Methods, devices, equipment and storage media for monitoring the current of power battery systems

Abstract

This application discloses a current monitoring method, apparatus, device, and storage medium for a power battery system, relating to the fields of battery management and current measurement technology. The power battery system is internally equipped with an electroacoustic generator and an acoustic measuring device. The method includes: acquiring a sound signal generated when current flows through the electroacoustic generator within the power battery system based on the acoustic measuring device; generating a target spectrum diagram based on the sound signal; inputting the target spectrum diagram into a current recognition neural network model to obtain current recognition data; and determining the current measurement value corresponding to the sound signal based on the current recognition data.

Description

Technical Field

[0001] This application relates to the field of battery management and current measurement technology, and in particular to a current monitoring method, device, equipment and storage medium for a power battery system. Background Technology

[0002] Real-time and accurate current monitoring of the power battery system is crucial for ensuring the safety and performance of electric vehicles. Currently, mainstream current measurement technologies mainly rely on direct measurement methods based on electromagnetic induction principles, such as Hall effect sensors or shunts. However, in the complex and highly variable high-voltage, high-current operating environment inside the power battery system, these sensors and their signal transmission lines are highly susceptible to electromagnetic interference, leading to measurement signal distortion, decreased accuracy, and even false alarms, thus directly affecting the accuracy and reliability of current monitoring. Summary of the Invention

[0003] The embodiments of this application provide a method, apparatus, device, and storage medium for monitoring the current of a power battery system. It can measure and monitor the current value of the power battery system based on the sound emitted when the current flows through the electroacoustic generator inside the power battery system, thereby improving the accuracy and reliability of the current monitoring of the power battery system.

[0004] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0005] This application specifically includes the following aspects: In a first aspect, this application proposes a current monitoring method for a power battery system, wherein the power battery system is internally equipped with an electroacoustic generator and an acoustic measuring instrument, the method comprising: The acoustic measuring device is used to obtain the sound signal generated when the current inside the power battery system flows through the electroacoustic generator; Based on the sound signal, generate a target spectrogram; The target spectrum is input into the current recognition neural network model to obtain current recognition data; Based on the current recognition data, the current measurement value corresponding to the sound signal is determined.

[0006] In one feasible implementation, generating the target spectrogram based on the sound signal includes: The audio signal is subjected to frame segmentation and windowing processing to obtain multiple windowed signal frames; Perform time-frequency transformation on the multiple windowed signal frames to obtain the initial spectrum of each windowed signal frame; The initial spectra are input into the Mel filter bank and processed to generate a log-Mel spectrogram as the target spectrogram.

[0007] In one feasible implementation, the current identification neural network model includes a convolutional neural network, a long short-term memory network, and an attention mechanism module, wherein the attention mechanism module includes a spectral attention module and a temporal attention module; the step of inputting the target spectrogram into the current identification neural network model to obtain current identification data includes: The convolutional neural network extracts local time-frequency features characterizing current changes from the target spectrogram. The local time-frequency features are subjected to frequency domain weighting processing by the spectral attention module to obtain frequency domain enhanced features. The frequency domain enhancement features are temporally encoded using the long short-term memory network to obtain temporally encoded features. The current identification data is obtained by performing time-domain weighting on the time-series encoded features through the time-series attention module.

[0008] In one feasible implementation, determining the current measurement value corresponding to the sound signal based on the current identification data includes: If the current recognition neural network model is a regression model, then the current recognition data will be used as the current measurement value; If the current recognition neural network model is a classification model, the current measurement value is determined based on the preset correspondence between current categories and current ranges, as well as the current recognition data.

[0009] In one feasible implementation, the current recognition neural network model is obtained through training via the following steps: The mixed sound signal samples in the training dataset are sequentially processed by framing, windowing, and Mel filter bank to generate corresponding sample log Mel spectrograms; wherein, the mixed sound signal samples in the training dataset include: sound signal samples generated by the electroacoustic generator under the drive of a known current value and collected by the acoustic measuring instrument, and the known current value labels corresponding to the sound signal samples; The initial neural network model is trained based on the sample log Mel spectrum and the corresponding known current value labels. During training, the model parameters are adjusted based on the performance evaluation results of the validation set and the test set; The current recognition neural network model is obtained based on the adjusted model parameters.

[0010] In one feasible implementation, the mixed sound signal sample further includes a noise signal sample for simulating the internal environmental noise of the power battery system.

[0011] Secondly, this application also proposes a current monitoring device for a power battery system, wherein an electroacoustic generator and an acoustic measuring instrument are installed inside the power battery system; the device includes: The sound acquisition module is used to acquire the sound signal generated when the current inside the power battery system flows through the electroacoustic generator based on the acoustic measuring instrument; A signal processing module is used to generate a target spectrogram based on the sound signal; A current identification module is used to input the target spectrum into a current identification neural network model to obtain current identification data; The current confirmation module is used to determine the current measurement value corresponding to the sound signal based on the current identification data.

[0012] Thirdly, this application also proposes a current monitoring device for a power battery system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the current monitoring method for the power battery system as described in any of the first aspects above.

[0013] Fourthly, this application also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the current monitoring method for a power battery system as described in any one of the first aspects above.

[0014] Fifthly, this application also proposes a current monitoring system for a power battery system, including the current monitoring device for the power battery system described in the third aspect above and the power battery system. The power battery system contains an electroacoustic generator and an acoustic measuring device. The electroacoustic generator is connected in series with the battery cells of the power battery system in the high-voltage circuit within the power battery system. The acoustic measuring device and the electroacoustic generator are housed within the same acoustic isolation structure. The electroacoustic generator is used to emit a sound signal based on the flowing current; the acoustic measuring device is used to acquire the sound signal.

[0015] This application proposes a current monitoring method for a power battery system. By acquiring the sound signal emitted by an electroacoustic generator based on the current flowing through it and converting it into a spectrum, the current value is then identified by a current recognition neural network model. By transforming the current measurement problem into an acoustic signal analysis problem, the accuracy and reliability of current monitoring of the power battery system are improved.

[0016] This application proposes a current monitoring method and related equipment for a power battery system. Other advantages, objectives and features of this application will be partly apparent from the following description, and partly understood by those skilled in the art through study and practice of this application. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a current monitoring method for a power battery system provided in an embodiment of this application; Figure 2 A time spectrum diagram of a digital audio signal provided in an embodiment of this application; Figure 3 A target spectrum diagram provided for an embodiment of this application; Figure 4 An example diagram of a current measurement value provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a power battery system provided in an embodiment of this application; Figure 6 A schematic diagram of the functional modules of a current monitoring device for a power battery system provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a current monitoring device for a power battery system provided in an embodiment of this application. Detailed Implementation

[0018] To better understand the technical solutions provided in the embodiments of this specification, the technical solutions of the embodiments of this specification will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this specification and the specific features in the embodiments are detailed descriptions of the technical solutions of the embodiments of this specification, rather than limitations on the technical solutions of this specification. In the absence of conflict, the embodiments of this specification and the technical features in the embodiments can be combined with each other.

[0019] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The term "two or more" includes two or more cases.

[0020] This application provides a method for monitoring the current of a power battery system, wherein the power battery system is internally equipped with an electroacoustic generator and an acoustic measuring instrument. Please refer to... Figure 1 This is a flowchart illustrating a current monitoring method for a power battery system provided in an embodiment of this application, which may specifically include: S110. Acquire the sound signal generated when the current inside the power battery system flows through the electroacoustic generator based on the acoustic measuring device.

[0021] Specifically, when the power battery system is working, the current generated by the charging and discharging of the battery cells flows through an electroacoustic generator connected in series in the high-voltage circuit. This electroacoustic generator has been specially optimized for frequency characteristics and has been designed with a specific frequency range based on the fixed laws of current magnitude, direction, and sound (e.g., low frequency bands correspond to small currents, and high frequency bands correspond to large currents). Furthermore, this frequency range is intentionally offset from the low-to-mid-frequency noise bands of components such as fans and pumps within the battery system, reducing the overlap between environmental noise and the frequency characteristics of current sounds at the source. Simultaneously, the electroacoustic generator uses directional sound-emitting materials such as piezoelectric ceramics, concentrating sound energy towards the acoustic measuring instrument, reducing diffusion in other directions, and minimizing reflected noise interference. The acoustic measuring instrument is positioned close to the electroacoustic generator, and both are encapsulated within the same housing. Through a double-layered sound-absorbing structure and a vibration-damping bracket (a rubber vibration-damping pad is installed between the electroacoustic generator and the battery compartment to reduce mechanical vibration transmission and minimize interference from vibration noise on the acoustic measurements), the design minimizes environmental noise interference from fans, pumps, and other sources, ensuring that the acquired analog sound signal is primarily characterized by current. In the double-layered housing design, the outer shell is filled with sound-absorbing material (such as foam or ultra-fine glass wool) to form an acoustic attenuation cavity. The inner shell uses a high-density material (such as metal) to block high-frequency noise, while the outer material absorbs mid-to-low-frequency noise.

[0022] Subsequently, as Figure 2As shown, the acoustic measuring instrument performs analog-to-digital conversion on the acquired analog sound signals, transforming the continuous analog signals into discrete digital sound signals, providing a digital signal foundation for subsequent signal processing and spectrogram generation that is compatible with computer operations.

[0023] S120. Generate the target spectrum diagram based on the sound signal.

[0024] Specifically, sound signals are time-domain signals, and direct analysis makes it difficult to extract key current-related features. Therefore, it is necessary to convert them into a frequency-domain spectrum. For example... Figure 3 As shown, step S120 uses signal processing technology to convert the continuous sound signal into a spectrum that can intuitively reflect the frequency distribution and intensity changes, realizing the conversion from "sound waveform" to "feature spectrum". This visualizes the abstract sound signal as a feature carrier that can be recognized by the neural network. The spectrum clearly presents the frequency characteristic differences corresponding to different currents, providing a clear analytical object for subsequent model recognition.

[0025] S130. Input the target spectrum into the current recognition neural network model to obtain current recognition data.

[0026] Specifically, the current recognition neural network model learns from a large number of corresponding "spectrum graph-current value" samples, enabling it to infer current information from spectral features. This effectively eliminates residual noise interference, accurately extracts current-related feature information, and outputs current recognition data. It avoids the subjectivity and limitations of manual analysis, significantly improving the accuracy and efficiency of current recognition.

[0027] S140. Based on the current recognition data, determine the current measurement value corresponding to the sound signal.

[0028] Specifically, such as Figure 4 As shown, step S140 determines the corresponding current measurement value based on the current recognition data output by the current recognition neural network model and the model type (regression or classification). If it is a regression model, it directly outputs a precise continuous current value; if it is a classification model, it determines the current range based on a preset correspondence between current categories and intervals. This allows the method to meet measurement needs in different scenarios, providing specific current values ​​for scenarios requiring precise control, and simplifying the output for scenarios requiring only interval judgment, thus improving the applicability of the method.

[0029] For example, such as Figure 5As shown, in the power battery system provided in this application embodiment, the cell pack 1, electroacoustic generator 4, acoustic measuring device 5, and battery management unit 3 (BMU) are all integrated inside the battery box 2. The electroacoustic generator 4 is connected in series with the cell pack 1 in the high-voltage circuit. When the power battery is charging and discharging during vehicle operation, current flows through the electroacoustic generator 4. The electroacoustic generator 4 emits a specific pattern of sound according to the magnitude of the current (e.g., 0.5A corresponds to low-frequency sound, 10A corresponds to high-frequency sound) and direction. The acoustic measuring device 5 is in close contact with the electroacoustic generator 4. Protected by a double-layer shell (inner layer of sealed metal, outer layer filled with foam sound-absorbing material) and rubber damping pads, it is not disturbed by fan noise (mid-to-low frequency) and mechanical vibration, accurately collects the sound signal, and transmits it to the battery management unit 3. The battery management unit 3 first converts the sound signal into a target spectrum, then inputs it into the trained current recognition neural network model. After removing residual environmental noise, the model outputs current recognition data. Finally, the battery management unit 3 determines the current measurement value based on the data (e.g., the regression model outputs 3.2A, or the classification model outputs "medium current range"), thereby realizing real-time monitoring of the power battery current.

[0030] In some examples, a target spectrogram is generated based on the sound signal, including: The audio signal is framed and windowed to obtain multiple windowed signal frames; Perform time-frequency transformation on multiple windowed signal frames to obtain the initial spectrum of each windowed signal frame; Each initial spectrum is input into a Mel filter bank for processing, generating a log-Mel spectrum as the target spectrum.

[0031] Specifically, framing and windowing are used to avoid spectral leakage of the signal. Continuous sound signals are divided into multiple overlapping short frames and a window function is applied to make the edges of each frame transition smoothly. Time-frequency transformation converts each short frame from the time domain to the frequency domain, revealing the frequency composition of the signal. Mel filtering is based on the human ear's perception of different frequencies of sound. It filters and compresses the frequency domain signal to highlight key frequency features related to current. The resulting logarithmic Mel spectrogram can effectively reduce noise interference and enhance the features of useful signals.

[0032] For example, the sound signal acquired by the acoustic measuring device 5 is a continuous time-domain signal (e.g., containing current characteristic sound and a small amount of fan noise). First, the battery management unit 3 performs frame segmentation on the signal, setting the frame length to 25ms and the frame shift to 10ms, dividing the continuous signal into multiple overlapping short frames. Then, a Hamming window is applied to each short frame to reduce spectral leakage at the edges of each frame, resulting in multiple smooth windowed signal frames. Next, a Fourier transform (a type of time-frequency transform) is performed on each windowed signal frame, transforming each short frame from a time-domain representation of "time-amplitude" to a frequency-domain representation of "frequency-intensity," obtaining the initial spectrum corresponding to each frame. At this point, the initial spectrum contains both the target frequency component corresponding to the current (e.g., 1kHz-3kHz) and the low-frequency component corresponding to the fan noise (<500Hz). Finally, all initial spectra are input into a preset Mel filter bank. This filter bank, based on the characteristics of human auditory perception, performs finer segmentation of high-frequency signals, compresses low-frequency signals, filters out some noise frequency components unrelated to current, and takes the logarithm of the filtered signal to generate a 128×128 log-Mel spectrogram as the target spectrogram. In this target spectrogram, the frequency characteristics corresponding to current are significantly enhanced, and noise components are effectively suppressed, making it more suitable for feature extraction by subsequent neural network models.

[0033] In some examples, the current recognition neural network model includes a convolutional neural network, a long short-term memory network, and an attention mechanism module, which comprises a spectral attention module and a temporal attention module. The target spectrogram is input into the current recognition neural network model to obtain current recognition data, including: Local time-frequency features characterizing current changes are extracted from the target spectrogram using a convolutional neural network. The frequency domain enhanced features are obtained by performing frequency domain weighting on the local time-frequency features using the spectral attention module; Temporally encoded features are obtained by temporally encoding frequency domain enhanced features using a long short-term memory network; The current identification data is obtained by performing time-domain weighting on the time-series encoded features through the time-series attention module.

[0034] Specifically, the current recognition neural network model provided in this application has a structure of convolutional neural network (CNN) + long short-term memory network (LSTM) + attention mechanism module. By combining the advantages of different network structures, it achieves accurate extraction of current features in the spectrum through multi-stage feature processing.

[0035] Furthermore, Convolutional Neural Networks (CNNs) excel at extracting local spatial features from two-dimensional image data (such as spectrograms), effectively capturing local time-frequency patterns related to current (such as intensity changes within specific frequency ranges) in the spectrogram. Long Short-Term Memory Networks (LSTMs) focus on modeling the dependencies of time-series data, capturing dynamic features such as sound frequency drift caused by current changes. The attention mechanism module is divided into spectral attention and temporal attention. Spectral attention can automatically learn the importance weights of different frequency bands, suppressing noise-dominated bands (such as the low-frequency bands where fan noise is concentrated) and strengthening target frequency bands (such as the high-frequency bands corresponding to current). Temporal attention focuses on key temporal segments related to current changes (such as sound peaks during sudden current changes), ignoring irrelevant stationary noise segments. Through the process of "CNN extracting local features, spectral attention weighting enhancement, LSTM modeling temporal dependencies, and temporal attention focusing on key segments," accurate identification of current features is achieved, outputting current identification data.

[0036] For example, in the current identification process, the generated 128×128 log-Melogram spectrum is input into a pre-defined current identification neural network model. First, the CNN part of the model contains three convolutional layers (using 32, 64, and 128 3×3 kernels respectively) and a 2×2 max-pooling layer. Through convolutional operations, it extracts local time-frequency features from the spectrum, such as capturing local features like intensity peaks and harmonic distributions within the 1kHz-3kHz frequency band corresponding to the current, while initially filtering out some scattered noise features. Next, the Spectrum Attention (SE) module processes the local time-frequency features extracted by the CNN. Through global average pooling, fully connected layers, and the sigmoid activation function, it automatically learns the weights for different frequency channels. Low weights are assigned to low-frequency channels where fan noise is concentrated (to suppress noise), while high weights are assigned to high-frequency channels corresponding to the current (to enhance useful features), resulting in enhanced frequency domain features. Then, the LSTM part consists of two 128-unit LSTM layers, which input the frequency domain enhancement features in chronological order to model the dependencies between spectral features of different time frames. For example, it captures the temporal change pattern of the gradually increasing high-frequency components in the spectrum when the current increases from 2A to 5A, and outputs the temporal encoded features. Finally, the temporal attention module performs weighted processing on the temporal encoded features, focusing on key time segments with significant current changes (such as the spectral segment corresponding to the current abrupt change during acceleration), ignoring segments with stable current and high noise, and finally outputting accurate current identification data.

[0037] As shown in Table 1 below, comparative experiments verify the anti-interference advantages of the current recognition neural network model (i.e., CNN-LSTM model) in power battery current monitoring in this application. The experiments used traditional CNN and pure LSTM as baseline models and tested three scenarios: no interference (pure current noise), fan noise (SNR=10dB), and pump noise (SNR=5dB). Classification accuracy and regression RMSE were used as evaluation indicators. The results show that CNN-LSTM achieves an accuracy of 95% in the no-interference scenario, 89% in the fan noise scenario, and 85% in the pump noise scenario (low SNR), with an average RMSE of only 0.20A. Compared with traditional CNN (65% accuracy and RMSE 0.35A in the pump noise scenario) and pure LSTM (70% accuracy and RMSE 0.30A in the pump noise scenario), this application performs better in electromagnetic interference, especially in low SNR scenarios, with an accuracy improvement of 10%-20% and a RMSE reduction of 33%. It can better adapt to the complex environment of fan and pump noise in the actual operation of power batteries, providing more stable and accurate support for current monitoring.

[0038]

[0039] Table 1 In some examples, the current measurement value corresponding to the sound signal is determined based on current identification data, including: If the current recognition neural network model is a regression model, then the current recognition data will be used as the current measurement value. If the current recognition neural network model is a classification model, then the current measurement value is determined based on the preset correspondence between current categories and current ranges, as well as the current recognition data.

[0040] Specifically, by adapting to two types of current recognition neural network models (regression model and classification model), this approach provides flexible output methods for measurement results to meet the application needs of different scenarios. The core of the regression model is to directly output continuous values ​​corresponding to the true current value through the fitting ability of the neural network, achieving accurate measurement. The classification model, on the other hand, divides the current into multiple preset intervals (such as low current, medium current, and high current) through model training, outputs corresponding category labels, and then, based on the preset correspondence between current categories and current intervals, deduces the measured current value (interval or representative value of the interval). This design caters to both scenarios requiring precise values ​​and those requiring rapid judgment but not high precision, thus improving the applicability of the method.

[0041] For example, in scenario one: the precise charging and discharging control scenario of the power battery system, accurate real-time current values ​​are needed to adjust the charging and discharging strategy. In this case, a regression-type current recognition neural network model is used. When the target spectrum is input into the model, the output current recognition data is a continuous value (e.g., 4.8A). This data is directly used as the current measurement value. The battery management unit 3 controls the voltage output of the charging and discharging circuit based on this precise value to ensure the stability and safety of the charging and discharging process. In scenario two: the fault warning scenario of the power battery system, it is only necessary to determine whether the current is within the normal range (e.g., low current 0.5A-3A, medium current 3A-7A, high current 7A-10A), without requiring precise values. In this case, a classification-type current recognition neural network model is used. The current recognition data output by the model is a classification label (e.g., "medium current"). The system determines that the current measurement value is within the 3A-7A range based on the preset correspondence ("medium current" corresponds to the 3A-7A range). If the label is "high current" and continues for more than a set time, the system triggers a fault warning to remind the driver to check the battery status.

[0042] In some examples, the current recognition neural network model is trained through the following steps: The mixed sound signal samples in the training dataset are sequentially processed by framing, windowing, and Mel filter bank to generate the corresponding sample log Mel spectrograms. The mixed sound signal samples in the training dataset include: sound signal samples generated by an electroacoustic generator under the drive of a known current value and collected by an acoustic measuring instrument, as well as the known current value labels corresponding to the sound signal samples. The initial neural network model is trained based on the sample log Mel spectrum and the corresponding known current value labels; During training, the model parameters are adjusted based on the performance evaluation results of the validation set and the test set; Based on the adjusted model parameters, a current recognition neural network model is obtained.

[0043] Specifically, the training process of the current recognition neural network model follows a standardized procedure of "data preprocessing, model training, parameter adjustment, and model determination" to ensure the model's ability to accurately identify current characteristics. Constructing a training dataset is fundamental; it must include mixed sound signal samples from real-world scenarios and their corresponding current labels to ensure the model's generalization ability. The data preprocessing steps are consistent with the spectrogram generation steps in practical applications, ensuring that the feature forms of the training samples match the actual input data and improving the model's adaptability. The model training process iteratively optimizes the model, allowing it to gradually learn the mapping relationship between the spectrogram and current values. Performance evaluation based on the validation and test sets is used to monitor the model's training effect, adjusting parameters promptly to avoid overfitting or underfitting, ultimately obtaining the optimal-performing current recognition neural network model.

[0044] For example, a training dataset is first constructed: The working scenario of a power battery pack is simulated in a laboratory environment. An electroacoustic generator is set to operate under different known current values ​​(0.5A-10A, in 0.5A steps), while noise sources such as fans and pumps (noise intensity 30dB-70dB) are introduced. Mixed sound signal samples containing current-characteristic sounds and environmental noise are collected using an acoustic measuring instrument, with 50-100 samples corresponding to each current value, for a total of 1000 real samples. Then, labeled simulated data (such as pure current sound superimposed with noise of different intensities) is generated using signal synthesis technology, expanding the dataset to 5000 samples, and labeling each sample with the corresponding known current value. Next, all mixed sound signal samples are preprocessed: frames are divided into 25ms frames with a 10ms frame shift, a Hamming window is applied, Fourier transform and Mel filtering are performed, generating a sample log-Mel spectrogram consistent with actual applications. Then, the dataset was divided into training, validation, and test sets in a 70%:15%:15% ratio. Based on the sample spectrograms and corresponding current labels in the training set, the initial neural network model (containing 3 layers of CNN, 2 layers of LSTM, and an attention mechanism module) was trained. Mean squared error (MSE) or cross-entropy (for classification tasks) was used as the loss function, and Adam was used as the optimizer (learning rate 0.001, weight decay 1e-5), for a total of 100 training rounds. During training, the model performance was evaluated using the validation set after each round (e.g., calculating the root mean squared error RMSE). If the validation set performance did not improve for several consecutive rounds, early stopping was triggered to avoid overfitting. Based on the validation set evaluation results, parameters such as the number of convolutional kernels and the number of LSTM units were dynamically adjusted. Finally, the model with adjusted parameters was validated using the test set. When the regression RMSE on the test set reached 0.2A and the classification accuracy reached 95%, the model was determined as the final current recognition neural network model and deployed to the battery management unit 3 of the power battery system.

[0045] In some examples, the mixed audio signal samples also include noise signal samples used to simulate the ambient noise inside the power battery system.

[0046] For example, there are various environmental noises inside the power battery system, such as fans, pumps, and mechanical vibrations. These noises will be superimposed on the sound of current characteristics. If the training data only contains pure current sound samples, the model will be unable to adapt to the noisy environment in practical applications, resulting in a decrease in recognition accuracy. However, by adding simulated environmental noise samples to the training data, the model can fully learn the characteristics of noise during the training process, while strengthening its ability to extract useful current signals. This will enable it to effectively eliminate noise interference and accurately identify current characteristics in practical applications.

[0047] Furthermore, this application also proposes a current monitoring device for a power battery system, wherein an electroacoustic generator and an acoustic measuring instrument are installed inside the power battery system, specifically as follows: Figure 6 The diagram shown is a functional module schematic of a current monitoring device for a power battery system proposed in this application. The device includes: The sound acquisition module 21 is used to acquire the sound signal generated when the current inside the power battery system flows through the electroacoustic generator based on the acoustic measuring instrument; Signal processing module 22 is used to generate a target spectrogram based on the sound signal; The current identification module 23 is used to input the target spectrum into the current identification neural network model to obtain current identification data; The current confirmation module 24 is used to determine the current measurement value corresponding to the sound signal based on the current recognition data.

[0048] It should be noted that the above embodiments are merely best examples and are not intended to limit the implementation of this application.

[0049] Furthermore, such as Figure 7 As shown, this application embodiment also provides a current monitoring device 300 for a power battery system, including a processor 310, a memory 320, and a computer program 321 stored in the memory 320 and executable on the processor. When the processor 310 executes the computer program 321, it implements the steps of any of the above-described current monitoring methods for a power battery system.

[0050] Furthermore, embodiments of this application also provide a current monitoring system for a power battery system, including a current monitoring device for the power battery system and a power battery system. For example... Figure 5 As shown, the power battery system is equipped with an electroacoustic generator 4 and an acoustic measuring device 5. The electroacoustic generator 4 is connected in series with the battery cell group 1 of the power battery system in the high-voltage circuit inside the power battery system. The acoustic measuring device 5 and the electroacoustic generator 4 are set in the same acoustic isolation structure. The electroacoustic generator 4 is used to emit sound signals according to the current flowing through it; the acoustic measuring device 5 is used to acquire sound signals.

[0051] For example, the power battery system includes a cell pack 1, a battery housing 2, a battery management unit 3, an electroacoustic generator 4, and an acoustic measuring device 5. The cell pack 1, battery management unit 3, electroacoustic generator 4, and acoustic measuring device 5 are all located inside the battery housing 2. The electroacoustic generator 4 is connected in series with the cell pack 1 in the high-voltage circuit inside the power battery system. When the cell pack 1 is charged and discharged through the voltage circuit, current flows through the electroacoustic generator 4, and the electroacoustic generator 4 emits sound according to the flowing current. The sound changes according to the magnitude and direction of the current in a fixed pattern.

[0052] The electroacoustic generator 4 and the acoustic measuring instrument 5 are housed in the same enclosure to avoid external noise interference. The acoustic measuring instrument 5 is positioned close to the electroacoustic generator 4, thereby converting the collected sound into a digital sound signal that is transmitted to the battery management unit 3.

[0053] Since the electronic device described in this embodiment is the device used to implement the current monitoring method of a power battery system in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application is within the scope of protection of this application.

[0054] In practical implementation, when the computer program 321 is executed by the processor, it can achieve the following: Figure 1 Any of the corresponding implementation methods in the embodiments.

[0055] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0057] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0058] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0059] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0060] This application also provides a computer program product, which includes computer software instructions that, when executed on a processing device, cause the processing device to execute a process for a current monitoring method for a power battery system.

[0061] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0062] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0063] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0064] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0065] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0067] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

[0068] Although preferred embodiments have been described in this specification, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this specification.

[0069] Obviously, those skilled in the art can make various modifications and variations to this specification without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims and their equivalents, this specification is also intended to include such modifications and variations.

Claims

1. A method for monitoring the current of a power battery system, characterized in that, The power battery system is equipped with an electroacoustic generator and an acoustic measuring instrument. The method includes: The acoustic measuring device is used to obtain the sound signal generated when the current inside the power battery system flows through the electroacoustic generator; Based on the sound signal, generate a target spectrogram; The target spectrum is input into the current recognition neural network model to obtain current recognition data; Based on the current recognition data, the current measurement value corresponding to the sound signal is determined.

2. The method according to claim 1, characterized in that, The step of generating a target spectrogram based on the sound signal includes: The audio signal is subjected to frame segmentation and windowing processing to obtain multiple windowed signal frames; Perform time-frequency transformation on the multiple windowed signal frames to obtain the initial spectrum of each windowed signal frame; The initial spectra are input into the Mel filter bank and processed to generate a log-Mel spectrogram as the target spectrogram.

3. The method according to claim 1, characterized in that, The current recognition neural network model includes a convolutional neural network, a long short-term memory network, and an attention mechanism module, wherein the attention mechanism module includes a spectral attention module and a temporal attention module. The step of inputting the target spectrum into the current identification neural network model to obtain current identification data includes: The convolutional neural network extracts local time-frequency features characterizing current changes from the target spectrogram. The local time-frequency features are subjected to frequency domain weighting processing by the spectral attention module to obtain frequency domain enhanced features. The frequency domain enhancement features are temporally encoded using the long short-term memory network to obtain temporally encoded features. The current identification data is obtained by performing time-domain weighting on the time-series encoded features through the time-series attention module.

4. The method according to claim 1, characterized in that, The step of determining the current measurement value corresponding to the sound signal based on the current identification data includes: If the current recognition neural network model is a regression model, then the current recognition data will be used as the current measurement value; If the current recognition neural network model is a classification model, the current measurement value is determined based on the preset correspondence between current categories and current ranges, as well as the current recognition data.

5. The method according to claim 1, characterized in that, The current recognition neural network model is obtained through the following steps: The mixed sound signal samples in the training dataset are sequentially processed by framing, windowing, and Mel filter bank to generate corresponding sample log Mel spectrograms; wherein, the mixed sound signal samples in the training dataset include: sound signal samples generated by the electroacoustic generator under the drive of a known current value and collected by the acoustic measuring instrument, and the known current value labels corresponding to the sound signal samples; The initial neural network model is trained based on the sample log Mel spectrum and the corresponding known current value labels. During training, the model parameters are adjusted based on the performance evaluation results of the validation set and the test set; The current recognition neural network model is obtained based on the adjusted model parameters.

6. The method according to claim 5, characterized in that, The mixed sound signal sample also includes a noise signal sample used to simulate the internal environmental noise of the power battery system.

7. A current monitoring device for a power battery system, characterized in that, The power battery system is equipped with an electroacoustic generator and an acoustic measuring instrument; the device includes: The sound acquisition module is used to acquire the sound signal generated when the current inside the power battery system flows through the electroacoustic generator based on the acoustic measuring instrument; A signal processing module is used to generate a target spectrogram based on the sound signal; A current identification module is used to input the target spectrum into a current identification neural network model to obtain current identification data; The current confirmation module is used to determine the current measurement value corresponding to the sound signal based on the current identification data.

8. A current monitoring device for a power battery system, comprising: The memory and processor are characterized in that the processor is used to execute a computer program stored in the memory to implement the steps of the current monitoring method for the power battery system as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the current monitoring method for the power battery system as described in any one of claims 1 to 6.

10. A current monitoring system for a power battery system, comprising the current monitoring device for a power battery system as described in claim 8 and a power battery system, wherein an electroacoustic generator and an acoustic measuring device are disposed inside the power battery system, the electroacoustic generator is connected in series with the battery cell of the power battery system in the high-voltage circuit inside the power battery system, and the acoustic measuring device and the electroacoustic generator are disposed within the same acoustic isolation structure, wherein, The electroacoustic generator is used to emit sound signals based on the current flowing through it; the acoustic measuring instrument is used to acquire the sound signals.

Images