Lithium battery system fault detection method and system based on digital twin

By arranging multiple types of sensors in the lithium battery system and utilizing neural network models, combining physical and acoustic data, comprehensive monitoring of the status of the lithium battery system and identification of potential faults are achieved, and the problem of difficulty in capturing potential faults in the existing technology is solved, improving the problem of the problem of difficulty in capturing potential faults in the existing technology. The reliability and security of the system are achieved.

CN118673816BActive Publication Date: 2024-11-19NANTONG TAIPING TONGREN ELECTRONIC TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411113083.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-19
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

Existing lithium battery fault detection methods are difficult to capture potential hidden faults in lithium battery systems, such as polarization effects and material aging, resulting in system performance degradation or damage. Therefore, how to comprehensively detect the operating status of lithium battery systems in a timely manner and detect and diagnose it. Faults are a problem.

Method used

Using the fault detection method of lithium battery system based on digital twins, by arranging multiple types of sensors in the lithium battery system to collect physical data and acoustic data, using neural network models to train the acoustic recognition model, and combining physical data and acoustic data to generate the final detection result, realizing Comprehensive monitoring of the status of lithium battery systems and timely identification of potential faults.

Benefits of technology

The reliability and safety of the lithium battery system are improved. Through the combination of multiple sensors and neural network models, the characteristics of different types of data can be captured more accurately, enhanced resistance to environmental noise, and improved detection accuracy and comprehensiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118673816B_ABST
    Figure CN118673816B_ABST
Patent Text Reader

Abstract

The present application relates to the field of lithium battery detection technology, and in particular to a lithium battery system fault detection method and system based on digital twins, the method comprising arranging sensors in the lithium battery system, collecting physical data and acoustic data; calculating the evaluation value of the lithium battery system based on the physical data, comparing the evaluation value with the threshold, and generating a first detection result if it exceeds the threshold and is in an abnormal state; obtaining a training data set, including the state of the lithium battery system and the corresponding acoustic data, using the acoustic data as input and the lithium battery system state as output, training a preset model to obtain an acoustic recognition model; inputting the acoustic data into the acoustic recognition model, and the acoustic recognition model outputs the state and generates a second detection result; combining the first detection result and the second detection result to generate a final detection result. The present application can comprehensively detect the operating state of the lithium battery system, timely discover and diagnose potential faults, thereby improving the reliability and safety of the lithium battery system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of lithium battery detection technology, and in particular to a lithium battery system fault detection method and system based on digital twins. Background Art

[0002] With the continuous development of Industry 4.0 and Internet of Things technology, digital twin technology, as an emerging digital technology, is rapidly entering various industrial fields. Digital twin technology creates a virtual model of the physical entity and uses sensor data and historical data for real-time simulation and predictive analysis, thereby achieving real-time monitoring, optimization and fault prediction of the physical entity. As an important component of energy storage, lithium batteries are widely used in electric vehicles, smart grids, consumer electronics and other fields. With the continuous expansion of lithium battery applications, the requirements for their reliability, safety and service life are also constantly increasing. Therefore, the use of digital twin technology to realize fault detection of lithium battery systems has gradually become a hot topic in current research.

[0003] At present, the commonly used lithium battery fault detection methods are mainly based on the monitoring and analysis of sensor data. These sensors usually include voltage, current, temperature, etc., which collect various parameter data of the battery system in real time and transmit them to the monitoring system for processing and analysis. Among them, the voltage sensor is used to monitor the voltage change of the battery, the current sensor is used to detect the charging and discharging process of the battery, and the temperature sensor is used to monitor the temperature change of the battery. Once the data collected by the sensor exceeds the preset safety range or threshold, the monitoring system will issue an alarm to notify the staff or the system to automatically take corresponding measures, such as stopping charging and discharging, cutting off the power supply, etc., to prevent more serious failures or safety accidents in the battery system.

[0004] However, the above technologies have limited detection capabilities for potential hidden faults in lithium battery systems during implementation. Since the above methods mainly rely on monitoring a few parameters, it is difficult to capture some internal structural faults, such as polarization effects and material aging. These faults may not be easy to detect in the early stages, but will gradually accumulate and cause system performance to deteriorate or even damage. Therefore, how to comprehensively detect the operating status of the lithium battery system, timely discover and diagnose potential faults, and improve the reliability and safety of the lithium battery system is a difficult problem that needs to be solved at present. Summary of the invention

[0005] This application provides a lithium battery system fault detection method and system based on digital twins, which can comprehensively detect the operating status of the lithium battery system, timely discover and diagnose potential faults, thereby improving the reliability and safety of the lithium battery system. This application provides the following technical solutions:

[0006] In a first aspect, the present application provides a lithium battery system fault detection method based on digital twins, the method comprising:

[0007] Arranging multiple types of sensors in the lithium battery system to collect physical data and acoustic data of the lithium battery system;

[0008] Calculating an evaluation value of the lithium battery system based on the physical data, comparing the evaluation value with a preset threshold, and if the evaluation value exceeds the preset threshold, considering the system to be in an abnormal state, and generating a first detection result;

[0009] Acquire a training data set, the training data set including all states of the lithium battery system and their corresponding acoustic data, use the acoustic data as input and the corresponding state of the lithium battery system as output, train a preset neural network model to obtain an acoustic recognition model;

[0010] Inputting the acoustic data into the acoustic recognition model, the acoustic recognition model outputs a state corresponding to the lithium battery system and generates a second detection result;

[0011] The first detection result and the second detection result are combined to generate a final detection result.

[0012] In a specific implementation scheme, arranging multiple types of sensors in the lithium battery system to collect physical data and acoustic data of the lithium battery system includes:

[0013] The physical data includes current data, voltage data and temperature data;

[0014] The acoustic data includes audio data, vibration data, ultrasonic data and acoustic emission data.

[0015] In a specific implementation manner, calculating the evaluation value of the lithium battery system based on the physical data includes:

[0016] Use the following formula to calculate the evaluation value for the lithium battery system:

[0017] in, is the current data, is the voltage data, is the temperature data.

[0018] In a specific implementation manner, after calculating the evaluation value of the lithium battery system based on the physical data, the method further includes:

[0019] Obtain historical physical data for a period of time, obtain the historical current average, historical voltage average, and historical temperature average through the historical physical data for a period of time, and calculate the historical current change rate based on the historical current average, historical voltage average, and historical temperature average. , Historical voltage change rate And the historical temperature change rate , the calculation formula is as follows:

[0020] Calculate the historical current change rate , Historical voltage change rate And the historical temperature change rate Afterwards, the average rate of change of historical physical data is calculated using the following formula: :

[0021] According to the average change rate of historical physical data Correct the previously calculated evaluation value:

[0022] .

[0023] In a specific implementation scheme, the acquisition of a training data set includes all states of the lithium battery system and their corresponding acoustic data, and the acoustic data is used as input, and the corresponding state of the lithium battery system is used as output. The preset neural network model is trained to obtain an acoustic recognition model, which includes:

[0024] Acquire a first training data set, the first training data set includes all states of the lithium battery system and their corresponding audio data and vibration data, take a certain state of the lithium battery system as output, take the feature vector extracted and fused from the corresponding audio data and vibration data as input, train a preset neural network model to obtain a first acoustic recognition model;

[0025] After extracting the feature vectors of the audio data and vibration data, the feature vectors of the two are fused using the following formula:

[0026] in is the feature vector of audio data, is the feature vector of the vibration data.

[0027] In a specific implementation scheme, the acquisition of a training data set includes all states of the lithium battery system and their corresponding acoustic data, and the acoustic data is used as input, and the corresponding state of the lithium battery system is used as output. The preset neural network model is trained to obtain an acoustic recognition model, which includes:

[0028] Obtain a second training data set, the second training data set includes all states of the lithium battery system and their corresponding ultrasonic data and acoustic emission data, take a certain state of the lithium battery system as output, take the feature vector extracted and fused from the corresponding ultrasonic data and acoustic emission data as input, train the preset neural network model to obtain a second acoustic recognition model;

[0029] After extracting the feature vectors of ultrasonic data and acoustic emission data, the feature vectors of the two are fused using the following formula:

[0030] in, is the feature vector of ultrasonic data, is the feature vector of acoustic emission data.

[0031] In a specific implementation scheme, combining the first test result and the second test result to generate a final test result includes:

[0032] constructing a matrix based on the first detection result and the second detection result;

[0033] Perform singular value decomposition on the constructed matrix;

[0034] Select an appropriate number of singular values ​​to reconstruct the matrix;

[0035] The final detection results are extracted from the reconstructed matrix.

[0036] In the second aspect, the present application provides a lithium battery system fault detection system based on digital twins, which adopts the following technical solutions:

[0037] A lithium battery system fault detection system based on digital twins, comprising:

[0038] A data acquisition module, used to arrange multiple types of sensors in the lithium battery system to collect physical data and acoustic data of the lithium battery system;

[0039] A threshold comparison module, used to calculate an evaluation value of the lithium battery system based on the physical data, compare the evaluation value with a preset threshold, and if the evaluation value exceeds the preset threshold, it is considered to be in an abnormal state and a first detection result is generated;

[0040] A model training module is used to obtain a training data set, wherein the training data set includes all states of the lithium battery system and their corresponding acoustic data, and uses the acoustic data as input and the corresponding state of the lithium battery system as output to train a preset neural network model to obtain an acoustic recognition model;

[0041] A model output module, used for inputting the acoustic data into the acoustic recognition model, the acoustic recognition model outputting a state corresponding to the lithium battery system, and generating a second detection result;

[0042] The result generating module is used to combine the first detection result and the second detection result to generate a final detection result.

[0043] In a third aspect, the present application provides an electronic device, comprising a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement a lithium battery system fault detection method based on digital twins as described in the first aspect.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium, in which a program is stored. When the program is executed by a processor, it is used to implement a lithium battery system fault detection method based on digital twins as described in the first aspect.

[0045] In summary, the beneficial effects of this application include at least:

[0046] 1) Take audio data and vibration data, ultrasonic data and acoustic emission data as inputs, use two independent acoustic recognition models to detect the status of the lithium battery system, and then record the output status of the two models together as the second detection result. Audio data and vibration data have different frequency ranges and signal characteristics. Processing these data separately can better extract their respective features. The same is true for ultrasonic data and acoustic emission data. Their frequency ranges and signal characteristics are different from audio and vibration data. Separate processing helps improve the accuracy of feature extraction.

[0047] 2) By using two independent models, the characteristics of different types of data can be captured more finely, thereby improving the overall detection accuracy. Each model focuses on processing a specific set of acoustic data, which helps to reduce errors caused by differences in data characteristics. In addition, audio data and vibration data mainly involve low-frequency and medium-frequency ranges, while ultrasonic data and acoustic emission data mainly involve high-frequency ranges. By processing these data separately, a wider frequency range can be covered, thereby more comprehensively monitoring the status of the lithium battery system. Audio and vibration data are easily interfered by environmental noise, while ultrasonic and acoustic emission data are not easily affected by environmental noise due to their higher frequencies. By combining the two sets of data, the system's resistance to environmental noise can be enhanced.

[0048] 3) SVD can extract the main features of the data. By integrating the information of the two test results, the system status can be considered more comprehensively, and the information from different data sources can be comprehensively utilized, thereby improving the comprehensiveness and accuracy of the final result. Singular value decomposition can reduce the dimensionality of the original data. By selecting an appropriate number of singular values ​​to reconstruct the data, the dimension of the data is reduced while retaining the main information, thereby reducing the complexity of data processing and improving computational efficiency.

[0049] This application uses multiple types of sensors to collect physical data and acoustic data of the lithium battery system, combining the application of digital twin technology and neural network models. First, multiple types of sensors are arranged in the lithium battery system to collect physical data including current, voltage, temperature, and acoustic data such as audio, vibration, and ultrasound. Then, the evaluation value of the lithium battery system is calculated using the physical data and compared with the preset threshold to detect abnormal conditions. At the same time, the acoustic data is trained through the neural network model and associated with the state of the lithium battery system to realize the acoustic recognition model. Finally, the detection results of the physical data and acoustic data are combined to generate the final detection results, so as to comprehensively monitor the state of the lithium battery system and identify potential faults in a timely manner. Through this comprehensive method, the detection capability of the abnormal state of the lithium battery system can be improved, thereby enhancing the reliability and safety of the system.

[0050] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present application in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of a lithium battery system fault detection method based on digital twin in an embodiment of the present application.

[0052] Figure 2 It is a flowchart of using the singular value decomposition method to generate the final detection result in an embodiment of the present application.

[0053] Figure 3 It is a schematic diagram of the overall process of the lithium battery system fault detection method based on digital twin in the embodiment of the present application.

[0054] Figure 4 It is a structural block diagram of a lithium battery system fault detection system based on digital twins in an embodiment of the present application.

[0055] Figure 5 It is a block diagram of an electronic device for lithium battery system fault detection based on digital twin in an embodiment of the present application. DETAILED DESCRIPTION

[0056] The specific implementation methods of the present application are further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application but are not intended to limit the scope of the present application.

[0057] Optionally, the present application takes the digital twin-based lithium battery system fault detection method provided in each embodiment as an example of being used in an electronic device, where the electronic device is a terminal or a server. The terminal may be a computer, a tablet computer, etc. This embodiment does not limit the type of electronic device.

[0058] Reference Figure 1 , is a flow chart of a lithium battery system fault detection method based on digital twins provided in an embodiment of the present application, the method comprising at least the following steps:

[0059] Step S101: multiple types of sensors are arranged in the lithium battery system to collect physical data and acoustic data of the lithium battery system.

[0060] In step S101, multiple types of sensors are arranged in the lithium battery system to collect physical data and acoustic data of the lithium battery system, wherein the physical data include but are not limited to current data, voltage data and temperature data. Optionally, the physical data may also include humidity data, pressure data, etc. In the embodiment of the present application, only current data, voltage data and temperature data are used as examples for illustration, and the present application does not limit the specific types of physical data. In implementation, in the lithium battery system, a current sensor is arranged at the positive and negative connecting lines of the lithium battery to collect current data for real-time monitoring of the charge and discharge current of the battery. A voltage sensor is arranged at the positive and negative poles of the lithium battery cell or battery pack to collect voltage data for real-time monitoring of the voltage change of the lithium battery. A temperature sensor is arranged on the surface of the lithium battery to collect temperature data for real-time monitoring of the temperature change of the lithium battery. The acoustic data include but are not limited to audio data, vibration data, ultrasonic data and acoustic emission data. In implementation, in the lithium battery system, a high-sensitivity acoustic sensor is arranged to collect audio data for collecting sound signals inside and around the battery, such as possible gas release sound inside the lithium battery, material friction sound, etc. Arrange high-sensitivity vibration sensors to collect vibration data, which are used to collect vibration signals inside and on the surface of lithium batteries. Arrange ultrasonic sensors to collect ultrasonic signals of changes in the internal structure of lithium batteries. Arrange acoustic emission sensors to capture high-frequency stress waves generated by crack expansion or other dynamic processes inside the material.

[0061] In addition, as a preferred embodiment, in the embodiment of the present application, after collecting the physical data and acoustic data of the lithium battery system, the collected physical data and acoustic data are preprocessed, including steps such as data cleaning, filtering, denoising, and outlier processing. Subsequently, the preprocessed physical data and acoustic data are used to build a digital twin model of the lithium battery system. In this application, a ready-made digital twin platform is directly selected, and a digital twin model of the lithium battery system is directly established on the selected digital twin platform. Real-time data is input into the digital twin platform through an API or data interface, and a database is used to store data for subsequent analysis and query. A real-time data visualization panel is created on the platform, including curve graphs, bar charts, etc. of data such as current, voltage, and temperature. The digital twin model is mainly used to monitor the operating status of the lithium battery system in real time, and provides a simple and intuitive data display interface to facilitate users to view and manage sensor data.

[0062] Step S102: Calculate an evaluation value of the lithium battery system based on the physical data, and compare the evaluation value with a preset threshold. If the evaluation value exceeds the preset threshold, it is considered to be in an abnormal state, and a first detection result is generated.

[0063] In step S102, based on the previously collected physical data, the evaluation value for the lithium battery system is calculated using the following formula:

[0064] in, is the current data, is the voltage data, For temperature data, the above formula multiplies the reciprocal of the current with the voltage divided by the square of the temperature to obtain a comprehensive evaluation value index for the lithium battery system. Current is a key parameter in the lithium battery system, which reflects the charge and discharge state of the battery. Generally, the larger the current, the heavier the load on the battery system, which may accelerate the aging of the system or cause other abnormal conditions. Therefore, incorporating the reciprocal of the current into the formula can make the evaluation value more sensitive when the current is large, making it easier to detect abnormal conditions. Voltage and temperature are two other important parameters that directly affect the performance and safety of the battery system. Generally, the voltage and temperature of the battery should be kept within a suitable range, and too high or too low may cause system abnormalities. Dividing the voltage by the square of the temperature can take into account the effect of temperature on the battery voltage, because too high or too low temperature will affect the voltage output of the battery. The introduction of the square term can make the evaluation value more sensitive when the temperature changes greatly, making it easier to detect abnormal conditions. Multiplying the reciprocal of the current and the voltage divided by the square of the temperature can comprehensively consider the relationship between current, voltage and temperature. The formula designed above can comprehensively consider the three key parameters of current, voltage and temperature, and calculate the evaluation value of the lithium battery system based on their combined indicators. By introducing nonlinear terms such as the reciprocal of current and voltage divided by the square of temperature, the calculation of the evaluation value can be made more sensitive and accurate. Therefore, the formula can more effectively detect the abnormal state of the lithium battery system, take corresponding measures to deal with the abnormal situation in time, thereby improving the reliability and safety of the system.

[0065] In implementation, after the evaluation value of the lithium battery system is calculated by the above formula, the evaluation value is compared with the preset threshold value. If the evaluation value exceeds the preset threshold value, the lithium battery system is judged to be in an abnormal state. Otherwise, the lithium battery system is judged to be in a normal state. At the same time, if the lithium battery system is in an abnormal state, it is necessary to judge the specific abnormal type. Specifically, the abnormal physical data is first analyzed in detail. Compare the numerical values ​​and change trends of the evaluation values ​​on different variables (current, voltage, temperature). Observe the change pattern of the evaluation value to preliminarily judge the type of abnormality. Then analyze the correlation between the evaluation value and other variables. Observe the relationship between the evaluation value and the current, voltage, and temperature to infer the main cause of the abnormality. For example, if the evaluation value is positively correlated with the current but not with the temperature, it may mean that the current abnormality is the main type of abnormality. In addition, the abnormal data can be compared and analyzed with the historical data. Check whether similar abnormal situations have ever occurred in the historical data, as well as the treatment methods and results at that time. By comparing historical data, the types and possible causes of abnormalities can be better understood. After determining the abnormality type through the above operation, the abnormal state, abnormality type and specific physical data are recorded as the first detection result. If it is in a normal state, only the normal state and specific physical data need to be recorded as the first detection result.

[0066] In another feasible embodiment, the evaluation value for the lithium battery system may also be calculated using the following formula:

[0067] In the above formula, the ratio of voltage to current is first included in the formula, which can better comprehensively consider the impact of voltage and current on the system state. The ratio of voltage to current reflects the resistance characteristics of the battery and can better describe the working state inside the battery. At the same time, the square root ratio of temperature to voltage is introduced to take into account the nonlinear relationship between temperature and voltage. This ratio can more accurately reflect the impact of temperature on voltage, especially when the voltage is low. In addition, the logarithmic function relationship between temperature and voltage is calculated to better capture the comprehensive impact of temperature and voltage on the system state. The characteristics of the logarithmic function can compress a larger range of changes into a smaller range, making the evaluation value change smoother under different temperature and voltage conditions. This logarithmic function relationship can not only enhance the detection sensitivity of the evaluation value under abnormal conditions, but also improve the stability and robustness of the evaluation value.

[0068] In addition, as preferably, in the embodiment of the present application, the above two formulas can also be used to calculate the evaluation value for the lithium battery system respectively, which are recorded as the first evaluation value and the second evaluation value respectively, and then the first evaluation value and the second evaluation value are added and averaged, and the average value is used as the final evaluation value, and finally the subsequent threshold comparison operation is performed. Since the two formulas use different parameter combinations and mathematical operations, they may reflect the state of the lithium battery system from different angles. Adding and averaging the evaluation values ​​calculated by the two formulas is equivalent to comprehensively considering the information of these two different aspects, so that the final evaluation value is more comprehensive and comprehensive. At the same time, each formula may have certain limitations, but averaging their results can reduce the impact of the deviation of individual formulas on the final evaluation results. This comprehensive evaluation can improve the accuracy of the evaluation and reduce the misjudgment caused by the lack of a single indicator. For example, the first formula is mainly based on the relationship between the three parameters of current, voltage and temperature for evaluation, but does not directly consider the resistance characteristics inside the battery, and does not perform special processing on the relationship between temperature and voltage, which may lack sufficient sensitivity when the voltage is low. The second formula introduces the ratio of voltage to current, which is equivalent to considering the resistance inside the battery, so that the state of the battery system can be more comprehensively evaluated. The square root ratio of temperature to voltage is also introduced. This processing method can improve the sensitivity of the evaluation value when the voltage is low and better reflect the abnormal conditions of the battery system. In addition, by combining the evaluation results of multiple indicators, the false alarm rate can be reduced, that is, the possibility of misjudging normal conditions as abnormal conditions can be reduced. Because different indicators often produce errors in different situations, comprehensive consideration of these indicators can reduce the accumulation of such errors, thereby improving the credibility of the evaluation.

[0069] In addition, as a preferred embodiment of the present application, after using any of the above methods to calculate the evaluation value of the lithium battery system, the evaluation value can also be corrected. Specifically, first obtain the historical physical data of the past period of time. The past period of time in this application is generally three hours. Through the historical physical data of the past period of time, obtain the historical current average, historical voltage average and historical temperature average. Based on the historical current average, historical voltage average and historical temperature average, calculate the historical current change rate , Historical voltage change rate And the historical temperature change rate , the calculation formula is as follows:

[0070] Calculate the historical current change rate , Historical voltage change rate And the historical temperature change rate Afterwards, the average rate of change of historical physical data is calculated using the following formula: :

[0071] Finally, according to the average change rate of historical physical data Correct the previously calculated evaluation value:

[0072] Through the calculation and correction process of the above formula, the changing trend of the physical data of the lithium battery system in the past period of time is further considered, so that the evaluation value is adjusted more objectively. By calculating the rate of change of historical physical data, it is possible to understand the state change trend of the system in the past period of time. This enables the present application to more accurately evaluate the state of the current system because the system's performance in the past is taken into account. The average rate of change of historical data is used for correction so that the evaluation value is based more on objective data rather than subjective judgment, which helps to reduce the impact of personal subjective factors on the evaluation results and improve the objectivity and credibility of the evaluation.

[0073] Step S103, obtaining a training data set, the training data set includes all states of the lithium battery system and their corresponding acoustic data, taking the acoustic data as input and the corresponding state of the lithium battery system as output, training a preset neural network model to obtain an acoustic recognition model.

[0074] In step S103, firstly, a first training data set is obtained, and the first training data set includes all states of the lithium battery system and their corresponding audio data and vibration data. A certain state of the lithium battery system is used as the output, and the feature vector extracted and fused from the corresponding audio data and vibration data is used as the input, and the preset neural network model is trained to obtain a first acoustic recognition model. Optionally, in the embodiment of the present application, a recurrent neural network is selected as the preset neural network model, and the acoustic data is usually time-series, for example, the audio data and the vibration data are both changing over time. RNN can effectively process such time series data because they retain previous information when processing each time step, which makes them very suitable for capturing long-term dependencies in the sequence. Specifically, audio signal processing technology is used to extract the feature vector of the audio data, such as short-time Fourier transform, Mel frequency cepstral coefficient and other technologies, and the audio data is converted into a feature vector with lower dimension and richer information. The feature vector of the vibration data is extracted using methods such as time domain analysis, frequency domain analysis or time-frequency domain analysis. After extracting the feature vectors of the audio data and the vibration data, the feature vectors of the two are feature-fused using the following formula:

[0075] in is the feature vector of audio data, The feature vector of the vibration data, the dot product of the feature vector can multiply the information of the two vectors element by element, and add the results, so as to integrate the information of the audio data and the vibration data. The square root operation can map the result to a smaller range to maintain the stability and consistency of the data. Optionally, the cosine similarity calculation method can also be used to fuse the feature vectors of the audio data and the vibration data, and this application does not limit the fusion method of the feature vectors.

[0076] In step S103, a second training data set is then obtained, and the second training data set includes all states of the lithium battery system and their corresponding ultrasonic data and acoustic emission data. A certain state of the lithium battery system is used as the output, and the corresponding ultrasonic data and acoustic emission data are extracted and fused. The feature vector is used as the input to train the preset neural network model to obtain a second acoustic recognition model. Optionally, a recurrent neural network is selected as the preset neural network model in an embodiment of the present application. Acoustic data is usually time-series, for example, ultrasonic data and acoustic emission data change over time. RNN can effectively process such time series data because they retain previous information when processing each time step, which makes them very suitable for capturing long-term dependencies in the sequence. Specifically, the feature vectors of ultrasonic data and acoustic emission data are extracted using methods such as time domain analysis, frequency domain analysis, or time-frequency domain analysis. After extracting the feature vectors of ultrasonic data and acoustic emission data, the feature vectors of the two are feature-fused using the following formula:

[0077] in, is the feature vector of ultrasonic data, is the feature vector of acoustic emission data. The above formula achieves the averaging of the information of both ultrasonic data and acoustic emission data by adding and averaging the feature vectors of the ultrasonic data and acoustic emission data. This method can ensure that the two types of data are treated equally in the feature fusion process without favoring either type of data. The method of adding and averaging can integrate the information of ultrasonic data and acoustic emission data to form a comprehensive feature vector. Doing so can enable the model to comprehensively consider information from different data sources and improve the comprehensiveness and accuracy of the model. The square root operation can map the result to a smaller range to maintain the stability and consistency of the data. Optionally, the cosine similarity calculation method can also be used to fuse the feature vectors of ultrasonic data and acoustic emission data. This application does not limit the fusion method of the feature vectors.

[0078] Step S104: input the acoustic data into the acoustic recognition model, the acoustic recognition model outputs the state corresponding to the lithium battery system, and generates a second detection result.

[0079] In step S104, after the feature vector extracted and fused from the audio data and the vibration data is input into the first acoustic recognition model, the first acoustic recognition model outputs the first state of the corresponding lithium battery system. After the feature vector extracted and fused from the ultrasonic data and the acoustic emission data is input into the second acoustic recognition model, the second acoustic recognition model outputs the second state of the corresponding lithium battery system. The first state and the second state of the lithium battery system are recorded together as the second detection result. The audio data and the vibration data, the ultrasonic data and the acoustic emission data are respectively used as inputs, and two independent acoustic recognition models are used to detect the state of the lithium battery system, and then the output states of the two models are recorded together as the second detection result. The audio data and the vibration data have different frequency ranges and signal characteristics. Processing these data separately can better extract their respective features. The same is true for ultrasonic data and acoustic emission data. Their frequency ranges and signal characteristics are different from those of audio and vibration data. Separate processing helps to improve the accuracy of feature extraction. At the same time, by using two independent models, the features of different types of data can be captured more finely, thereby improving the overall detection accuracy. Each model focuses on processing a specific set of acoustic data, which helps to reduce errors caused by differences in data characteristics. In addition, audio data and vibration data mainly involve low-frequency and medium-frequency ranges, while ultrasonic data and acoustic emission data mainly involve high-frequency ranges. By processing these data separately, a wider frequency range can be covered, thereby more comprehensively monitoring the status of the lithium battery system. Audio and vibration data are easily interfered by environmental noise, while ultrasonic and acoustic emission data are not easily affected by environmental noise due to their higher frequencies. By combining the two sets of data, the system's resistance to environmental noise can be enhanced.

[0080] In addition, as a preferred embodiment, in the present application, a loss function is designed for the acoustic recognition model, and based on the loss function, the acoustic recognition model parameters are updated using the gradient descent method until the loss function is minimized. as follows:

[0081] in, represents the loss function based on time information, represents the loss function based on spectral information, and is the weight parameter used to balance the two types of information. In the formula, represents the length of the sequence, and Represents the predicted sequence and the true sequence at time step The value of . In the formula, and Represent the spectral distribution of the predicted sequence and the true sequence respectively. Acoustic data usually has temporal and spectral characteristics. Temporal characteristics mean that acoustic data is continuous and correlated in time. Therefore, considering temporal information in the loss function can help the model better understand the time series structure of the data. The spectral characteristics reflect the distribution of acoustic data in the frequency domain. Information of different frequencies is also important for acoustic tasks. Therefore, considering spectral information can help the model better understand the frequency characteristics of acoustic data. Combining temporal information and spectral information in the loss function can comprehensively utilize data features at different levels. The temporal information loss function focuses on the temporal changes of acoustic data, while the spectral information loss function focuses on the spectral distribution of acoustic data. The combination of the two can guide the training of the model more comprehensively. By designing a loss function that comprehensively considers temporal and spectral characteristics, the model can learn the characteristics of acoustic data more comprehensively during the training process and better fit the real lithium battery system state. This helps to improve the performance and generalization ability of the acoustic recognition model.

[0082] In step S105, the first detection result and the second detection result are combined to generate a final detection result.

[0083] In addition, as a preferred embodiment, in the embodiment of the present application, a singular value decomposition method can be used to combine the first detection result and the second detection result to generate a final detection result, referring to Figure 2 , the method comprises at least the following steps:

[0084] S1051. Construct a matrix based on the first detection result and the second detection result.

[0085] Specifically, the first detection result and the second detection result are first represented as two matrices and In this embodiment of the present application, it is assumed that the matrix The dimension is , matrix The dimension is ,in is the number of states of the first detection result, is the number of states of the second test result, is the number of features. To combine the two results, we stack the two matrices together to get a The matrix .

[0086] S1052. Perform singular value decomposition on the constructed matrix.

[0087] In the implementation, the constructed matrix is ​​processed using the standard SVD algorithm. Perform singular value decomposition and get three matrices , , ,in is a The orthogonal matrix of is a The diagonal matrix of is a An orthogonal matrix.

[0088] S1053. Select an appropriate number of singular values ​​to reconstruct the matrix.

[0089] Specifically, from Select an appropriate number of singular values, usually the first few singular values, where is the number of singular values ​​selected, and then the matrix is ​​reconstructed by the following formula :

[0090] in, yes The matrix of yes The diagonal matrix of yes The matrix of .

[0091] S1054. Extract the final detection result from the reconstructed matrix.

[0092] Specifically, from the reconstructed matrix Select specific rows or columns as the final test results. This can be determined according to the actual situation, which may be certain rows or columns of the entire matrix, or certain specific combinations of rows or columns. SVD can extract the main features of the data. By integrating the information of the two test results, the state of the system can be considered more comprehensively, and information from different data sources can be comprehensively utilized, thereby improving the comprehensiveness and accuracy of the final result. Singular value decomposition can reduce the dimensionality of the original data. By selecting an appropriate number of singular values ​​to reconstruct the data, the dimension of the data is reduced while retaining the main information, thereby reducing the complexity of data processing and improving computational efficiency.

[0093] In summary, combined with Figure 3, this application utilizes multiple types of sensors to collect physical data and acoustic data of the lithium battery system, combining the application of digital twin technology and neural network models. First, multiple types of sensors are arranged in the lithium battery system to collect physical data including current, voltage, temperature, and acoustic data such as audio, vibration, and ultrasound. Then, the evaluation value of the lithium battery system is calculated using the physical data and compared with the preset threshold to detect abnormal conditions. At the same time, the acoustic data is trained through a neural network model and associated with the state of the lithium battery system to realize an acoustic recognition model. Finally, the detection results of the physical data and acoustic data are combined to generate the final detection results, thereby comprehensively monitoring the state of the lithium battery system and identifying potential faults in a timely manner. Through this comprehensive approach, the ability to detect abnormal conditions of the lithium battery system can be improved, thereby enhancing the reliability and safety of the system.

[0094] Figure 4 This is a structural block diagram of a lithium battery system fault detection system based on digital twins provided by an embodiment of the present application. The device includes at least the following modules:

[0095] A data acquisition module is used to arrange multiple types of sensors in the lithium battery system to collect physical data and acoustic data of the lithium battery system;

[0096] A threshold comparison module is used to calculate an evaluation value of the lithium battery system based on physical data, and compare the evaluation value with a preset threshold value. If the evaluation value exceeds the preset threshold value, it is considered to be in an abnormal state and a first detection result is generated;

[0097] A model training module is used to obtain a training data set, which includes all states of the lithium battery system and their corresponding acoustic data. The acoustic data is used as input and the corresponding state of the lithium battery system is used as output to train a preset neural network model to obtain an acoustic recognition model.

[0098] A model output module, used to input the acoustic data into the acoustic recognition model, the acoustic recognition model outputs the state corresponding to the lithium battery system, and generates a second detection result;

[0099] The result generation module is used to combine the first detection result and the second detection result to generate a final detection result.

[0100] For relevant details, refer to the above method embodiment.

[0101] Figure 5 4 is a block diagram of an electronic device provided by an embodiment of the present application. The device at least includes a processor 401 and a memory 402.

[0102] The processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 401 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 401 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 401 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 401 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0103] The memory 402 may include one or more computer-readable storage media, which may be non-transitory. The memory 402 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 402 is used to store at least one instruction, which is used to be executed by the processor 401 to implement the lithium battery system fault detection method based on digital twins provided in the method embodiment of the present application.

[0104] In some embodiments, the electronic device may further optionally include: a peripheral device interface and at least one peripheral device. The processor 401, the memory 402 and the peripheral device interface may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface via a bus, a signal line or a circuit board. Schematically, the peripheral devices include but are not limited to: a radio frequency circuit, a touch display screen, an audio circuit, and a power supply.

[0105] Of course, the electronic device may also include fewer or more components, which is not limited in this embodiment.

[0106] Optionally, the present application also provides a computer-readable storage medium, in which a program is stored, and the program is loaded and executed by a processor to implement the lithium battery system fault detection method based on digital twins of the above method embodiment.

[0107] Optionally, the present application also provides a computer product, which includes a computer-readable storage medium, in which a program is stored, and the program is loaded and executed by a processor to implement the lithium battery system fault detection method based on digital twins of the above method embodiment.

[0108] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0109] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A lithium battery system fault detection method based on digital twin, characterized in that: The method comprises: Arranging multiple types of sensors in the lithium battery system to collect physical data and acoustic data of the lithium battery system; the physical data includes current data, voltage data and temperature data, and the acoustic data includes audio data, vibration data, ultrasonic data and acoustic emission data; Calculating an evaluation value of the lithium battery system based on the physical data, comparing the evaluation value with a preset threshold, and if the evaluation value exceeds the preset threshold, considering the system to be in an abnormal state, and generating a first detection result; A training data set is obtained, wherein the training data set includes all states of the lithium battery system and their corresponding acoustic data, and the acoustic data is used as input and the corresponding state of the lithium battery system is used as output to train a preset neural network model to obtain an acoustic recognition model, including: Acquire a first training data set, the first training data set includes all states of the lithium battery system and their corresponding audio data and vibration data, take a certain state of the lithium battery system as output, take the feature vector extracted and fused from the corresponding audio data and vibration data as input, train a preset neural network model to obtain a first acoustic recognition model; After extracting the feature vectors of the audio data and vibration data, the feature vectors of the two are fused using the following formula: Obtain a second training data set, the second training data set includes all states of the lithium battery system and their corresponding ultrasonic data and acoustic emission data, take a certain state of the lithium battery system as output, take the feature vector extracted and fused from the corresponding ultrasonic data and acoustic emission data as input, train the preset neural network model to obtain a second acoustic recognition model; After extracting the feature vectors of ultrasonic data and acoustic emission data, the feature vectors of the two are fused using the following formula: Inputting the acoustic data into the first acoustic recognition model and the second acoustic recognition model, and recording the output states of the first acoustic recognition model and the second acoustic recognition model together as a second detection result; The first detection result and the second detection result are combined to generate a final detection result.

2. The lithium battery system fault detection method based on digital twin according to claim 1 is characterized in that: Calculating the evaluation value of the lithium battery system based on the physical data includes: The first evaluation value and the second evaluation value for the lithium battery system are calculated using the following two formulas: The first evaluation value and the second evaluation value are added together to obtain an average value.

3. The lithium battery system fault detection method based on digital twin according to claim 2 is characterized in that: After calculating the evaluation value of the lithium battery system based on the physical data, the method further includes: Obtain historical physical data for a period of time. Obtain historical current averages, historical voltage averages, and historical temperature averages through historical physical data for a period of time. Calculate historical current change rates K1, historical voltage change rates K2, and historical temperature change rates K3 based on the historical current averages, historical voltage averages, and historical temperature averages. The calculation formulas are as follows: After calculating the historical current change rate K1, the historical voltage change rate K2, and the historical temperature change rate K3, the average change rate K of the historical physical data is calculated using the following formula: Correct the previously calculated evaluation value based on the average change rate K of the historical physical data: 。 4. The lithium battery system fault detection method based on digital twin according to claim 1, characterized in that: Combining the first detection result and the second detection result to generate a final detection result includes: constructing a matrix based on the first detection result and the second detection result; Perform singular value decomposition on the constructed matrix; Select an appropriate number of singular values ​​to reconstruct the matrix; The final detection results are extracted from the reconstructed matrix.

5. A lithium battery system fault detection system based on digital twins, characterized in that: include: A data acquisition module, used to arrange multiple types of sensors in the lithium battery system to collect physical data and acoustic data of the lithium battery system; the physical data includes current data, voltage data and temperature data, and the acoustic data includes audio data, vibration data, ultrasonic data and acoustic emission data; A threshold comparison module, used to calculate an evaluation value of the lithium battery system based on the physical data, compare the evaluation value with a preset threshold, and if the evaluation value exceeds the preset threshold, it is considered to be in an abnormal state and a first detection result is generated; The model training module is used to obtain a training data set, wherein the training data set includes all states of the lithium battery system and their corresponding acoustic data, and uses the acoustic data as input and the corresponding state of the lithium battery system as output to train a preset neural network model to obtain an acoustic recognition model, including: Acquire a first training data set, the first training data set includes all states of the lithium battery system and their corresponding audio data and vibration data, take a certain state of the lithium battery system as output, take the feature vector extracted and fused from the corresponding audio data and vibration data as input, train a preset neural network model to obtain a first acoustic recognition model; After extracting the feature vectors of the audio data and vibration data, the feature vectors of the two are fused using the following formula: Obtain a second training data set, the second training data set includes all states of the lithium battery system and their corresponding ultrasonic data and acoustic emission data, take a certain state of the lithium battery system as output, take the feature vector extracted and fused from the corresponding ultrasonic data and acoustic emission data as input, train the preset neural network model to obtain a second acoustic recognition model; After extracting the feature vectors of ultrasonic data and acoustic emission data, the feature vectors of the two are fused using the following formula: A model output module, used for inputting the acoustic data into the first acoustic recognition model and the second acoustic recognition model, and recording the output states of the first acoustic recognition model and the second acoustic recognition model together as a second detection result; The result generating module is used to combine the first detection result and the second detection result to generate a final detection result.

6. An electronic device, characterized in that: The device includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement a lithium battery system fault detection method based on digital twins as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The storage medium stores a program, which, when executed by a processor, is used to implement a lithium battery system fault detection method based on digital twins as described in any one of claims 1 to 4.

Citation Information

Patent Citations

  • Fault diagnosis method and alarm system based on digital twin lithium battery system

    CN116184293A

  • Bearing fault detection system and method

    CN118408743A

  • Sensor data anomaly detection method and device, equipment and storage medium

    CN118425863A

  • Method and system for predicting remaining service life of lithium battery

    WO2022257211A1