessing module to construct a current-voltage correlation data set. Specifically, a data feature extraction algorithm is used to perform time partitioning on the current signal and the voltage signal, and the signals are frequency-domain decomposed by discrete Fourier transform (DFT) to separate the fluctuation component and the pulse component, thereby completely restoring the dynamic change characteristics of the current and voltage.

[0044] For example, for a running battery sample, after discrete Fourier transform, it is found that the fluctuation component reflects the periodic fluctuation characteristics in the current signal, and the pulse component clearly captures the rapid response peak of the current transient change.

[0045] Calculate the short-term fluctuation trend based on the fluctuation component, analyze the current fluctuation characteristics based on the short-term fluctuation trend, and construct a current transient feature map using the pulse component.

[0046] By performing time series fitting on the fluctuation component, calculate the short-term fluctuation trend, and compare the fluctuation trend with the original data to extract the fluctuation characteristics of the current signal. Specifically, perform short-term windowing processing on the fluctuation component through a moving average algorithm, and use a polynomial fitting method to generate a fluctuation trend curve; at the same time, process the pulse component through a sparse modeling method to construct a current transient feature map to capture the high dynamic change points of the signal.

[0047] For example, in practical applications, through the extraction of current fluctuation characteristics, it is found that a significant increase in fluctuations in a certain area indicates a load mutation, and the feature map constructed using the pulse component effectively locates the abnormal position of the instantaneous high current rise.

[0048] Perform noise reduction filtering on the current signal and voltage signal based on the current fluctuation characteristics and the current transient feature map to obtain a denoised current signal and a denoised voltage signal.

[0049] Through the constructed current fluctuation characteristics and current transient feature map, use a joint filtering algorithm to perform noise reduction processing on the current signal and voltage signal. Specifically, perform multi-scale decomposition on the current fluctuation characteristics using wavelet transform to extract the low-frequency stationary component, perform amplitude envelope analysis on the transient feature map using Hilbert transform to generate transient feature parameters; combine these parameters into a multi-input filter model to achieve signal noise reduction.

[0050] For example, after being processed by the multi-input filter, the voltage signal that was originally unable to clearly show a change pattern due to noise pollution is significantly purified, and both the fluctuation characteristics and the transient response are clearer.

[0051] Among them, the step of performing noise reduction filtering on the current signal and voltage signal based on the current fluctuation characteristics and the current transient feature map to obtain a denoised current signal and a denoised voltage signal includes: performing frequency domain decomposition on the current fluctuation characteristics using wavelet transform to obtain a low-frequency stationary component and a high-frequency fluctuation component, and performing envelope analysis on the current transient feature map using Hilbert transform to obtain transient feature parameters.

[0052] By performing wavelet decomposition on the current fluctuation characteristics, the signal is separated in the frequency domain into a low-frequency steady component and a high-frequency fluctuation component. A smoothing filter is applied to the low-frequency part, and a noise suppression algorithm is applied to the high-frequency part. Specifically, the Hilbert transform is used to analyze the envelope of the transient feature map to extract peak information and amplitude distribution, and key transient feature parameters are generated for noise reduction.

[0053] For example, through the Hilbert transform, the instantaneous characteristics of sharp changes in a certain section of the current signal are captured, and the time point where the corresponding amplitude is the largest reveals the time when a potential abnormal event occurs, providing a basis for subsequent evaluation.

[0054] The steady-state trend parameters are calculated based on the low-frequency steady component, the short-term fluctuation parameters are calculated based on the high-frequency fluctuation component, and the pulse timing parameters are calculated based on the transient feature parameters.

[0055] By performing trend analysis on the low-frequency steady component, a fitting algorithm is used to calculate the steady-state trend parameters of current and voltage, and the frequency-domain characteristics are extracted from the high-frequency fluctuation component to calculate the short-term fluctuation parameters; at the same time, by performing statistical analysis on the transient feature parameters, the pulse timing parameters at each time point are calculated. Specifically, a low-pass filter is used to extract the steady component and fit to generate a trend curve, the amplitude-frequency characteristics of high-frequency fluctuations are analyzed by the fast Fourier transform (FFT), and the pulse distribution is statistically analyzed in combination with the transient envelope characteristics.

[0056] For example, in the sample data, the trend curve of the low-frequency steady component shows a gradually increasing trend of current over time, while the high-frequency fluctuation component reflects the characteristics of instantaneous load changes, and the pulse frequency of the transient envelope is concentrated in the peak interval of the system operating current.

[0057] A short-term current change feature set is constructed through the steady-state trend parameters and the short-term fluctuation parameters. Time series analysis is performed on the short-term current change feature set to obtain a short-term steady component and a short-term change component, and a current denoising filter matrix is calculated based on the short-term steady component and the short-term change component.

[0058] By jointly mapping the steady-state trend parameters and the short-term fluctuation parameters, a short-term current change feature set is constructed, and time series decomposition is performed on the feature set to obtain a short-term steady component and a short-term change component. Specifically, a sliding window method is used to partition the feature set, the steady and change characteristics are extracted by the difference algorithm, and a current denoising filter matrix is constructed by the feature projection method to optimize the representation of the current characteristics in each partition.

[0059] For example, for the short-term current change feature set, time series analysis reveals significant fluctuations in the short-term change component in a certain stage, indicating that the battery is disturbed by an external load, while the steady component always remains within the safe operating trend range.

[0060] Calculate the transient impact factor based on the short - time current change feature set and pulse timing parameters. Calculate the transient impact correction parameter through the transient impact factor, and use the transient impact correction parameter to adjust the feature weights of the current denoising filter matrix to obtain the corrected current denoising filter matrix.

[0061] By statistically analyzing the correlation between the short - time current change feature set and pulse timing parameters, calculate the transient impact factor and apply it to calculate the transient impact correction parameter for adjusting the feature weights in the current denoising filter matrix. Specifically, use principal component analysis to perform feature stratification on the transient impact factor and feature set, and assign higher weight correction factors to high - dynamic regions to optimize the denoising performance of the matrix.

[0062] For example, during the analysis, it is found that the regions with higher transient impact correction parameters are concentrated near the load switching points. The adjusted current denoising filter matrix can more accurately retain these key characteristics, significantly improving the sensitivity of anomaly capture.

[0063] Perform wavelet decomposition on the voltage signal to obtain the low - frequency voltage component and high - frequency voltage component, and calculate the current - voltage joint denoising parameter using the low - frequency voltage component, high - frequency voltage component, and current denoising filter matrix.

[0064] By performing wavelet decomposition on the voltage signal, it is divided into the low - frequency voltage component and high - frequency voltage component. Combining with the corrected current denoising filter matrix, calculate the joint denoising parameter that describes the dynamic characteristics between current and voltage. Specifically, use discrete wavelet transform to extract signal components and adjust the voltage noise suppression coefficient using the weight characteristics in the current matrix to enhance the joint denoising effect.

[0065] For example, after applying this method to the operation data of a certain battery, the smoothness of the low - frequency voltage component is significantly enhanced, while the noise interference of the high - frequency voltage component is effectively suppressed, demonstrating the synergistic effect of the current - voltage joint denoising parameter.

[0066] Perform joint filtering on the low - frequency voltage component and high - frequency voltage component based on the current - voltage joint denoising parameter to obtain the filtered voltage component, and calculate the filtered current signal using the filtered voltage component and the corrected current denoising filter matrix.

[0067] By applying the current - voltage joint denoising parameter to the joint filtering process of the low - frequency voltage component and high - frequency voltage component, generate the filtered voltage component, and calculate the filtered current signal in combination with the corrected current denoising filter matrix. Specifically, use a weighted linear filter to filter the voltage component and optimize the filtered current signal through a feature - matching algorithm to ensure the integrity and noise reduction quality of the signal.

[0068] For example, the filtered voltage component exhibits good smoothing characteristics, while the filtered current signal also retains key dynamic change characteristics during sudden load changes, significantly reducing the signal distortion degree.

[0069] By comparing the characteristics of the filtered voltage component and the filtered current signal, a denoised current signal and a denoised voltage signal are obtained.

[0070] By performing multi-dimensional feature comparison on the filtered voltage component and current signal, and using the consistency evaluation results thereof to generate the final denoised current signal and denoised voltage signal. Specifically, a feature cross-validation algorithm is adopted to perform synchronous analysis of the filtered signal in the time domain and frequency domain to ensure the reliability of the original characteristics of the signal after denoising.

[0071] For example, through feature comparison, the dynamic characteristics of the denoised current signal are clearer, and the stability of the denoised voltage signal is significantly improved, verifying that the quality of the signal after comparison can meet the detection requirements.

[0072] In step S200, the instantaneous current change rate is calculated using the denoised current signal, the instantaneous power change rate is calculated using the denoised voltage signal, and a short-term current-voltage change pair is generated based on the instantaneous current change rate and the instantaneous power change rate.

[0073] By extracting the time derivative of the denoised current signal, the instantaneous current change rate is calculated, and the instantaneous power change rate is calculated by multiplying the voltage signal and the denoised current signal. Specifically, the instantaneous current change rate is solved by a difference algorithm, and the instantaneous power change rate is calculated by performing point-by-point multiplication on the synchronous sampling data of the current and voltage. By combining the instantaneous current change rate and the instantaneous power change rate, a short-term current-voltage change pair is constructed to form a feature basis for describing the dynamic change relationship between current and power.

[0074] For example, during the operation of a specific battery, when the current fluctuates rapidly, the instantaneous current change rate shows high-frequency oscillation, and at the same time, the power change rate exhibits a significant peak during this period, reflecting the significant characteristics of short-term load changes inside the battery.

[0075] Based on the short-term current-voltage change pair, a short-term correlation matrix is constructed, principal component analysis is performed on the short-term correlation matrix, and a current-voltage time series feature matrix is generated based on the analysis results.

[0076] By performing matrix operation on the data sequence of the short-term current-voltage change pair, using rows and columns to represent time and feature dimensions, a short-term correlation matrix is constructed. Specifically, standardization processing is adopted to eliminate unit differences, and principal component analysis (PCA) is used to reduce the dimension of the correlation matrix, extract the most characteristic principal component vectors, and generate a current-voltage time series feature matrix.

[0077] For example, in the experiment, the time-series feature matrix after PCA analysis retained more than 90% of the signal energy, demonstrating the dominant role of the main features in the short-term dynamic changes of the battery, while the secondary noise features were effectively filtered out.

[0078] Feature aggregation is performed on the current-voltage time-series feature matrix to obtain a short-term current change map and a short-term voltage change map. The short-term current change map and the short-term voltage change map are subjected to map mapping to obtain a short-term current-voltage feature mapping matrix.

[0079] By performing clustering analysis on the current and voltage features in the current-voltage time-series feature matrix respectively, a short-term current change map and a short-term voltage change map are generated. Specifically, the density peak clustering algorithm is used to cluster and group the data points in the feature matrix to capture the change trend. The current change map and the voltage change map are mapped based on the time axis to generate a short-term current-voltage feature mapping matrix reflecting their dynamic correlation.

[0080] For example, in a certain short-term window, the current change map shows multiple concentrated change intervals, and the feature mapping matrix formed after mapping with the voltage change map presents obvious high-correlation regions, indicating the common influence of load fluctuations on the current and voltage.

[0081] The short-term current-voltage feature mapping matrix is used for feature screening. Based on the screening results, a short-circuit candidate feature set is generated. Clustering analysis is performed on the short-circuit candidate feature set, and the classification confidence is calculated by combining the analysis results with historical short-circuit data. The short-circuit feature classification data set and the short-circuit anomaly factor are generated according to the classification confidence. Among them, the historical short-circuit data are the current, voltage, temperature, and time of the battery when a short-circuit event occurs in the historical working state.

[0082] By setting feature screening rules (such as high-correlation threshold and time continuity threshold), potential abnormal features are extracted from the short-term current-voltage feature mapping matrix to generate a short-circuit candidate feature set. Specifically, the K-means clustering algorithm is used to analyze the short-circuit candidate feature set, and the confidence of feature classification is verified by combining historical short-circuit data to generate a short-circuit feature classification data set and a short-circuit anomaly factor.

[0083] For example, when analyzing the working data of a battery, through clustering analysis and historical comparison, it is found that the current features in certain time periods highly match the features of historical short-circuit events, and the classification confidence exceeds 95%. The anomaly factor is successfully marked as a high-risk level.

[0084] Among them, the steps of constructing a short-term correlation matrix based on short-term current and voltage changes, performing principal component analysis on the short-term correlation matrix, and generating a current-voltage time-series feature matrix based on the analysis results include: partitioning the short-term current and voltage change pairs through a time-sliding window to obtain an initial feature set, and extracting the feature clustering centers of the initial feature set based on the neighborhood density estimation algorithm.

[0085] Divide the short-term current and voltage change pairs into several time periods through the sliding window technique to form an initial feature set. Specifically, use the neighborhood density estimation algorithm to determine the feature clustering centers in the data-intensive regions, providing core feature points for subsequent matrix construction.

[0086] For example, in the feature set within a certain time period, multiple clustering centers of current and voltage are found, and these centers correspond to specific load change points, laying the foundation for the construction of the short-term correlation matrix.

[0087] Calculate the current change gradient and voltage change gradient in the region corresponding to the feature clustering centers, calculate the current-voltage correlation parameters using the current change gradient and voltage change gradient, and construct a short-term correlation matrix based on the current-voltage correlation parameters.

[0088] By calculating the gradient change rates of current and voltage around each clustering center, use these gradient parameters to calculate the current-voltage correlation parameters, and use them as the weight basis of the short-term correlation matrix. Specifically, construct the correlation parameters through the least squares fitting model to reflect the dynamic relationship between different feature regions.

[0089] For example, when calculating the change gradient of a certain clustering center, the current gradient shows a gradual increase, while the voltage gradient shows a rapid decline. After combining the parameter calculations, it is found that this region is an early sign of potential abnormality.

[0090] Perform principal component analysis on the short-term correlation matrix to obtain the principal component feature vectors, calculate the current change principal component and voltage change principal component based on the principal component feature vectors, construct a current change mapping matrix using the current change principal component, and construct a voltage change mapping matrix using the voltage change principal component.

[0091] By performing principal component analysis on the short-term correlation matrix, extract the principal component feature vectors, and separate the current and voltage change principal components. Specifically, map the current change principal component to the current change mapping matrix, and at the same time map the voltage change principal component to the voltage change mapping matrix for further feature analysis.

[0092] For example, in the current change matrix obtained by this mapping method, the main features show periodic fluctuations along the time axis, while the voltage change matrix captures several significant mutation points.

[0093] Calculate the current timing characteristics through the current change mapping matrix, calculate the voltage timing characteristics through the voltage change mapping matrix, perform feature matching on the current timing characteristics and the voltage timing characteristics, and obtain the current-voltage timing characteristic matrix.

[0094] Calculate the current timing characteristics by analyzing the rows and columns of the current change mapping matrix, calculate the voltage timing characteristics for the voltage change mapping matrix using a similar method, and finally perform feature matching on the two through a time window and a dynamic similarity metric algorithm to generate a complete current-voltage timing characteristic matrix.

[0095] For example, the finally generated timing characteristic matrix shows the synchronous change law of the current and voltage characteristics. The high matching in multiple time periods reflects the normal operating state of the battery, while the low matching area indicates potential abnormalities.

[0096] In step S300, obtain the historical distribution of the short-circuit anomaly factor, calculate the anomaly factor drift rate of the short-circuit anomaly factor based on the historical distribution, generate an anomaly factor time evolution model according to the anomaly factor drift rate, calculate the anomaly sensitivity of the short-circuit characteristic classification data set using the anomaly factor time evolution model, and perform hierarchical screening based on the anomaly sensitivity to obtain the short-circuit characteristic screening set and the short-circuit anomaly characteristic set.

[0097] Through the analysis of a large amount of historical short-circuit data, obtain the historical distribution of the short-circuit anomaly factor, including the change trends of current, voltage, and other environmental parameters; use the kernel density estimation method to model the anomaly factor distribution, and further calculate the anomaly factor drift rate to capture the change law of the short-circuit characteristics under different usage conditions. Specifically, dynamically associate the drift rate with time to generate an anomaly factor time evolution model and establish a time-dependent expression of the anomaly distribution. Combine the time evolution model to calculate the anomaly sensitivity of each feature in the short-circuit characteristic classification data set one by one, and perform hierarchical screening based on the sensitivity to finally generate the short-circuit characteristic screening set and the short-circuit anomaly characteristic set.

[0098] For example, in the experimental analysis, it is found through the time evolution model that the anomaly sensitivity of a specific feature increases significantly in a high-temperature environment, while other features remain stable under various conditions, thereby screening out the key anomaly features with environmental dependence.

[0099] Perform statistical analysis on the short-circuit characteristic screening set to obtain the feature stability parameter, calculate the feature anomaly intensity using the feature stability parameter and the short-circuit anomaly characteristic set, and dynamically correct the weights in the short-time current-voltage characteristic mapping matrix according to the feature anomaly intensity to obtain the short-circuit characteristic correction matrix.

[0100] By statistically analyzing the features in the short - circuit feature screening set, the stability parameter of each feature is calculated, which reflects the sensitivity and stability of the feature with respect to environmental changes. Using the feature stability parameter in combination with the short - circuit abnormal feature set, a weighted coefficient model is adopted to calculate the abnormal intensity of each feature. Specifically, the abnormal intensity is used as a weight correction factor for the short - time current - voltage feature mapping matrix to dynamically adjust the weights in the matrix and generate a new short - circuit feature correction matrix.

[0101] For example, when analyzing a set of short - circuit features, it is found that the features with low stability have significantly reduced weights after dynamic correction, while the features with high stability and high intensity obtain higher weights, making the corrected matrix more suitable for the actual operating conditions.

[0102] Based on the short - circuit feature correction matrix, the preset short - circuit detection threshold is adjusted to obtain an adaptive short - circuit detection threshold. Using the adaptive short - circuit detection threshold to perform real - time matching on the short - circuit feature classification data set, a short - circuit matching error matrix is obtained.

[0103] Through the short - circuit feature correction matrix, the original fixed short - circuit detection threshold is adjusted to make it adaptable to different operating environments, generating an adaptive short - circuit detection threshold. Specifically, in combination with the real - time input short - circuit feature classification data set, the feature data and the threshold are dynamically matched, the matching error value is calculated, and the error results are organized into a short - circuit matching error matrix as the basis for subsequent abnormal analysis.

[0104] For example, in real - time analysis, the adaptive short - circuit detection threshold successfully captures a slight short - circuit trend that does not reach the alarm standard of the fixed threshold, indicating that the dynamic threshold adjustment improves the detection sensitivity.

[0105] Perform an abnormal trend analysis on the short - circuit matching error matrix, and based on the analysis results, calculate the abnormal determination value in combination with the short - circuit abnormal factor, and generate a short - circuit abnormal determination result according to the abnormal determination value.

[0106] Through time - series analysis of the short - circuit matching error matrix, identify the abnormal trend of its evolution over time, and in combination with the real - time changes of the short - circuit abnormal factor, comprehensively calculate the abnormal determination value. Specifically, use a recurrent neural network (RNN) to predict the trend of the error matrix and integrate the abnormal factor through a weighted summation method to generate the final abnormal determination value. If the abnormal determination value exceeds the set safety threshold, a short - circuit abnormal determination result is generated.

[0107] For example, in a certain experimental scenario, the peak value of the short - circuit matching error matrix shows the rapid development of an abnormality. Combining the change trend of the abnormal factor, it is finally determined that there is a high short - circuit risk in this stage and an alarm signal is sent in real - time.

[0108] In step S400, a short - circuit time series is constructed based on the short - circuit anomaly determination result. The frequency and intensity of the short - circuit event are calculated based on the short - circuit time series, and the short - circuit time evolution feature is generated according to the frequency and intensity of the short - circuit event.

[0109] By time - stamping the short - circuit anomaly determination result, relevant data is extracted to construct the short - circuit time series. Specifically, through statistical analysis of the time distribution of the anomaly determination value, the occurrence frequency and intensity of the short - circuit event are calculated, and trend fitting is performed on the frequency and intensity data to generate the short - circuit time evolution feature with time dependence, which is used to describe the dynamic change law of the short - circuit behavior.

[0110] For example, in a historical operation data segment, the short - circuit time series shows that the frequent intervals and high - intensity concentrations of short - circuit events occur under the high - load state of battery use, forming a clear trend of time evolution characteristics.

[0111] The short - circuit influence range is calculated using the short - circuit time evolution feature. According to the short - circuit influence range, the short - circuit trend correction is performed on the short - circuit feature correction matrix to obtain the short - circuit correction trend data. The short - circuit correction trend data is decomposed by multi - scale time series decomposition to obtain the short - term trend component, long - term trend component, and abnormal shock component.

[0112] By fitting the spatial distribution of the frequency and intensity of the short - circuit time evolution feature, the potential short - circuit influence range is calculated, and the influence range parameter is applied to the weight re - distribution of the short - circuit feature correction matrix to perform the short - circuit trend correction operation and generate the short - circuit correction trend data. Specifically, the multi - scale time series decomposition algorithm is used for the corrected data to decompose the short - term trend component, long - term trend component, and abnormal shock component, which are used to reflect different time dimensions of the short - circuit trend.

[0113] For example, through multi - scale decomposition and analysis of the correction trend data of a certain battery, the short - term trend component shows a rapid response of current spikes, the long - term trend component reflects the cumulative change of load periodicity, and the abnormal shock component captures the high - amplitude abnormal characteristics caused by unexpected short - circuit events.

[0114] The short - term change rate is calculated according to the short - term trend component, the cumulative offset is calculated according to the long - term trend component, and the short - circuit trend change parameter is generated based on the short - term change rate and the cumulative offset.

[0115] By dynamically fitting the slope of the short - term trend component, the short - term change rate is calculated to reflect the rapid evolution speed of the short - circuit risk; combined with the long - term trend component, the cumulative offset is obtained using the cumulative sum algorithm, which represents the overall deviation degree of the trend. Specifically, through the weighted calculation method, the short - term change rate and the cumulative offset are combined to generate the short - circuit trend change parameter, which serves as the basis for quantitatively evaluating the change of the short - circuit risk.

[0116] For example, during the actual calculation process, the short-term change rate of a certain battery system rapidly climbs at the moment of overload, while the cumulative offset shows a stable increase in the long-term trend, indicating that the system is in a potentially high-risk operating state.

[0117] Calculate the future change probability of a short circuit using the short circuit trend change parameter and the abnormal impact component, perform trend fusion on the future change probability of the short circuit and the short circuit feature classification data set to obtain short circuit trend prediction data, calculate the short circuit safety boundary using the short circuit trend prediction data, and perform risk matching on the short circuit safety boundary to obtain the short circuit assessment result.

[0118] By combining the short circuit trend change parameter with the abnormal impact component, use a time series prediction model to calculate the probability distribution of the future change of the short circuit, and fuse it with the short circuit feature classification data set to form short circuit trend prediction data. Specifically, use the key parameters in the prediction data to calculate the short circuit safety boundary, and dynamically match the safety boundary with the short circuit feature classification data input in real time to evaluate whether the current state is within the safe range, and finally obtain the short circuit assessment result.

[0119] For example, in a certain actual assessment case, through the prediction of the future change probability of a short circuit, it is found that the short circuit risk of a certain node increases significantly within the next 10 minutes, and the short circuit assessment result timely triggers an alarm signal and activates corresponding protection measures.

[0120] In this embodiment, current signals and voltage signals of the battery in the operating state are collected by sensors, a current-voltage correlation data set is constructed, and the signals are decomposed and analyzed at multiple levels to extract steady-state components, fluctuation components, and pulse components, realizing a comprehensive expression of signal characteristics. Combining short-term current-voltage change pairs, a short-term correlation matrix is constructed, and technologies such as principal component analysis and feature mapping are used to gradually generate a current-voltage time series feature matrix and a short-term current-voltage feature mapping matrix, screen and cluster the features, construct a short circuit candidate feature set, form a short circuit feature classification data set and a short circuit abnormal factor, providing a data basis for short circuit detection. At the feature level, by dynamically correcting the short-term current-voltage feature mapping matrix, adjusting the weight of the current denoising filter matrix, and combining the short circuit time evolution characteristics to correct the short circuit trend, short-term trend components, long-term trend components, and abnormal impact components are generated. Further, based on these components, the short-term change rate, cumulative offset, and short circuit trend change parameter are calculated, the future change probability of the short circuit is calculated in combination with a time series prediction model, and it is fused with the short circuit feature classification data set to predict the short circuit trend and generate the short circuit assessment result. This embodiment realizes the in-depth mining of battery short circuit characteristics from multiple angles such as dynamic trend, time series components, and future prediction through multi-level and multi-scale signal processing and data analysis technologies, significantly improving the accuracy and reliability of short circuit detection, and providing comprehensive protection for the safety of battery operation.

[0121] Example 3: As Figure 2 shown, the present application also provides a battery short - circuit test device 10, including an acquisition module 11, a calculation module 12, an adjustment module 13, a decomposition module 14, and a generation module 15.

[0122] The acquisition module 11 is mainly used to acquire the current signal and voltage signal during the operation of the battery, and perform noise reduction processing on the current signal and voltage signal to obtain a denoised current signal and a denoised voltage signal.

[0123] The calculation module 12 is mainly used to construct a current - voltage dynamic characteristic map based on the denoised current signal and denoised voltage signal, generate a short - term current - voltage characteristic mapping matrix based on the current - voltage dynamic characteristic map, and perform hierarchical clustering analysis on the short - term current - voltage characteristic mapping matrix to obtain a short - circuit feature classification data set and a short - circuit anomaly factor.

[0124] The adjustment module 13 is mainly used to perform dynamic threshold adjustment on the denoised current signal and denoised voltage signal by using the short - circuit anomaly factor to obtain an adaptive short - circuit detection threshold, and perform real - time matching on the short - circuit feature classification data set according to the adaptive short - circuit detection threshold to obtain a short - circuit anomaly determination result.

[0125] The decomposition module 14 is mainly used to perform multi - scale time - series decomposition on the short - circuit anomaly determination result to obtain a short - term trend component, a long - term trend component, and an abnormal impact component, and calculate the short - term trend component, the long - term trend component, and the abnormal impact component to obtain a short - circuit evaluation result.

[0126] The generation module 15 is mainly used to compare the short - circuit evaluation result with a preset risk level. If the short - circuit trend of the short - circuit evaluation result exceeds the safe range preset by the risk level, a short - circuit warning signal is generated.

[0127] In this embodiment, the acquisition module 11 collects the current signal and voltage signal during the operation of the battery in real time, and performs efficient denoising processing on them to generate accurate and reliable denoised current signal and denoised voltage signal, laying a stable data foundation for subsequent analysis; the calculation module 12 constructs a current-voltage dynamic feature map based on the denoised current signal and denoised voltage signal, and performs calculation and hierarchical clustering analysis on the map to form a short-time current-voltage feature mapping matrix, and finally extracts a short-circuit feature classification data set and a short-circuit anomaly factor, realizing the precise capture of the battery short-circuit feature; the adjustment module 13 uses the short-circuit anomaly factor to perform dynamic threshold adjustment on the denoised current signal to generate an adaptive short-circuit detection threshold, and combines it with the short-circuit feature classification data set for real-time comparison to output a short-circuit anomaly determination result, improving the sensitivity and accuracy of detection; the decomposition module 14 performs multi-scale time series decomposition on the short-circuit anomaly determination result, separates the short-term trend component, long-term trend component and abnormal impact component, and calculates them to generate a comprehensive short-circuit evaluation result, revealing the battery short-circuit risk from multiple dimensions; the generation module 15 compares the short-circuit evaluation result with the preset risk level. If it is found that the short-circuit trend exceeds the safe range, a short-circuit warning signal is immediately generated to ensure timely response to potential short-circuit risks. This embodiment adopts a modular design, covering all key links of the battery short-circuit test from data acquisition, dynamic analysis to evaluation generation, significantly improving the accuracy, real-time performance and adaptability of detection, and providing an efficient and reliable technical guarantee for the safe operation of the battery.

[0128] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described device and each module can refer to the corresponding processes in the foregoing Embodiment 1, and will not be repeated here.

[0129] The structures, proportions, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those skilled in this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have a substantial technical meaning. Any modification of the structure, change of the proportional relationship or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that can be covered by the technical content disclosed by the present invention.

[0130] The above is the case. The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A battery short - circuit detection method, characterized in that, Including: Obtain the current signal and voltage signal during the operation of the battery, and perform denoising processing on the current signal and the voltage signal to obtain a denoised current signal and a denoised voltage signal; Construct a current-voltage dynamic feature map based on the denoised current signal and the denoised voltage signal, generate a short-time current-voltage feature mapping matrix based on the current-voltage dynamic feature map, and perform hierarchical clustering analysis on the short-time current-voltage feature mapping matrix to obtain a short-circuit feature classification data set and a short-circuit anomaly factor; Use the short-circuit anomaly factor to perform dynamic threshold adjustment on the denoised current signal and the denoised voltage signal to obtain an adaptive short-circuit detection threshold, and perform real-time matching on the short-circuit feature classification data set according to the adaptive short-circuit detection threshold to obtain a short-circuit anomaly determination result; Perform multi-scale time series decomposition on the short-circuit anomaly determination result to obtain a short-term trend component, a long-term trend component, and an abnormal impact component, and calculate the short-term trend component, the long-term trend component, and the abnormal impact component to obtain a short-circuit evaluation result; Compare the short-circuit evaluation result with a preset risk level. If the short-circuit trend of the short-circuit evaluation result exceeds the safe range preset by the risk level, generate a short-circuit warning signal.

2. The battery short-circuit detection method according to claim 1, characterized in that The step of obtaining the current signal and voltage signal during the operation of the battery, and performing denoising processing on the current signal and the voltage signal to obtain a denoised current signal and a denoised voltage signal includes: Collect the current signal and voltage signal of the battery in the operating state through a sensor, construct a current-voltage correlation data set based on the current signal and the voltage signal, and perform signal decomposition on the current-voltage correlation data set to obtain a fluctuation component and a pulse component; Calculate the short-time fluctuation trend according to the fluctuation component, analyze the current fluctuation characteristics based on the short-time fluctuation trend, and construct a current transient feature mapping using the pulse component; Perform noise reduction filtering on the current signal and the voltage signal based on the current fluctuation characteristics and the current transient feature mapping to obtain a denoised current signal and a denoised voltage signal.

3. The battery short-circuit detection method according to claim 2, wherein, The step of performing noise reduction filtering on the current signal and the voltage signal based on the current fluctuation characteristics and the current transient feature mapping to obtain a denoised current signal and a denoised voltage signal includes: Perform frequency domain decomposition on the current fluctuation characteristics using wavelet transform to obtain a low-frequency steady component and a high-frequency fluctuation component, perform envelope analysis on the current transient feature mapping using Hilbert transform to obtain transient feature parameters; Calculate steady-state trend parameters based on the low-frequency steady component, calculate short-time fluctuation parameters based on the high-frequency fluctuation component, and calculate pulse timing parameters based on the transient feature parameters; Construct a short-time current change feature set through the steady-state trend parameters and the short-time fluctuation parameters, perform time series analysis on the short-time current change feature set to obtain a short-time steady component and a short-time change component, and calculate a current denoising filter matrix based on the short-time steady component and the short-time change component; Calculate the transient influence factor according to the short-time current change feature set and the pulse timing parameters, calculate the transient influence correction parameter through the transient influence factor, and use the transient influence correction parameter to adjust the feature weights of the current denoising filter matrix to obtain the corrected current denoising filter matrix; Perform wavelet decomposition on the voltage signal to obtain a low-frequency voltage component and a high-frequency voltage component, and calculate the current-voltage joint denoising parameter by using the low-frequency voltage component, the high-frequency voltage component, and the current denoising filter matrix; Perform joint filtering on the low-frequency voltage component and the high-frequency voltage component based on the current-voltage joint denoising parameter to obtain a filtered voltage component, and calculate a filtered current signal by using the filtered voltage component and the corrected current denoising filter matrix; Perform feature comparison on the filtered voltage component and the filtered current signal to obtain a denoised current signal and a denoised voltage signal.

4. The battery short circuit detection method according to claim 1, wherein The steps of constructing a current-voltage dynamic feature map based on the denoised current signal and the denoised voltage signal, generating a short-time current-voltage feature mapping matrix based on the current-voltage dynamic feature map, and performing hierarchical clustering analysis on the short-time current-voltage feature mapping matrix to obtain a short-circuit feature classification data set and a short-circuit anomaly factor include: Calculate the instantaneous current change rate by using the denoised current signal, calculate the instantaneous power change rate by using the denoised voltage signal, and generate a short-time current-voltage change pair based on the instantaneous current change rate and the instantaneous power change rate; Construct a short-time correlation matrix based on the short-time current-voltage change pair, perform principal component analysis on the short-time correlation matrix, and generate a current-voltage time series feature matrix based on the analysis result; Perform feature aggregation on the current-voltage time series feature matrix to obtain a short-time current change map and a short-time voltage change map, map the short-time current change map and the short-time voltage change map to obtain a short-time current-voltage feature mapping matrix; Use the short-time current-voltage feature mapping matrix for feature screening, generate a short-circuit candidate feature set based on the screening result, perform clustering analysis on the short-circuit candidate feature set, and calculate the classification confidence by combining the analysis result with historical short-circuit data. Generate a short-circuit feature classification data set and a short-circuit anomaly factor according to the classification confidence; wherein, the historical short-circuit data is the current, voltage, temperature, and time of the battery when a short-circuit event occurs in the historical working state.

5. The battery short-circuit detection method according to claim 4, characterized in that, The steps of constructing a short-time correlation matrix based on the short-time current-voltage change pair, performing principal component analysis on the short-time correlation matrix, and generating a current-voltage time series feature matrix include: Partition the short-time current-voltage change pair through a time sliding window to obtain an initial feature set, and extract the feature clustering center of the initial feature set based on the neighborhood density estimation algorithm; Calculate the current change gradient and the voltage change gradient of the area corresponding to the feature clustering center, calculate the current-voltage correlation parameter by using the current change gradient and the voltage change gradient, and construct a short-time correlation matrix based on the current-voltage correlation parameter; Perform principal component analysis on the short-term correlation matrix to obtain principal component eigenvectors, calculate the principal components of current change and the principal components of voltage change based on the principal component eigenvectors, construct a current change mapping matrix using the principal components of current change, and construct a voltage change mapping matrix using the principal components of voltage change; Calculate the current time series characteristics through the current change mapping matrix, calculate the voltage time series characteristics through the voltage change mapping matrix, and perform feature matching on the current time series characteristics and the voltage time series characteristics to obtain a current-voltage time series feature matrix.

6. The battery short-circuit detection method according to claim 1, characterized in that, The step of using the short-circuit anomaly factor to perform dynamic threshold adjustment on the denoised current signal and the denoised voltage signal to obtain an adaptive short-circuit detection threshold, and performing real-time matching on the short-circuit feature classification dataset according to the adaptive short-circuit detection threshold to obtain a short-circuit anomaly determination result includes: Obtain the historical distribution of the short-circuit anomaly factor, calculate the anomaly factor drift rate of the short-circuit anomaly factor based on the historical distribution, generate an anomaly factor time evolution model according to the anomaly factor drift rate, calculate the anomaly sensitivity of the short-circuit feature classification dataset using the anomaly factor time evolution model, and perform hierarchical screening based on the anomaly sensitivity to obtain a short-circuit feature screening set and a short-circuit anomaly feature set; Perform statistical analysis on the short-circuit feature screening set to obtain a feature stability parameter, calculate the feature anomaly intensity using the feature stability parameter and the short-circuit anomaly feature set, and dynamically correct the weights in the short-term current-voltage feature mapping matrix according to the feature anomaly intensity to obtain a short-circuit feature correction matrix; Adjust the preset short-circuit detection threshold based on the short-circuit feature correction matrix to obtain an adaptive short-circuit detection threshold, and perform real-time matching on the short-circuit feature classification dataset using the adaptive short-circuit detection threshold to obtain a short-circuit matching error matrix; Perform an anomaly trend analysis on the short-circuit matching error matrix, calculate an anomaly determination value based on the analysis result in combination with the short-circuit anomaly factor, and generate a short-circuit anomaly determination result according to the anomaly determination value.

7. The battery short-circuit detection method according to claim 6, wherein The step of performing multi-scale time series decomposition on the short-circuit anomaly determination result to obtain a short-term trend component, a long-term trend component, and an anomaly impact component, and calculating the short-circuit evaluation result by calculating the short-term trend component, the long-term trend component, and the anomaly impact component includes: Construct a short-circuit time series according to the short-circuit anomaly determination result, calculate the frequency and intensity of the occurrence of short-circuit events based on the short-circuit time series, and generate a short-circuit time evolution feature according to the frequency and intensity of the occurrence of short-circuit events; Calculate the short-circuit influence range using the short-circuit time evolution feature, perform short-circuit trend correction on the short-circuit feature correction matrix according to the short-circuit influence range to obtain short-circuit correction trend data, and perform multi-scale time series decomposition on the short-circuit correction trend data to obtain a short-term trend component, a long-term trend component, and an anomaly impact component; Calculate the short-term change rate according to the short-term trend component, calculate the cumulative offset according to the long-term trend component, and generate a short-circuit trend change parameter based on the short-term change rate and the cumulative offset; Calculate the short-circuit future change probability by using the short-circuit trend change parameter and the abnormal shock component, perform trend fusion on the short-circuit future change probability and the short-circuit feature classification data set to obtain short-circuit trend prediction data, calculate the short-circuit safety boundary by using the short-circuit trend prediction data, and perform risk matching on the short-circuit safety boundary to obtain a short-circuit evaluation result.

8. A battery short - circuit test device, characterized in that, Including: An acquisition module, configured to acquire current signals and voltage signals during the operation of the battery, and perform denoising processing on the current signals and the voltage signals to obtain denoised current signals and denoised voltage signals; A calculation module, configured to construct a current-voltage dynamic feature map based on the denoised current signal and the denoised voltage signal, generate a short-term current-voltage feature mapping matrix based on the current-voltage dynamic feature map, and perform hierarchical clustering analysis on the short-term current-voltage feature mapping matrix to obtain a short-circuit feature classification data set and a short-circuit abnormal factor; An adjustment module, configured to perform dynamic threshold adjustment on the denoised current signal and the denoised voltage signal by using the short-circuit abnormal factor to obtain an adaptive short-circuit detection threshold, and perform real-time matching on the short-circuit feature classification data set according to the adaptive short-circuit detection threshold to obtain a short-circuit abnormal determination result; A decomposition module, configured to perform multi-scale time series decomposition on the short-circuit abnormal determination result to obtain a short-term trend component, a long-term trend component, and an abnormal shock component, and perform calculations on the short-term trend component, the long-term trend component, and the abnormal shock component to obtain a short-circuit evaluation result; A generation module, configured to compare the short-circuit evaluation result with a preset risk level, and if the short-circuit trend of the short-circuit evaluation result exceeds the safety range preset by the risk level, generate a short-circuit warning signal.

Images